Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks490 from public data
- Named entity recognition using an HMM-based chunk tagger809
A Hidden Markov Model and an HMM-based chunk tagger is proposed, from which a named entity (NE) recognition system is built to recognize and classify names, times and numerical quantities, and the NER problem can be resolved effectively.
- Using Cross-Entity Inference to Improve Event Extraction294
This paper proposes a new method of event extraction by well using cross-entity inference that regard entity-type consistency as key feature to predict event mentions and adopts this inference method to improve the traditional sentence-level event extraction system.
- Recognizing names in biomedical texts: a machine learning approach287
The PowerBioNE system is the first system which deals with the cascaded entity name phenomenon and the HMM and the k-NN algorithm outperform other models, such as back-off HMM, linear interpolated H MM, support vector machines, C4.5 rules and RIPPER, by effectively capturing the local context dependency and resolving the data sparseness problem.
- A composite kernel to extract relations between entities with both flat and structured features286
This study illustrates that the composite kernel can effectively capture both flat and structured features without the need for extensive feature engineering, and can also easily scale to include more features.
- Multi-criteria-based active learning for named entity recognition261
A multi-criteria-based active learning approach is proposed and effectively applied to named entity recognition and includes all the criteria using two selection strategies, both of which result in less labeling cost than single-criterion-based method.
- RETRACTED: Multi-modal Graph Fusion for Named Entity Recognition with Targeted Visual Guidance259
A unified multi-modal graph fusion (UMGF) approach for MNER that achieves an attention-based multi- modal representation for each word and performs entity labeling with a CRF decoder.
- Modeling both Context- and Speaker-Sensitive Dependence for Emotion Detection in Multi-speaker Conversations239
This paper proposes a conversational graph-based convolutional neural network that represents each utterance and each speaker as a node and can be symbolized as a large heterogeneous graph and the emotion detection task can be recast as a classification problem of the utterance nodes in the graph.
- Tree Kernel-Based Relation Extraction with Context-Sensitive Structured Parse Tree Information221
Evaluation on the ACE RDC corpora shows that the dynamic context-sensitive tree span is much more suitable for relation extraction than SPT and the tree kernel outperforms the state-of-the-art Collins and Duffy’s convolution tree kernel.
- Semi-Supervised Learning for Imbalanced Sentiment Classification191
This paper investigates a more common case of semi-supervised learning for imbalanced sentiment classification, in which various random subspaces are dynamically generated to deal with the imbalanced class distribution problem.
- Effective adaptation of a Hidden Markov Model-based named entity recognizer for biomedical domain169
A general Hidden Markov Model-based named entity recognizer effectively to biomedical domain is adapted and a simple algorithm to solve the abbreviation problem and a rule-based method to deal with the cascaded phenomena in biomedical domain are presented.
- Joint Multi-modal Aspect-Sentiment Analysis with Auxiliary Cross-modal Relation Detection168
This paper is the first to jointly perform multi-modal ATE (MATE) and multi- modal ASC (MASC), and it proposes a multi-Modal joint learning approach with auxiliary cross-modAL relation detection for multi- Modal aspect-level sentiment analysis (MALSA).
- Coreference resolution using competition learning approach164
The experimental results show that the proposed competition learning approach to coreference resolution can outperform those based on the single-candidate model and applies a candidate filter to reduce the computational cost and data noises during training and resolution.
- Modeling Source Syntax for Neural Machine Translation150
This paper shows that source syntax can be explicitly incorporated into NMT effectively to provide further improvements, and linearize parse trees of source sentences to obtain structural label sequences.
- Sentiment Classification and Polarity Shifting142
A feature selection method is adopted to automatically generate the training data for a binary classifier on polarity shifting detection of sentences, and each document in the original polarity classification training data is split into two partitions which are used to train two base classifiers respectively for further classifier combination.
- Exploiting constituent dependencies for tree kernel-based semantic relation extraction140
Evaluation on the ACE RDC 2004 corpus shows that the dynamic syntactic parse tree outperforms all previous tree spans, and the composite kernel combining this tree kernel with a linear state-of-the-art feature-based kernel, achieves the so far best performance.
- Enhancing HMM-based biomedical named entity recognition by studying special phenomena128
This work proposes a rich set of features, including orthographic, morphological, part-of-speech, and semantic trigger features, integrated via a Hidden Markov Model with back-off modeling for biomedical abbreviation recognition and cascaded named entity recognition.
- Stance Detection with Hierarchical Attention Network125
This work presents a neural model to fully employ various linguistic information to construct the document representation and proposes a hierarchical attention network to weigh the importance ofVarious linguistic information, and learn the mutual attention between the document and the linguistic information.
- Employing Personal/Impersonal Views in Supervised and Semi-Supervised Sentiment Classification124
An ensemble method and a co-training algorithm are explored to employ the two views, personal and impersonal views, in both supervised and semi-supervised sentiment classification respectively.
- Recognition of protein/gene names from text using an ensemble of classifiers109
An ensemble of classifiers for biomedical name recognition in which three classifiers, one Support Vector Machine and two discriminative Hidden Markov Models, are combined effectively using a simple majority voting strategy is proposed.
- Aspect Sentiment Classification with both Word-level and Clause-level Attention Networks107
The need for incorporating the importance degrees of both words and clauses inside a sentence is highlighted and a hierarchical network with both word-level and clause-level attentions to aspect sentiment classification is proposed.
- Modeling Graph Structure in Transformer for Better AMR-to-Text Generation105
This paper proposes a novel structure-aware self-attention approach to better model the relations between indirectly connected concepts in the state-of-the-art seq2seq model, i.e. the Transformer.
- Cache-based Document-level Statistical Machine Translation101
This paper presents three kinds of caches to store relevant document-level information: a dynamic cache, which stores bilingual phrase pairs from the best translation hypotheses of previous sentences in the test document; a static cache,which stores relevantilingual phrase pairs extracted from similar bilingual document pairs in the training parallel corpus; and a topic cache,Which stores the target-side topic words related with the test documents in the source-side.
- Chemical-induced disease relation extraction via convolutional neural network98
This article describes the work on the BioCreative-V chemical–disease relation (CDR) extraction task, which employed a maximum entropy (ME) model and a convolutional neural network model for relation extraction at inter- and intra-sentence level, respectively.
- Active Learning for Imbalanced Sentiment Classification95
This paper proposes a novel active learning approach, named co-selecting, that employs two feature subspace classifiers to collectively select most informative minority-class samples for manual annotation by leveraging a certainty measurement and an uncertainty measurement, to reduce human-annotation efforts.
- Building Chinese Discourse Corpus with Connective-driven Dependency Tree Structure94
Preliminary evaluation justifies the appropriateness of the CDT scheme to Chinese discourse analysis and the usefulness of the manually annotated CDTB corpus.
- An NP-cluster based approach to coreference resolution94
This paper proposes a supervised learning-based approach which does coreference resolution by exploring the relationships between NPs and coreferential clusters and shows that this approach outperforms the baseline NP-NP based approach in both recall and precision.
- Discovering Relations Between Named Entities from a Large Raw Corpus Using Tree Similarity-Based Clustering88
This work proposes a tree-similarity-based unsupervised learning method to extract relations between Named Entities from a large raw corpus and is the first work to use a tree similarity metric in relation clustering.
- Tree kernel-based semantic relation extraction with rich syntactic and semantic information87
An evaluation on the Automatic Content Extraction/Relation Detection and Characterization (ACE RDC) corpora shows that the proposed tree kernel-based method outperforms other state-of-the-art methods.
- Speculation and Negation Scope Detection via Convolutional Neural Networks83
A Convolutional Neural Network (CNN)-based model with probabilistic weighted average pooling to address speculation and negation scope detection and extracts meaningful features from various syntactic paths be-tween the cues and the candidate tokens in both constituency and dependency parse trees.
- 82
- Exploring syntactic structured features over parse trees for relation extraction using kernel methods78
A composite kernel for relation extraction by combining the convolution tree kernel with a simple linear kernel is proposed that can effectively capture both flat and structured features without extensive feature engineering, and easily scale to include more features.
- An Emotion Cause Corpus for Chinese Microblogs with Multiple-User Structures76
This study presents a pilot study of emotion cause detection which deals with Chinese microblogs using a complicated structure and chooses two popular classification methods (SVM and LSTM) to do emotion cause Detection.
- A Tree Kernel-Based Unified Framework for Chinese Zero Anaphora Resolution75
Experimental results on a Chinese zero anaphora corpus show that the proposed tree kernel-based methods significantly outperform the feature-based ones, which indicates the critical role of the structural information in zeroAnaphora resolution and the necessity of tree kernel -based methods in modeling such structural information.
- Protein-Protein Interaction Extraction: A Supervised Learning Approach}73
This paper proposes using Maximum Entropy to extract protein-protein interaction information from the literature, which overcomes the limitation of the state of art co-occurrence based and rule-based approaches and incorporates corpus statistics of various lexical, syntactic and semantic features.
- Self-regulation: Employing a Generative Adversarial Network to Improve Event Detection72
A self-regulated learning approach is proposed by utilizing a generative adversarial network to generate spurious features and employing a recurrent network to eliminate the fakes.
- Improving AMR Parsing with Sequence-to-Sequence Pre-training71
This paper proposes a seq2seq pre-training approach to build pre-trained models in both single and joint way on three relevant tasks, i.e., machine translation, syntactic parsing, and AMR parsing itself, and extends the vanilla fine-tuning method to a multi-task learning fine- Tuning method that optimizes for the performance of AMR parse while endeavors to preserve the response of pre- trained models.
- Aspect Sentiment Classification with Document-level Sentiment Preference Modeling71
A Cooperative Graph Attention Networks (CoGAN) approach for cooperatively learning the aspect-related sentence representation and the importance of the document-level sentiment preference information to ASC is justified and the effectiveness of the approach capturing such information is demonstrated.
- Transformer-based Label Set Generation for Multi-modal Multi-label Emotion Detection69
A unified approach, namely multi-modal emotion set generation network (MESGN) to generate an emotion set for an utterance and shows that the proposed MESGN architecture can effectively improve the performance of multi- modal multi-label emotion detection.
- Chemical-induced disease relation extraction with various linguistic features69
A machine learning based system that utilized simple yet effective linguistic features to extract relations with maximum entropy models and the hypernym relations between entity concepts derived from the Medical Subject Headings (MeSH)-controlled vocabulary to obtain more accurate classification models and better extraction performance.
- Sentence-level Emotion Classification with Label and Context Dependence69
This paper recast sentence-level emotion classification as a factor graph inferring problem in which the label and context dependence are modeled as various factor functions.
- Hierarchical Modeling of Global Context for Document-Level Neural Machine Translation68
This paper proposes a hierarchical model to learn the global context for document-level neural machine translation (NMT), which feedback the extracted global document context to each word in a top-down fashion to distinguish different translations of a word according to its specific surrounding context.
- Active learning for cross-domain sentiment classification67
This paper proposes an novel active learning approach for cross-domain sentiment classification by actively selecting a small amount of labeled data in the target domain by fully exploiting the unlabeled data in the target domain with the label propagation algorithm.
- Cross language dependency parsing using a bilingual lexicon67
It is shown that a translated English treebank helps a Chinese parser obtain a state-of-the-art result.
- Multilingual dependency learning: a huge feature engineering method to semantic dependency parsing65
It is illustrated that semantic dependency parsing can be transformed into a word-pair classification problem and implemented as a single-stage machine learning system.
- A bilingual attention network for code-switched emotion prediction64
A Bilingual Attention Network (BAN) model is proposed to aggregate the monolingual and bilingual informative words to form vectors from the document representation, and integrate the attention vectors to predict the emotion.
- Extracting relation information from text documents by exploring various types of knowledge61
This study illustrates that the base phrase chunking information is very effective for relation extraction and contributes to most of the performance improvement from syntactic aspect while current commonly used features from full parsing give limited further enhancement, suggesting that a cheap and robust solution in relation extraction can be achieved without decreasing too much in performance.
- Multi-modal Multi-label Emotion Detection with Modality and Label Dependence60
This paper proposes a multi-modal sequence-to-set approach to effectively model both kinds of dependence in multi- modal multi-label emotion detection.
- What reviews are satisfactory60
This paper proposes that user preferences are more explicit clues to infer the opinions of users on the review helpfulness, and implements a binary helpfulness based review classification system to divide helpful reviews and useless, and builds a Ranking SVM based automatic helpfulness voting system (AHV) which rank reviews based on their helpfulness.
- Tree kernel-based protein–protein interaction extraction from biomedical literature60
This paper proposes a novel approach to tree kernel-based PPI extraction, where the tree representation generated from a constituent syntactic parser is further refined using the shortest dependency path between two proteins derived from a dependency parser.
- Multi-modal Multi-label Emotion Recognition with Heterogeneous Hierarchical Message Passing59
This paper proposes a heterogeneous hierarchical message passing network to effectively model above dependencies in a multi-label scenario and proposes a new multi-modal multi- label emotion dataset based on partial time-series content to show predominant generalization of the model.
- Employing Compositional Semantics and Discourse Consistency in Chinese Event Extraction56
This paper proposes two novel inference mechanisms to explore special characteristics in Chinese via compositional semantics inside Chinese triggers and discourse consistency between Chinese trigger mentions.
- Improving Noun Phrase Coreference Resolution by Matching Strings55
A noun phrase coreference resolution system which aims to enhance the identification of the coreference realized by string matching by introducing an additional set of features to capture the different matching patterns between noun phrases.
- Tree Kernel-based Negation and Speculation Scope Detection with Structured Syntactic Parse Features53
A new approach for tree kernel-based scope detection by using the structured syntactic parse information and the way of selecting compatible features for different part-of-speech cues is explored.
- Learning the Scope of Negation via Shallow Semantic Parsing52
Evaluation on the BioScope corpus shows that structured syntactic information is effective in capturing the domination relationship between a negation signal and its dominated arguments, and shows that the simplified shallow semantic parsing approach much outperforms the state-of-the-art chunking ones.
- A Discrete CVAE for Response Generation on Short-Text Conversation51
A discrete latent variable with an explicit semantic meaning to improve the CVAE on short-text conversation and proposes a two-stage sampling approach to enable efficient diverse variable selection from a large latent space assumed in the short- text conversation task.
- Stance detection via sentiment information and neural network model50
This work proposes a joint neural network model to predict the stance and sentiment of a post simultaneously, and uses a neural stacking model to leverage sentimental information for the stance detection task.
- Recognizing names in biomedical texts using mutual information independence model and SVM plus sigmoid48
A support vector machine (SVM) plus sigmoid is proposed to resolve the data sparseness problem in the MIIM and a biomedical name recognition system with better performance and better portability can be achieved.
- Independent spanning trees on twisted cubes46
The vertex conjecture is confirmed for the n-dimensional twisted cube TQ"n by providing an O(NlogN) algorithm to construct n vertex-independent spanning trees rooted at any vertex, where N denotes the number of vertices in T Q"n.
- Document re-ranking using cluster validation and label propagation46
A novel document re-ranking approach in information retrieval, which is done by a label propagation-based semi-supervised learning algorithm to utilize the intrinsic structure underlying in the large document data.
- Joint learning on sentiment and emotion classification43
An extra data set that is annotated with both sentiment and emotion labels are employed to estimate the transformation probability between the two kinds of labels and this is leveraged to transfer the classification labels to benefit the two tasks from each other.
- Sentiment Classification in Customer Service Dialogue with Topic-Aware Multi-Task Learning42
This study proposes a topic-aware multi-task learning (TML) approach which learns topic-enriched utterance representations in customer service dialogue by capturing various kinds of topic information.
- A Top-down Neural Architecture towards Text-level Parsing of Discourse Rhetorical Structure42
This paper casts discourse parsing as a recursive split point ranking task, where a split point is classified to different levels according to its rank and the elementary discourse units associated with it are arranged accordingly, and proposes a top-down neural architecture toward text-level DRS parsing.
- Aspect Sentiment Classification Towards Question-Answering with Reinforced Bidirectional Attention Network41
A Reinforced Bidirectional Attention Network (RBAN) approach is proposed to address two inherent challenges in ASC-QA, i.e., semantic matching between question and answer, and data noise.
- Improving pronoun resolution by incorporating coreferential information of candidates41
A trainable model which incorporates coreferential information of candidates into pronoun resolution is presented and preliminary experiments show that the model will boost the resolution performance given the right antecedents of the candidates.
- Joint Syntactic and Semantic Parsing of Chinese39
Evaluation on Chinese TreeBank, Chinese PropBank, and Chinese NomBank shows that the integrated parsing approach outperforms the pipeline parsing approach on n-best parse trees, a natural extension of the widely used pipeline parse approach on the top- best parse tree.
- Diagnosable evaluation of DCC linear congruential graphs under the PMC diagnostic model39
It was proved that G(F,n) is a 2t-regular graph and has connectivity 2t, where t=|F| and 1==3 and 2=2.
- Imbalanced sentiment classification38
Evaluation across different datasets shows the effectiveness of both the under-sampling framework and the smoothing strategy in handling the imbalanced problems in real sentiment classification applications.
- 37
- Multimodal Topic-Enriched Auxiliary Learning for Depression Detection36
This work argues that the global topic information derived from user-generated contents is crucial to boost the performance of the depression detection task, and proposes a new Multimodal Topic-enriched Auxiliary Learning (MTAL) approach, aiming at capturing the topic information inside different modalities for depression detection.
- Improving nominal SRL in Chinese language with verbal SRL information and automatic predicate recognition36
Evaluation on Chinese NomBank shows that research in integrating various features derived from verbal SRL significantly improves the performance and also shows that the nominal SRL system much outperforms the state-of-the-art ones.
- Adversarial Attention Modeling for Multi-dimensional Emotion Regression35
A neural network-based approach, namely Adversarial Attention Network, to the task of multi-dimensional emotion regression, which automatically rates multiple emotion dimension scores for an input text, achieves notable improvements in r-values on both EMOBANK Reader's and Writer's multi- dimensional emotion regression tasks over the state-of-the-art baselines.
- Negation and Speculation Identification in Chinese Language35
This paper constructs a Chinese corpus which consists of three sub-corpora from different resources, and presents a new syntactic structure-based framework to identify the linguistic scope of a cue, instead of the traditional chunking- based framework.
- Hybrid text chunking35
Compared with standard HMM-based tagger, this tagger incorporates more contextual information into a lexical entry and an error-driven learning approach is adopted to decrease the memory requirement by keeping only positive lexical entries and makes it possible to further incorporate more context-dependentLexical entries.
- Modeling both Intra- and Inter-modal Influence for Real-Time Emotion Detection in Conversations34
A bidirectional dynamic dual influence network for real-time emotion detection in conversations is proposed, which can simultaneously model both intra- and inter-modal influence with biddirectional information propagation for current utterance and its historical utterances.
- N-gram-based Tense Models for Statistical Machine Translation34
This paper proposes n-gram-based tense models for SMT and successfully integrate them into a state-of-the-art phrase-based SMT system via two additional features.
- Error-driven HMM-based chunk tagger with context-dependent lexicon34
This paper proposes a new error-driven HMM-based text chunk tagger with context-dependent lexicon which uses a new Hidden Markov Modelling approach which incorporates more contextual information into a lexical entry.
- Opinion Target Extraction Using a Shallow Semantic Parsing Framework33
Evaluation on two datasets shows that structured syntactic information plays a critical role in capturing the domination relationship between an opinion expression and its targets and shows that the simplified shallow semantic parsing approach much outperforms the state-of-the-art sequence labeling one.
- Effective Sentiment-relevant Word Selection for Multi-modal Sentiment Analysis in Spoken Language33
This paper proposes a novel approach to selecting effective sentiment-relevant words for multi-modal sentiment analysis with focus on both the textual and acoustic modalities and employs a deep reinforcement learning mechanism to do so.
- Using active learning to expand training data for implicit discourse relation recognition33
Experimental results show that expanding the training set with small-scale carefully-selected external data yields substantial performance gain, with the improvements of about 4% for accuracy and 3.6% for F-score, which allows a weak classifier to achieve a comparable performance against the state-of-the-art systems.
- Adversarial Feature Adaptation for Cross-lingual Relation Classification33
A feature adaptation approach for cross-lingual relation classification is proposed, which employs a generative adversarial network (GAN) to transfer feature representations from one language with rich annotated data to another language with scarce annotation data.
- Adversarial Learning for Discourse Rhetorical Structure Parsing32
A novel method to transform both gold standard and predicted constituency trees into tree diagrams with two color channels and an adversarial bot between gold and fake tree diagrams to estimate the generated DRS trees from a global perspective is introduced.
- 32
- Emotion Detection in Code-switching Texts via Bilingual and Sentimental Information32
This paper uses a term-document bipartite graph to incorporate both bilingual and sentimental information, and proposes a label propagation based approach to learn and predict in the bipartites graph.
- A Grammar-driven Convolution Tree Kernel for Semantic Role Classification32
A grammardriven convolution tree kernel for semantic role classification by introducing more linguistic knowledge into the standard tree kernel is proposed and a composite kernel to integrate feature-based and tree kernel-based methods is presented.
- CNN-Based Broad Learning for Cross-Domain Emotion Classification31
The experimental results show that the proposed convolutional neural network based broad learning method can improve the performance of emotion classification more effectively than those baseline methods.
- Unified Dependency Parsing of Chinese Morphological and Syntactic Structures31
A unified dependency parsing approach for Chinese which takes unsegmented sentences as input and outputs both morphological and syntactic structures with a single model and algorithm and proves the effectiveness of the unified model and algorithms in parsing structures of words, phrases and sentences simultaneously.
- 30
- Multi-Modal Sentiment Classification With Independent and Interactive Knowledge via Semi-Supervised Learning29
The key idea is to leverage the semi-supervised variational autoencoders to mine more information from unlabeled data for multi-modal sentiment analysis, which includes both the independent knowledge within single modality and the interactive knowledge among different modalities.
- Event Factuality Identification via Generative Adversarial Networks with Auxiliary Classification29
This paper proposes a two-step framework, first extracting essential factors related with event factuality from raw texts as the input, and then identifying the factuality of events via a Generative Adversarial Network with Auxiliary Classification (AC-GAN).
- Natural Language Processing and Chinese Computing29
This talk will introduce a new entity-based search paradigm and various knowledge mining techniques to construct knowledge graphs from different types of data and show how the knowledge can be used to understand queries, enable semantic matching, and provide direct answers to natural language queries.
- Argument Inference from Relevant Event Mentions in Chinese Argument Extraction29
A novel global argument inference model is proposed to explore specific relationships, such as Coreference, Sequence and Parallel, among relevant event mentions to recover those intersentence arguments in the sentence, discourse and document layers which represent the cohesion of an event or a topic.
- Dependency-Driven Feature-based Learning for Extracting Protein-Protein Interactions from Biomedical Text29
This paper incorporates dependency information as well as other lexical and syntactic knowledge in a feature-based framework and shows that its system achieves comparable performance with other state-of-the-art PPI extraction systems, yet the best performance among all the feature- based ones.
- A Unified Framework for Scope Learning via Simplified Shallow Semantic Parsing29
Evaluation on the BioScope corpus shows that the structural information plays a critical role in capturing the relationship between a cue and its dominated arguments, and the parsing approach significantly outperforms the state-of-the-art chunking ones.
- 28
- Sentiment Classification towards Question-Answering with Hierarchical Matching Network28
A high-quality annotated corpus with specially-designed annotation guidelines for QA-style sentiment classification is created and a three-stage hierarchical matching network is proposed to explore deep sentiment information in a QA text pair.
- A Constituent-Based Approach to Argument Labeling with Joint Inference in Discourse Parsing28
A novel constituent-based approach to argument labeling is proposed, which integrates the advantages of both linear tagging and subtree extraction and unifies intra- and intersentence cases by treating the immediately preceding sentence as a special constituent.
- Research on Chinese negation and speculation: corpus annotation and identification27
A feature-based sequence labeling model to detect the negative or speculative cues and a cross-lingual cue expansion strategy is proposed to increase the coverage in cue detection.
- Bilingual Active Learning for Relation Classification via Pseudo Parallel Corpora26
Experimental results on the ACE RDC 2005 Chinese and English corpora show that bilingual active learning for relation classification significantly outperforms monolingual active learning.
- Semi-supervised learning for semantic relation classification using stratified sampling strategy26
This paper presents a new approach to selecting the initial seed set using stratified sampling strategy in bootstrapping-based semi-supervised learning for semantic relation classification, and investigates different augmentation strategies in iteratively adding reliable instances to the labeled set.
- 25
- 25
- Interactively-Propagative Attention Learning for Implicit Discourse Relation Recognition24
A propagative attention learning model using a cross-coupled two-channel network that yields substantial improvements over the baselines (BiLSTM and BERT).
- Topic Tensor Network for Implicit Discourse Relation Recognition in Chinese24
A topic tensor network is proposed to recognize Chinese implicit discourse relations with both sentence-level and topic-level representations and significantly outperforms several state-of-the-art baselines in both micro and macro F1-scores.
- Human-Like Decision Making: Document-level Aspect Sentiment Classification via Hierarchical Reinforcement Learning24
This paper proposes a new Hierarchical Reinforcement Learning (HRL) approach to DASC that incorporates clause selection and word selection strategies to tackle the data noise problem in the task of DASC.
- Negation Focus Identification with Contextual Discourse Information24
Evaluation on the *SEM 2012 shared task corpus indicates the usefulness of contextual discourse information in negation focus identification and justifies the effectiveness of the proposed graph model in capturing such global information.
- 24
- Employing the centering theory in pronoun resolution from the semantic perspective24
The use of both the semantic role features and the relative pronominal ranking feature in pronoun resolution is guided by extending the centering theory from the grammatical level to the semantic level in tracking the local discourse focus.
- 22
- MCDTB: A Macro-level Chinese Discourse TreeBank22
This paper describes the relevant experiments on the construction of the Macro Chinese Discourse Treebank (MCDTB), a higher-level Chinese discourse corpus, and annotates the macro discourse information, including discourse structure, nuclearity and relationship, and the additional discourse information to make the micro discourse annotation more objective and accurate.
- Document-Level Machine Translation Evaluation with Gist Consistency and Text Cohesion22
Two superior yet low-cost quantitative objective methods to enhance traditional MT metric by modeling document-level phenomena from the perspectives of gist consistency and text cohesion are proposed.
- Bilingual Event Extraction: a Case Study on Trigger Type Determination22
A uniform text representation with bilingual features to represent the samples and handle the difficulty of locating the triggers in the translated text from both monolingual and bilingual perspectives is proposed.
- Dependency-driven Anaphoricity Determination for Coreference Resolution22
Evaluation on the ACE 2003 corpus shows that the D-DSPT outperforms all previous parse tree structures on anaphoricity determination, and that applying the anaphricity determination module in coreference resolution achieves the so far best performance.
- Global learning of noun phrase anaphoricity in coreference resolution via label propagation22
A global learning method is employed in determining the anaphoricity of noun phrases via a label propagation algorithm to improve learning-based coreference resolution.
- XLPT-AMR: Cross-Lingual Pre-Training via Multi-Task Learning for Zero-Shot AMR Parsing and Text Generation20
A novel cross-lingual pre-training approach via multi-task learning (MTL) for both zeroshot AMR parsing and AMR-to-text generation of multiple non-English languages is proposed, demonstrating that this approach significantly outperforms a strong baseline of pre- training approach, and greatly advances the state of the art.
- Using Syntactic Head Information in Hierarchical Phrase-Based Translation20
The Head-Driven HPB (HD-HPB) model is presented, which incorporates head information in translation rules to better capture syntax-driven information in a derivation and allows improved reordering between any two neighboring non-terminals to explore a larger reordering search space.
- CropQuant-Air: an AI-powered system to enable phenotypic analysis of yield- and performance-related traits using wheat canopy imagery collected by low-cost drones19
CropQuant-Air is presented, an AI-powered software system that combines state-of-the-art deep learning models and image processing algorithms to enable the detection of wheat spikes and phenotypic analysis using wheat canopy images acquired by low-cost drones, and represents valuable advances in yield-based field phenotyping and phenotypesic analysis.
- Global Encoding for Long Chinese Text Summarization19
This article constructs a first large-scale, long Chinese text summarization corpus and proposes a sequence-to-sequence (Seq2Seq) model that incorporates a global encoding process with an attention mechanism that achieves a competitive result on the LCSPIRT corpus.
- User Classification with Multiple Textual Perspectives.19
Empirical studies on two basic user classification tasks, i.e., gender classification and age classification, demonstrate the effectiveness of the proposed ensemble LSTM approach to user classification with multiple textual perspectives.
- Improving Semantic Parsing with Enriched Synchronous Context-Free Grammar19
This article introduces both structure and syntax informed nonterminals, better guiding the parsing in favor of well-formed structure, instead of using a uninformed nonterminal in SCFGs, and addresses the unknown word translation issue via synthetic translation rules.
- Context-Sensitive Convolution Tree Kernel for Pronoun Resolution19
Evaluation on the ACE 2003 corpus shows that the dynamic-expansion tree span scheme can well cover necessary structured information in the parse tree for pronoun resolution and the context-sensitive tree kernel much outperforms previous tree kernels.
- A CDT-Styled End-to-End Chinese Discourse Parser18
This article presents an end-to-end Chinese discourse parser with the Connective-Driven Dependency Tree scheme, which consists of multiple components in a pipeline architecture, such as the elementary discourse unit (EDU) detector, discourse relation recognizer, discourse parse tree generator, and attribution labeler.
- Semi-Stacking for Semi-supervised Sentiment Classification18
This paper applies metalearning to predict the unlabeled data given the outputs from the member algorithms and proposes N-fold cross validation to guarantee a suitable size of the data for training the meta-classifier.
- Query Expansion in the Age of Pre-trained and Large Language Models: A Comprehensive Survey17
This survey reviews QE methods in the PLM/LLM era and provides a unified view of the emerging landscape, summarizing how different model families enable new expansion behaviors, including stronger contextualization, more controllable generation, and instruction-following.
- On-Chip Isolation and Reciprocal Signal Amplification Detection of Tumor-Derived Exosomes in Dual-Control Microfluidic Device17
A dual-control microchip integrated with a surface-enhanced Raman scattering (SERS) signal amplification detection method that achieves magnetic separation control and DNA cascade signal amplification with electrokinetic enrichment detection and enables precise and highly efficient assay of exosomes.
- Sherlock: Towards Multi-scene Video Abnormal Event Extraction and Localization via a Global-local Spatial-sensitive LLM17
A Global-local Spatial-sensitive Large Language Model (LLM) named Sherlock, i.e., acting like Sherlock Holmes to track down the criminal events, for this M-VAE task, and designs a Global-local Spatial-enhanced MoE (GSM) module and a Spatial Imbalance Regulator (SIR) to address the two challenges respectively.
- Extraction of causal relations based on SBEL and BERT model17
A simplified form of BEL statements is proposed to facilitate BEL extraction and BERT (Bidirectional Encoder Representation from Transformers) is employed to improve the performance of causal relation extraction (RE).
- Document-Level Event Factuality Identification via Adversarial Neural Network17
This work presents an LSTM neural network based on adversarial training with both intra- and inter-sequence attentions to identify document-level event factuality in English and Chinese texts.
- Towards Emotion-enriched Text-to-Motion Generation via LLM-guided Limb-level Emotion Manipulating16
This paper constructs a coarse-grained Emotional Text-to-Motion (EmotionalT2M) dataset and a fine-grained Limb -level Emotional Text-to-Motion (Limb-ET2M) dataset to justify the effectiveness of the proposed L3EM approach and demonstrates the significant advantage of the L3EM approach over the state-of-the-art baselines.
- 16
- CVAE-based Re-anchoring for Implicit Discourse Relation Classification16
This work uses Conditional VAE (CVAE) to estimate the risk of erroneous sampling, and develops a re-anchoring method which migrates the anchor of sampling area of VAE to re-duce the risk.
- Chemical-induced disease relation extraction via attention-based distant supervision16
The experiments demonstrate that distant supervision is promising for extracting chemical disease relations from biomedical literature, and capturing both local and global attention features simultaneously is effective in attention-based distantly supervised learning.
- Joint Modeling of Trigger Identification and Event Type Determination in Chinese Event Extraction16
Evaluation on the ACE 2005 Chinese corpus justifies the effectiveness of the joint model of trigger identification and event type determination over a strong baseline.
- Semi-Supervised Learning for Relation Extraction.16
Evaluation on the ACE RDC 2003 corpus shows that the proposed semi-supervised learning method outperforms the normal LP algorithm via all the available labeled data without SVM bootstrapping and can largely reduce the computational burden.
- Machine learning-based named entity recognition via effective integration of various evidences16
Through the HMM and an effective constraint relaxation algorithm to deal with the data sparseness problem, PowerNE is able to effectively apply and integrate various internal and external evidences of entity names and the named entity recognition problem is resolved effectively.
- 15
- Exploring Various Linguistic Features for Stance Detection15
Different from ordinary features, four linguistic features including lexical features, morphology features, semantic features and syntax features in Chinese micro-blogs in stance classifier are explored, and the system gets a good performance, which ranks the third place among sixteen systems.
- Semi-supervised Sentiment Classification with Self-training on Feature Subspaces15
This paper proposes a novel self-training approach where multiple feature subspace-based classifiers are utilized to explore a set of good features for better classification decision and to select the informative samples for automatically labeling.
- Head-Driven Hierarchical Phrase-based Translation15
An extension of Chiang's hierarchical phrase-based (HPB) model is presented, called Head-Driven HPB (HD-HPB), which incorporates head information in translation rules to better capture syntax-driven information, as well as improved reordering between any two neighboring non-terminals at any stage of a derivation to explore a larger reordering search space.
- Active Learning for Chinese Word Segmentation15
Experimental results show that employing the WBA model and the diversity measurement into active learning on CWS can save much annotation cost with little loss in the perfor mance.
- Kernel-based semantic relation detection and classification via enriched parse tree structure15
Evaluation on the ACE RDC corpora shows that both the new tree kernel and the enriched parse tree structure contribute significantly to RDC and the tree kernel method much outperforms the state-of-the-art ones.
- 14
- Two Heads are Better than One: Zero-shot Cognitive Reasoning via Multi-LLM Knowledge Fusion14
This work argues that the power of a single Large Language Model (LLM) is typically finite since it may not have learned some relevant knowledge about the question, and proposes a Multi-LLM Knowledge Fusion (MLKF) approach, which resorts to heterogeneous knowledge emerging from multiple LLMs, for zero-shot cognitive reasoning tasks.
- Cache-based Document-level Neural Machine Translation.14
A cache-based approach to document-level neural machine translation by capturing contextual information either from recently translated sentences or the entire document, which achieves substantial improvements over several state-of-the-art SMT and NMT baselines.
- Dual word and document seed selection for semi-supervised sentiment classification14
A novel selection strategy is proposed to simultaneously select good words and documents for manual annotation by considering both of their annotation costs and informativeness.
- 13
- Overview of NLPCC 2018 Shared Task 1: Emotion Detection in Code-Switching Text13
This paper presents the overview of the shared task, emotion detection in code-switching text, in NLPCC 2018, and introduces the task, the corpus, the participating teams, and the evaluation results.
- Two-View Label Propagation to Semi-supervised Reader Emotion Classification13
This paper proposes a two-view label propagation approach to semi-supervised reader emotion classification by exploiting two views, namely source text and response text in a label propagation algorithm that depends on two word-document bipartite graphs to model the relationship among the samples in the two views respectively.
- Interactive gender inference with integer linear programming13
This paper proposes a joint inference approach which well incorporates label correlations among the instances in ILP to achieve global optimization with various kinds of intra-task and extra-task constraints in interactive gender inference.
- Cross-argument inference for implicit discourse relation recognition13
Evaluation on PDTB proves the effectiveness of the inference mechanism in implicit relation recognition to the four level-1 relations and shows that the mechanism significantly outperforms other alternatives.
- 13
- 12
- 12
- Sentiment Forecasting in Dialog12
A Neural Sentiment Forecasting (NSF) model is proposed to address inherent challenges of sentiment forecasting in dialog, and a neural simulation model is employed to simulate the next utterance based on the context (previous utterances encountered).
- Event Factuality Detection in Discourse12
This work proposes a document-level approach on event factuality detection, which employs Bi-directional Long Short-Term Memory (BiLSTM) neural networks to learn contextual information of the event in sentences and utilizes a double-layer attention mechanism to capture the latent correlation features among event sequences in the discourse.
- The SoNLP-DP System in the CoNLL-2015 shared Task12
This paper describes the submitted discourse parsing system of the natural language group of Soochow University (SoNLP-DP) to the CoNLL 2015 shared task and achieves the 18.51% in F1-measure on the official blind test set.
- Using compositional semantics and discourse consistency to improve Chinese trigger identification12
Various morphological structures are explored to better represent the compositional semantics inside Chinese triggers and automatically identify the head morpheme as the governing sememe of a trigger in inferring unknown triggers.
- Hierarchical learning strategy in semantic relation extraction12
Two classifier learning approaches, i.e. the simple perceptron algorithm and the state-of-the-art Support Vector Machines, are applied using the hierarchical learning strategy to deal with the data sparseness problem in semantic relation extraction.
- 11
- Pre-trained Token-replaced Detection Model as Few-shot Learner11
This paper proposes a novel approach to few-shot learning with pre-trained token-replaced detection models like ELECTRA, and demonstrates that this approach outperforms few- shot learners withPre-trained masked language models in both one-sentence and two- sentence learning tasks.
- Coupling Context Modeling with Zero Pronoun Recovering for Document-Level Natural Language Generation11
This work proposes a highly adaptive two-stage approach to couple context modeling with ZP recovering to mitigate the ZP problem in NLG tasks and shows that this approach can improve the performance to a great extent.
- 11
- Cross-Lingual Emotion Classification with Auxiliary and Attention Neural Networks11
This paper uses machine translation services to eliminate the language gap between Chinese and English data and proposes a joint learning framework to leverage both Chinese andEnglish data, which develops auxiliary representations from several auxiliary emotion classification tasks.
- 11
- Emotion Corpus Construction on Microblog Text11
A novel emotion taxonomy which contains both basic emotions and complex emotions for annotating emotions expressed in microblog text is proposed, providing a good foundation for further emotion recognition task.
- 11
- Toward a Unified Framework for Standard and Update Multi-Document Summarization11
A unified framework for extracting standard and update summaries from a set of documents using a topic modeling approach for salience determination and a dynamic modeling approach is proposed for redundancy control is presented.
- Clustering-Based Stratified Seed Sampling for Semi-Supervised Relation Classification11
A clustering-based stratified seed sampling approach to semi-supervised learning that achieves the best F1-score of 75.9 on the sub-task of semantic relation classification, approaching the one with golden clustering.
- Tree kernel-based semantic role labeling with enriched parse tree structure11
This paper proposes a tree kernel method for a particular shallow semantic parsing task, called semantic role labeling (SRL), with an enriched parse tree structure that significantly outperforms the state-of-the-art tree kernel methods.
- Chinese Word Segmentation and Named Entity Recognition Based on a Context-Dependent Mutual Information Independence Model11
Evaluation shows that the system performs well on all the word segmentation closed tracks and achieves very good scalability across different corpora, and the use of the same strategy in named entity recognition shows promising performance.
- Word association and MI-Trigger-based language modeling11
It is concluded that the preferred relationships between words are useful to language disambiguation and can be modeled efficiently by the MI-Trigger-based modeling approach.
- OpenBA: an open-sourced 15B bilingual asymmetric Seq2Seq model pre-trained from scratch10
This report presents OpenBA, an open-sourced 15B bilingual asymmetric Seq2Seq model, to contribute an LLM variant to the Chinese-oriented open-source model community and enhances OpenBA with effective and efficient techniques.
- Emotion across Modalities and Cultures: Multilingual Multimodal Emotion-Cause Analysis with Memory-inspired Framework10
A memory-inspired multilingual multimodal framework (M3F) based on large language model (LLM), which can effectively capture the temporal and global informative features of non-linguistic modalities through memory bank module, which simulates the way memory is stored in human cognitive processes.
- Multi-layer Joint Learning of Chinese Nested Named Entity Recognition Based on Self-attention Mechanism10
A multi-layer joint learning model for Chinese named entities recognition based on self-attention aggregation mechanism where a series of multi-layered sequence labeling sub-models are joined to recognize named entities in a bottom-up fashion is proposed.
- Explicitly Modeling Word Translations in Neural Machine Translation10
It is shown that word translations can be explicitly incorporated into NMT effectively to avoid wrong translations, and proposed cross-lingual encoders are able to improve the translation accuracy for both traditional RNN- based NMT and recent self-attention-based NMT.
- Joint Learning on Relevant User Attributes in Micro-blog10
A joint learning approach, namely Aux-LSTM, is proposed, which first learns a proper auxiliary representation between the related tasks and then leverages the auxiliary representation to integrate the learning process in both tasks.
- Unsupervised Negation Focus Identification with Word-Topic Graph Model10
This work proposes an unsupervised word-topic graph model to represent and measure the focus candidates from both lexical and topic perspectives, and proposes a document-sensitive biased PageRank algorithm to optimize the ranking scores of focus candidates.
- 10
- Sentiment Classification Method of Chinese Micro-blog Based on Emotional Knowledge10
An unsupervised method of sentiment classification is proposed and applied to perform sentiment classification on Sina micro-blog and achieves a decent performance on sentiment classification for Chinese micro- blog.
- 10
- Employing Constituent Dependency Information for Tree Kernel-Based Semantic Relation Extraction between Named Entities10
Evaluation on the ACE RDC 2004 English and 2005 Chinese benchmark corpora shows that the dynamic syntactic parse tree much outperforms all previous tree spans, indicating its effectiveness in well representing the structural nature of relation instances while removing redundant information.
- Hedge detection and scope finding by sequence labeling with normalized feature selection10
A system which adopts a standard sequence labeling technique for hedge detection and scope finding, which achieves stable and competitive results for all the closed tasks and can be much further improved using a sufficient feature selection.
- Hedge Detection and Scope Finding by Sequence Labeling with Procedural Feature Selection10
A system which adopts a standard sequence labeling technique for hedge detection and scope finding, which achieves stable and competitive results for all the closed tasks and can be much further improved using a sufficient feature selection.
- 9
- 9
- Joint Learning-based Causal Relation Extraction from Biomedical Literature9
A joint learning model is proposed, which combines entity relation extraction and entity function detection to exploit their commonality and capture their inter-relationship, so as to improve the performance of biomedical causal relation extraction.
- Negative Focus Detection via Contextual Attention Mechanism9
An attention-based neural network to model contextual information is introduced which consists of a Bidirectional Long Short-Term Memory (BiLSTM) neural network and a Conditional Random Fields (CRF) layer to effectively encode the order information and the long-range context dependency in a sentence.
- Cross-media User Profiling with Joint Textual and Social User Embedding9
This paper addresses cross-media user profiling by bridging the knowledge between the source and target media with a uniform user embedding learning approach and demonstrates the effectiveness of the proposed approach to two cross- media user profiling tasks, i.e., cross- Media gender classification and cross-Media age regression.
- Global Inference to Chinese Temporal Relation Extraction9
This paper proposes a discourse-level global inference model to mine those temporal relations between event mentions in document-level, especially in nonadjacent sentences, and provides various kinds of discourse- level constraints, which derived from event semantics, to further improve the model.
- 9
- 9
- 9
- A Chinese efficient analyser integrating word segmentation, part-of-speech tagging, partial parsing and full parsing9
An efficient analyser for the Chinese language, which efficiently and effectively integrates word segmentation, part-of-speech tagging, partial parsing and full parsing based on a Hidden Markov Model and an HMM-based tagger.
- Hawkeye: Discovering and Grounding Implicit Anomalous Sentiment in Recon-videos via Scene-enhanced Video Large Language Model8
A new Scene-enhanced Video Large Language Model named Hawkeye, i.e., acting like a raptor (e.g., a Hawk) to discover and locate prey, for the IasDig task, and designs a graph-structured scene modeling module and a balanced heterogeneous MoE module to address the above two challenges.
- Emotion Detection with Neural Personal Discrimination8
Experimental results show the usefulness of personal attributes, and the effectiveness of the proposed NPD approach in capturing such personal attributes with significant gains over the state-of-the-art models.
- Event Factuality Identification via Hybrid Neural Networks8
A hybrid neural network model which combines Bidirectional Long Short-Term Memory networks and Convolutional Neural Networks for event factuality identification, and considers lexical and syntactic features is proposed.
- Domain-specific Named Entity Recognition with Document-Level Optimization8
This article proposes a document-level optimization approach to NER and applies it in a domain-specific document- level NER task to incorporate various entity recognition knowledge in the document level via Integer Linear Programming (ILP).
- Recognizing Macro Chinese Discourse Structure on Label Degeneracy Combination Model8
A Label Degeneracy Combination Model (LD-CM) is proposed to find the solution of structure recognition in the solution space of nuclearity recognition to improve the accuracy of the model on the Macro Chinese Discourse TreeBank.
- Active Learning for Age Regression in Social Media8
The key idea of the active learning approach is first to divide the whole feature space into several disjoint feature subspaces and then leverage them to learn a committee of regressors to reduce the annotation effort for age regression via active learning.
- Semi-supervised Gender Classification with Joint Textual and Social Modeling.8
This paper proposes a semi-supervised approach to gender classification by leveraging textual features and a specific kind of indirect links among the users which they call “ same-interest” links and proposes a factor graph, namely Textual and Social Factor Graph (TSFG), to model both the textual and the “same- interest” link information.
- Employing Event Inference to Improve Semi-Supervised Chinese Event Extraction8
Various kinds of linguistic knowledge-driven event inference mechanisms to semi-supervised Chinese event extraction can capture linguistic knowledge from four aspects, i.e. semantics of argument role, compositional semantics of trigger, consistency on coreference events and relevant events, to further recover missing event mentions from unlabeled texts.
- Pronoun Resolution in English and Chinese Languages Based on Tree Kernel8
This paper proposes a tree kernel method to anaphora resolution of pronouns in both English and Chinese by a dynamic-expansion scheme to automatically determine a proper tree span for pronoun resolution by the centering theory, competitor-related information, and semantic role- related information.
- A Unified Framework for Discourse Argument Identification via Shallow Semantic Parsing8
Evaluation using Penn Discourse Treebank shows that the current research’s parsing approach significantly outperforms the state-of-the-art chunking alternatives.
- Semantic Role Labeling in Chinese Language for Nominal Predicates8
The nominal SRL performance has been improved by properly integrating features that were derived from a state-of-the-art verbal SRL system, and the nominal predicate recognition is explained in detail.
- 8
- 7
- Comment-aided Video-Language Alignment via Contrastive Pre-training for Short-form Video Humor Detection7
A novel two-branch hierarchical model for short-form video humor detection, named Comment-aided Video-Language Alignment (CVLA) via data-augmented multi-modal contrastive pre-training, which demonstrates that CVLA dramatically outperforms state-of-the-art and several competitive baseline approaches.
- 7
- NUT-RC: Noisy User-generated Text-oriented Reading Comprehension7
A noisy user-generated text-oriented RC model that integrates the extractive and the generative RC model by a multi-task learning mechanism and an answer selection module and demonstrates that it significantly outperforms the state-of-the-art social media- oriented RC models.
- 7
- Chinese Event Factuality Detection7
This work introduces three kinds of linguistic features to represent event factuality, including factuality cue, event polarity, and tense, and employs a CNN-based feature encoder to capture their latent feature representations automatically and integrates them with word embeddings to identify the factuality label of target event.
- Employing Text Matching Network to Recognise Nuclearity in Chinese Discourse.7
A novel text matching network (TMN) is proposed that encodes the discourse units and the paragraphs by combining Bi-LSTM and CNN to capture both global dependency information and local n-gram information.
- Joint Modeling of Structure Identification and Nuclearity Recognition in Macro Chinese Discourse Treebank7
A macro discourse structure presentation schema is explored to present the macro level discourse structure, and a corresponding corpus is proposed, named Macro Chinese Discourse Treebank, to achieve global optimization with various kinds of constraints.
- One vs. Many QA Matching with both Word-level and Sentence-level Attention Network7
A novel QA matching approach, namely One vs. Many Matching, which aims to address the novel scenario where one question sentence often has an answer with multiple sentences is proposed, and is improved by employing both word-level and sentence-level attentions.
- Building a Chinese discourse topic corpus with a micro-topic scheme based on theme-rheme theory7
A micro-topic scheme (MTS) representation is proposed based on functional grammar theory, with elementary discourse topic unit (EDTU) as the node and referent of theme-rheme as link to represent the discourse topic structure in the Chinese language according to theme- rheme theory.
- Feature Engineering for CRFs Based Opinion Target Extraction7
Experimental results show that semantic role is a good indicator for opinion target and this paper proposes using semantic role as a specific feature in CRFs based opinion target extraction.
- Unified Semantic Role Labeling for Verbal and Nominal Predicates in the Chinese Language7
Evaluation of the first reported work of unified verbal and nominal SRL on Chinese PropBank and NomBank shows that the unified approach significantly improves the performance, in particular that of nominal S RL.
- 7
- 6
- 6
- 6
- 6
- 6
- Towards Discourse-Aware Document-Level Neural Machine Translation6
This work proposes a document-level NMT system enhanced with the discourse-aware document context, which is named Disco2NMT, which significantly surpasses both context-agnostic and context-aware baseline systems on multiple evaluation indicators.
- Feature-level Incongruence Reduction for Multimodal Translation6
This work proposes to extend MNMT architecture with a harmonization network, which harmonizes multimodal features(linguistic and visual features) by unidirectional modal space conversion by leading to the competitive performance to the-state-of-the-art.
- Breaking the Corpus Bottleneck for Context-Aware Neural Machine Translation with Cross-Task Pre-training6
This paper aims to improve context-aware NMT by taking the advantage of the availability of both large-scale sentence-level parallel dataset and source-side monolingual documents by proposing two pre-training tasks that are jointly and simultaneously learned via the same model.
- More than Text: Multi-modal Chinese Word Segmentation6
A time-dependent multi-modal interactive model based on Transformer framework to integrate multi- modal information for word sequence labeling is proposed and the experimental results show the effectiveness of this approach with fusing text and audio.
- Syntax-Guided Sequence to Sequence Modeling for Discourse Segmentation6
A sequence-to-sequence model along with a GCN based encoder to well utilize intra-sentence dependencies and a multi-head biaffine attention based decoder to predict EDU boundaries is proposed and Experimental results show that the syntax information used is significantly useful and the resulting model is competitive when compared with the state-of-the-art.
- Improving neural sentence alignment with word translation6
Three cross-lingual encoders to incorporate word translation into neural sentence alignment are proposed: Mixed Encoder that learns words and their translation annotation vectors over sequences wherewords and their translations are mixed alternatively.
- Chinese Zero Pronoun Resolution6
This article proposes a chain-to-chain approach to improve the performance of ZP resolution in three aspects: consecutive ZPs are clustered into coreferential chains, each working as one independent anaphor as a whole; additional features from ZP and common NP coreFerential chains are employed to better represent anaphors and their antecedent candidates, respectively.
- Employing Multiple Decomposable Attention Networks to Resolve Event Coreference6
A novel neural network is introduced, MDAN (Multiple Decomposable Attention Networks), to resolve document-level event coreference from different views, i.e., event mention, event arguments and trigger context, and applies a document- level global inference mechanism to further resolve the coreference chains.
- Two-Target Stance Detection with Target-Related Zone Modeling6
This study addresses a more challenging version of stance detection on Twitter, namely two-target stance detection, where two targets are involved and proposes a hierarchical neural network to perform stance detection for each target and introduces a target-related attention mechanism in the hierarchical network.
- Joint Sentiment and Emotion Classification with Integer Linear Programming6
Empirical study verifies the appropriateness and effectiveness of the proposed approach to joint sentiment and emotion classification in Integer Linear Programming.
- Leveraging Interactive Knowledge and Unlabeled Data in Gender Classification with Co-training6
Experimental results on a large data set from micro-blog demonstrate the appropriateness of leveraging interactive knowledge inGender classification and the effectiveness of the proposed co-training approach in gender classification.
- A Convolutional Deep Neural Network for Coreference Resolution via Modeling Hierarchical Features6
This work explored a convolutional deep neural network CDNN to extract discourse level features automatically and obtains a competitive score of average F1 over MUC, B3, and CEAF, which places it above the mean score of other systems on the dataset of CoNLL-2012 Shared Task.
- Collective Personal Profile Summarization with Social Networks6
A collective factor graph (CoFG) model is proposed to incorporate all these resources of knowledge to summarize personal profiles with local textual attribute functions and social connection factors and demonstrates the effectiveness of the proposed approach.
- Phrase-Based Evaluation for Machine Translation6
High-level abstract information such as semantic similarity and topic model are introduced into this phrase-based evaluation metric, which achieves comparable correlation with human judgements at segment-level and significant higher correlation at document-level.
- 6
- 6
- Automatic Expansion of Abbreviations in Chinese News Text6
An n-gram based approach to Chinese abbreviation expansion using a dictionary of non-reduced abbreviation/full-form pairs and a disambiguation technique to select a proper expansion based on bigram word segmentation is presented.
- DialoguePFM:Prompt-based Fusion Model for Emotion Recognition in Conversation5
The Prompt-based Fusion Model (DialoguePFM), which innovatively introduces an emotion representation that conveys the emotion label via mask token in a pre-defined prompt, and refine both emotion and utterance representations by capturing comprehensive dialogue information using a novel speaker-aware attention mechanism.
- ECFCON: Emotion Consequence Forecasting in Conversations5
This work proposes a new task called Emotion Consequence Forecasting in CONversations (ECFCON), and proposes a corresponding dialogue-level dataset, which is benchmarked by conducting experiments from the perspectives of traditional methods, generalized LLMs prompting methods, and clue-driven hybrid methods.
- 5
- One Type Context Is Not Enough: Global Context-aware Neural Machine Translation5
This article proposes to model global context for source language from both sentence level and word level, and shows that global context at both sentence level and word level significantly improve translation performance.
- Cognition-driven multimodal personality classification5
To tackle the inherent noise challenges in this CMPC task, this work proposes a tailored reinforcement learning approach, namely, multi-agent SelectNet, aiming to integrate the opinion-word and image-region selection strategies to select informative opinion- word andimage-region features for CMPC.
- Semantic Simplification for Sentiment Classification5
This work enhances the original text with a sentiment-driven simplified clause to intensify its sentiment and employs Abstract Meaning Representation (AMR) for generating simplified clauses, since AMR explicitly provides core semantic knowledge and potentially offers core concepts and explicit structures of original texts.
- Document-Level Neural Machine Translation with Hierarchical Modeling of Global Context5
A hierarchical model to learn the global context for document-level neural machine translation (NMT) is proposed through a sentence encoder to capture intra-sentence dependencies and a document encoders to model document- level inter-sentences consistency and coherence.
- Towards better entity linking5
A simple but effective coarse-to-fine unsupervised knowledge base (KB) extraction approach to improve the quality of KB and a highway network framework to bridge key words and sequential information captured with a self-attention mechanism to better represent both local and global information.
- Neural Co-training for Sentiment Classification with Product Attributes5
This work uses the attention-based bi-directional Gated Recurrent Unit (Att-BiGRU) to model the semantic content of each review and regard different categories of the target product as different views and uses a co-training framework to learn and predict the unlabeled reviews with different views.
- Many vs. Many Query Matching with Hierarchical BERT and Transformer5
This paper proposes a novel query matching approach for informal text, namely Many vs. Many Matching with hierarchical BERT and transformer and employs fine-tuned BERT (bidirectional encoder representation from transformers) to capture the pair-wise sentence matching representations.
- Semi-supervised Sentiment Classification Based on Auxiliary Task Learning5
A semi-supervised sentiment classification based on auxiliary task learning, namely Aux-LSTM, is proposed, which is used to assist learning the sentiment classification task with a small amount of human-annotated samples by training auto-annotation samples.
- Disease name recognition based on syntactic and semantic features5
Experiments show the CRF model, combined with these syntactic and semantic features, can significantly improve the state-of-the-art performance of disease entity recognition, achieving an F1 score of 85.3%.
- Minimally Supervised Chinese Event Extraction from Multiple Views5
A novel minimally supervised model is proposed for Chinese event extraction from multiple views and a semantic relationship view is introduced to capture the relevant event mentions from relevant documents.
- Elementary Discourse Unit in Chinese Discourse Structure Analysis5
The experimental result shows that the definition of clause corresponds with Chinese character in theory and has strong distinction ability in calculation.
- Classifier-Based Tense Model for SMT5
This work proposes to use a classifier-based tense model to keep the main tense in target side consistent with the one in source side, and presents a simple and effective way to help this model by expanding more phrase pairs with different tenses.
- 5
- 5
- Research on Tree Kernel-Based Entity Semantic Relation Extraction5
An improved tree kernel-based approach to entity semantic relation extraction is described, where the performance is improved by incorporation of entity-related semantic information into, the structured representation of relation instances and the pruning of redundant information.
- 5
- Relation Extraction Using Convolution Tree Kernel Expanded with Entity Features5
This study indicates that not only can this convolution tree kernel-based approach for relation extraction effectively capture both syntactic structure and entity information of relation instances, but also can avoid the difficulty with tuning the parameters in composite kernels.
- 4
- TopicDiff: A Topic-enriched Diffusion Approach for Multimodal Conversational Emotion Detection4
A model-agnostic Topic-enriched Diffusion (TopicDiff) approach for capturing multimodal topic information in MCE tasks that integrates the diffusion model into neural topic model to alleviate the diversity deficiency problem of neural topic model in capturing topic information.
- 4
- Linguistic synesthesia detection: Leveraging culturally enriched linguistic features4
This study incorporates comprehensive linguistic features into a neural model to detect linguistic synesthetic usages in a sentence automatically and proposes a joint model to detect the original and synesthetic modalities of the sensory words collectively.
- 4
- Real-time Emotion Pre-Recognition in Conversations with Contrastive Multi-modal Dialogue Pre-training4
Through detailed experiments and extensive analysis, it is demonstrated that the proposed multi-modal contrastive pre-training architecture effectively enhances the performance of multi- modal real-time emotion pre-recognition in conversations.
- 4
- Emotion Recognition with Conversational Generation Transfer4
An Emotion Recognition with Conversational Generation Transfer (ERCGT) framework to model the interaction among utterances by transfer learning is proposed and empirical studies illustrate the effectiveness of the proposed framework over several strong baselines on three benchmark emotion classification datasets.
- Sentiment classification via user and product interactive modeling4
This work proposes a novel interactive model that outperforms previous state-of-the-art methods significantly by learning the local and global interactions among users’ preferences, product characteristics, and reviews.
- Language Adaptation for Entity Relation Classification via Adversarial Neural Networks4
This paper introduces two kinds of adversarial structures, dual-channel AdvNN and single-channel AdvNN, and finds that both dual-channel and single-channel AdvNN significantly improve the performances of cross-lingual entity relation classification.
- Document structure model for survey generation using neural network4
A neural document structure model is presented that takes paragraphs as units, and model sentences in paragraphs, and employs a hierarchical model to learn structure among sentences, which can be used to select important and informative sentences to generate survey.
- Improving Context-Aware Neural Machine Translation with Source-side Monolingual Documents4
This work proposes a novel self-supervised pre-training approach with Global Context (PGC), and designs a universal model for PGC which consists of a global context encoder, a sentence encoder and a decoder, with similar architecture to typical context-aware NMT models.
- Speech-Driven End-to-End Language Discrimination toward Chinese Dialects4
Evaluation of two benchmark Chinese dialect corpora shows the appropriateness and effectiveness of the proposed speech-driven approach to fine-grained Chinese dialect discrimination compared to the state-of-the-art methods.
- Hierarchical sequence labeling for extracting BEL statements from biomedical literature4
A framework for hierarchical relation extraction using hierarchical sequence labeling on the instance-level training corpus derived from the original sentence-level corpus via word alignment is proposed and has the potential to be enhanced by adopting more advanced machine learning approaches.
- Active Learning for Chinese Word Segmentation on Judgements4
The empirical study demonstrates that the proposed approach effectively reduces the annotation cost and performances better than other baseline sample selection strategies under the same scale of annotation.
- 4
- Personal summarization from profile networks4
This paper proposes a collective ranking approach, called SocialRank, to combine textual information in an individual profile and social context information from relevant profiles in generating a personal profile summary, and a collective factor graph model, called CoFG, to summarize personal profiles with local textual attribute functions and social connection factors.
- Semantic Role Labeling of Chinese Nominal Predicates with Dependency-Driven Constituent Parse Tree Structure4
Evaluation on Chinese NomBank corpus shows that the tree kernel based method on D-CPT performs significantly better than other tree kernel-based ones and achieves comparable performance with the state-of-the-art feature- based ones, indicating the effectiveness of the novel D- CPT structure in representing various kinds of dependency relations in a CPT-style structure.
- Survey of Discourse Analysis Methods4
This paper surveys the state-of-the-art researches in Chinese and English discourse analysis under the perspective of computational linguistics, including the applications of Chinese andEnglish discourse analysis, the process of constructing a full Chinese and England discourse parser according to different discourse theories, discourse corpus and evaluation, as well as algorithms and detailed implementation.
- 4
- Exploring Local and Global Semantic Information for Event Pronoun Resolution4
A predicate -argument structure is proposed to represent the local semantic informati on about an event while the global semantic information is represented b y the entity coreference chains related with vario us arguments in the predicate-argument structure to complement its locality.
- Combining Dependency and Constituent-based Syntactic Information for Anaphoricity Determination in Coreference Resolution4
Two ways to combine dependency and constituent-based syntactic information to explore their complementary advantage are proposed, one is a dependency-driven constituent- based structured representation, and the other uses a composite kernel.
- Learning Noun Phrase Anaphoricity in Coreference Resolution via Label Propagation4
A global learning method in determining the anaphoricity of noun phrases via a label propagation algorithm to improve learning-based coreference resolution and demonstrates the great effectiveness of this method on the ACE2003 corpus.
- Chinese Dependency Parse Based Semantic Role Labeling4
In the system, effective pruning algorithm and useful features are adopted for Chinese dependency tree, and the semantic role identification and classification are accomplished by a maximum entropy classifier.
- Survey on Coreference Resolution4
This paper interprets the concepts of coreference resolution, and introduces the state-of-the-arts in coreferenceresolution from three aspects:Corpus resources, evaluation measures and resolution algorithms.
- Feature engineering for predicate identification and classification in semantic analysis4
This paper picks out 7 basic features and over 30 new features with different combinations of useful combinations into the baseline system with the maximum entropy classifier and shows that those new features and the combination of them can much improve the performance of the system.
- 4
- A Unified Framework for Text Analysis in Chinese TTS4
A robust text analysis system for Chinese text-to-speech synthesis that can achieve correct interpretation for most words and a post-processing for interpreting phonetic tags and fine-tuning pronunciation order for some special NSWs if necessary is presented.
- A Collaborative Ability Measurement for Co-training4
A new measurement (CA) is proposed for representing the collaborative ability of co-training classifiers based on the overlapping proportion between certain and uncertain instances and indicates whether two classifiers can co-train effectively.
- Forward-backward search method4
In order to simulate the process of man’s “thinking forward and backward”, a new search method-forward-backward search is described, from which the characteristics and efficiency of this method can be seen.
- 3
- Omni-SILA: Towards Omni-scene Driven Visual Sentiment Identifying, Locating and Attributing in Videos3
An Implicit-enhanced Causal MoE (ICM) approach is proposed for addressing the Omni-SILA task, aiming to interactively and precisely identify, locate and attribute visual sentiments through both explicit and implicit scene information.
- 3
- 3
- Document-Level Event Factuality Identification via Reinforced Semantic Learning Network3
A reinforced multi-granularity hierarchical network model: Reinforced Semantic Learning Network (RSLN), which means it can learn semantics from sentences and tokens at various levels of granularity and hierarchy, integrated with hierarchical reinforcement learning (HRL).
- ChatASU: Evoking LLM's Reflexion to Truly Understand Aspect Sentiment in Dialogues3
A Trusted Self-reflexion Approach (TSA) with ChatGLM as backbone to ChatASU, aiming to explore LLMs’ ability in understanding aspect sentiments in dialogue scenarios and justifying the effectiveness of TSA to ChatASU and the importance of considering the coreference and hallucination issues in ChatASU.
- 3
- 3
- 3
- Clinical Study of Sentinel Lymph Node Detection to Evaluate Pelvic Lymph Node Metastasis to Determine the Prognosis of Patients with Early Cervical Cancer3
The identification of SLN combined with 99mtechnetium-labeled sulfur colloid and methylthioninium chloride has a good accuracy and is safe for the assessment of the status of pelvic nodes in patients with initial stage cervical cancer.
- 3
- Word-Pair Relevance Modeling with Multi-View Neural Attention Mechanism for Sentence Alignment3
This work proposes to deal with the problem of sentence alignment by exploring the semantic interactionship among fine-grained word pairs within the framework of neural network, and first employs various relevance measures to capture various kinds of semantic interactions among word pairs by using a word-pair relevance network.
- Combining relation extraction with function detection for BEL statement extraction3
This work decomposes the problem into two subtasks: entity relation extraction and entity function detection, and proposes a deep learning-based approach to extract BEL statements.
- Employing Auto-annotated Data for Person Name Recognition in Judgment Documents3
This paper develops an auxiliary Long Short-Term Memory (LSTM) representation by training the auto-annotated data and then leverages the auxiliary LSTM representation to boost the performance of classifier trained on the human-annotation data.
- SoNLP-DP System for ConLL-2016 English Shallow Discourse Parsing3
This paper describes the submitted English shallow discourse parsing group of SoNLP-DP to the CoNLL-2016 shared task and uses a pipeline platform to conduct every subtask to form an end-to-end shallow discourse parser in the Penn Discourse Treebank (PDTB).
- 3
- Chinese Comma Disambiguation on K-best Parse Trees3
This paper proposes a joint approach combining K-best parse trees to Chinese comma disambiguation to reduce the dependent on syntactic parsing and releases a Chinese comma corpus which adds a layer of annotation to the manually-parsed sentences in the CTB (Chinese Treebank) 6.0 corpus.
- 3
- Semi-supervised Sentiment Classification with Random Feature Subspace Method3
This paper proposes a novel approach to semi-supervised learning for sentiment classification based on random subspace method that clearly outperforms the static way of the subspace generation and achieves much better performances than many other existing approaches for semi- supervised sentiment classification.
- Dependency-directed Tree Kernel-based Protein-Protein Interaction Extraction from Biomedical Literature3
Evaluation on multiple PPI corpora shows that the effectiveness of tree kernelbased methods for PPI extraction, in particular the advantage of dependency-directed constituent structure representation, justifies the effectiveness of tree kernelbased methods for PPI extraction.
- 3
- 3
- A Hybrid Approach to Chinese Abbreviation Expansion3
This paper presents a hybrid approach to Chinese abbreviation expansion that was evaluated on an abbreviation-expanded corpus built from the Peking University Corpus and showed that a recall of 83.8% and a precision of 86.3% can be achieved on average for different types of Chinese abbreviations.
- A Novel Hierarchical Document Clustering Algorithm Based on a kNN Connection Graph3
Evaluation on a news document corpus shows that the kNN connection graph-based hierarchical document clustering algorithm can achieve better performance than the famous k-means clustering algorithms while reducing the iteration number sharply in comparison with normal hierarchical document clusters.
- Direct modelling of output context dependence in discriminative hidden Markov model3
A dynamic back-off modelling algorithm using constraint relaxation principle is proposed to resolve the data sparseness problem in the discriminative HMM due to the direct modelling of the output context dependence in its output model.
- Using Subtext to Enhance Generative IDRR2
This paper leverage LLaMA to generate subtexts for argument pairs, and verify the effectiveness of subtext-based IDRR, and constructs an IDRR baseline using the decoder-only backbone LLaMA, and enhances it with subtext-aware relation reasoning.
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- 2
- Discourse Cohesion Evaluation for Document-Level Neural Machine Translation2
The DCoEM method is proposed and a new test suite that considers four cohesive manners (reference, conjunction, substitution, and lexical cohesion) to measure the cohesiveness of document translations is contributed.
- Domanial and dimensional adversarial learning for emotion regression2
Empirical evaluation shows that the proposed approach achieves notable improvements in r-values in the cross-domain multi-dimensional emotion regression task over the state-of-the-art baselines.
- Event Detection with Document Structure and Graph Modelling2
A graph model for event detection based on document structure used to connect sentences and words in a document and outperforms all baselines significantly and verifies the validity of document information is proposed.
- Adversarial BiLSTM-CRF Architectures for Extra-Propositional Scope Resolution2
This paper attempts to systematically address the above challenges via an adversarial BiLSTM-CRF model, to jointly model the potential extra-propositions and their contexts via the superiority of sequential architecture in effectively encoding order information and long-range context dependency.
- A Recursive Information Flow Gated Model for RST-Style Text-Level Discourse Parsing2
This paper proposes a recursive model to fuse multiple information flows and strengthen the representation of text spans through shift-reduce operations, and shows that the proposed parser can outperform the state-of-the-art in nuclearity detection under stringent discourse parsing evaluations.
- Multi-grain Representation Learning for Implicit Discourse Relation Recognition2
This paper proposes the multi-grain representation learning method, which uses different convolution filters to form larger-grain linguistic units to support the recognition of semantic relations between them.
- Focus-sensitive relation disambiguation for implicit discourse relation detection2
A novel task named implicit discourse relation disambiguation (IDRD) is created and a focus-sensitive relation disAmbiguation model is proposed that affirms a truly-correct relation when it is triggered by focal sentence constituents.
- Three-Layer Joint Modeling of Chinese Trigger Extraction with Constraints on Trigger and Argument Semantics2
A novel three-layer joint model to integrate three components in trigger extraction, i.e., trigger identification, event type determination, and event subtype determination is proposed and significantly outperforms the state-of-the-art Chinese trigger extraction and Chinese event extraction as a whole.
- A Novel Approach for Relation Extraction with Few Labeled Data2
A novel approach for relation extraction which uses the relation definition as a guide and only needs a hundred of high-quality mention examples for training model, which achieves competitive performance and can be incorporated with existing approaches to boost performance.
- Corpus Fusion for Emotion Classification2
Empirical studies show the effectiveness of the proposed approach to corpus fusion for emotion classification, and an Integer Linear Programming (ILP) optimization is proposed to refine the classification results.
- Parallel Corpus-based Bilingual Co-training for Relation Classification2
Experimental results on the Chinese and English parallel corpus show that bilingual co-training can improve relation classification in both languages, especially in English.
- SoNLP-DP System for ConLL-2016 Chinese Shallow Discourse Parsing2
This paper describes the submission to the CoNLL-2016 shared task on end-to-end Chinese shallow discourse parsing and defines a syntactically heuristic algorithm to identify elementary discourse units (EDUs) and further to recognize valid EDU pairs.
- Interactive Gender Inference in Social Media2
This paper proposes a two stage approach to interactive gender inference, which first applies a standard four-category classification algorithm to get a preliminary result, and then proposes a global optimization algorithm to achieve better performance.
- Syntactic Parsing with Hierarchical Modeling2
Evaluation on the Penn WSJ Treebank and the Penn Chinese Treebank using maximum entropy models shows that the proposed hierarchical model can achieve a good performance with more flexibility for future improvement.
- Document-Level Machine Translation Evaluation Metrics Enhanced with Simplified Lexical Chain2
This paper proposes to improve traditional MT evaluation metrics by simplified lexical chain, modeling document-level phenomena from the perspectives of text cohesion, and shows the effectiveness of such method on evaluating document- level translation quality and its potential of integrating with traditionalMT evaluation metrics to achieve higher correlation with human judgments.
- 2
- A Joint Learning Approach to Explicit Discourse Parsing via Structured Perceptron2
This paper focuses on building an end-to-end PDTB-style explicit discourse parser via structured perceptron by decomposing it into two components, i.e., a connective labeler which identifies connectives from a text and determines their senses in classifying discourse relationship and an argument labeler, which identifies corresponding arguments for a given connective.
- An improved ADRC and its Application2
A new ADRC structure of configurable zero point, and analysis the effect of strong interference on the extended state observer, and proposed the selection conditions of ESO parameter can effectively improve the response speed of the system and simplifies the tuning process parameters.
- Improving Syntactic Parsing of Chinese with Empty Element Recovery2
This paper explores the problem of empty element (EE) recovery in Chinese from the syntactic parsing perspective, which has been largely ignored in the literature and proposes two ways to automatically recover EEs: a joint constituent parsing approach and a chunk-based dependency parsing approach.
- Natural Language Processing and Chinese Computing: First CCF Conference, NLPCC 2012, Beijing, China, October 31-November 5, 2012. Proceedings2
Experimental results show that users’ historical behavior plays an important role in user preference modeling and the proposed method improves the final predication performance in the field of technical papers recommendation.
- 2
- Chinese Sentiment Classification on Imbalanced Data Distribution2
This paper collects product reviews from four domains and finds that the positive samples are much more than negative ones, and proposes a novel approach to combine both sampling and classification algorithms under an ensemble learning framework.
- Transductive Minimum Error Rate Training for Statistical Machine Translation2
Experimental results show that the transductive MERT method could yield significant performance improvements over a strong baseline on a large-scale Chineseto-English translation task.
- Exploiting Dependency Information for Feature-Based Protein-Protein Interaction Extraction2
This study illustrates that dependency information as well as base phrase chunking information is very effective for feature-based PPI extraction while the full parse feature have less effect on the overall performance.
- 2
- Semantic Role Labeling Based on Dependency Relationship2
Aiming at the bottlenecks of syntactic tree-based Semantic Role Labeling, this paper explores dependency relationship-based semantic role labeling which achieves the F1 measure of 80.94% on the WSJ portion of the CoNLL2008 SRL Shared Task, using automatic dependency parsing.
- Automatic identification of coordinate structure based on statistics and rules2
A statistical model and several novel rules are combined to automatically identify boundaries of coordinate structures and show that the combination of maximum entropy model and rules achieve performance 78.1% in F1, and that the rules bring 3.4% improvement in F2.
- Information Retrieval Technology : 4th Asia Information Retrieval Symposium, AIRS 2008: Harbin, China, January 15-18, 2008: Revised Selected Papers2
Improving Expertise Recommender Systems by Odds Ratio and Improving Spamdexing Detection Via a Two-Stage Classification Strategy are explored.
- Syntactic Parsing with Hierarchical Modeling2
Evaluation on the Penn WSJ Treebank and the Penn Chinese Treebank using maximum entropy models shows that the proposed hierarchical model can achieve a good performance with more flexibility for future improvement.
- Dependency tree-based SRL with proper pruning and extensive feature engineering2
This paper proposes a dependency tree-based SRL system with proper pruning and extensive feature engineering, and presents the unofficial system by applying a new effective pruning algorithm; including additional features; and adopting a better dependency parser, MSTParser.
- Mutual Information Independence Model Using Kernel Density Estimation for Segmenting and Labeling Sequential Data2
The evaluation on shallow parsing shows that MIIM can effectively capture the long context dependence to segment and label sequential data and it is interesting to note that using kernel density estimation leads to increased performance over using a classifier-based approach.
- Resolution of Data Sparseness in Named Entity Recognition Using Hierarchical Features and Feature Relaxation Principle2
A feature relaxation principle is proposed to resolve the data sparseness problem in MIIM-based named entity recognition via hierarchical features and suggest that the hierarchical features provide a potential for much better portability.
- DeepSVU: Towards In-depth Security-oriented Video Understanding via Unified Physical-world Regularized MoE1
A new Unified Physical-world Regularized MoE (UPRM) approach that incorporates two key components: the Unified Physical-world Enhanced MoE (UPE) Block and the Physical-world Trade-off Regularizer (PTR) to address the above two challenges, respectively.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Top-down Text-Level Discourse Rhetorical Structure Parsing with Bidirectional Representation Learning1
It is argued that combining local and global information for discourse parsing is more sensible, and a top-down discourse parser with bidirectional representation learning capabilities is introduced, to prove this.
- 1
- 1
- 1
- Neural Emotion Detection via Personal Attributes1
Experimental results show the usefulness of personal attributes in emotion detection, and the effectiveness of the proposed JPA and NPD approaches in capturing personal attributes over the state-of-the-art statistic and neural models.
- 1
- 1
- Discourse Parsing Enhanced by Discourse Dependence Perception1
A joint top-down framework that learns from both discourse dependency and constituency parsing through one shared encoder and two independent decoders and also explores a constituency-to-dependency conversion scheme tailored for the Chinese discourse corpus to ensure the high quality of the joint learning process.
- EDTC: A Corpus for Discourse-Level Topic Chain Parsing1
This research presents insight on discourse-level topic chain (DTC) parsing which aims at discovering new topics and in-vestigating how these topics evolve over time within an article to address the lack of data.
- 1
- Review Rating with Joint Classification and Regression Model1
This study proposes a novel approach to review rating, namely joint LSTM, by exploiting the advantages of both review classification and regression models, and employs an auxiliary Long-Short Term Memory (LSTM) layer to learn the auxiliary representation from the classification setting, and simultaneously join the Auxiliary representation into the main L STM layer for the review regression setting.
- 1
- Opinion Target Understanding in Event-Level Sentiment Analysis1
A neural 2-sequences-to-1-sequence framework is proposed to jointly leverage both texts in an event description and an comment, and apply a word-by-word attention mechanism to capture the alignment between the two texts.
- Supervised Domain Adaptation for Sentiment Regression1
This study proposes a novel approach to domain adaptation for sentiment regression that employs an auxiliary Long Short-Term Memory (LSTM) layer to learn the auxiliary representation from the source domain, and simultaneously join the auxiliary represented into the main LSTM layer for the target domain regression setting.
- 1
- A Case Study on Active Learning for Event Extraction1
This work introduces the uncertainty of argument classification into the active learning for pipeline and joint extraction models, and drives active learning to identify and annotate the most informative instances at each extraction stage.
- A Micro-topic Model for Coreference Resolution Based on Theme-Rheme Structure1
A micro-topic model based on the theme-rheme structure is proposed for coreference resolution and compared with the traditional mention object recognition in text space, this model reduces problem space and complexity.
- Expanding Native Training Data for Implicit Discourse Relation Classification1
A semantic frame based vector model towards unsupervised acquisition of semantically and relationally parallel data is proposed, aiming to enlarge natively implicit relation corpus so as to optimize the training effect.
- Random Walks for Opinion Summarization on Conversations1
A graph-based framework to opinion summarization on conversations is proposed, in particular, a random walk model is proposed to globally rank the utterances in a conversation and the relationship among them.
- Improved PSO Algorithm and its Application in Function Optimization1
The Grads-PSO algorithm (regulated by grad method) was proposed in this paper and the simulation results show that the improved PSO algorithm increase the computing speed than the standard PSO algorithms.
- Negation and Speculation Target Identification1
A new layer of the target information is incorporated over the BioScope corpus and a machine learning algorithm is proposed to automatically identify this new information to identify the target of a negative or speculative expression in biomedical texts.
- Comparative analysis of impact of lexical semantic information on Chinese entity relation extraction1
- The Role of Dependency Information for Protein-Protein Interaction Extraction1
Experimental evaluation on multiple PPI corpora reveals that dependency information as well as base phrase chunking information is very effective for feature-based PPI extraction, and the method achieves a promising performance of 54.7 in F-measure on the AIMed corpus, surpassing other state-of-the-art feature- based ones.
- 1
- Research on tree kernel for Chinese pronominal coreference resolution1
This paper proposes a tree kernel-based approach to anaphora resolution of pronoun in Chinese and combines syntax tree as a feature with other base features to achieve coreference resolution success.
- Hybrid Method to Chinese Base Noun Phrase Recognition1
A hybrid method to recognize Chinese Base Noun Phrase (BaseNP), including the use of grammer rules, statistical approach and classification combination, which generates a combination classification and improves the precision.
- 1
- Hierarchical Learning Strategy in Relation Extraction Using Support Vector Machines1
The state-of-the-art Support Vector Machines is applied as the basic classifier learning approach using the hierarchical learning strategy and much improves the performance on least- and medium- frequent relations.
- 1
- TREC 2005 Genomics Track at I2R1
This paper describes the methods used for the Ad Hoc task of TREC Genomics Track, where Synonym dictionary for genes and pseudo relevance feedback are used to expand queries and BM25 model is implemented to retrieve relevant documents.
- Chunking-based Chinese word tokenization1
Experiments show that such a system can well deal with the unknown word problem in Chinese word tokenization through HMM-based chunking.
- Modeling of long distance context dependency in Chinese1
Evaluation on the XINHUA new corpus of 29 million words shows that inclusion of the best 1,600,000 word pairs decreases the perplexity of the MI-Trigram model compared with the trigram model, and it is found that MI-Ngram modeling has much better performance than ngram modeling.
- MI-trigger-based Language Modelling1
It is concluded that the preferred relationships between words are useful to language disambiguation and can be modeled efficiently by the MI-Trigger-based modeling approach.
- Negative focus detection via hypergraph attention LSTM network and LLM-driven data augmentation–
A novel HyperGraph attention LSTM network (LSTM-HyG) is proposed to capture high-level semantics for sentences, and negated verbs, negative cues, by a fine-grained label scheme that can learn adequate sequential dependency relationship of words.
- Towards Closed-Loop Embodied Empathy Evolution: Probing LLM-Centric Lifelong Empathic Motion Generation in Unseen Scenarios–
An Emotion-Transferable and Scenario-Adapted Mixture of Experts (ES-MoE) approach which designs a causal-guided emotion decoupling block and a scenario-adapted expert constructing block to address the two challenges in the L2-EMG task are proposed.
- Aligning Acoustic and Semantic Representations for Speech Aspect-Based Sentiment Analysis–
This work introduces a novel speech aspect-based sentiment analysis task, taking raw speech signals as input and adopting a cross-modal mixup approach based on optimal transport to jointly extract sentiment elements from both speech and ASR-generated text and effectively align acoustic and semantic representations.
- –
- –
- –
- Don’t Just Listen, Try Planning: Graph-based Retrieval-Generation Agent for Long-form Audio Meeting Understanding–
The LongAudioQA dataset is constructed and the GRGA model is proposed, which models heterogeneous audio features into a multi-dimensional graph and leverages agent planning for retrieval and answer generation.
- –
- A Dataset and Model for Personalized Chinese Stance Detection–
The first dataset for Chinese stance detection that includes user profile information (PC-STANCE) is created, containing 31,033 Chinese microblogs annotated with stances toward 8 targets, as well as user gender, age, and location information.
- –
- –
- –
- –
- –
- –
- MRC and Transfer Learning Framework for Document-level Event Factuality Identification with Heterogeneous Spectral Attention Networks–
Experimental performance on ExDLEF corpus shows that the span-extraction MRC model is superior to several state-of-the-art baselines, and the effectiveness of transfer learning under MRC paradigms is proved.
- –
- –
- –
- –
- –
- Enhancing aspect sentiment quad prediction with syntactic information in generation model–
This study proposes an innovative structure-aware framework that explicitly encodes the syntactic structure into the pre-trained generation model while preserving its original distributional knowledge.
- –
- A process-parameter-driven cross-attention convolutional autoencoder for high-precision critical dimension prediction–
This study proposes a process-parameter-driven cross-attention convolutional autoencoder (PCA-CAE) based on the wafer height map (WHM), which is a data source often overlooked in manufacturing yet of great value.
- –
- –
- Learning to Differentiate Pairwise-Argument Representations for Implicit Discourse Relation Recognition–
A joint learning framework that combines prototypical and adversarial learning as well as hub-migration based redistribution to enable encoders to produce clearly distinguishable representations is proposed.
- –
- –
- –
- –
- –
- A Starter Performance Testing System Based of PIDNN Controller–
A new PID neural network(PIDNN) controller is designed which integrated the advantages of PID and intelligent controller and has simple structure as PID and self-learning ability as NN to optimize parameters.
- –
- –
- How to Understand "Support"? An Implicit-enhanced Causal Inference Approach for Weakly-supervised Phrase Grounding–
An Implicit-Enhanced Causal Inference approach to address the challenges of modeling the implicit relations and highlighting them beyond the explicit and a high-quality implicit-enhanced dataset is annotated to evaluate IECI and detailed evaluations show the great advantages of IECI over the state-of-the-art baselines.
- A multi-view heterogeneous and extractive graph attention network for evidential document-level event factuality identification–
A novel EvDEFI model named Heterogeneous and Extractive Graph Attention Network (HEGAT), which can update representations of events and sentences by multi-view graph attentions based on tokens and various lexical features from both local and global levels is proposed.
- –
- –
- –
- –
- –
- –
- –
- A Text-Image Pair Is Not Enough: Language-Vision Relation Inference with Auxiliary Modality Translation–
This paper introduces a concept of 012 modality translation with two potential direc-013 tions to generate additional modalities, and pro-014 pose the auxiliary modality translation frame-015 work (AMT) for LVRI.
- –
- –
- Response Generation via Structure-Aware Constraints–
This work proposes a response-generation model with structure-aware constraints to capture the structure of dialog and generate a better response with various constraints of the act, sentiment, and topic.
- –
- Coreference Resolution: Are the Eliminated Spans Totally Worthless?–
Experimental results on the document-level CoNLL-2012 Shared Task English dataset show that eliminated spans are indeed much effective and the proposed mention representation refining strategy can achieve competitive results when compared with previous state-of-the-art in coreference resolution.
- –
- –
- –
- –
- –
- –
- Triple-Joint Modeling for Question Generation Using Cross-Task Autoencoder–
This model employs a cross-task autoencoder to incorporate QG, QA and CSE into a joint learning process, which could better utilize the correlation between the contexts of different tasks in learning representations and provide more task-specific information.
- Hierarchical-Gate Multimodal Network for Human Communication Comprehension–
A novel neural architecture for understanding human communication called the Hierarchical-gate Multimodal Network (HGMN), which shows state-of-the-art performance on five publicly available datasets for multimodal sentiment analysis, emotion recognition and speaker trait recognition.
- Incorporating Image Matching Into Knowledge Acquisition for Event-Oriented Relation Recognition–
This paper speculates the relation class of the target events based on the relation-class distributions on the similar events, incorporating image processing into the acquisition of similar event instances, and using the neural network based image matching for approximate calculation between events.
- –
- Compressed sensing image restoration algorithm based on improved SURF operator–
The experimental results show that the proposed compressed sensing image restoration algorithm has good restoration performance, and the restored image has finer edge and texture structure and better visual effect.
- Cursor momentum for fascination measurement–
A deep insight is provided into the goal-directed cursor movement that occurs within a remarkably short period of time (<30 milliseconds), which is the interval between a user’s click-through and decision-making behaviors.
- –
- –
- –
- Bilingual Parallel Active Learning Between Chinese and English–
A Bilingual Parallel Active Learning paradigm is proposed, where an instance-level parallel Chinese and English corpus adapted from OntoNotes is augmented for relation extraction and both the seeds and jointly selected unlabeled instances at each iteration are parallel between two languages in order to enhance active learning.
- –
- –
- –
- Research on Chinese discourse rhetorical structure representation scheme and corpus annotation–
This talk will present a novel discourse structure representation scheme for Chinese, called Connectivedriven Dependency Tree (CDT), and describe the adventure in corpus annotation of the Chinese Discourse Treebank (CDTB) of 500 documents, using a top-down strategy to keep consistent with Chinese native’s cognitive habit.
- –
- Chinese cross document co-reference resolution based on SVM classification and semantics–
This paper transformed CDCR as a classification problem by using an Support Vector Machine (SVM) classifier to resolve both name disambiguation and variant consolidation, both of which were prevalent in information extraction.
- –
- Research on Bilingual Dependency Relationship Mapping for Chinese-English Lexicon Construction–
A dependency relationship mapping model is proposed, which constructs bilingual lexicon by mapping dependency context words, dependency relationship types and directions simultaneously simultaneously, and significantly outperforms a state-of-the-art system inilingual lexicon construction from both Chinese-English and English-Chinese.
- –
- Sentence Compression Based on ILP Decoding Method–
This paper presents a decoding method based on Integer Linear Programming (ILP), which considers sentence compression as the selection of the optimal compressed target sentence and has the advantage of speed and using fewer features in the case of similar results obtained.
- Chinese Semantic Role Labeling with Dependency-Driven Constituent Parse Tree Structure–
This is the first research of tree kernel-based SRL on effectively exploring dependency relationship information, which achieves comparable performance with the state-of-the-art feature-based ones.
- –
- –
- EMG Signal Decoded Based Virtual Artificial Intelligence Hand Control System–
An electromyography (EMG) decoded based virtual artificial intelligence control system has been developed to quantify the performance of real-time control of a multifunctional myoelectric prosthesis.
- Tree Kernel-based Zero Pronouns Recognition–
This paper introduced a learning approach of Tree Kernel-based Zero Pronouns Recognition, and presented a system based-upon this approach, which shows that the recognition rate of the approach based on tree kernel is better than the one of the System based on rules.
- Semantic Role Labeling in Chinese Language Based on Tree Kernel Function–
This paper implements a semantic role labeling in Chinese language using the convolution tree kernel of Support Vector Machine(SVM) to properly express the structural representation between predicates and arguments on dependency tree and explores two methods to prune the dependency tree: Shortest Path Tree(SPT) and Minimum Tree(MT).
- –
- Semantic role labeling for dependency relations–
This paper presents a dependency relations-based semantic role labeling system, which takes dependency relations as the labeled units, and a dependency tree is created from a sequence of labeled dependency relations, from which a number of features are extracted.
- Soochow University: Description and Analysis of the Chinese Word Sense Induction System for CLP2010–
This paper proposes a feature-based approach using the spectral clustering algorithm to solve the problem of Chinese word sense induction, and compares various clustering algorithms and similarity metrics.
- Research on Semantic Role Information in Anaphora Resolution–
This paper proposes a machine learning-based approach to coreference resolution with special focus on the semantic role labeling information of the anaphor and the antecedent candidate, and analyzes the integration of semantic role feature with detailed pronoun type knowledge.
- –
- Establishing Credit Mechanism of Network ARP to Defend ARP Deception–
The establishment of the credit mechanism of network ARP and the multiple defensive approach from the root causes of ARP deception are proposed.
- Tree Kernel Function-based Pronoun Coreference Resolution–
A tree kernel-based approach to anaphora resolution of pronoun is proposed on the basis of structured information automatically captured by convolve kernel of SVM, which combines with semantic role feature and verb-driving feature which are acquired from ASSERT.
- Tree Kernel-based Relation Extraction with Various Entity-Related Features *–
This study indicates that not only can this convolution tree kernel-based approach for relation extraction effectively capture both syntactic structure and entity information of relation instances in a single tree kernel, but also can avoid the difficulty with tuning the parameters in composite kernels.
- Building a Collocation Net–
This paper presents an approach to build a novel two-level collocation net, which enables calculation of the collocation relationship between any two words, from a large raw corpus, which is efficient and effective in solving the data sparseness problem and determining the collocation relationship between any two words.
Publication data from OpenAlex, with missing venues and authors filled in from Crossref; citation counts are the higher of OpenAlex and Semantic Scholar, last synced 2026-10-12. One-sentence summaries under some papers are written by Semantic Scholar’s model. Citation counts may be lower than on Google Scholar, which indexes more sources.
Report an error
Wrong papers, two people merged into one, or a profile that should not be here? Tell us and we will fix or hide it.