Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileWorks210 from public data
- Math-Shepherd: Verify and Reinforce LLMs Step-by-step without Human Annotations1,249
An innovative process-oriented math process reward model called Math-Shepherd, which assigns a reward score to each step of math problem solutions, which holds significant potential for the future evolution of LLMs.
- A Survey on In-context Learning1,205
This paper presents a formal definition of ICL and clarify its correlation to related studies, and organizes and discusses advanced techniques, including training strategies, prompt designing strategies, and related analysis.
- Knowledge Neurons in Pretrained Transformers785
This paper examines the fill-in-the-blank cloze task for BERT and proposes a knowledge attribution method to identify the neurons that express the fact, finding that the activation of such knowledge neurons is positively correlated to the expression of their corresponding facts.
- Why Can GPT Learn In-Context? Language Models Implicitly Perform Gradient Descent as Meta-Optimizers420
Inspired by the understanding of meta-optimization, a momentum-based attention is designed by analogy with the momentum- based gradient descent algorithm and its consistently better performance over vanilla attention supports the understanding from another aspect.
- Table-to-Text Generation by Structure-Aware Seq2seq Learning286
The attention visualizations and case studies show that the novel structure-aware seq2seq architecture which consists of field-gating encoder and description generator with dual attention is capable of generating coherent and informative descriptions based on the comprehensive understanding of both the content and the structure of a table.
- 277
- CBLUE: A Chinese Biomedical Language Understanding Evaluation Benchmark262
The first Chinese Biomedical Language Understanding Evaluation Evaluation (CBLUE) benchmark is presented: a collection of natural language understanding tasks including named entity recognition, information extraction, clinical diagnosis normalization, single-sentence/sentence-pair classification, and an associated online platform for model evaluation, comparison, and analysis.
- Jointly Extracting Event Triggers and Arguments by Dependency-Bridge RNN and Tensor-Based Argument Interaction261
A novel dependency bridge recurrent neural network (dbRNN) is proposed that simultaneously applying tree structure and sequence structure in RNN brings much better performance than only uses sequential RNN.
- Speculative Decoding: Exploiting Speculative Execution for Accelerating Seq2seq Generation217
Experimental results on various seq2seq tasks including machine translation and abstractive summarization show the proposed Speculative Decoding approach can achieve around $5\times$ speedup for the popular Transformer architectures with comparable generation quality to beam search decoding, refreshing the impression that the draft-then-verify paradigm introduces only $1.4\times$$\sim $ speedup.
- A Token-level Reference-free Hallucination Detection Benchmark for Free-form Text Generation210
This work proposes a novel token-level, reference-free hallucination detection task and an associated annotated dataset named HaDeS (HAllucination DEtection dataSet), and utilizes an iterative model-in-loop strategy to mitigate label imbalance during annotation.
- 190
- Towards Time-Aware Knowledge Graph Completion184
A novel time-aware knowledge graph completion model that is able to predict links in a KG using both the existing facts and the temporal information of the facts and achieves the state-of-the-art on temporal facts consistently.
- A Soft-label Method for Noise-tolerant Distantly Supervised Relation Extraction150
This work introduces an entity-pair level denoise method which exploits semantic information from correctly labeled entity pairs to correct wrong labels dynamically during training, and proposes a joint score function which combines the relational scores based on the entity- Pair representation and the confidence of the hard label to obtain a new label.
- 147
- Exploiting collaborative filtering techniques for automatic assessment of student free-text responses144
The experimental results show that the feature-based matrix factorization model outperforms the baseline models and the model with a re-ranking phase can achieve a better and competitive performance -- 63.6% overall accuracy on the Beetle dataset.
- Encoding Temporal Information for Time-Aware Link Prediction136
A novel time-aware KB embedding approach taking advantage of the happening time of facts is proposed, which uses temporal order constraints to model transformation between time-sensitive relations and enforce the embeddings to be temporally consistent and more accurate.
- Synthetic Data (Almost) from Scratch: Generalized Instruction Tuning for Language Models123
This work introduces Generalized Instruction Tuning (called GLAN), a general and scalable method for instruction tuning of Large Language Models (LLMs) that exclusively utilizes a pre-curated taxonomy of human knowledge and capabilities as input and generates large-scale synthetic instruction data across all disciplines.
- 117
- Order-Planning Neural Text Generation From Structured Data117
This paper proposes an order-planning text generation model, where order information is explicitly captured by link-based attention and a self-adaptive gate combines the link- based attention with traditional content-based Attention.
- Implicit Discourse Relation Classification via Multi-Task Neural Networks117
This work designs related discourse classification tasks specific to a corpus, and proposes a novel Convolutional Neural Network embedded multi-task learning system to synthesize these tasks by learning both unique and shared representations for each task.
- Incorporating Glosses into Neural Word Sense Disambiguation107
GAS models the semantic relationship between the context and the gloss in an improved memory network framework, which breaks the barriers of the previous supervised methods and knowledge-based methods.
- Making Large Language Models Better Reasoners with Alignment94
The constraint alignment loss has two objectives: a) Alignment, which guarantees that positive scores surpass negative scores to encourage answers with high-quality COTs, and b) Constraint, which keeps the negative scores confined to a reasonable range to prevent the model degradation.
- XGPT: Cross-modal Generative Pre-Training for Image Captioning86
XGPT is a new method of Cross-modal Generative Pre-Training for Image Captioning that is designed to pre-train text-to-image caption generators through three novel generation tasks, including Image-conditioned Masked Language Modeling (IMLM), Image- Conditioned Denoising Autoencoding (IDA), and Text-conditioning Image Feature Generation (TIFG).
- Stabilizing MoE Reinforcement Learning by Aligning Training and Inference Routers73
Rollout Routing Replay (R3), a method that records routing distributions from the inference engine and replays them during training significantly reduces training-inference policy KL divergence and mitigates extreme discrepancies without compromising training speed, is proposed.
- An Enhanced Span-based Decomposition Method for Few-Shot Sequence Labeling73
SD decomposes the span matching problem into a series of span-level procedures, mainly including enhanced span representation, class prototype aggregation and span conflicts resolution, which achieves the new state-of-the-art results on two popular FSSL benchmarks, FewNERD and SNIPS.
- Reading and Thinking: Re-read LSTM Unit for Textual Entailment Recognition68
A new L STM unit called re-read LSTM (rLSTM) is designed to implement this “thinking” process, which means to read the premise again and again while reading the hypothesis to get a better understanding of the premise.
- Leveraging Gloss Knowledge in Neural Word Sense Disambiguation by Hierarchical Co-Attention66
This paper introduces a co-attention mechanism to generate co-dependent representations for the context and gloss, and extends the attention mechanism in a hierarchical fashion in order to capture both word-level and sentence-level information.
- 58
- Learning to Control the Fine-grained Sentiment for Story Ending Generation54
To the best of the knowledge, this is the first endeavor to control the fine-grained sentiment for story ending generation without manually annotating sentiment labels.
- 50
- 47
- 47
- Towards Fine-grained Text Sentiment Transfer47
This paper proposes a novel Seq2SentiSeq model, where the numeric sentiment intensity value is incorporated into the decoder via a Gaussian kernel layer to finely control the sentiment intensity of the output.
- RBPB: Regularization-Based Pattern Balancing Method for Event Extraction46
Inspired by the progress in representation learning, trigger embedding, sentence-level embedding and pattern features together as their features for trigger classification are used so that the effect of patterns and other useful features can be balanced.
- Hierarchical Encoder with Auxiliary Supervision for Neural Table-to-Text Generation: Learning Better Representation for Tables45
A two-level hierarchical encoder with coarse-to-fine attention to handle the attribute-value structure of the tables, and proposes 3 joint tasks apart from the prime encoder-decoder learning, namely auxiliary sequence labeling task, text autoencoder and multi-labeling classification, as the auxiliary supervisions for the table encoder.
- 43
- Self-Boosting Large Language Models with Synthetic Preference Data41
This work introduces SynPO, a self-boosting paradigm that leverages synthetic preference data for model alignment and trains LLMs to autonomously learn the generative rewards for their own outputs and eliminates the need for large-scale annotation of prompts and human preferences.
- Chinese Semantic Role Labeling with Bidirectional Recurrent Neural Networks41
Bidirectional recurrent neural network with long-short-term memory (LSTM) with RNN to capture bidirectional and long-range dependencies in a sentence with minimal feature engineering is introduced.
- Reinforcement Pre-Training40
This work reframe next-token prediction as a reasoning task trained using RL, where it receives verifiable rewards for correctly predicting the next token for a given context, and positions RPT as an effective and promising scaling paradigm to advance language model pre-training.
- Towards Faithfulness in Open Domain Table-to-text Generation from an Entity-centric View38
Two methods for faithful generation are proposed: augmented training by incorporating the auxiliary entity information, including both an augmented plan-based model and an unsupervised model and 2) training instance selection based on faithfulness ranking which improve generation fidelity in both full dataset setting and few shot setting.
- Chinese semantic role labeling with shallow parsing38
This paper evaluates SRL methods that take partial parses as inputs and implements SRL systems which cast SRL as the classification of syntactic chunks with IOB2 representation for semantic roles (i.e. semantic chunks).
- 37
- 37
- Pun-GAN: Generative Adversarial Network for Pun Generation35
Experiments show that the proposed Pun-GAN can generate sentences that are more ambiguous and diverse in both automatic and human evaluation.
- ERSOM: A Structural Ontology Matching Approach Using Automatically Learned Entity Representation35
An ontology matching approach, named ERSOM, which mainly includes an unsupervised representation learning method based on the deep neural networks to learn the general representation of the entities and an iterative similarity propagation method that takes advantage of more abundant structure information of the ontology to discover more mappings.
- Joint Learning Templates and Slots for Event Schema Induction31
A joint entity-driven model to learn template and slots simultaneously based on the constraints of templates and slots in the same sentence is proposed and the entities' semantic information is also considered for the inner connectivity of the entities.
- A Multi-View Fusion Neural Network for Answer Selection29
A Multi-View Fusion Neural Network, where each attention component generates a ``view'' of the QA pair and a fusion RNN integrates the generated views to form a more holistic representation.
- Large-Scale Simple Question Generation by Template-Based Seq2seq Learning28
A 28M Chinese Q&A corpora based on the Chinese knowledge base provided by NLPCC2017 KBQA challenge is presented and a novel neural network architecture which combines template-based method and seq2seq learning to generate highly fluent and diverse questions is proposed.
- AdaMMS: Model Merging for Heterogeneous Multimodal Large Language Models with Unsupervised Coefficient Optimization27
As the first model merging method capable of merging heterogeneous MLLMs without labeled data, extensive experiments on various model combinations demonstrated that AdaMMS outperforms previous model merging methods on various vision-language benchmarks.
- Towards Accurate Distant Supervision for Relational Facts Extraction27
This paper points out and analyzes some critical factors in DS which have great impact on accuracy, including valid entity type detection, negative training examples construction and ensembles, and proposes an approach to handle these factors.
- An Empirical Study on Model-agnostic Debiasing Strategies for Robust Natural Language Inference26
This paper benchmarks prevailing neural NLI models including pretrained ones on various adversarial datasets and tries to combat distinct known biases by modifying a mixture of experts (MoE) ensemble method and shows that it’s nontrivial to mitigate multiple NLI biases at the same time, and that model-level ensemble method outperforms MoE ensemble method.
- Bring you to the past: Automatic Generation of Topically Relevant Event Chronicles25
Experimental results demonstrate the first novel approach to automatically generate a topically relevant event chronicle during a certain period given a reference chronicles during another period is promising to tackle this new problem.
- Learning Robust Representations for Continual Relation Extraction via Adversarial Class Augmentation24
- Towards Comprehensive Description Generation from Factual Attribute-value Tables24
This work proposes force attention (FA) method to encourage the generator to pay more attention to the uncovered attributes to avoid potential key attributes missing, and proposes reinforcement learning for information richness to generate more informative as well as more loyal descriptions for tables.
- Event Detection with Burst Information Networks22
Two fast centroid-aware event detection models based on a novel text stream representation – Burst Information Networks (BINets) and BINet-based approaches are proposed for addressing the challenge of retrospective event detection.
- Modeling Consumer Buying Decision for Recommendation Based on Multi-Task Deep Learning21
This paper proposes a multi-task learning model with long short-term memory networks (LSTM) to learn consumer buying decision process and corresponding stages and can capture user intentions and preferences, predicts the conversion rate of each candidate product, and makes recommendations accordingly.
- PeriodicLoRA: Breaking the Low-Rank Bottleneck in LoRA Optimization20
PeriodicLoRA (PLoRA), which accumulates low-rank update matrices multiple times to achieve a higher update rank and is introduced, introduces a momentum-based unloading strategy for PLoRA to mitigate the training instability.
- Inductively Representing Out-of-Knowledge-Graph Entities by Optimal Estimation Under Translational Assumptions20
This work proposes a simple and effective method that inductively represents OOKG entities by their optimal estimation under translational assumptions and shows that it outperforms the previous methods by a large margin with higher efficiency.
- Affinity-Preserving Random Walk for Multi-Document Summarization20
This paper introduces affinity-preserving random walk to the summarization task, which preserves the affinity relations of sentences by an absorbing random walk model to enforce the diversity constraint of summarization in the random walk process.
- Large Language Model for Science: A Study on P vs. NP19
This work proposes Socratic reasoning, a general framework that promotes in-depth thinking with LLMs for complex problem-solving that encourages LLMs to recursively discover, solve, and integrate problems while facilitating self-evaluation and refinement.
- Hierarchical Curriculum Learning for AMR Parsing17
A Hierarchical Curriculum Learning (HCL) framework with Structure-level (SC) and Instance-level Curricula (IC) that reduces the difficulty of learning complex structures, thus the flat model can better adapt to the AMR hierarchy.
- Large Language Models Struggle with Unreasonability in Math Problems16
This work proposes the Unreasonable Math Problems (UMP) benchmark, designed to evaluate LLMs' ability to detect and respond to unreasonable math problem statements, and finds that prompting and fine-tuning enhance the detection of unreasonable inputs, with minor and acceptable trade-offs, making them practical solutions in this challenging setting.
- 16
- Behind the Scenes16
To cope with the context-bypassing problem in FSEC models, adversarial training and trigger reconstruction techniques are introduced and experiments show these techniques help not only improve the performance, but also enhance the generalization ability of models.
- Formal Schema of Diagrammatic Chinese Syntactic Analysis16
A new formal schema of diagrammatic Chinese syntactic analysis based on sentence pattern structure, including diagrammatic style and its XML structure is put forward.
- A Spectral Method for Unsupervised Multi-Document Summarization15
A spectral-based hypothesis is proposed, which states that the goodness of summary candidate is closely linked to its so-called spectral impact, and an accelerated greedy solution based on a surrogate of spectral impact is proposed.
- Premise-based Multimodal Reasoning: Conditional Inference on Joint Textual and Visual Clues14
Inspired by the designs of both visual commonsense reasoning and natural language inference tasks, a new task is proposed termed “Premise-based Multi-modal Reasoning” (PMR) where a textual premise is the background presumption on each source image.
- News Stream Summarization using Burst Information Networks14
Extensive experiments show that the BINet-based approaches are not only efficient and can be used in a real-time online summarization setting, but also can generate high-quality summaries, outperforming the state-of-the-art approach.
- Application of cascade binary pointer tagging in joint entity and relation extraction of Chinese medical text13
Two models based on the cascade binary pointer tagging network (CBPTN) with conditional layer normalization and biaffine transformation-based CBPTN with multi-head selection are proposed and tested on two Chinese medical datasets: CMeIE and CEMRDS.
- Decompose, Fuse and Generate: A Formation-Informed Method for Chinese Definition Generation13
This paper builds a formation-informed dataset, and proposes a model DeFT, which Decomposes words into formation features, dynamically Fuses different features through a gating mechanism, and generaTes word definitions.
- The CIPS-SIGHAN CLP 2014 Chinese Word Segmentation Bake-off13
This paper summarizes the SIGHAN 2014 Chinese Word Segmentation bakeoff in several aspects such as dataset, evaluation results, and analyze errors of segmentation by instance and make a suggestion for improving segmentation systems.
- 12
- 12
- 12
- 12
- Coarse-to-Fine Entity Representations for Document-Level Relation Extraction10
This work proposes the coarse-to-fine CFER model, a graph-based model that adopts a coarse- to-fine strategy involving two phases that combines the entity representations from both two levels into more comprehensive representations for relation extraction.
- Less is More: Rethinking State-of-the-art Continual Relation Extraction Models with a Frustratingly Easy but Effective Approach10
This paper proposes a Frustratingly easy but Effective Approach (FEA) method with two learning stages for CRE, and finds that two strong CRE baselines can be subsumed into the proposed training pipeline.
- 10
- Event-Based Time Label Propagation for Automatic Dating of News Articles10
The experiments show that the event-based propagation model can predict document timestamps in high accuracy and the model combined with a MaxEnt classifier outperforms the state-ofthe-art method for this task especially when the size of the training set is small.
- 9
- Statistical Dataset Evaluation: Reliability, Difficulty, and Validity9
This paper proposes a model-agnostic dataset evaluation framework for automatic dataset quality evaluation and seeks the statistical properties of the datasets and addresses three fundamental dimensions: reliability, difficulty, and validity, following a classical testing theory.
- 9
- Revisiting Distant Supervision for Relation Extraction9
The primary aim is to re-examine the distant supervision-based approaches under the experimental settings without the above issues by training models on the same dataset and creating a new testing dataset annotated by the workers on Amzaon Mechanical Turk.
- 9
- Recognizing Textual Entailment Using Probabilistic Inference9
This paper proposes to use the predicate-argument structures to represent the discourse commitments extracted from text and borrows the distant supervision technique to mine the implicit facts from the text.
- 9
- Extracting Hyponymy Relation between Chinese Terms9
This paper treats the task of identifying hyponymy relation as two separate problems and solve them based on the following three techniques: term type's commonality, sequential patterns, property nouns and domain verbs.
- 8
- 8
- Discriminatively-Tuned Generative Classifiers for Robust Natural Language Inference8
This paper proposes GenNLI, a generative classifier for NLI tasks, and empirically characterize its performance by comparing it to five baselines, including discriminative models and large-scale pretrained language representation models like BERT.
- 8
- 8
- 8
- Prediction of maximal projection for semantic role labeling8
This paper makes a new definition of MP in S-structure and proposes two methods to predict it: the anchor group approach and the single anchor approach and results indicate that the prediction of MP improves semantic role labeling.
- A Study on Terminology Extraction Based on Classified Corpora8
Three algorithms are provided and evaluated to investigate termhood based on classified corpora and the effects of these different termhood measurement features are investigated, to find which features are more effective and also, how to improve these different features to achieve the best performance.
- 8
- Chain-of-Thought Tokens are Computer Program Variables7
It is found that preserving only tokens that store intermediate results would achieve comparable performance, suggesting that CoT tokens may function like variables in computer programs but with potential drawbacks like unintended shortcuts and computational complexity limits between tokens.
- 7
- WSD-GAN: Word Sense Disambiguation Using Generative Adversarial Networks7
This paper proposes a generative adversarial framework WSD-GAN which combines two mainstream methods in WSD and achieves competitive results on several English all-words WSD datasets.
- A Progressive Learning Approach to Chinese SRL Using Heterogeneous Data7
A novel progressive learning model is proposed which augments the Progressive Neural Network with Gated Recurrent Adapters and can accommodate heterogeneous inputs and effectively transfer knowledge between them, to improve Chinese SRL by using heterogeneous corpora together.
- The integration of dependency relation classification and semantic role labeling using bilayer maximum entropy Markov models7
A bilayer directed graph is presented to express probabilistic relationships between syntactic and semantic relations and Maximum Entropy Markov Models are implemented to estimate conditional probability distribution and to do inference.
- Study on Evolution of Domain Ontology7
This paper reconstructs and uses such recognized ontology knowledge as professional thesaurus, professional dictionary and textbook and realizes knowledge collection and manipulation to build auto-learning system of depended text's ontology finally to set up automatic construction of domain ontology’s concept description system.
- 7
- 7
- Chinese Function Tag Labeling7
It is shown that good performance for Chinese function tagging can be achieved by using labeling method, extending the work of Blaheta (2004), in which the objects being modeled are syntax trees which require some mechanism to convert them into feature vectors.
- 6
- FSM: A Finite State Machine Based Zero-Shot Prompting Paradigm for Multi-Hop Question Answering6
FSM improves LLMs' ability to follow specified output format requirements, significantly reducing the difficulty of answer interpretation and the need for reformatting, and improves LLMs' ability to follow specified output format requirements.
- Statistical dataset evaluation: A case study on named entity recognition6
This paper proposes a model-agnostic dataset evaluation framework for automatic dataset quality evaluation and introduces nine statistical metrics for a statistical dataset evaluation framework, following a Classical Test Theory (CTT).
- ImageNetVC: Zero- and Few-Shot Visual Commonsense Evaluation on 1000 ImageNet Categories6
This work proposes ImageNetVC, a human-annotated dataset specifically designed for zero- and few-shot visual commonsense evaluation across 1,000 ImageNet categories, and benchmark the fundamental visual commonsens knowledge of both unimodal LLMs and VaLMs.
- 6
- 6
- 6
- 6
- 5
- Recognizing Textual Entailment via Multi-task Knowledge Assisted LSTM5
A deep neural network architecture called Multi-task Knowledge Assisted LSTM (MKAL) is proposed, which aims to conduct implicit inference with the assistant of KB and use predicate-to-predicate attention to detect the entailment between predicates.
- 5
- 5
- Research on the Key Technology Based-NLP Of Chinese Medicine Pulse's Mathematical Quantifying5
A mathematical model of ancient pulse is constructed by utilizing the technique and method of natural language processing (NLP), according to traditional Chinese medicine theory and pulse-symptom knowledge, so as to form the open system of pulsepsila theory and to realize the development of it.
- A Collocation-Based WSD Model: RFR-SUM5
A WSD Model RFR-SUM Model is put forward to disambiguate polysemous Chinese word sense to evaluate the strength of collocation, and it reveals that leaving only 5% important collocation information can keep almost the same precision.
- 4
- 4
- 4
- Will Repeated Reading Benefit Natural Language Understanding?4
It is found that repeated reading mechanism do improve performance on some tasks but not on others (sentiment analysis, semantic relation classification, event extraction), and how these differences may be caused in each of the tasks is discussed.
- Study on Modality Annotation Framework of Modern Chinese4
A Chinese modality annotation framework aimed for deep semantic understanding is preliminarily practiced, which constructed a modal meaning classification system on the basis of existing research results, built the modal operator dictionary, established rules for annotation, and annotated modal operators of a sentence which have been tagged basic proposition arguments.
- Discovering Concept-Level Event Associations from a Text Stream4
This work investigates the importance and challenge of this task and proposes a novel solution by using event sequential patterns, which can discover important event associations implicitly expressed.
- Reducing Hallucinations in Entity Abstract Summarization with Facts-Template Decomposition3
Experimental results demonstrate that SlotSum could generate summaries that are significantly more factual with credible external knowledge and benefit from the facts-template decomposition.
- Be a Multitude to Itself: A Prompt Evolution Framework for Red Teaming3
This paper proposes RTPE, a scalable evolution framework to evolve red teaming prompts across both breadth and depth dimensions, facilitating the automatic generation of numerous high-quality and diverse red teaming prompts.
- 3
- 3
- SeRI: A Dataset for Sub-event Relation Inference from an Encyclopedia3
A novel dataset called SeRI (Sub-event Relation Inference), which includes 3,917 event articles from English Wikipedia and the annotations of their sub-events and can be used for training or evaluating a model that mines sub-event relation from encyclopedia-style texts.
- Distantly Supervised Neural Network Model for Relation Extraction3
A novel approach to alleviate the problem of distant supervision with representation learning in the framework of deep neural network that constructs the more powerful mention level representation by tensor-based transformation and further learns the entity pair level representation which aggregates and denoises the features of associated mentions.
- Research on Chinese Semantic Role Labeling Based on Shallow Parsing3
In shallow parsing stage, this paper makes use of word formation to get fake head morpheme information of the target verb, which alleviates the problem of data sparseness, and imporves the performance of the parser with the F-score up to 0.93.
- 3
- Prediction of thematic rank for structured semantic role labeling3
Experimental results show that automatic prediction of thematic hierarchy can help semantic role classification and a loglinear model is proposed to accurately identify thematic rank between two arguments.
- 3
- 2
- 2
- Plug-and-Play Training Framework for Preference Optimization2
A novel training framework is proposed, which employs multiple sampling to analyze output distributions, assign different weights to samples, and incorporate these weights into the preference optimization process, which enables LLMs to prioritize challenging examples during training, improving learning efficiency.
- 2
- 2
- 2
- 2
- 2
- An Anchor-Based Automatic Evaluation Metric for Document Summarization2
A new protocol for designing reference-based metrics that require the endorsement of source document(s) is considered and an anchored ROUGE metric fixing each summary particle on source document, which bases the computation on more solid ground is proposed.
- Fine-grained Coordinated Cross-lingual Text Stream Alignment for Endless Language Knowledge Acquisition2
Experiments show this approach not only outperforms previous approaches on bilingual lexicon extraction from coordinated text streams but also can harvest high-quality alignments from large amounts of streaming data for endless language knowledge mining, which makes it promising to be a new paradigm for automatic language knowledge acquisition.
- A Research on Construction of Knowledge Fusion Network in Chinese and English Languages2
A prototype of Chinese-English lexical and encyclopedic knowledge fusion network CELK-Net is created by an automatic mapping algorithm inspired by Babelnet, a localized bilingual semantic network with large coverage in Chinese language.
- 2
- One Tense per Scene: Predicting Tense in Chinese Conversations2
A set of novel sentence-level features are proposed using rich linguistic resources and a new hypothesis of “One tense per scene” is proposed to incorporate scene-level (global) evidence to enhance the performance.
- A semi-supervised method for opinion target extraction2
A semi-supervised self-learning method, which is based on a Naive Bayes classifier exploiting context features and PMI scores, to extract opinion targets is proposed, which outperforms the state-of-the-art models on COAE2008 dataset2, especially in precision.
- 2
- Event Schema Induction Based on Relational Co-occurrence over Multiple Documents2
Experiments show that co-occurrence information over multiple documents can help to com- bine similar elements of event schema as well as to alleviate incoherence problems.
- Learning to Extract Attribute Values from a Search Engine with Few Examples2
An attribute value extraction method based on analysing snippets from a search engine where a pattern based detector is applied to locate the candidate attribute values in snippets and a classifier is used to predict whether a candidate value is correct.
- Semantic Role Labeling System Based on Dependency Tree Distance Method for Arguments Identification2
A novel rule based method for the semantic role identification according to the dependency tree distance, which recognizes 98.5% of arguments on CoNLL 2009 Chinese dataset and obtains best candidate arguments related to the verb.
- 2
- 2
- 2
- Automatic Extraction of Ontology Attribute Value Based on WWW2
A method of extracting Ontology attribute values automatically based on WWW is proposed and an interactive method is described to unilize the interaction between the attribute-value-related sentence selection and the attribute values extraction.
- 2
- The Acquisition of Chinese Verb's Subcategorization Frame Types Based on Linguistic Theory and Statistical Algorithm2
A semi-supervise method is applied for subcategorization frame type acquisition with linguistic theory and statistical algorithm and gains better precision and coverage.
- 2
- An Information-Theoretic Empirical Analysis of Dependency-Based Feature Types for Word Prediction Models2
A quantitative analysis of the information gain and the information redundancy for various combinations of feature types inspired by both dependency structure and bigram structure is described, using a Chinese treebank and taking word prediction as the object.
- 1
- 1
- 1
- 1
- RepCL: Exploring Effective Representation for Continual Text Classification1
This work formally analyzes representation bias from an information bottleneck perspective and suggests that exploiting representations with more class-relevant information could alleviate the bias.
- 1
- 1
- Distinguishing Specific and Daily Topics1
This paper investigates several numeric features that describe useful statistical information for this task, and proposes a novel Bayesian model for distinguishing specific and daily topics from a collection of documents based on documents’ content.
- Semantic Class Learning with Deep Coordinate Structures in Web Pages1
This paper introduces a kind of descriptive structure, the deep coordinate structure, which universally exists in web pages, which is used for the task of semantic class learning and compares it with several state-of-art systems.
- Weakly-Supervised Extraction of Ontology Concept Instances and Concept Attributes from the Web1
A weakly-supervised method of extracting Ontology concept instances and attributes from the Web is proposed and the precision of the top 500 and 1000 results reaches 94% and 93%,respectively.
- An Integrated Approach for Concept Learning and Relation Extraction1
This paper proposes an integrated approach to do the two tasks together, where the attribute values of concepts are used to evaluate the extracted hierarchical relations and the weight of intermediate results of both tasks are introduced into the iteration to ensure the accuracy of results.
- A WSD Model for Corpus Construction1
RFR-SUM is an approach to the automatic disambiguation of polysemous Chinese words that calculates and then sums the collocational strengths of words in the local context of a given word.
- 1
- 1
- Domain Knowledge Engineering Based on Encyclopedias and the Web Text1
This paper intends to draw a basic framework for the construction of domain knowledge base through technologies in natural language text analysis and machine learning, using encyclopedia resources and text information resources on the Web.
- –
- –
- Exploring Activation Patterns of Parameters in Language Models–
This work utilizes a gradient-based metric to assess the activation level of model parameters and finds that two sentences with consistent semantics tend to share similar parameter activation patterns in deep layers, which aligns with the third finding.
- Enhancing Reliability across Short and Long-Form QA via Reinforcement Learning–
This work introduces a targeted RL framework designed to mitigate both intrinsic and extrinsic hallucinations across short and long-form question answering, paving the way for more capable and reliable large language models.
- –
- –
- Achilles-Bench: A Challenging Benchmark for Low-Resource Evaluation–
This work finds that there exists a set of “hard examples” in low-resource settings that challenge neural networks but are not well evaluated, which causes over-estimated performance.
- Chip-Tuning: Classify Before Language Models Say–
Chip-tuning is proposed, a simple and effective structured pruning framework specialized for classification problems that significantly outperforms previous state-of-the-art baselines in both accuracy and pruning ratio, achieving a pruning ratio of up to 50%.
- Bi-Drop: Enhancing Fine-tuning Generalization via Synchronous sub-net Estimation and Optimization–
Bi-Drop, a fine-tuning strategy that selectively updates model parameters using gradients from various sub-nets dynamically generated by dropout, demonstrates excellent generalization ability and robustness for domain transfer, data imbalance, and low-resource scenarios.
- –
- –
- –
- –
- –
- –
- –
- Construction of the Contemporary Chinese Common Verbs’ Semantic Framework Dictionary–
The lexical framework of common Chinese verbs is defined, which contains a detailed description and corresponding example sentences of 2,782 common verbs with 4,516 meanings.
- Soap: Soaking Capacity Optimization for Multi-Document Summarization–
This paper considers the MDS task as an optimization problem with a novel measure named soaking capacity being the objective function and proposes a greedy algorithm named Soap by adopting a surrogate of soaking capacity to accelerate the computation.
- Study on the Annotation Framework of Chinese Logic Complement Semantics–
A Chinese logic complement semantics annotation framework aimed for deep semantic comprehension is preliminarily practiced, which constructed a classification system including negation, degree, tense and aspect, and mood on the basis of existing research results, and built the operator dictionary and established rules for annotation.
- Multi-domain Chinese-English Terminology Database–
This release of the multi-domain Chinese-English terminology database contains four sample of domain specialized terminology databases: Sports, Tourism, Information Technology, Computational Linguistics and English, Chinese, Japanese, Russian, German.
- –
- –
- –
- –
- The Construction of language Resource and Knowledge Base for Chinese Language Computing–
This talk will introduce the building of the multi-level annotated Chinese language resource, the ontology engineering based on encyclopedias and the Web, and the construction of the mapping resource between conceptual relations and their natural language expressions to link linguistic knowledge and the world knowledge together.
- To Extract Concept Instances and Concept Attributes Based on Coordinate Structure–
Experimental results show that the coverage of this method of extracting concept instances and concept attributes based on the coordinate structure is significantly improved without reducing the precision.
- –
- –
- –
- Research on Automatic Chinese Multi-word Term Extraction Based on Term Component–
This paper presents an automatic Chinese multi-word term extraction method based on the unithood and the termhood measure, Inspired by the economical law of term generating, which proposes two measures of a candidate term to be a true term.
- –
- –
- A New Weighted Evidence Combination Method–
The new formula makes up the deficiency of the formula combined by absorption method, and makes the result of conflicted evidence combination more effective and proves the validity of the new method.
- An information-theory-based feature type analysis for the modelling of statistical parsing–
An information-theory-based method for feature types analysis in probabilistic evaluation modelling for statistical parsing uses entropy and conditional entropy to measure whether a feature type grasps some of the information for syntactic structure prediction.
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.