Is this you? Claim this profile to correct it, add a bio and choose the work people see first.
Claim this profileAcademic lineage
Possible advisorsa guess from early papers, not confirmed
- Stephen Charles ClarkSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works334 from public data
- A Survey on Evaluation of Large Language Models3,896
This paper presents a comprehensive review of these evaluation methods for LLMs, focusing on three key dimensions: what to evaluate, where to evaluate, and how to evaluate, and offers invaluable insights to researchers in the realm of LLMs evaluation.
- 🧜Siren’s Song in the AI Ocean: A Survey on Hallucination in Large Language Models1,188
This paper presents taxonomies of the LLM hallucination phenomena and evaluation benchmarks, analyzes existing approaches aiming at mitigating LLm hallucination, and discusses potential directions for future research.
- Template-Based Named Entity Recognition Using BART407
A template-based method for NER is proposed, treating NER as a language model ranking problem in a sequence-to-sequence framework, where original sentences and statement templates filled by candidate named entity span are regarded as the source sequence and the target sequence, respectively.
- Transition-based Dependency Parsing with Rich Non-local Features374
This paper shows that it can improve the accuracy of transition-based dependency parsers by considering even richer feature sets than those employed in previous systems by improving the accuracy in the standard Penn Treebank setup and rivaling the best results overall.
- PromptRobust: Towards Evaluating the Robustness of Large Language Models on Adversarial Prompts351
This study uses a plethora of adversarial textual attacks targeting prompts across multiple levels: character, word, sentence, and semantic to demonstrate that contemporary LLMs are not robust to adversarial prompts.
- DialogSum: A Real-Life Scenario Dialogue Summarization Dataset338
Experimental results show unique challenges in dialogue summarization, such as spoken terms, special discourse structures, coreferences and ellipsis, pragmatics and social common sense, which require specific representation learning technologies to better deal with.
- A Tale of Two Parsers: Investigating and Combining Graph-based and Transition-based Dependency Parsing338
A beam-search-based parser that combines both graph-based and transition-based parsing into a single system for training and decoding is proposed, showing that it outperforms both the pure graph- based and the pure transition- based parsers.
- Gated Neural Networks for Targeted Sentiment Analysis311
A sentence-level neural model is proposed to address the limitation of pooling functions, which do not explicitly model tweet-level semantics and gives significantly higher accuracies compared to the current best method for targeted sentiment analysis.
- A Graph-to-Sequence Model for AMR-to-Text Generation269
This work introduces a neural graph-to-sequence model, using a novel LSTM structure for directly encoding graph-level semantics, and shows superior results to existing methods in the literature.
- Neural Networks for Open Domain Targeted Sentiment247
This work empirically studies the effect of word embeddings and automatic feature combinations on the open domain targeted sentiment task by extending a CRF baseline using neural networks, and proposes a novel integration of neural and discrete features which combines their relative advantages, leading to significantly higher results compared to both baselines.
- Sentence-State LSTM for Text Representation226
This work investigates an alternative LSTM structure for encoding text, which consists of a parallel state for each word, and shows that the proposed model has strong representation power, giving highly competitive performances compared to stacked BiLSTM models with similar parameter numbers.
- Syntactic Processing Using the Generalized Perceptron and Beam Search221
It is argued that the conceptual and computational simplicity of the framework, together with its language-independent nature, make it a competitive choice for a range of syntactic processing tasks and one that should be considered for comparison by developers of alternative approaches.
- Tweet Sarcasm Detection Using Deep Neural Network217
This work investigates the use of neural network for tweet sarcasm detection, and compares the effects of the continuous automatic features with discrete manual features.
- Automatic Features for Essay Scoring – An Empirical Study212
This paper employs convolutional neural network for the effect of automatically learning features, and compares the result with the state-of-art discrete baselines.
- Lexicon Enhanced Chinese Sequence Labeling Using BERT Adapter204
This paper proposes Lexicon Enhanced BERT (LEBERT) for Chinese sequence labeling, which integrates external lexicon knowledge into BERT layers directly by a Lexicon Adapter layer, and facilitates deep lexiconknowledge fusion at the lower layers of BERT.
- N-ary Relation Extraction using Graph-State LSTM201
This work proposes a graph-state LSTM model, which uses a parallel state to model each word, recurrently enriching state values via message passing, and speeds up computation by allowing more parallelization.
- Evaluating Commonsense in Pre-Trained Language Models200
This work studies the commonsense ability of GPT, BERT, XLNet, and RoBERTa by testing them on seven challenging benchmarks, finding that language modeling and its variants are effective objectives for promoting models' commonsens ability while bi-directional context and larger training set are bonuses.
- MuTual: A Dataset for Multi-Turn Dialogue Reasoning182
MuTual is introduced, a novel dataset for Multi-Turn dialogue Reasoning, consisting of 8,860 manually annotated dialogues based on Chinese student English listening comprehension exams, which shows that there is ample room for improving reasoning ability.
- CoSDA-ML: Multi-Lingual Code-Switching Data Augmentation for Zero-Shot Cross-Lingual NLP179
A data augmentation framework to generate multi-lingual code-switching data to fine-tune mBERT, which encourages model to align representations from source and multiple target languages once by mixing their context information.
- The INTERSPEECH 2017 Computational Paralinguistics Challenge: Addressee, Cold & Snoring174
These sub-challenges, their conditions, and the baseline feature extraction and classifiers are described, which include data-learnt feature representations by end-to-end learning with convolutional and recurrent neural networks, and bag-of-audiowords for the first time in the challenge series.
- 170
- 164
- Leveraging Context Information for Natural Question Generation162
This work proposes a model that matches the answer with the passage before generating the question and shows that this model outperforms the existing state of the art using rich features.
- Context-Sensitive Twitter Sentiment Classification Using Neural Network151
A context-based neural network model for Twitter sentiment analysis is proposed, incorporating contextualized features from relevant Tweets into the model in the form of word embedding vectors.
- Joint Word Segmentation and POS Tagging Using a Single Perceptron144
This paper proposes a joint segmentation and POS tagging model that does not impose any hard constraints on the interaction between word and POS information, and uses a discriminative statistical model, trained using the generalized perceptron algorithm.
- Latent Emotion Memory for Multi-Label Emotion Classification143
A Latent Emotion Memory network (LEM) is proposed for multi-label emotion classification that can learn the latent emotion distribution without external knowledge, and can effectively leverage it into the classification network.
- Chinese Segmentation with a Word-Based Perceptron Algorithm141
Open tests on the first and second SIGHAN bakeoffs show that the proposed word-based segmentor is competitive with the best in the literature, achieving the highest reported F-scores for a number of corpora.
- Inducing Target-Specific Latent Structures for Aspect Sentiment Classification140
This work proposes gating mechanisms to dynamically combine information from word dependency graphs and latent graphs which are learned by self-attention networks to complement supervised syntactic features with latent semantic dependencies inpect-level sentiment analysis.
- Cross-Domain NER using Cross-Domain Language Modeling135
This work considers using cross- domain LM as a bridge cross-domains for NER domain adaptation, performing cross-domain and cross-task knowledge transfer by designing a novel parameter generation network and shows that this method can effectively extract domain differences from cross- domains LM contrast, allowing unsupervised domain adaptation while also giving state-of-the-art results among supervised domain adaptation methods.
- 129
- Transition-Based Neural Word Segmentation126
Experimental results demonstrate that word features lead to comparable performance to the best systems in the literature, and a further combination of discrete and neural features obtains top accuracies on several benchmarks.
- What Have We Achieved on Text Summarization?122
It is found that under similar settings, extractive summarizers are in general better than their abstractive counterparts thanks to strength in faithfulness and factual-consistency, and pre-training techniques, and in particular sequence-to-sequence pre- training, are highly effective for improving text summarization, with BART giving the best results.
- 121
- YEDDA: A Lightweight Collaborative Text Span Annotation Tool119
Yedda provides a systematic solution for text span annotation, ranging from collaborative user annotation to administrator evaluation and analysis, and overcomes the low efficiency of traditional text annotation tools by annotating entities through both command line and shortcut keys.
- Does it Make Sense? And Why? A Pilot Study for Sense Making and Explanation116
A benchmark to directly test whether a system can differentiate natural language statements that make sense from those that do not make sense is released and models trained over large-scale language modeling tasks as well as human performance are evaluated, showing that there are different challenges for system sense-making.
- GLUE-X: Evaluating Natural Language Understanding Models from an Out-of-distribution Generalization Perspective114
The findings confirm the need for improved OOD accuracy in NLP tasks, as significant performance degradation was observed in all settings compared to in-distribution (ID) accuracy.
- SemEval-2020 Task 4: Commonsense Validation and Explanation107
This paper presents SemEval-2020 Task 4,CommonsenseValidation andExplanation (ComVE), which includes three subtasks, aiming to evaluate whether a system can distinguish anatural language statement that makes senseto humans from one that does not, and provide thereasons.
- Transition-based parsing of the Chinese treebank using a global discriminative model105
A global discriminative model based on the shift-reduce parsing process, combined with a beam-search decoder, obtaining competitive accuracies on CTB2 and obtaining the highest scores in the literature for a dependency-based evaluation.
- Can Generative Pre-trained Language Models Serve As Knowledge Bases for Closed-book QA?97
A new dataset of closed-book QA is constructed using SQuAD, and the performance of BART is investigated, showing that it is challenging for BART to remember training facts in high precision, and also challenging to answer closed- book questions even if relevant knowledge is retained.
- G-Transformer for Document-Level Machine Translation97
G-Transformer is proposed, introducing locality assumption as an inductive bias into Transformer, reducing the hypothesis space of the attention from target to source, and achieving new state-of-the-art BLEU scores for both nonpretraining and pre-training settings on three benchmark datasets.
- Solving Aspect Category Sentiment Analysis as a Text Generation Task97
This work considers a more direct way of making use of pre-trained language models, by casting the ACSA tasks into natural language generation tasks, using natural language sentences to represent the output.
- A Neural Probabilistic Structured-Prediction Model for Transition-Based Dependency Parsing95
A neural probabilistic structured-prediction model for transition-based dependency parsing, which integrates search and learning is proposed, giving performance comparable to the best linear parser.
- A Fast Decoder for Joint Word Segmentation and POS-Tagging Using a Single Discriminative Model92
It is shown that the standard beam-search algorithm can be used as an efficient decoder for the global linear model of Zhang and Clark (2008) for joint word segmentation and POS-tagging, achieving a significant speed improvement.
- 85
- Structural Adapters in Pretrained Language Models for AMR-to-Text Generation83
The benefits of explicitly encoding graph structure into PLMs using StructAdapt are empirically shown, outperforming the state of the art on two AMR-to-text datasets, training only 5.1% of the PLM parameters.
- Modeling Global and Local Node Contexts for Text Generation from Knowledge Graphs83
This work gathers both encoding strategies, proposing novel neural models that encode an input graph combining both global and local node contexts, in order to learn better contextualized node embeddings.
- Sentiment-aware volatility forecasting83
Experimental results suggest that this model not only outperforms pure statistical models, e.g., GARCH and its variants, Gaussian-process volatility model, but also outperforms the state-of-the-art autoregressive deep neural nets architectures, such as the variational recurrent neural network and the neural stochastic volatility model.
- Financial Sentiment Analysis: An Investigation into Common Mistakes and Silver Bullets81
Investigating the error patterns of some widely acknowledged sentiment analysis methods in the finance domain finds that those methods belonging to the same clusters are prone to similar error patterns, and there are six types of linguistic features that are pervasive in the common errors.
- 80
- Porous Lattice Transformer Encoder for Chinese NER80
PLTE is proposed, an extension of transformer encoder that is tailored for ChineseNER, which models all the characters and matched lexical words in parallel with batch process-ing and introduces a porous mechanism to augment localness modeling and maintain the strength of capturing the rich long-term dependencies.
- A Pilot Study for Chinese SQL Semantic Parsing80
A Spider dataset for Chinese is built, showing that word-based semantic parser is subject to segmentation errors and cross-lingual word embeddings are useful for text-to-SQL.
- In-Order Transition-based Constituent Parsing80
A novel parsing system based on in-order traversal over syntactic trees, designing a set of transition actions to find a compromise between bottom-up constituent information and top-down lookahead information is proposed.
- AMR-to-text Generation with Synchronous Node Replacement Grammar77
This paper addresses the task of AMR-to-text generation by leveraging synchronous node replacement grammar by leveragingynchronous node Replacement grammar and gives the state-of-the-art result.
- Character-Level Chinese Dependency Parsing76
This paper presents novel adaptations of two major shift-reduce dependency parsing algorithms to character-level parsing, and demonstrates improved performances over word-based parsing methods on the Chinese Treebank.
- Entity Enhanced BERT Pre-training for Chinese NER75
A semi-supervised entity enhanced BERT pre-training method that extracts an entity lexicon from the relevant raw text using a new-word discovery method and integrates the entity information into BERT using Char-Entity-Transformer, which augments the self-attention using a combination of character and entity representations.
- Combining Discrete and Neural Features for Sequence Labeling74
The results on standard benchmarks show that state-of-the-art neural models can give accuracies comparable to the best discrete models in the literature for most tasks and combing discrete and neural features unanimously yield better results.
- 74
- Domain Adaptation for CRF-based Chinese Word Segmentation using Free Annotations73
Experimental results show that the Chinese word segmentation model benefits from free partially annotated data on the SIGHAN Bakeoff 2010 data, and different sources of free annotations are transformed into a unified form of partial annotation.
- MuCGEC: a Multi-Reference Multi-Source Evaluation Dataset for Chinese Grammatical Error Correction72
This paper presents MuCGEC, a multi-reference multi-source evaluation dataset for Chinese Grammatical Error Correction (CGEC), consisting of 7,063 sentences collected from three Chinese-as-a-Second-Language (CSL) learner sources, and conducts experiments with two mainstream CGEC models, both enhanced with large pretrained language models.
- Extracting Entities and Events as a Single Task Using a Transition-Based Neural Model72
This work builds a first model to this end using a neural transition-based framework, incrementally predicting complex joint structures in a state-transition process, and shows the benefits of the joint model, which gives the best result in the literature.
- 70
- Open Domain Event Extraction Using Neural Latent Variable Models67
This work considers open domain event extraction, the task of extracting unconstraint types of events from news clusters, and proposes a novel latent variable neural model, which is scalable to very large corpus.
- 66
- NumHTML: Numeric-Oriented Hierarchical Transformer Model for Multi-Task Financial Forecasting65
A numeric-oriented hierarchical transformer model (NumHTML) to predict stock returns, and financial risk using multi-modal aligned earnings calls data by taking advantage of the different categories of numbers and their magnitude.
- 64
- Emoji-Based Sentiment Analysis Using Attention Networks64
This article investigates a sentiment analysis model based on bidirectional long short-term memory, and the model has two advantages compared with the existing work: it does not need feature engineering and utilizes the attention approach to model the impact of emojis on text.
- 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.
- Tables as Texts or Images: Evaluating the Table Reasoning Ability of LLMs and MLLMs63
This study provides insights into the effective use of LLMs on table-related tasks and introduces for the first time the assessment of LLMs' performance on image-based table representations.
- Natural Language Inference in Context - Investigating Contextual Reasoning over Long Texts62
ConTRoL is a new dataset for ConTextual Reasoning over Long texts, a passage-level NLI dataset with a focus on complex contextual reasoning types such as logical reasoning, derived from competitive selection and recruitment test for police recruitment with expert level quality.
- Subword Encoding in Lattice LSTM for Chinese Word Segmentation62
This work investigates subword information for Chinese word segmentation, by integrating sub word embeddings trained using byte-pair encoding into a Lattice LSTM (LaLSTM) network over a character sequence.
- Structural Information Preserving for Graph-to-Text Generation59
This work introduces two types of autoencoding losses, each individually focusing on different aspects (a.k.a. views) of input graphs, that can guide the model for preserving input information via multi-task training.
- 58
- Exploring the Efficacy of Automatically Generated Counterfactuals for Sentiment Analysis56
This work proposes an alternative by describing and evaluating an approach to automatically generating counterfactual data for the purpose of data augmentation and explanation and demonstrates how this approach can achieve significant improvements in model performance when compared to models training on the original data and even whenCompared to models trained with the benefit of human-generated augmented data.
- Coarse-to-Fine Pre-training for Named Entity Recognition55
A NER-specific pre-training framework to in-ject coarse-to-fine automatically mined entity knowledge into pre-trained models and devise a self-supervisedauxiliary task to mine the fine-grained namedentity knowledge via clustering.
- Leveraging Local and Global Patterns for Self-Attention Networks55
The extensive analyses verify that the two types of contexts are complementary to each other, and the proposed hybrid attention mechanism gives highly effective improvements in their integration.
- Contrastive Attention Mechanism for Abstractive Sentence Summarization55
The proposed contrastive attention mechanism is more focused on the relevant parts for the summary than the conventional attention mechanism, and greatly advances the state-of-the-art performance on the abstractive sentence summarization task.
- Multi-prototype Chinese Character Embedding55
Evaluation on character similarity shows that multi-prototype embeddings are significantly better than a single-prototype baseline, and used as features in the Chinese NER task, the embedded characters result in a 1.74% F-score improvement over a state-of-the-art baseline.
- On Compositional Generalization of Neural Machine Translation54
This paper quantitatively analyze effects of various factors using compound translation error rate, then demonstrates that the NMT model fails badly on compositional generalization, although it performs remarkably well under traditional metrics.
- Unsupervised Domain Adaptation for Joint Segmentation and POS-Tagging52
An empirical investigation on type-supervised domain adaptation for joint Chinese word segmentation and POS-tagging, making use of domainspecific tag dictionaries and only unlabeled target domain data to improve target-domain accuracies, given a set of annotated source domain sentences.
- Word-Context Character Embeddings for Chinese Word Segmentation51
This work investigates training character embeddings on a word-based context in a similar way, showing that the simple method improves state-of-the-art neural word segmentation models significantly, beating tri-training baselines for leveraging auto-segmented data.
- Transition-Based Disfluency Detection using LSTMs51
This model incrementally constructs and labels the disfluency chunk of input sentences using a new transition system without syntax information, which can capture non-local chunk-level features and is free for noise in syntax.
- Survey on Factuality in Large Language Models50
This survey offers a structured guide for researchers aiming to fortify the factual reliability of LLMs, and examines the mechanisms through which LLMs store and process facts, seeking the primary causes of factual errors.
- Contrastive Data and Learning for Natural Language Processing50
This tutorial intends to help researchers in the NLP and computational linguistics community to understand this emerging topic and promote future research directions of using contrastive learning for NLP applications.
- Alignment-Enhanced Transformer for Constraining NMT with Pre-Specified Translations50
This work introduces a dedicated head in the multi-head Transformer architecture to capture external supervision signals and shows that this method is highly effective in constraining NMT with pre-specified translations, consistently outperforming previous methods in translation quality.
- Shift-Reduce Constituent Parsing with Neural Lookahead Features50
A bidirectional LSTM model is built, which leverages full sentence information to predict the hierarchy of constituents that each word starts and ends and gives the highest reported accuracies for fully-supervised parsing.
- ZORE: A Syntax-based System for Chinese Open Relation Extraction49
A syntax-based Chinese ORE system, ZORE, for extracting relations and semantic patterns from Chinese text by identifying relation candidates from automatically parsed dependency trees and then extracts relations with their semantic patterns iteratively through a novel double propagation algorithm.
- Leveraging Dependency Forest for Neural Medical Relation Extraction45
A graph neural network is used to represent the forests, automatically distinguishing the useful syntactic information from parsing noise, and shows that this method outperforms the standard tree-based methods, giving the state-of-the-art results in the literature.
- Cross-Lingual Dependency Parsing Using Code-Mixed TreeBank44
Experiments on University Dependency Treebanks show that code-mixed treebanks are more effective than translated treebanks, giving highly competitive performances among cross-lingual parsing methods.
- 43
- A span-graph neural model for overlapping entity relation extraction in biomedical texts42
Experimental results on two biomedical entity relation extraction tasks, including drug-drug interaction detection and protein-protein interaction detection, show that the proposed method outperforms previous models by a substantial margin, demonstrating the effectiveness of span-graph based method for overlapping relation extraction in biomedical texts.
- Personality Alignment of Large Language Models40
Inspired by psychometrics, the Personality Alignment with Personality Inventories (PAPI) dataset is created, which includes data from over 320,000 real subjects across multiple personality assessments, including both the Big Five Personality Factors and Dark Triad traits.
- Learning 6-DoF Fine-Grained Grasp Detection Based on Part Affordance Grounding39
This paper designs a two-stage fine-grained robotic grasping framework, including a novel 3D part language grounding model and a part-aware grasp pose detection model, in which explicit language input from human or large language models (LLMs) could guide a robot to generate part-level 6-DoF grasping pose with textual explanation.
- Understanding In-Context Learning from Repetitions39
This work quantitatively investigate the role of surface features in text generation, and empirically establish the existence of token co-occurrence reinforcement, a principle that strengthens the relationship between two tokens based on their contextual co-Occurrences.
- Natural Language Processing39
This gentle introduction to natural language processing develops fundamental mathematical and deep learning models for NLP under a unified framework, equipping students with the tools needed to establish a deep understanding of NLP solutions, adapt existing models, and confidently develop innovative models of their own.
- Tree Communication Models for Sentiment Analysis39
A tree communication model using graph convolutional neural network and graph recurrent neural network, which allows rich information exchange between phrases constituent tree, and outperforms existing work on bidirectional tree-LSTMs in both accuracy and efficiency.
- Entity Relation Extraction as Dependency Parsing in Visually Rich Documents37
This paper adapts the popular dependency parsing model, the biaffine parser, to this entity relation extraction task, and identifies relations between groups of words with layout information instead.
- Transition-Based Syntactic Linearization37
This work extends the transition system so that word ordering is performed in addition to syntactic parsing, resulting in a linearization system that runs in guaranteed quadratic time and runs an order of magnitude faster than a state-of-the-art baseline using best-first search.
- Natural Language Processing Meets Quantum Physics: A Survey and Categorization36
This survey reviews representative methods at the intersection of NLP and quantum physics in the past ten years, categorizing them according to the use of quantum theory, the linguistic targets that are modeled, and the downstream application.
- 35
- 35
- Head-Lexicalized Bidirectional Tree LSTMs35
This paper proposes a method for automatic head-lexicalization for tree-structure LSTMs, propagating head words from leaf nodes to every constituent node, and builds a tree LSTM in the top-down direction, which corresponds to bidirectional sequential LSTm in structure.
- Neural Reranking for Named Entity Recognition35
A neural reranking system for named entity recognition (NER), leverages recurrent neural network models to learn sentence-level patterns that involve named entity mentions, and can significantly improve the NER accuracies over two different baselines.
- AMR-to-text generation as a Traveling Salesman Problem35
A Maximum Entropy classifier is trained to estimate the traveling costs, and a TSP solver is used to find the optimized solution, which reports a BLEU score of 22.44 on the SemEval-2016 Task8 dataset.
- Word Segmentation for Chinese Novels35
A novel double-propagation algorithm is designed that mines nounentities together with common contextual patterns, and uses them as plug-in features to a model trained on the source domain to improve a state-of-the-art segmentation model for the novel.
- Discriminative Syntax-Based Word Ordering for Text Generation35
This article studies word ordering using a syntax-based approach and a discriminative model, and develops a learning-guided search framework, based on best-first search, and investigates several alternative training algorithms.
- Neural Network for Heterogeneous Annotations34
This paper empirically investigate heterogeneous annotations using neural network models, building a neural network counterpart to discrete stacking and multi-view learning, respectively, finding that neural models have their unique advantages thanks to the freedom from manual feature engineering.
- Syntax-Aware Opinion Role Labeling with Dependency Graph Convolutional Networks33
This work introduces multi-task learning (MTL) to train the parser and the ORL model simultaneously and proposes dependency graph convolutional networks (DEPGCN) to encode parser information at different processing levels.
- Learning Target-Specific Representations of Financial News Documents For Cumulative Abnormal Return Prediction32
A novel target-specific abstract-guided news document representation model that uses a target-sensitive representation of the news abstract to weigh sentences in the news content, so as to select and combine the most informative sentences for market modeling.
- Event-Driven Headline Generation32
This work proposes an event-driven model for headline generation that can be viewed as a novel combination of extractive and abstractive headline generation, combining the advantages of both methods using event structures.
- 31
- Shift-Reduce CCG Parsing with a Dependency Model31
This paper presents the first dependency model for a shift-reduce CCG parser, and develops a novel training technique using a dependency oracle, in which all derivations are hidden.
- Supervised Knowledge Makes Large Language Models Better In-context Learners30
Using the proposed plug-in method, enhanced versions of Llama 2 and ChatGPT surpass their original versions regarding generalizability and factuality and the potential of the methodology in fostering more reliable LLMs is highlighted.
- Lattice LSTM for Chinese Sentence Representation30
This work investigates a neural representation that simultaneously encodes character and word information without the need for segmentation, and compares lattice LSTM structure with its sequential L STM counterpart, namely ELMo.
- Multiscale Collaborative Deep Models for Neural Machine Translation30
This paper presents a MultiScale Collaborative (MSC) framework to ease the training of NMT models that are substantially deeper than those used previously and provides empirical evidence showing that the MSC nets are easy to optimize and can obtain improvements of translation quality from considerably increased depth.
- Syntax grounded graph convolutional network for joint entity and event extraction29
This work proposes to use syntax features in an implicit approach, by adopting Soft Head Vectors drawn from a well-trained parser as an adjacency matrix, densely connecting all words with the probabilistic head scores.
- Syntax-Based Grammaticality Improvement using CCG and Guided Search29
This paper studies grammaticality improvement using a syntax-based algorithm based on ccg that gives a BLEU score higher than the previous result of 33.7 achieved by a dependency-based system.
- Out-of-Distribution Generalization in Natural Language Processing: Past, Present, and Future28
This paper presents the first comprehensive review of recent progress, methods, and evaluations on the generalization challenge from an OOD perspective in natural language understanding, and hopes this survey will encourage future research in this area.
- AMR Parsing with Latent Structural Information28
This work investigates parsing AMR with explicit dependency structures and interpretable latent structures and fuse both dependency and latent structure via an extended graph neural networks to generate the latent soft structure without additional annotations.
- A Transition-based Model for Joint Segmentation, POS-tagging and Normalization28
Experimental results show that the proposed transition-based model can help improve the performance of word segmentation on microblogs, giving an error reduction in segmentation accuracy of 12.02%, compared to the traditional approach.
- 27
- Using Context-to-Vector with Graph Retrofitting to Improve Word Embeddings27
This paper aims to improve word embeddings by incorporating more contextual information from existing pre-trained models into the Skip-gram framework, which it calls Context-to-Vec, and proposing a post-processing retrofitting method for static embedDings independent of training by employing priori synonym knowledge and weighted vector distribution.
- Roadmap toward the metaverse: An AI perspective27
A roadmap toward the metaverse: An AI perspective from an AI perspective is presented for the first time in what is likely to be a landmark moment in the development of artificial intelligence.
- A Unified Span-Based Approach for Opinion Mining with Syntactic Constituents27
The experimental results show that the proposed unified span-based approach achieves significant improvements over previous works in the exact F1 score and reduces the number of wrongly-predicted opinion expressions and roles, showing the effectiveness of the method.
- 27
- DialogSum Challenge: Summarizing Real-Life Scenario Dialogues26
This work carefully annotates a large-scale dialogue summarization dataset based on multiple public dialogue corpus, opening the door to all kinds of summarization models.
- Dialogue State Induction Using Neural Latent Variable Models26
This work proposes the task of dialogue state induction, building two neural latent variable models that mine dialogue states automatically from unlabeled customer service dialogue records and shows that the models can effectively find meaningful dialogue states.
- Knowledge Enhanced Fine-Tuning for Better Handling Unseen Entities in Dialogue Generation25
This work introduces two auxiliary training objectives: Interpret Masked Word, which conjectures the meaning of the masked entity given the context; and Hypernym Generation, which predicts the hypernym of the entity based on the context.
- Does BERT Solve Commonsense Task via Commonsense Knowledge25
This work proposes two attention-based methods to analyze commonsense knowledge inside BERT, and finds that attention heads successfully capture the structured commonsenseknowledge encoded in ConceptNet, which helps BERT solve commonsense tasks directly.
- Joint Word Segmentation, POS-Tagging and Syntactic Chunking25
A joint model that performs segmentation, POS-tagging and chunking simultaneously, and a semi-supervised method to derive chunk cluster features from large-scale automatically-chunked data to address the sparsity of full chunk features.
- UniSumm and SummZoo: Unified Model and Diverse Benchmark for Few-Shot Summarization24
Experimental results and analysis show that UniSumm outperforms strong baselines by a large margin across all sub-tasks in SummZoo under both automatic and human evaluations and achieves comparable results in human evaluation compared with a GPT-3.5 model.
- SemGloVe: Semantic Co-Occurrences for GloVe From BERT24
SemGloVe is proposed, which distills semantic co-occurrences from BERT into static GloVe word embeddings and can define the co- Occurrence weights by directly considering the semantic distance between word pairs.
- News-driven stock prediction via noisy equity state representation24
To the knowledge, this work is the first to explicitly model both events and noise over a fundamental stock value state for news-driven stock movement prediction and shows that the proposed model outperforms strong baselines.
- Exploring Generalization Ability of Pretrained Language Models on Arithmetic and Logical Reasoning24
This research finds that the PLMs can easily generalize when the distribution is the same, however, it is still difficult for them to generalize out of the distribution.
- Syntax-Based Word Ordering Incorporating a Large-Scale Language Model24
This work takes the Zhang and Clark system as the baseline, and incorporates an N-gram model by applying online large-margin training, which significantly improved on the baseline by 3.7 BLEU points.
- See What LLMs Cannot Answer: A Self-Challenge Framework for Uncovering LLM Weaknesses23
A Self-Challenge evaluation framework with human-in-the-loop to demonstrate that LLMs can autonomously identify their inherent flaws and provide insights for future dynamic and automatic evaluation.
- ZPR2: Joint Zero Pronoun Recovery and Resolution using Multi-Task Learning and BERT23
Joint modeling significantly outperforms the baseline that already beats the previous state of the arts and investigates their interaction by solving both tasks together, while the previous work treats them separately.
- Active Learning for Dependency Parsing with Partial Annotation23
This paper is the first to apply a probabilistic model to active learning for dependency parsing, which can provide tree probabilities and dependency marginal probabilities as principled uncertainty metrics, and directly learn parameters from PA based on a forest-based training objective.
- Towards Event-level Causal Relation Identification22
This work proposes a model that achieves the best performance under both mention- and event-level settings, outperforming a number of strong baselines by at least 2.8% on F1 score.
- Overview of the NLPCC 2019 Shared Task: Cross-Domain Dependency Parsing22
An overview of the NLPCC 2019 shared task on cross-domain dependency parsing, including (1) the data annotation process, (2) task settings, methods, results, and analysis of submitted systems and the authors' recent work (Li+19), and discussions on related works and future directions.
- Semi-Supervised Feature Transformation for Dependency Parsing22
This paper proposes a novel semi-supervised approach to addressing the problem of data sparseness problem by transforming the base features into high-level features (i.e. meta features) with the help of a large amount of automatically parsed data.
- Spotting AI's Touch: Identifying LLM-Paraphrased Spans in Text21
A novel detection framework, paraphrased text span detection (PTD), aiming to identify paraphrased text spans within a text, and shows that PTD models can generalize to versatile paraphrasing prompts and multiple paraphrased text spans.
- Towards Fine-grained Causal Reasoning and QA21
A novel fine-grained causal reasoning dataset is introduced and a series of novel predictive tasks in NLP, such as causality detection, event causality extraction, and Causal QA are presented.
- Syntax-Enhanced Self-Attention-Based Semantic Role Labeling21
The experiment results demonstrate that with proper incorporation of the high quality syntactic information, the model achieves a new state-of-the-art performance for the Chinese SRL task on the CoNLL-2009 dataset.
- Expectation-Regulated Neural Model for Event Mention Extraction21
A LSTM-based neural model is proposed that learns tweet-level features automatically and is more effective compared with a state-of-the-art discrete baseline for disambiguating subtle semantic differences between true event mentions and false cases that use similar wording patterns.
- 21
- 20
- Chart Pruning for Fast Lexicalised-Grammar Parsing20
This paper investigates two forms of chart pruning, and develops a novel method for pruning complete cells in a parse chart, resulting in a wide-coverage CCG parser that can process almost 100 sentences per second, with little or no loss in accuracy over the baseline with no pruning.
- 19
- End-to-End AMR Corefencence Resolution19
This work introduces the first end-to-end AMR coreference resolution model in order to build multi-sentence AMRs, which alleviates error propagation and is more robust for both in-domain and out-domain situations.
- Investigating Self-Attention Network for Chinese Word Segmentation19
This work investigates self-attention network (SAN) for Chinese word segmentation, making comparisons between BiLSTM-CRF models and the influence of contextualized character embeddings using BERT, and proposes a method for integrating word information into SAN segmentation.
- Multi-task and multi-view training for end-to-end relation extraction19
Experimental results show that the MTMVL model outperforms the state-of-the-art methods, which indicates the effectiveness of multi-auxiliary information for knowledge transferring.
- Exploiting meta features for dependency parsing and part-of-speech tagging19
A learning framework of feature transformation is presented, addressing the sparseness problem by transforming sparse conventional base features into less sparse high-level features (i.e. meta features) with the help of a large amount of automatically annotated data.
- 18
- Online Back-Parsing for AMR-to-Text Generation18
This work proposes a decoder that back predicts projected AMR graphs on the target sentence during text generation and can better preserve the input meaning than standard decoders.
- Joint Morphological Generation and Syntactic Linearization18
Joint morphological generation and linearization is studied, making use of word order and inflections information for both tasks and reducing error propagation, demonstrating significantly outperforms a strong pipelined baseline.
- Challenges to Open-Domain Constituency Parsing17
This work analyzes challenges to open-domain constituency parsing using a set of linguistic features on various strong constituency parsers and finds that BERT significantly increases parsers’ cross-domain performance by reducing their sensitivity on the domain-variant features.
- 17
- SUDA-Alibaba at MRP 2019: Graph-Based Models with BERT16
This paper describes participating systems in the shared task on Cross- Framework Meaning Representation Parsing (MRP) at the 2019 Conference for Computational Language Learning (CoNLL), which includes five frameworks for graph-based meaning representations, i.e., DM, PSD, EDS, UCCA, and AMR.
- RefuteBench: Evaluating Refuting Instruction-Following for Large Language Models15
A comprehensive benchmark, RefuteBench, is proposed, covering tasks such as question answering, machine translation, and email writing, to assess whether models can positively accept feedback in form of refuting instructions and whether they can consistently adhere to user demands throughout the conversation.
- 15
- 15
- Multi-domain Spoken Language Understanding Using Domain- and Task-aware Parameterization15
This article proposes to improve the parameterization of this method by using domain-specific and task-specific model parameters for fine-grained knowledge representation and transfer and shows its transferability when adapting to a new domain with little data, outperforming the prior best model.
- On Commonsense Cues in BERT for Solving Commonsense Tasks15
Using two different measures, it is found that BERT does use relevant knowledge for solving the task, and the presence of commonsense knowledge is positively correlated to the model accuracy.
- Topic-informed neural approach for biomedical event extraction15
A variational neural approach for biomedical event extraction is proposed, which can take advantage of latent topics underlying documents and is able to produce more meaningful and event-indicative words compare to prior topic models.
- Dependency Parsing with Partial Annotations: An Empirical Comparison15
The results show that LLG par is most effective in directly learning from PA, and other parsers can achieve best performance when PAs are completed into full trees by LLGPar.
- LibN3L:A Lightweight Package for Neural NLP15
A light-weight machine learning tool for NLP research that provides several basic layers which mainly aims for single-layer linear and non-linear transformations, facilitating implementation of linear models as well as neural models.
- 14
- 14
- Learning How to Self-Learn: Enhancing Self-Training Using Neural Reinforcement Learning14
A deep reinforcement learning method is proposed to learn self-training strategy automatically based on the neural representation of sentences and the hidden features from classifiers, designed to capture their linguistic characteristics and learn an optimal policy for instance selection automatically.
- MCRanker: Generating Diverse Criteria On-the-Fly to Improve Pointwise LLM Rankers13
This work proposes to build a zero-shot pointwise ranker that first recruits a virtual annotation team to generate query-based criteria from various perspectives and then uses these criteria to conduct an ensemble passage evaluation.
- 13
- Smelting Gold and Silver for Improved Multilingual AMR-to-Text Generation13
It is found that combining both complementary sources of information further improves multilingual AMR-to-text generation and surpass the previous state of the art for German, Italian, Spanish, and Chinese by a large margin.
- Enhancing Generalization in Natural Language Inference by Syntax13
This work investigates the use of dependency trees to enhance the generalization of BERT in the NLI task, leveraging on a graph convolutional network to represent a syntax-based matching graph with heterogeneous matching patterns.
- Syntactic Dependencies and Distributed Word Representations for Analogy Detection and Mining13
It is shown that distributed representations of dependency structure can be used for measuring relational similarities, thereby help analogy mining and empirical investigate the use of syntactic dependencies on improving Chinese analogy detection based on distributed word representations.
- Syntactic Dependencies and Distributed Word Representations for Chinese Analogy Detection and Mining13
It is shown that distributed representations of dependency structure can be used for measuring relational similarities, thereby help analogy mining and empirical investigate the use of syntactic dependencies on improving Chinese analogy detection based on distributed word representations.
- Punctuation Processing for Projective Dependency Parsing13
This work proposes an alternative approach to addressing punctuation in dependency parsing that treats punctu- ations as properties of their neighbour- ing words, used as features to guide the parser to build the dependency graph.
- Quantum data parallelism in quantum neural networks12
The effective application of quantum parallelism is demonstrated, via quantum superposition and entanglement, to achieve data parallelism in generic quantum neural network models, and it is shown that parallel computing remains viable, and quantum interference among data becomes notable.
- Consistency Regularization Training for Compositional Generalization12
This work improves the capability of Transformer on compositional generalization through consistency regularization training, which promotes representation consistency across samples and prediction consistency for a single sample.
- 12
- 12
- A Neural Network Approach to Verb Phrase Ellipsis Resolution12
Experimental results show that the neural models outperform the state-of-the-art baselines in both subtasks and the end-to-end results, comparing the differences between statistical and neural models.
- Joint Location and Velocity Estimation and Fundamental CRLB Analysis for Cell-Free MIMO-ISAC11
A logarithmic likelihood function (LLF) is formulated for both single- and multiple-target sensing under deterministic, unknown, and spatially varying radar cross-section (RCS) models and closed-form Cramér-Rao lower bounds (CRLBs) for joint location and velocity estimation are derived.
- Deep Learning Approaches to Text Production. By Shashi Narayan and Claire Gardent (University of Edinburgh; CNRS/LORIA, Nancy). Morgan & Claypool (Synthesis Lectures on Human Language Technologies, edited by Graeme Hirst, volume 43), 2020, xxvi+175 pp; paperback, ISBN 978-1-68173-758-4, $79.95; ebook, ISBN 978-1-68173-759-1, $63.96; hardcover, ISBN 978-1-68173-760-7, $99.95; doi:10.2200/S00979ED1V01Y201912HLT04411
This tutorial will cover the fundamentals and the state-of-the-art research on neural models for text production, and outline the constraints specific to each subtasks and examine how the existing neural models account for them.
- 11
- An Empirical Comparison Between N-gram and Syntactic Language Models for Word Ordering11
An empirical comparison between N-gram and syntactic language models on word order task is given and it is shown that the quality of automatically-parsed training data has a relatively small impact on syntactic models.
- Intelligent substation virtual circuit verification method combining knowledge graph and deep learning10
This paper proposes a virtual circuit automatic verification method that combines knowledge graphs with deep learning, and utilizes expert knowledge and relevant standard specifications to construct a knowledge graph of virtual circuits.
- Improving (Dis)agreement Detection with Inductive Social Relation Information From Comment-Reply Interactions10
It is found social relations can boost the performance of the (dis)agreement detection model, especially for the long-token comment-reply pairs, demonstrating the effectiveness of the social relation graph.
- 10
- 9
- 9
- 9
- The Cross-lingual Conversation Summarization Challenge9
A new benchmark is constructed, covering 2 real-world scenarios and 3 language directions, including a low-resource language, that can motivate researches to go beyond English and break the barrier for non-English speakers to benefit from recent advances of conversation summarization.
- Inference-Masked Loss for Deep Structured Output Learning9
This work empirically shows the inference-masked loss combined with the negative log-likelihood loss improves the performance on different tasks, namely entity relation recognition on CoNLL04 and ACE2005 corpora, and spatial role labeling on CLEF 2017 mSpRL dataset.
- Multi-Granular Text Encoding for Self-Explaining Categorization9
This work defines multi-granular ngrams as basic units for explanation, and organizes all n Grammars into a hierarchical structure, so that shorter n grams can be reused while computing longer n Grammar, and can extract intuitive multi- granular evidence to support its predictions.
- Neural Transition-based Syntactic Linearization9
Neural syntactic linearization is studied, building a transition-based syntacticlinearizer leveraging a feed forward neural network, observing significantly better results compared to LSTM language models on this task.
- A Neural Probabilistic Structured-Prediction Method for Transition-Based Natural Language Processing9
A neural probabilistic structured-prediction method for transition-based natural language processing, which integrates beam search and contrastive learning, which achieves significant accuracy improvements over the locally normalized greedy baseline on the two tasks, respectively.
- 8
- Lost in Context? On the Sense-Wise Variance of Contextualized Word Embeddings8
This work quantifies how much the contextualized embeddings of each word sense vary across contexts in typical pre-trained models and proposes a prompt-augmentation method to alleviate such bias in distance-based word sense disambiguation settings.
- Investigating Non-local Features for Neural Constituency Parsing8
This paper investigates injecting non-local features into the training process of a local span-based parser, by predicting constituent n-gram non- local patterns and ensuring consistency between non-Local patterns and local constituents.
- Categorizing Semantic Representations for Neural Machine Translation8
The main idea is to enhance generalization by reducing sparsity and overfitting, which is achieved by finding prototypes of token representations over the training set and integrating their embeddings into the source encoding.
- 8
- Multi-Point Semantic Representation for Intent Classification8
A simple and novel multi-point semantic representation framework with relatively low annotation cost to leverage the fine-grained factor information, decomposing queries into four factors, i.e., topic, predicate, object/condition, query type and a compositional intent bi-attention model with three kinds of attention mechanisms among queries, labels and factors.
- DRTS Parsing with Structure-Aware Encoding and Decoding8
This work proposes a structural-aware model at both the encoder and decoder phase to integrate the structural information, where graph attention network (GAT) is exploited for effectively modeling.
- 8
- Latent Variable Sentiment Grammar8
Two formalisms with deep sentiment representations that capture sentiment subtype expressions by latent variables and Gaussian mixture vectors are investigated, respectively.
- A Comparison between Pre-training and Large-scale Back-translation for Neural Machine Translation7
MT engines trained with pre-trained BERT and back-translation with incrementally larger amounts of data are compared, implementing the two most widely-used monolingual paradigms and BERT has limited advantages compared with large-scale back-translation in accuracy and consistency.
- Does Chinese BERT Encode Word Structure?7
This work investigates Chinese BERT using both attention weight distribution statistics and probing tasks, finding that word information is captured by BERT; word-level features are mostly in the middle representation layers; and downstream tasks make different use of word features in BERT.
- From Genesis to Creole Language7
The enlarged Singlish treebank has achieved significant relative error reduction and the state-of-the-art Singlish POS tagging and dependency parsing accuracies have been improved to 91.16% and 85.57%, respectively.
- Using Dynamic Embeddings to Improve Static Embeddings.7
This work uses contextualized embeddings to facilitate training of static embedding lookup tables, and shows that the resultingembeddings outperform existingstatic embedding methods on various lexical semantics tasks.
- 7
- Deep Learning in Lexical Analysis and Parsing7
In this tutorial, a review of each line of work is given, by contrasting them with traditional statistical methods, and organizing them in consistent orders.
- Transition-Based Deep Input Linearization7
A transition-based model is constructed to jointly perform linearization, function word prediction and morphological generation, which considerably improves upon the accuracy compared to a pipelined baseline system.
- 6
- 6
- 6
- 6
- SciMine: An Efficient Systematic Prioritization Model Based on Richer Semantic Information6
A novel framework SciMine is proposed to accelerate this screening process by capturing semantic feature representations from both background and the corpus by utilizing an autoencoder-based classifier and a feature-dependent classification module to extract general document-level and phrase-level information.
- CSynGEC: Incorporating Constituent-based Syntax for Grammatical Error Correction with a Tailored GEC-Oriented Parser6
This work investigates the integration of constituent-based and dependency-based syntax for GEC in two ways: intra-model combination, which means using separate GCNs to encode both kinds of syntax for decoding in a single model, and an inter-model combinations, which finds that the former method improves recall over using one standalone syntax formalism while the latter improves precision.
- 6
- Implicit Syntactic Features for Target-dependent Sentiment Analysis6
This work studies a method to leverage syntactic information without explicitly building the parser outputs, by training an encoder-decoder structure parser model on standard syntactic treebanks, and then leveraging its hidden encoder layers when analysing tweets.
- Syntactic SMT Using a Discriminative Text Generation Model6
A novel architecture for syntactic SMT is studied, which does not rely on translation rules, but treat translation as an unconstrained target sentence generation task, using soft features to capture lexical and syntactic correspondences between the source and target languages.
- On WordNet Semantic Classes and Dependency Parsing6
Improvements in dependency parsing using WordNet semantic classes are small and not significant using automatic POS tags, contrary to previously published results using gold POS tags.
- LOGicalThought: Logic-Based Ontological Grounding of LLMs for High-Assurance Reasoning5
A novel neurosymbolically-grounded architecture called LogT is proposed that uses an advanced logical language and reasoner in conjunction with an LLM to construct a dual symbolic graph context and logic-based context that transforms the problem from inference over long-form guidelines into a compact grounded evaluation.
- 5
- 5
- DC-MBR: Distributional Cooling for Minimum Bayesian Risk Decoding5
A simple and effective method, Distributional Cooling MBR (DC-MBR), which manipulates the entropy of output distributions by tuning down the Softmax temperature is proposed, and it is theoretically proved the equivalence between the pre-tuning label smoothing factor and distributional cooling.
- 5
- 5
- Dependency Tree Representations of Predicate-Argument Structures5
A novel annotation framework for representing predicate-argument structures that uses dependency trees to encode the syntactic and semantic roles of a sentence simultaneously is presented, which eliminates the structural gap between syntax and shallow semantics, making them compatible.
- 4
- OpenFact: Factuality Enhanced Open Knowledge Extraction4
This work focuses on the factuality property during the extraction of an OpenIE corpus named OpenFact, which contains more than 12 million high-quality knowledge triplets and proposes a comprehensive framework to handle both expressiveness and groundedness.
- 4
- Advancing Chinese Event Detection via Revisiting Character Information4
A novel method to model character information in both the encoding and decoding stages to advance the neural event detection model that can encode both words and characters and predict their event types jointly and further leverage interactions between word and its characters to optimize the inference.
- Exploiting Unlabeled Data for Target-Oriented Opinion Words Extraction4
This paper proposes a novel Multi-Grained Consistency Regularization (MGCR) method to make use of unlabeled data and design two filters specifically for TOWE to filter noisy data at different granularity.
- On the Role of Pre-trained Language Models in Word Ordering: A Case Study with BART4
To explain why BART helps word ordering, analysis is extended with probing and empirically identified that syntactic dependency knowledge in BART is a reliable explanation.
- When Computational Representation Meets Neuroscience: A Survey on Brain Encoding and Decoding4
The task of brain decoding from the perspective of simple concepts and complete sentences is discussed, and the main limitations in this field are described and put forward with possible solutions.
- 4
- Layer-Wise Multi-View Learning for Neural Machine Translation4
This work proposes layer-wise multi-view learning to solve neural machine translation, circumventing the necessity to change the model structure, and can maintain the same inference speed as the original model.
- Improving Shift‐Reduce Phrase‐Structure Parsing with Constituent Boundary Information4
This article proposes two different models to capture constituent boundary information, based on which two sets of novel features are designed for a shift‐reduce parser, and proposes a boundary prediction model that uses a classifier to predict the boundaries of constituents.
- EPINN-GF: An enhanced physics-informed neural network integrating geometric features for stress prediction in complex spinal geometries3
EPINN-GF accurately predicts stress in the geometrically complex spinal regions and multi-physical environments, offering the potential clinical value for the spinal disorders by enabling personalized treatment planning, optimizing surgical strategies, and supporting early AIS progression prediction.
- The value for money of artificial intelligence-empowered precision medicine: a systematic review and regression analysis3
Although risk-of-bias assessment indicated potential systematic optimism, AI-PM was cost-saving or cost-effective in 89% of base-case analyses, with incremental cost-effectiveness ratios ranging from dominant to $129,174/quality-adjusted life-year (QALY).
- [Early classification and recognition algorithm for sudden cardiac arrest based on limited electrocardiogram data trained with a two-stages convolutional neural network].3
An early prediction and classification algorithm for SCA based on deep transfer learning that helps address the requirement of large training datasets for deep learning models and enables early and accurate detection and identification of high-risk ECG signs before the onset of SCA.
- Dynamics of Instruction Fine-Tuning for Chinese Large Language Models3
This research provides a granular analysis of how data volume, parameter size, and data construction methods influence the development of each underlying ability of LLM, such as creative writing, code generation, and logical reasoning.
- Knowledgeable Salient Span Mask for Enhancing Language Models as Knowledge Base3
This work develops two solutions to help the model learn more knowledge from unstructured text in a fully self-supervised manner and is the first to explore fullySelf Supervised learning of knowledge in continual pre-training.
- 3
- 3
- 3
- Domain Representation for Knowledge Graph Embedding3
This work proposed to learn a Domain Representations over existing knowledge graph embedding models, such that entities that have similar attributes are organized into the same domain, so that hierarchical knowledge of domains can give further evidence in link prediction.
- Domain Adaptive Neural Sentence Compression by Tree Cutting3
This work investigates syntax tree based noise pruning methods for neural sentence compression and identifies the most informative regions in a syntactic dependency tree by self attention over context nodes and maximum density subtree extraction.
- 3
- 3
- A Semantics Oriented Grammar for Chinese Treebanking3
This paper introduces a semantics oriented grammar for Chinese, designed to provide basic supports for tasks such as automatic semantic parsing and sentence generation, and is the first Chinese grammar representation capable of direct transformation into logical forms.
- Syntactic Processing Using Global Discriminative Learning and Beam-Search Decoding3
This tutorial discusses a framework for incremental left-to-right structured predication, which makes use of global discriminative learning and beam-search decoding, and gives an introduction to the algorithms and efficient implementations.
- ViZPar: A GUI for ZPar with Manual Feature Selection3
ViZPar is presented, a tool that provides an enhacing of the usability of ZPar, including parameter selection and output visualization, which makes the tool very useful for people interested in obtaining the best parser through feature engineering.
- Recent Advances in Dependency Parsing3
This work focuses on dependency-based as opposed to constituency-based approaches to syntactic parsing, with application to a wide range of research areas and different languages.
- 2
- Constituency Parsing Using LLMs2
Two strategies to guide LLMs to generate more accurate constituent trees by learning from erroneous samples and refining outputs in a multi-agent collaboration way are proposed, thereby enhancing overall parsing performance and achieving promising results across different learning paradigms.
- 2
- Non-Autoregressive Document-Level Machine Translation2
A comprehensive examination of typical NAT models in the context of document-level MT is conducted and a simple but effective design of sentence alignment between source and target is proposed.
- Mining Error Templates for Grammatical Error Correction2
Experimental results on the newly proposed CTC-2021 Chinese GEC benchmark show that combing the error templates can effectively improve the performance of a strong GEC system, especially on two error types with very little training data.
- Building Dialogue Understanding Models for Low-resource Language Indonesian from Scratch2
A Bi-Confidence-Frequency Cross-Lingual transfer framework (BiCF) is proposed, which consists of “BiCF Mixing”, “Latent Space Refinement” and “Joint Decoder’, respectively, to overcome the lack of low-resource language dialogue data and performs reliably and cost-efficiently on different scales of manually annotated Indonesian data.
- 2
- 2
- Cross-Lingual Dependency Parsing via Self-Training2
Results show that self-training can boost the dependency parsing performances on the target languages and the POS tagger assistant instance selection can achieve further improvements consistently.
- Investigating Lexical and Semantic Cognition by Using Neural Network to Encode and Decode Brain Imaging2
Results show that by using neural networks for fMRI encoding and decoding accuracies are improved compared to a linear model on the same input.
- 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.
- 2
- 2
- Enhancing Shift-Reduce Constituent Parsing with Action N-Gram Model2
It is shown that incorporating action n-gram models into a state-of-the-art parsing framework could achieve parsing accuracy improvements on three datasets across two languages.
- ViZPar: A GUI for ZPar with Manual Feature Selection ViZPar: una GUI para ZPar con Selecci on Manual de Features2
ViZPar is presented, a tool that provides an enhacing of the usability of ZPar, including parameter selection and output visualization, which makes the tool very useful for people interested in obtaining the best parser through feature engineering.
- 2
- 1
- 1
- 1
- Robust water meter reading recognition in challenging scenarios based on YOLOv111
A comprehensive solution to improve recognition accuracy in challenging water meter reading scenarios by integrating improved feature extraction and multi-scale fusion strategies, and introduces two specialized evaluation metrics: Transition-Tolerant Meter Reading Rate (TTMRrate) and Weight-Aware Digit Recognition Rate (WADRrate).
- 1
- 1
- 1
- 1
- 1
- Cross-domain Constituency Parsing by Leveraging Heterogeneous Data1
Experimental results show that the knowledge transfer model outperforms various baseline models, including conventional chart-based and transition-based parsers and the current large-scale language model for zero-shot and few-shot settings.
- EASE: An Easily-Customized Annotation System Powered by Efficiency Enhancement Mechanisms1
The results show that EASE, an Easily-Customized Annotation System Powered by Efficiency Enhancement Mechanisms, can meet the diverse needs of NLP researchers and significantly accelerate the annotation process.
- Token-Level Fitting Issues of Seq2seq Models1
This work identifies three major factors that influence token-level fitting, which include token frequency, parts-of-speech, and prediction discrepancy, and finds that external factors such as language, model size, domain, data scale, and pretraining can also influence the fitting of tokens.
- 1
- 1
- 1
- 1
- 1
- 1
- 1
- Research on Pair Learning Method and Pattern Based on Pair Programming1
This paper explains the method of pair learning from the aspects of pedagogy, psychology and game theory and uses the model that two individuals make up a pair and two pairsMake up a large group, based on the distributed pair system XPairtise.
- 1
- 1
- Heterogeneous Text Style Control Using Prompts–
This work introduces a unified prompt-driven approach to paraphrase generation that leverages diverse prompts, enabling fine-grained user control over aspects such as syntax and sentiment, and incorporates translation to enable sophisticated cross-lingual text controls.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Exploration of Sex Differences in the Brain based on Structure and fMRI Techniques–
Gender differences were shown in both the spatial distribution of spontaneous fluctuation power and the functional connectivity of resting states and in the whole measured cortex.
- –
- –
- –
- –
- –
- –
- THE SEMANTIC POTENTIAL OF INTERROGATIVE SENTENCES WITH PRONOUN GDE–
This study analyzes semantic potential of interrogative sentences with pronoun «gde», and illustrates contexts, conditions on which depends realization each one of men-tioned meanings.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Exploring Hierarchical Interaction Between Review and Summary for Better Sentiment Analysis.–
A hierarchically-refined attention network is proposed for better exploiting multi-interaction between a review and its summary for sentiment analysis, and is also more effective compared to existing methods when the user summary is replaced with summary generated by an automatic summarization system.
- –
- Highly Reliable Product Code for Error Correction–
A new product code is proposed which combines CRC code with RSPC which is called as CRSPC and can significantly reduce the probabilities of these drawbacks, thereby improving the performance of erasures-and-errors decoding.
- Lists and Mutability–
This chapter introduces lists, a sequential container type that is mutable, the first mutable type introduced in the book; they can be regarded as modifiable counterparts of tuples.
- DISCRIMINATIVE LEARNING APPROACHES FOR THE STATISTICAL PROCESSING OF CHINESE–
It is concluded that the discriminative method is a competitive choice for the statistical processing of the Chinese language using a global linear model trained by the generalized perceptron algorithm together with beam-search decoding.
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.