Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks279 from public data
- Neural Document Summarization by Jointly Learning to Score and Select Sentences348
This paper presents a novel end-to-end neural network framework for extractive document summarization by jointly learning to score and select sentences, which significantly outperforms the state-of-the-art extractive summarization models.
- 184
- Knowledge-Based Question Answering as Machine Translation136
A translation-based approach to solve knowledge-based question answering tasks in one unified framework based on CYK parsing, which achieves better results than previous methods.
- Document-Level Relation Extraction with Reconstruction135
A novel encoder-classifier-reconstructor model that manages to reconstruct the ground-truth path dependencies from the graph representation to ensure that the proposed DocRE model pays more attention to encode entity pairs with relationships in the training.
- 120
- Table-to-Text: Describing Table Region With Natural Language116
A generative model to generate a natural language sentence describing a table region, e.g., a row, and a flexible copying mechanism that selectively replicates contents from the table in the output sequence is presented.
- Automated image quality evaluation of T2‐weighted liver MRI utilizing deep learning architecture101
To develop and test a deep learning approach named Convolutional Neural Network for automated screening of T2‐weighted (T2WI) liver acquisitions for nondiagnostic images, and compare this automated approach to evaluation by two radiologists.
- 98
- Syntax-Directed Attention for Neural Machine Translation96
This paper extends the local attention with syntax-distance constraint, which focuses on syntactically related source words with the predicted target word to learning a more effective context vector for predicting translation, to provide more translation performance for NMT from source representation.
- Chinese Named Entity Recognition with a Sequence Labeling Approach: Based on Characters, or Based on Words?87
The results show that without the global features the person names and the location names have good result based on characters, but the organization names are more suitable based on words when global features are added, and the performance ofbased on words improved significantly.
- 85
- Discriminative Reasoning for Document-level Relation Extraction76
A novel discriminative reasoning framework is proposed to explicitly model the paths of these reasoning skills between each entity pair in this document to outperforms the previous state-of-the-art performance on the large-scale DocRE dataset.
- Deep Attention Neural Tensor Network for Visual Question Answering70
A novel deep attention neural tensor network (DA-NTN) for visual question answering, which can discover the joint correlations over images, questions and answers with tensor-based representations and integrates into the state-of-the-art VQA models.
- 64
- Iterative GRAPPA (iGRAPPA) for improved parallel imaging reconstruction61
Both phantom and in vivo MRI experiments demonstrated that the iterative approach reduces parallel imaging artifacts and permits high‐quality image reconstruction with a relatively small number of calibration lines and slight changes of GRAPPA weights.
- Knowledge Distillation for Multilingual Unsupervised Neural Machine Translation53
A simple method to translate between thirteen languages using a single encoder and a single decoder is introduced, making use of multilingual data to improve UNMT for all language pairs and proposes two knowledge distillation methods to further enhance multilingual UNMT performance.
- 53
- Exploring the sensitivity of magnetic resonance fingerprinting to motion50
Different kinds of motion have distinct effects on the precision and effective resolution of the parametric maps measured with MRF, showing that MRF-based acquisitions can be relatively robust to motion effects occurring at the beginning or end of the sequence.
- High‐resolution sodium imaging of human brain at 7 T50
The feasibility of high‐resolution sodium magnetic resonance imaging on human brain at 7 T was demonstrated in this study and the quantification of sodium concentration on the phantom and brain images enabled better accuracy at the high nominal resolutions than at the low nominal resolution.
- Learning to Decouple Relations: Few-Shot Relation Classification with Entity-Guided Attention and Confusion-Aware Training48
CTEG is proposed, a model equipped with two novel mechanisms to learn to decouple these easily-confused relations, and the results show that the proposed model achieves comparable and even much better results to strong baselines in terms of accuracy.
- Attention-Fused Deep Matching Network for Natural Language Inference47
An attention-fused deep matching network (AF-DMN) for natural language inference that takes two sentences as input and iteratively learns the attention-aware representations for each side by multi-level interactions and adds a self-attention mechanism to fully exploit local context information within each sentence.
- Issues with Entailment-based Zero-shot Text Classification46
It is found that pre-trained BERT without any fine-tuning can yield competitive performance against BERT fine- tuned for NLI, and with the concern that these models heavily rely on spurious lexical patterns for prediction, preliminary approaches for more robust NLI are experiments, but the results are in general negative.
- 46
- Sodium imaging of human brain at 7 T with 15‐channel array coil46
The 15‐channel array coil is an efficient tool to substantially improve SNR in sodium imaging on human brain by performing self‐calibrated sensitivity encoding parallel image reconstruction and uniform image intensity across entire field of view.
- 44
- Lost in Benchmarks? Rethinking Large Language Model Benchmarking with Item Response Theory43
A critical analysis of benchmark effectiveness is provided, examining mainstream prominent LLM benchmarks using results from diverse models and proposes Pseudo-Siamese Network for Item Response Theory (PSN-IRT), an enhanced Item Response Theory framework that incorporates a rich set of item parameters within an IRT-grounded architecture.
- Unsupervised Bilingual Word Embedding Agreement for Unsupervised Neural Machine Translation43
Empirical results show that the proposed methods significantly outperform conventional UNMT, and two methods that train UNMT with UBWE agreement are proposed.
- Computational and experimental evaluation of the Tic-Tac-Toe RF coil for 7 Tesla MRI41
This study extensively compare the designs of two distributed-circuit based RF coils: the Tic-Tac-Toe (TTT) head coil and TEM head coil on multiple anatomically detailed head models and in-vivo and examines RF consistency across multiple head models for different RF excitation strategies.
- Whole brain perfusion measurements using arterial spin labeling with multiband acquisition41
The multiband (MB) excitation and reconstruction technique was both developed and evaluated for accelerated data acquisition of arterial spin labeling (ASL) to cover whole brain perfusion maps.
- Diagnostic evaluation of machine translation systems using automatically constructed linguistic check-points40
A method that automatically extracts check-points from parallel sentences that can monitor a MT system in translating important linguistic phenomena to provide diagnostic evaluation is presented.
- HIT at TREC 2012 Microblog Track35
The query expansion and document expansion techniques are explored to address the retrieval of short tweet texts and the microblog filtering via comparing the new tweet against top m relevant tweet retrieved in the history is accomplished.
- Evaluating o1-Like LLMs: Unlocking Reasoning for Translation through Comprehensive Analysis32
Results show that o1-Like LLMs establish new multilingual translation benchmarks, with DeepSeek-R1 surpassing GPT-4o in contextless tasks and three key insights are revealed.
- 30
- Dual‐echo arteriovenography imaging with 7T MRI30
To implement a dual‐echo sequence MRI technique at 7T for simultaneous acquisition of time‐of‐flight MR angiogram (MRA) and blood oxygenation level‐dependent MR venogram (MRV) in a single MR acquisition and to compare the image qualities with those acquired at 3T.
- 29
- 29
- Compound Embedding Features for Semi-supervised Learning29
This paper built compound features from continuous word embeddings based on clustering and showed that the compound features not only improved the performances on several NLP tasks, but also ran faster, suggesting the potential ofembeddings.
- CATSLU: The 1st Chinese Audio-Textual Spoken Language Understanding Challenge28
The 1st Chinese Audio-Textual Spoken Language Understanding Challenge (CATSLU) is described which introduces a new dataset with audio-textual information, multiple domains and domain knowledge and introduces two scenarios of audio- textual SLU in which participants are encouraged to utilize data of other domains or not.
- Selection Bias Explorations and Debias Methods for Natural Language Sentence Matching Datasets27
This paper investigates the problem of selection bias on six NLSM datasets and finds that four out of them are significantly biased, and proposes a training and evaluation framework to alleviate the bias.
- Efficient Disfluency Detection with Transition-based Parsing26
An efficient disfluency detection approach based on right-to-left transitionbased parsing is proposed, which can efficiently identify disfluencies and keep ASR outputs grammatical.
- Understanding Data Augmentation in Neural Machine Translation: Two Perspectives towards Generalization25
Inspired by recent theoretic advances in deep learning, the paper understands DA from two perspectives towards the generalization ability of a model: input sensitivity and prediction margin, which are defined independent of specific test set thereby may lead to findings with relatively low variance.
- Improving Neural Machine Translation with Neural Syntactic Distance25
It is shown that it is not trivial to improve NMT with NSD; however, the proposed strategies are shown to improve translation performance of the baseline model.
- 25
- A Unified Tagging Approach to Text Normalization23
A unified tagging approach to perform the text normalization task using Conditional Random Fields (CRF) and Experimental results on email data cleaning show that the proposed method significantly outperforms the approach of using cascaded models and that of employing independent models.
- Statistics based hybrid approach to Chinese base phrase identification23
A statistics based hybrid model for identifying 7 types of Chinese base phrases in view is presented and experiments show the efficiency of the proposed method in simplifying sentence structure.
- Modeling Future Cost for Neural Machine Translation21
A future cost representation is learned based on the current generated target word and its contextual information to compute an additional loss to guide the training of the NMT model and is used to help the generation of the next target word in the decoding.
- A Survey on Human Preference Learning for Aligning Large Language Models20
This survey reviews the progress in human preference learning within a unified framework, aiming to provide a comprehensive perspective on established methodologies while exploring avenues to further advance LLM alignment.
- Unsupervised Neural Machine Translation for Similar and Distant Language Pairs20
A simple, yet effective and efficient approach that (like UNMT) relies solely on monolingual corpora: pseudo-data-based unsupervised neural machine translation is proposed.
- Sentence-Level Agreement for Neural Machine Translation20
The proposed sentence-level agreement module can be integrated into NMT as an additional training objective function and can also be used to enhance the representation of the source sentences.
- Deterministic Attention for Sequence-to-Sequence Constituent Parsing20
A deterministic attention mechanism is proposed that deterministically selects the important context and is not affected by the sequence length and when combined with a novel bottom-up linearization method, this parser demonstrated better performance than that achieved by the sequences-to-sequence parser with probabilistic attention mechanism.
- Improving Pivot-Based Statistical Machine Translation Using Random Walk20
A novel approach that utilizes a machine learning method to improve pivot-based statistical machine translation (SMT) by utilizing Markov random walks to connect possible translation phrases between source and target language.
- Automatic Translation Template Acquisition Based on Bilingual Structure Alignment20
A method for automatically acquiring translation templates from bilingual corpora by combining a language parsing with a statistical bilingual language model is proposed in this paper.
- Evaluation of spiral acquisition variants for functional imaging of human superior colliculus at 3T field strength19
High‐resolution functional magnetic resonance imaging of human subcortical brain structures is challenging because of their deep location in the cranium, and their comparatively weak blood oxygen level dependent responses to strong stimuli.
- Exploring Deep Belief Network for Chinese Relation Extraction18
A novel model based on Deep Belief Network is first presented to detect and classify the relations among Chinese entities to demonstrate that the proposed approach is effective in handling high dimensional feature space including character N-grams, entity types and the position information.
- 17
- Interactive Conversational Head Generation16
A new conversation head generation benchmark for synthesizing behaviors of a single interlocutor in a face-to-face conversation and proposes corresponding baseline solutions to three novel tasks targeting the interaction modeling during the face-to-face conversation.
- 16
- 15
- Improving Pivot-Based Statistical Machine Translation by Pivoting the Co-occurrence Count of Phrase Pairs15
This paper presents a novel approach to calculate the translation probability by pivoting the co-occurrence count of source-pivot and pivot-target phrase pairs, and shows that this method leads to significant improvements over the baseline systems.
- Learning Chinese bracketing knowledge based on a bilingual language model15
The preliminary experiments show automatically learned knowledge accords well with manually annotated brackets, and is particularly useful to acquire bracketing knowledge for a less studied language that lacks tools and resources found in a second language more studied.
- 14
- Self-Training for Unsupervised Neural Machine Translation in Unbalanced Training Data Scenarios14
This paper defines and analyzes the unbalanced training data scenario for UNMT, and proposes UNMT self-training mechanisms to train a robust UNMT system and improve its performance in this case.
- 14
- Subcategorization acquisition and evaluation for Chinese verbs14
Evaluation on the acquisition of 20 multi-pattern verbs shows that the experiment achieved the similar precision and recall with former researches, which indicates great potentialities of subcategorization information in the fields of NLP.
- Dual Instruction Tuning with Large Language Models for Mathematical Reasoning13
This work proposes a dual instruction tuning strategy to meticulously model mathematical reasoning from both forward and reverse directions, and introduces the Intermediate Reasoning State Prediction task (forward reasoning) and the Instruction Reconstruction task (reverse reasoning) to enhance the LLMs' understanding and execution of instructions.
- 13
- DesignProbe: A Graphic Design Benchmark for Multimodal Large Language Models12
The DesignProbe is established, a benchmark to investigate the capability of MLLMs in design and considers both the attribute recognition and semantic understanding tasks, and indicates that refining prompts can enhance the performance of MLLMs.
- 12
- 11
- Cross Copy Network for Dialogue Generation11
This paper proposes a novel network architecture - Cross Copy Networks (CCN) to explore the current dialog context and similar dialogue instances' logical structure simultaneously, and proves that the proposed algorithm is superior to existing state-of-art content generation models.
- The Extraction of Trajectories from Real Texts Based on Linear Classification11
This study proposes a method where the extraction of trajectories is modeled as a binary classification problem and resolved based on a linear classifier with syntactic features, and presents an analysis of how the space relations are described in Chinese.
- Enhancement of functional MRI signal at high‐susceptibility regions of brain using simultaneous multiecho multithin‐slice summation imaging technique10
To increase temporal‐signal‐to‐noise ratio (tSNR) and blood oxygen level‐dependent (BOLD) sensitivity at high‐susceptibility regions of the brain by means of a simultaneous multiecho multithin‐slice summation imaging technique.
- 10
- 10
- 10
- Chinese term extraction using minimal resources10
A term candidate extraction algorithm is proposed to identify features of the relatively stable and domain independent term delimiters rather than that of the terms to ensure the efficiency and relative domain independent nature of the approach.
- 10
- 10
- 9
- 9
- StyleInject: Parameter Efficient Tuning of Text-to-Image Diffusion Models8
Comprehensive experiments show that StyleInject surpasses traditional LoRA in both text-image semantic consistency and human preference evaluation, all while ensuring greater parameter efficiency.
- 8
- Research on Domain Term Extraction Based on Conditional Random Fields8
An automatic domain term extraction method is proposed based on condition random fields, which treats domain terms extraction as a sequence labeling problem, and terms' distribution characteristics as features of the CRF model.
- 8
- LLM-based Discriminative Reasoning for Knowledge Graph Question Answering7
This work proposes READS to reformulate the KGQA process into discriminative subtasks, which simplifies the search space for each subtasks, thereby alleviating hallucination and ungrounded reasoning issues in LLMs.
- Combining Constituent and Dependency Syntactic Views for Chinese Semantic Role Labeling7
This paper presents a novel feature-based semantic role labeling (SRL) method which uses both constituent and dependency syntactic views and outperforms the traditional constituent-based or dependency-based SRL methods.
- 7
- 6
- Enhancing Bilingual Lexicon Induction via Bi-directional Translation Pair Retrieving6
This work proposes a novel and effective method to improve translation pair retrieval in cross-lingual WEs that considers both source-side and target-side perspectives throughout the retrieval process to alleviate false positive word pairings that emanate from a single perspective.
- Learning and Evaluating Human Preferences for Conversational Head Generation6
Experimental results validate the superiority of Preference Score in aligning with human perception, and also demonstrate robustness and generalizability to unseen data, making it a valuable tool for advancing conversation head generation.
- 6
- Document-Level Relation Extraction with Path Reasoning6
This article proposes a novel method to capture this reasoning path from one entity to another entity, thereby better simulating reasoning skills to classify relation between two entities and introduces an additional attention layer to summarize multiple reasoning paths for further enhancing the performance of the DocRE model.
- 6
- Improving Vector Space Word Representations Via Kernel Canonical Correlation Analysis6
This work proposes to use kernel canonical correlation analysis to capture a non-linear relationship between word embeddings of two languages and demonstrates that the proposed approach achieves essentially better performances than previous linear methods on all of the three tasks, especially for language pairs with substantial typological difference.
- 6
- A Lexicalized Reordering Model for Hierarchical Phrase-based Translation6
This paper proposes a novel lexicalized reordering model which is built directly on synchronous rules for hierarchical phrase-based translation and achieves good results on NIST machine translation test sets.
- Train the machine with what it can learn: corpus selection for SMT6
This paper first identifies literally translated sentence pairs via lexical and grammatical compatibility, and then uses these data to train SMT models, and indicates that larger training corpora do not always lead to higher decoding performance when the added data are not literal translations.
- 6
- Recent advances on NLP research in Harbin Institute of Technology6
This paper gives an introduction to the achievements on NLP in HIT, which has become one of the most important natural language processing bases for teaching and scientific research in China now.
- UCFA‐Net: A U‐shaped cross‐fusion network with attention mechanism for enhanced polyp segmentation5
A new framework for polyp segmentation called U‐shaped cross‐fusion network with attention mechanism (UCFA‐Net), which employs a pyramid vision transformer as encoder to extract image features at multiple scales, which achieves superior average scores across the five datasets and exhibits greater robustness in the face of diverse and demanding scenarios.
- Real-time Image Enhancement with Attention Aggregation5
This article carefully re-formulate the curve function to improve learning stability, and aggregate different semantic attention into the curve regression process, which can overcome the major problems of curve-based methods that generate moderate results with low contrast.
- EviDR: Evidence-Emphasized Discrete Reasoning for Reasoning Machine Reading Comprehension5
An evidence-emphasized discrete reasoning approach (EviDR), in which sentence and clause level evidence is first detected based on distant supervision, and then used to drive a reasoning module implemented with a relational heterogeneous graph convolutional network to derive answers.
- Syntax-based Transformer for Neural Machine Translation5
A syntax-based Transformer for MT is proposed, which incorporates source-side syntax structures generated by the parser into the self-attention and positional encoding of the encoder, and evaluates on two language pairs show that the syntax- based Transformer outperforms the conventional (non-syntactic) Transformer.
- CN-HIT-MI.T at SemEval-2019 Task 6: Offensive Language Identification Based on BiLSTM with Double Attention5
An attention-based two layers bidirectional longshort memory neural network (BiLSTM) for semantic feature extraction and a residual connection mechanism is used to synthesize two different deep features, and an emoji attention mechanisms are used to extract semantic information of emojis in text.
- Context-Aware Smoothing for Neural Machine Translation5
This work proposes a novel context-aware smoothing method to dynamically learn a sentence-specific vector for each word depending on its local context words in a sentence to improve the translation performance.
- Cross-lingual Projections between Languages from Different Families5
This work enhances the projection method based on word alignments by introducing target-language word representations as features and proposing a novel noise removing methodbased on these word representations.
- Syllable-based Machine Transliteration with Extra Phrase Features5
Phrase-based model is utilized to solve machine transliteration with the mapping between Chinese characters and English syllables rather than English characters rather thanEnglish characters.
- 5
- 5
- 5
- 5
- ToolSample: Dual Dynamic Sampling Methods with Curriculum Learning for RL-based Tool Learning4
Dynamic Sampling with Curriculum Learning (DSCL) is introduced, a framework specifically designed to address the unique characteristics of tool learning: its multiple interdependent sub-tasks and multi-valued reward functions.
- Make Imagination Clearer! Stable Diffusion-based Visual Imagination for Multimodal Machine Translation4
A stable diffusion-based imagination network integrated into a multimodal large language model (MLLM) to explicitly generate an image for each source sentence, thereby advancing multimodal MT.
- Empowering LLMs in Task-Oriented Dialogues: A Domain-Independent Multi-Agent Framework and Fine-Tuning Strategy4
A Domain-Independent Multi-Agent Framework (DIMF), which contains Intent Classification Agent, Slot Filling Agent and Response Agent, and a simple and effective Data Distribution Adaptation (DDA) method to mitigate degradation issues during DPO training is designed.
- 4
- CAN-GRU: A Hierarchical Model for Emotion Recognition in Dialogue4
The results show that the proposed model(CAN-GRU) and its variants achieve better performance than baselines and the effects of unidirectional and bidirectional networks.
- An Information Retrieval-Based Approach to Table-Based Question Answering4
This work proposes a simple yet effective information retrieval based approach to answer complex questions with open domain web tables by ranking all table cells based on their representations, and selecting the cells of the highest ranking score as the answer.
- Improving Unsupervised Dependency Parsing with Knowledge from Query Logs4
S syntactic knowledge is acquired from query logs to help estimate better probabilities in dependency models with valence and achieves improvements of 8.07% on CoNLL 2007 English using the AOL query logs.
- 4
- 4
- Phrase Table Combination Deficiency Analyses in Pivot-Based SMT4
Experimental results indicated that both covering rate of phrase pairs and translation probability accuracy affect the quality of translation.
- 4
- Construction Technology of Ontology Knowledge Base in Multiple Minority Languages4
This paper mainly focuses on the construction idea of multi-language ontology with unified criteria and interface and the key problems are described, such as the description method of basic rules of the languages based on concept semantic similarity in multi-languages ontology.
- 4
- Research and Development of Biomedical Text Mining4
Main approaches and relevant achievements in biomedical text mining are introduced, including machine learning methods to named entity recognition, abbreviation and synonym recognition, relation extraction, as well as relevant resource constructions, international evaluations and academic gatherings.
- HIT-WSD: using search engine for multilingual Chinese-English lexical sample task4
This system disambiguates senses of Chinese words and finds the correct translation in English by using the web as WSD knowledge source and does not require any sense-tagged corpus.
- Discriminative Models for Automatic Acquisition of Translation Equivalences4
Discriminative features including translation literality, phrase alignment probability, and phrase length ratio are used to evaluate equivalences and results indicate that its precision is 87.8%, and recall is 89.8% for support vector machine.
- 4
- 3
- 3
- Speculative Decoding Meets Quantization: Compatibility Evaluation and Hierarchical Framework Design3
A hierarchical framework that employs a small model as an intermediate stage to turn tree-style drafts into sequence drafts, leveraging the memory access benefits of the target quantized model and achieving a 2.78-times speedup across various tasks for the 4-bit weight Llama-3-70B model on an A100 GPU.
- 3
- HopPG: Self-Iterative Program Generation for Multi-Hop Question Answering over Heterogeneous Knowledge3
This work proposes a self-iterative framework for multi-hop program generation (HopPG) over heterogeneous knowledge, which leverages the previous execution results to retrieve supporting facts and generate subsequent programs hop by hop.
- Toward automatic support for leading court debates: a novel task proposal & effective approach of judicial question generation3
An innovative end-to-end model called ’Judicial Questioning Aid’ is proposed which is capable of proactively leading the court debate by asking useful questions to a certain litigant given previous context, and encodes the contextual utterance information with respect to global legal knowledge and local case judicial factors.
- 3
- Incorporating Phrase-Level Agreement into Neural Machine Translation3
A novel phrase-level agreement method that directly reduces the difference between the representations of the source-side and target-side phrase and achieves significant improvements over state-of-the-art baselines.
- 3
- 3
- Chinese Terminology Extraction Using EM-Based Transfer Learning Method3
Experimental results show the proposed Chinese terminology extraction with EM-based transfer learning method outperforms traditional statistical terminology extraction method significantly.
- 3
- Research on Automatic Pattern Acquisition Based on Construction Extension3
This paper is intended to discuss specific naming rules for the entities and to assign the methods into computers in order for them to automatically obtain new patterns of term.
- Distributed English Text Chunking Using Multi-agent Based Architecture3
By applying and testing the approach on the public training and test corpus, the F score for arbitrary phrases identification using Distributed Multi-Agent strategy achieves 95.70% compared to the previous best F score of 94.17%.
- 2
- 2
- 2
- 2
- Visual-Aware Text-to-Speech*2
A new visual-aware text-to-speech (VA-TTS) task to synthesize speech conditioned on both textual inputs and sequential visual feedback of the listener in face- to-face communication and devise a baseline model to fuse phoneme linguistic information and listener visual signals for speech synthesis.
- An efficient confusing choices decoupling framework for multi-choice tasks over texts2
A query-guided attention (QGA) mechanism is designed to automatically filter the text contents causing confusion by measuring the syntactic relevance between different contents and the query and a confusion-aware training (CAT) mechanism is designed to learn the correct and easily confused choices asynchronously and perform a pushing-away operation between their selection processes.
- 2
- HITMI&T at SemEval-2021 Task 5: Integrating Transformer and CRF for Toxic Spans Detection2
This paper introduces a single model built on the model of multi-layer bidirectional transformer encoder, which aims to accurately locate toxic spans within a text using BIO tagging scheme and introduces conditional random field (CRF) to make the model learn the constraints between tags.
- CN-HIT-MI.T at SemEval-2020 Task 8: Memotion Analysis Based on BERT2
This paper adopts BERT and ResNet for evaluation of detecting the emotions of Internet memes and uses RandAugment to enhance the data of the picture, and uses Training Signal Annealing to solve the impact of the imbalance of the label.
- Frequent Pattern-based Graph Exploration2
A visual exploration system which provides some frequent sub-graphs extracted from the whole graph, and the relationship between them so that users can raise their study questions more quickly when they first confront a graph dataset is proposed.
- Learning Domain Invariant Word Representations for Parsing Domain Adaptation2
It is shown that strong domain adaptation results for dependency parsing can be achieved using a conceptually simple method that learns domain-invariant word representations by fine-tuning pretrained word representations adversarially.
- Improving Dependency Parsing on Clinical Text with Syntactic Clusters from Web Text2
This paper proposes to gain syntactic knowledge from web text as syntactic cluster features to improve dependency parsing on clinical text and clusters words according to their distributed representation, and uses these syntactic clusters features to solve the data sparseness problem.
- 2
- Head-modifier Relation based Non-lexical Reordering Model for Phrase-Based Translation2
This work proposes a non-lexical head-modifier based reordering model on word level by utilizing constituent based parse tree in source side that significantly outperforms the baseline by 1.48% bleu score.
- A Dirichlet Process Mixture Based Name Origin Clustering and Alignment Model for Transliteration2
A coupled Dirichlet process mixture model (cDPMM) is exploited to address overfitting and names multiorigin cluster issues simultaneously in the transliteration sequence alignment step over the name pairs.
- 2
- Hierarchical Phrase Table Combination for Machine Translation2
Each phrase table is trained separately in each domain, and while computational overhead is significantly reduced by training them in parallel, the performance measured by BLEU is at least as comparable to the traditional batch training method.
- 2
- 2
- 2
- 2
- Combining Multiple Chinese Word Segmentation Results for Statistical Machine Translation2
The basic idea is to use multiple CWS results as additional language knowledge source and the best result shows 1.89 percentage gain in BLEU points over a start-of-the-art SMT system.
- 2
- Two-stage sentence selection approach for multi-document summarization2
The influence of the token chosen in the two-stage sentence selection approach on the quality of the generated summaries is analyzed and proves its validity, compared with the traditional method of sentence selection.
- Weakly Supervised SVM for Chinese- English Cross-lingual Subcategorization Lexicon Acquisition2
Results of experiments indicate statistically significant improvement in the general cross-lingual acquisition performance and the weakly supervised SVM method was also used successfully for bilingual corpus preparation.
- 2
- 2
- Research on a Novel Word Co-occurrence Model and Its Application2
In this study, word sense ontology is firstly constructed by context multi-elements, and then, the usage of word co-occurrence in content was gotten in using part of speech, semantic, location, average co-Occurrence transition probabilities, and was expressed as word co/occurrence feature.
- Incorporating Passage Feature Within Language Model Framework for Information Retrieval2
Experimental results on the TREC collections indicate that the Jelinek-Mercer smoothing scheme significantly outperforms the unigram language model and gets better performance than passage language model in collections whose documents are not very long.
- Two-Fold Filtering for Chinese Subcategorization Acquisition with Diathesis Alternations Used as Heuristic Information2
This paper proposes a two-fold filtering method, which in experiments improved the performance of a Chinese acquisition system remarkably, with an increased precision rate of 76.94% and a recall rate of 83.83%, making the acquired lexicon much more practical for further manual proofreading and other NLP uses.
- Two-Fold Filtering for Chinese Subcategorization Acquisition with Diathesis Alternations Used as2
This paper proposes a two-fold filtering method, which in experiments improved the performance of a Chinese acquisition system remarkably, with an increased precision rate of 76.94% and a recall rate of 83.83%, making the acquired lexicon much more practical for further manual proofreading and other NLP uses.
- 2
- 2
- Automatic Detection of Prosody Phrase Boundaries for Text-to-Speech System2
This work introduced a method based on the transform-based error-driven learning to get the rules for detecting prosodic phrase boundaries, and then used trees to organize the rules in the TTS system.
- 1
- 1
- 1
- Enhancing isomorphism between word embedding spaces for distant languages bilingual lexicon induction1
This work proposes a novel BLI method incorporating synonymous knowledge that significantly outperforms the strong BLI systems in word translation and stabilizes the distance between words to optimize the monolingual THE AUTHORS space, yielding higher isomorphism.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Improving Feature-Rich Transition-Based Constituent Parsing Using Recurrent Neural Networks1
This framework incorporates RNN-based scores into the feature template used by a feature-rich parser and achieves state-of-the-art performance on English WSJ treebank and SPMRL 2014 German treebank, without requiring any additional unlabeled data.
- Re-ranking for Bilingual Lexicon Extraction with Bi-directional Linear Transformation from Comparable Corpora1
A bi-directional linear transformation based re-ranking method by combining the two direction linear score that can achieve a significant improvement of 69% in the precision at Top-1 over the unidirectional baseline approach on the English-to-Chinese bilingual lexicon extraction task.
- 1
- Bilingual Lexicon Extraction with Forced Correlation from Comparable Corpora1
A simple solution based on canonical correlation analysis CCA which forces the bilingual word embedding for training the transformation to be maximally linearly correlated onto the projection subspaces is proposed.
- 1
- 1
- Soft Dependency Matching for Hierarchical Phrase-based Machine Translation1
A soft dependency matching model for hierarchical phrase-based (HPB) machine translation that characterize the structure consistency by three features and integrate them into the standard SMT log-linear model to guide the translation process.
- 1
- The HIT-LTRC Machine Translation System for IWSLT 20121
HIT-LTRC took part in the Olympics Task which required the participants to translate Chinese to English with limited data, and mainly used the phrase-based models to carry out experiments, and factored- based models were also performed in comparison.
- Inferring the Location of an Event in Large Corpus1
A method for automatic discovery of pairs of an event and a place term related by the location, which is based on the analogy with the problem of inferring semantic information from text corpus statistical method of Dunning's likelihood ratio is used.
- 1
- Reexamination on Potential for Personalization in Web Search1
It is quantitatively demonstrated that individual user information is more beneficial than common user information and potential for personalization is calculated to present an overview of what queries can benefit more from individualuser information.
- 1
- PKU_HIT: An Event Detection System Based on Instances Expansion and Rich Syntactic Features1
This paper builds a Naive Bayesian classifier which uses additional training instances expanded from an untagged Chinese corpus automatically for target verb WSD and event detection and uses a feature-based machine learning method which makes combined use of both constituent-based and dependency-based features.
- 1
- 1
- 1
- Application of a Mixed Kernel in the Oil Water-flooded Layer Identification1
A B-spline kernel combined with RBF is developed, a mixed kernel has high recognition accuracy with the advantages of the short running time and is used to automatically identify the water-flooded status of oilsaturated stratum.
- 1
- 1
- 1
- Subcategorization Acquisition Based on Weakly Supervised SVM for Chinese Verbs1
This paper proposes a weakly supervised method for Chinese subcategorization acquisition, where the unsupervised hypothesis generator is replaced with a SVM classifier.
- Chinese Information Processing and Its Prospects1
The paper presents some main progresses and achievements in Chinese information processing and focuses on six aspects, i.e., Chinese syntactic analysis, Chinese semantic analysis, machine translation, information retrieval, information extraction, and speech recognition and synthesis.
- A Hybrid Chinese Language Model based on a Combination of Ontology with Statistical Method1
The structure of a hybrid Chinese language model based on a combination of ontology with statistical method is determined and the proposed method improved the precision of nature language processing.
- 1
- 1
- ASMem: Anchor sparse memory for multi-domain knowledge editing of large language models–
This study proposes anchor sparse memory (ASMem), which is a plug-and-play editing module that isolates edits into parallel memory-specific parameters and introduces a novel anchor prototype routing mechanism to enable precise query-to-memory alignment and reduce semantic interference during editing.
- –
- –
- –
- –
- –
- Enhancing Large Language Models'Machine Translation via Dynamic Focus Anchoring–
This work dynamically analyze and identify translation challenges, then incorporate them into LLMs in a structured manner to mitigate mistranslations or misunderstandings of CSUs caused by information flattening and enhances LLMs' performance across multiple NLP tasks with minimal resource consumption.
- –
- –
- –
- –
- –
- Unsupervised Cross-Lingual Mapping for Phrase Embedding Spaces–
This paper explores the ways to map phrase embeddings of two languages into a common embedding space without supervision and extracted bilingual phrase translation as a gold standard to evaluate the result.
- –
- –
- Building A Case-based Semantic English-Chinese Parallel Treebank–
A case-based English-to-Chinese semantic constituent parallel Treebank is constructed by labelling each node of the Deep Syntactic Tree (DST) with refined semantic cases to represent bilingual shallow semantic relations more precisely.
- A Context-Aware Model Using Distributed Representations for Chinese Zero Pronoun Resolution–
A context-aware model is built and a key words extraction strategy is proposed and a classification model is designed by using distributed representations as context feature in Chinese zero pronoun resolution, the first work of its kind.
- A Maximum Entropy model for Automatic Summarization–
The automatic summarization method based on maximum entropy model is designed and the results of experiments and examples show that the new method has a good practical effect.
- –
- Bilingual Lexicon Extraction with Temporal Distributed Word Representation from Comparable Corpora–
The experimental results confirm that the proposed solution can offer better performance on the English-to-Chinese bilingual lexicon extraction task and exploits the temporal distributed word representation for polysemous words.
- –
- Impact Analysis of the Person in Topic Event Mining.–
Experimental results reported for four well-known topic events in China show that the model could achieve desirable result in analyzing person’s impact on the topic event.
- –
- –
- –
- Filtering Training Corpus and Improving Word Alignment for Statistical Machine Translation–
This paper proposes a corpus filtering approach based on alignment perplexity, and also proposes an improved discriminative word alignment algorithm which can produce alignments with lower alignment error rate.
- Phrasal Syntactic Category Sequence Model for Phrase-Based MT–
A novel phrasal syntactic category sequence (PSCS) model is proposed which allows a PBMT decoder to prefer more grammatical translations and yields a 0.7 BLEU point improvement when compared to the baseline P BMT system.
- Decorated Phrase Model and Syntax-Based Reordering Model for Statistical Machine Translation–
A novel phrase model is introduced, in which treebank tags are employed to decorate the bilingual phrase pairs, and tag sequences, instead of phrase pairs are used to train the lexicalized reordering model.
- Improved response to face recognition in UHF fMRI through parallel transmission RF pulse design–
It is successfully demonstrated that PTX 3DTRF design can be used for improving the fMRI signal response in amygdala and the proposed methodology was found to be robust across volunteers, especially when the 5-spoke trajectory was used.
- –
- Research on the Translation Model Combination Based on Hypergraph–
A translation model combination is introduced, which combines two mainstream translation models (Hiero and MaxEnt-based BTG) during decoding, to solve the spurious ambiguity and consensus problem.
- –
- –
- –
- –
- –
- A Deterministic Method to Predict Phrase Boundaries of a Syntactic Tree–
A deterministic model to predict all the phrase boundaries of a syntactic tree, including base constituent boundaries and nested constituent boundaries is presented, which shows that phrase boundary identification could be done without phrase labels and could achieve comparable result to Stanford parser.
- Acquisition of Argument Correspondence between Chinese and English Verb Subcategorization–
A novel method to obtain SCF argument crrespondence between Chinese and English based on active learning and it is shown that the performance of SMT combined bilingual argument relationships has significant improvement, which indicates the validity of argument corresponding relationships automatically obtained.
- Statistical Analysis for Chinese-English Verb Subcategorization–
Based on large scale Chinese-English parallel corpus, a systematic experiment of statistical analysis for bilingual verb subcategorization is described, with lexical and grammatical compatibility as heuristics andlinguistic classification according to Chinese and English syntax.
- –
- –
- –
- –
- –
- Comparative Study on Hierarchical Phrase Structures and Linguistic Phrase Structures.–
This paper proposes a framework for analysis of SMT translations output from a hierarchical phrase decoder that will explore automatic or semi-automatic identification and correction of some translation errors based on comparison between hierarchical phrase structures and linguistic phrase structures.
- –
- –
- Spatial Semantic Analysis Based on a Cognitive Approach–
In order to extract and represent spatial semantics from linguistic expressions, a method is proposed based on an extension of Kinstch’s Construction-Integration model that can implicitly integrate the contextual and general knowledge into the calculating process.
- Multi-agent distributed English chunking on real public corpus–
Through testing on the real public corpus, F score of English chunking using multi-agent model achieves to 92%, which almost satisfies the practical need.
- –
- –
- A Phrase Combination Approach to Patent SMT–
A phrase combination approach to patent SMT (Statistical Machine Translation) for Japanese to English to minimize the segmentation problems caused by the rich OOV words in the patent texts.
- Comparative Study of Statistical Translation Models on Chinese-English Speech Translation–
By analyzing and comparing the translation results, it's found that the performance of the phrase-based translation model is the best and the syntax-based model is not the best, but it uses less parameters than the other two.
- Improving English subcategorization acquisition with diathesis alternations as heuristic information–
A new filtering method is proposed, which improved the performance of Korhonen's acquisition system remarkably, with the precision increased to 91.18% and recall unchanged, making the acquired lexicon much more practical for further manual proofreading and other NLP uses.
- –
- Linguistic Knowledge Representation and Automatic Acquisition Based on a Combination of Ontology with Statistical Method–
A combination of ontology with statistical method is presented for linguistic knowledge representation and acquisition from training data and the proposed method improves the precision of nature language processing.
- Relation Links Extracted Approach Based on Blocking Links–
A novel approach is proposed to extract relation links from page based on link block with above 85% precise rate and about 70% recall rate and the result of experiment shows that the method is effective.
- –
- –
- Research on identification technology of basic noun phrases in English–
The typical methods that how to identify English noun phrase are summarized and the results of these methods are compared and the trend in the research of noun phrase identification is poin-(ted) out.
- Decision tree based Chinese base noun phrase identification–
A self-learning mechanism is integrated into the model which includes the following steps: auto-extraction of POS string sequences (BNP rules) and their context information from the corpora and ID3 algorithm based tree training.
- –
- Domain-Specific Term-List Expansion Using Existing Linguistic Resources–
The finding is that the use of a bilingual term list for the expansion proces does not provide a significant improvement over theUse of a simpler, more easily produced, monoligual term list.
- Automatic information transfer between English and Chinese–
The translation choice and transfer modules in an English Chinese machine translation system are introduced and the Bayes minimal error probability is taken as the evaluation function of the candidate translation.
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 2026-10-15. 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.