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
- Ralph GrishmanSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works296 from public data
- Executable Code Actions Elicit Better LLM Agents699
This work proposes to use executable Python code to consolidate LLM agents' actions into a unified action space (CodeAct), and collects an instruction-tuning dataset CodeActInstruct, which can be used with existing data to improve models in agent-oriented tasks without compromising their general capability.
- Cross-lingual Name Tagging and Linking for 282 Languages595
This work develops a cross-lingual name tagging and linking framework for 282 languages that exist in Wikipedia that is able to identify name mentions, assign a coarse-grained or fine- grained type to each mention, and link it to an English Knowledge Base (KB) if it is linkable.
- Overview of the TAC 2010 knowledge base population track583
An overview of the task definition and annotation challenges associated with KBP2010 is provided and the evaluation results and lessons that are learned are discussed based on detailed analysis.
- Efficient Attentions for Long Document Summarization488
Hepos, a novel efficient encoder-decoder attention with head-wise positional strides to effectively pinpoint salient information from the source is proposed, able to process ten times more tokens than existing models that use full attentions.
- Tool Learning with Foundation Models471
A systematic investigation and comprehensive review of tool learning and discusses several open problems that require further investigation, such as ensuring trustworthy tool use, enabling tool creation with foundation models, and addressing personalization challenges.
- Large Language Models on Graphs: A Comprehensive Survey367
A systematic review of scenarios and techniques related to large language models on graphs, including LLM as Predictor, LLM as Encoder, and LLM as Aligner, and compare the advantages and disadvantages of different schools of models is provided.
- Training Software Engineering Agents and Verifiers with SWE-Gym360
This work presents SWE-Gym, the first environment for training real-world software engineering (SWE) agents, and uses SWE-Gym to train language model based SWE agents, achieving up to 19% absolute gains in resolve rate on the popular SWE-Bench Verified and Lite test sets.
- A Survey of Knowledge-enhanced Text Generation355
A comprehensive review of the research on knowledge-enhanced text generation over the past five years is presented, which includes two parts: (i) general methods and architectures for integrating knowledge into text generation; (ii) specific techniques and applications according to different forms of knowledge data.
- A Dependency-Based Neural Network for Relation Classification238
A new structure, termed augmented dependency path (ADP), is proposed, which is composed of the shortest dependency path between two entities and the subtrees attached to the shortest path, and a dependency-based neural networks (DepNN) are developed to exploit the semantic representation behind the ADP.
- Text2Mol: Cross-Modal Molecule Retrieval with Natural Language Queries219
This work constructs a paired dataset of molecules and their corresponding text descriptions, which it uses to learn an aligned common semantic embedding space for retrieval and extends this to create a cross-modal attention-based model for explainability and reranking by interpreting the attentions as association rules.
- Text Classification Using Label Names Only: A Language Model Self-Training Approach205
This paper uses pre-trained neural language models both as general linguistic knowledge sources for category understanding and as representation learning models for document classification, and achieves around 90% accuracy on four benchmark datasets.
- Artificial intelligence for science in quantum, atomistic, and continuum systems189
This work aims to provide a technically thorough account of a subarea of AI4Science; namely, AI for quantum, atomistic, and continuum systems, and provides an in-depth yet intuitive account of techniques to achieve equivariance to symmetry transformations.
- Language Models with Image Descriptors are Strong Few-Shot Video-Language Learners175
The goal of this work is to build flexible video-language models that can generalize to various video-to-text tasks from few examples, such as domain-specific captioning, question answering, and future event prediction, and to demonstrate the power of language models in understanding videos on a wide variety of video- language tasks.
- 162
- CLIP-Event: Connecting Text and Images with Event Structures159
A contrastive learning framework to enforce vision-language pretraining models to comprehend events and associated argument (participant) roles is proposed, which takes advantage of text information extraction technologies to obtain event structural knowledge, and utilizes multiple prompt functions to contrast difficult negative descriptions by manipulating event structures.
- 156
- RM-R1: Reward Modeling as Reasoning149
It is hypothesized and validated that integrating reasoning into reward modeling significantly enhances RM's interpretability and performance, and a new class of generative reward models, Reasoning Reward Models (ReasRMs), are introduced, which formulate reward modeling as a reasoning task.
- Open Vocabulary Electroencephalography-to-Text Decoding and Zero-Shot Sentiment Classification146
It is proposed that the human brain functions as a special text encoder and a novel framework leveraging pre-trained language models (e.g., BART) is proposed, which achieves a 40.1% BLEU-1 score on EEG-To-Text decoding and a 55.6% F1 scoreon zero-shot EEG-based ternary sentiment classification, which significantly outperforms supervised baselines.
- Mobile-Agent-E: Self-Evolving Mobile Assistant for Complex Tasks145
This work introduces Mobile-Agent-E, a hierarchical multi-agent framework capable of self-evolution through past experience capable of achieving 22% absolute improvement over previous state-of-the-art approaches across three foundation model backbones.
- ERICA: Improving Entity and Relation Understanding for Pre-trained Language Models via Contrastive Learning143
Experimental results demonstrate that ERICA can improve typical PLMs (BERT and RoBERTa) on several language understanding tasks, including relation extraction, entity typing and question answering, especially under low-resource settings.
- 139
- If LLM Is the Wizard, Then Code Is the Wand: A Survey on How Code Empowers Large Language Models to Serve as Intelligent Agents133
An overview of the various benefits of integrating code into LLMs' training data and how these profound capabilities of LLMs, brought by code, have led to their emergence as intelligent agents in situations where the ability to understand instructions, decompose goals, plan and execute actions, and refine from feedback are crucial to their success on downstream tasks are traced.
- COVID-19 Literature Knowledge Graph Construction and Drug Repurposing Report Generation130
A novel and comprehensive knowledge discovery framework, COVID-KG, to extract fine-grained multimedia knowledge elements (entities, relations and events) from scientific literature and exploit the constructed multimedia knowledge graphs (KGs) for question answering and report generation.
- A Multi-lingual Multi-task Architecture for Low-resource Sequence Labeling117
A multi-lingual multi-task architecture to develop supervised models with a minimal amount of labeled data for sequence labeling that proves to be particularly effective for low-resource settings, when there are less than 200 training sentences for the target task.
- Linking Tweets to News: A Framework to Enrich Short Text Data in Social Media117
This work proposes a graph based latent variable model that models the inter short text correlations (text-to-text information) and shows that using tweet specific feature and news specific feature as well as temporal constraints, it is able to extract text- to-text correlations, and thus completes the semantic picture of a short text.
- Overview of TAC-KBP2015 Tri-lingual Entity Discovery and Linking.110
This new end-to-end Tri-lingual entity discovery and linking task, which requires a system to take raw texts from three languages as input, automatically extract entity mentions, link them to an English knowledge base, and cluster NIL mentions across languages, is given an overview.
- Graph-based event coreference resolution109
This paper models event coreference resolution as a spectral graph clustering problem and evaluates the clustering algorithm on ground truth event mentions using ECM F-Measure and obtains the ECM-F scores of 0.8363 and 0.8312 respectively.
- CREATOR: Tool Creation for Disentangling Abstract and Concrete Reasoning of Large Language Models107
The proposed CREATOR, a novel framework that enables LLMs to create their own tools using documentation and code realization, disentangles abstract tool creation and concrete decision execution, resulting in improved performance.
- Language specific issue and feature exploration in Chinese event extraction104
This paper points out a language specific issue in Chinese trigger labeling, and then commits to discussing the contributions of lexical, syntactic and semantic features applied in trigger labeling and argument labeling.
- Abstract Meaning Representation Guided Graph Encoding and Decoding for Joint Information Extraction103
This work proposes a novel AMR-guided framework for joint information extraction to discover entities, relations, and events with the help of a pre-trained AMR parser to convert natural language texts into structured semantic representations.
- Unsupervised Entity Linking with Abstract Meaning Representation99
Experimental results show that AMR captures contextual properties discriminative enough to make linking decisions, without the need for EL training data, and that system with AMR parsing output outperforms hand labeled traditional semantic roles as context representation for EL.
- GAIA: A Fine-grained Multimedia Knowledge Extraction System98
The system, GAIA, enables seamless search of complex graph queries, and retrieves multimedia evidence including text, images and videos, and achieves top performance at the recent NIST TAC SM-KBP2019 evaluation.
- Word Embeddings Are Steers for Language Models93
The learned LM-Steer serves as a lens in text styles: it reveals that word embeddings are interpretable when associated with language model generations and can highlight text spans that most indicate the style differences.
- 93
- Learning Shared Semantic Space for Speech-to-Text Translation92
By projecting audio and text features to a common semantic representation, Chimera unifies MT and ST tasks and boosts the performance on ST benchmarks, MuST-C and Augmented Librispeech, to a new state-of-the-art.
- Connecting the Dots: Event Graph Schema Induction with Path Language Modeling91
This work proposes a new Event Graph Schema, where two event types are connected through multiple paths involving entities that fill important roles in a coherent story, and introduces Path Language Model, an auto-regressive language model trained on event-event paths, to select salient and coherent paths to probabilistically construct these graph schemas.
- ReviewRobot: Explainable Paper Review Generation based on Knowledge Synthesis91
A novel ReviewRobot is built to automatically assign a review score and write comments for multiple categories such as novelty and meaningful comparison, and can serve as an assistant for paper reviewers, program chairs and authors.
- From Pixels to Insights: A Survey on Automatic Chart Understanding in the Era of Large Foundation Models89
This survey paper provides a comprehensive overview of the recent developments, challenges, and future directions in chart understanding within the context of large foundation models, and examines various modeling strategies that enhance chart understanding performance.
- New Tools for Web-Scale N-grams88
A new set of search tools are described that make use of part-of-speech tags, and collectively lower the barrier for lexical learning and ambiguity resolution at web-scale, which will allow novel sources of information to be applied to long-standing natural language challenges.
- Aligning LLMs with Individual Preferences via Interaction87
This work trains LLMs that can ''interact to align'', essentially cultivating the meta-skill of LLMs to implicitly infer the unspoken personalized preferences of the current user through multi-turn conversations, and then dynamically align their following behaviors and responses to these inferred preferences.
- Cross-lingual Structure Transfer for Relation and Event Extraction83
It is found that language-universal symbolic and distributional representations are complementary for cross-lingual structure transfer, and the event argument role labeling model transferred from English to Chinese achieves similar performance as the model trained from Chinese.
- Distantly-Supervised Named Entity Recognition with Noise-Robust Learning and Language Model Augmented Self-Training82
A noise-robust learning scheme comprised of a new loss function and a noisy label removal step, for training NER models on distantly-labeled data, and a self-training method that uses contextualized augmentations created by pre-trained language models to improve the generalization ability of the NER model are proposed.
- PaperRobot: Incremental Draft Generation of Scientific Ideas80
A PaperRobot who performs as an automatic research assistant by conducting deep understanding of a large collection of human-written papers in a target domain and constructing comprehensive background knowledge graphs is presented.
- Neural-Answering Logical Queries on Knowledge Graphs79
This paper proposes an embedding based method (NewLook), which supports four types of logical operations and can answer queries with multiple variable nodes, and goes beyond the linear transformation assumption and thus consistently outperforms the existing methods.
- Can LLMs Produce Faithful Explanations For Fact-checking? Towards Faithful Explainable Fact-Checking via Multi-Agent Debate77
Experimental results demonstrate that MADR significantly improves the faithfulness of LLM-generated explanations to the evidence, advancing the credibility and trustworthiness of these explanations.
- InfoSurgeon: Cross-Media Fine-grained Information Consistency Checking for Fake News Detection77
This work contributes a novel benchmark for fake news detection at the knowledge element level, as well as a solution for this task which incorporates cross-media consistency checking to detect the fine-grained knowledge elements making news articles misinformative.
- 76
- The Future is not One-dimensional: Complex Event Schema Induction by Graph Modeling for Event Prediction74
A new concept of Temporal Complex Event Schema is introduced: a graph-based schema representation that encompasses events, arguments, temporal connections and argument relations and a Temporal Event Graph Model that predicts event instances following the temporal complex event schema is proposed.
- Joint Biomedical Entity and Relation Extraction with Knowledge-Enhanced Collective Inference73
KECI takes a collective approach to link mention spans to entities by integrating global relational information into local representations using graph convolutional networks and fuses the initial span graph and the knowledge graph into a more refined graph using an attention mechanism.
- A Pairwise Event Coreference Model, Feature Impact and Evaluation for Event Coreference Resolution73
The problem of event coreference resolution in the ACE program is formally state, an agglomerative clustering algorithm for the task is presented, the feature impact in theevent coreference model is explored, and three evaluation metrics that were previously adopted in entity coreferenceresolution are compared.
- CUNY-BLENDER TAC-KBP2010 Entity Linking and Slot Filling System Description70
In the KBP2010 evaluation, the CUNY-BLENDER entity linking system explored the slot filling attributes that may potentially help disambiguate entity mentions and results show that this feedback approach can achieve 9.1% absolute improvement on micro-average accuracy over the baseline using vector space model.
- Do LVLMs Understand Charts? Analyzing and Correcting Factual Errors in Chart Captioning69
A new task of Chart Caption Factual Error Correction is established and CHARTVE, a model for visual entailment that outperforms proprietary and open-source LVLMs in evaluating factual consistency is introduced, and an interpretable two-stage framework that excels at correcting factual errors is proposed.
- 68
- Semi-supervised New Event Type Induction and Event Detection65
A Semi-Supervised Vector Quantized Variational Autoencoder framework is designed to automatically learn a discrete latent type representation for each seen and unseen type and optimize them using seen type event annotations.
- A Survey on Post-training of Large Language Models59
This paper presents the first comprehensive survey of PoLMs, systematically tracing their evolution across five core paradigms: Fine-tuning, which enhances task-specific accuracy; Alignment, which ensures ethical coherence and alignment with human preferences; Reasoning, which advances multi-step inference despite challenges in reward design; Efficiency, which optimizes resource utilization amidst increasing complexity; Integration and Adaptation, which extend capabilities across diverse modalities while addressing coherence issues.
- Near-imperceptible Neural Linguistic Steganography via Self-Adjusting Arithmetic Coding59
A new linguistic steganography method which encodes secret messages using self-adjusting arithmetic coding based on a neural language model which outperforms the previous state-of-the-art methods on four datasets by 15.3% and 38.9% in terms of bits/word and KL metrics, respectively.
- Collective Tweet Wikification based on Semi-supervised Graph Regularization59
A novel semi-supervised graph regularization model is proposed to incorporate both local and global evidence from multiple tweets through three fine-grained relations in order to identify semanticallyrelated mentions for collective inference.
- Code4Struct: Code Generation for Few-Shot Event Structure Prediction58
This work proposes C ODE 4S TRUCT to leverage text-to-structure translation capability to tackle structured prediction tasks in NLP, and uses C ODE 4S TRUCT to tackle the EAE task us-ing code generation.
- Joint bilingual name tagging for parallel corpora58
Two novel approaches to jointly and consistently extract names from parallel corpora using standard linear-chain Conditional Random Fields as the learning framework, incorporating cross-lingual features propagated between two languages are proposed.
- Speech segmentation and spoken document processing57
Different levels of speech segmentation, approaches to automatically recovering segment boundary locations, and experimental results demonstrating impact on several language processing tasks show a need for optimizing segmentation for the end task rather than independently.
- The Science of Precision Prevention55
This state-of-the-art review is based on a workshop convened to identify current gaps in knowledge and methods used in precision prevention intervention research, discuss opportunities to expand trials of implementation science to close the health equity gaps, and expand the education and training of a diverse precision prevention workforce.
- Overview of TAC-KBP2017 13 Languages Entity Discovery and Linking.55
An overview of the Tri-lingual Entity Discovery and Linking task at the Knowledge Base Population (KBP) track at TAC2017, and of the Ten Low Resource Language EDL Pilot is given.
- Gender and Animacy Knowledge Discovery from Web-Scale N-Grams for Unsupervised Person Mention Detection54
A simple approach to discover gender and animacy knowledge for person mention detection by learning noun-gender and noun-animacy pair counts from web-scale n-grams using specific lexical patterns and applying confidence estimation metrics to filter noise.
- Using semantic relations to refine coreference decisions54
A novel mechanism for improving reference resolution by using the output of a relation tagger to rescore coreference hypotheses is presented, which can improve performance on two quite different languages - English and Chinese.
- Deep Learning on Graphs for Natural Language Processing51
This tutorial will cover relevant and interesting topics on applying deep learning on graph techniques to NLP, including automatic graph construction for NLP), graph representation learning for N LP, advanced GNN based models (e.g., graph2seq, graph2tree, and graph2graph) for NLS, and the applications of GNNs in various NLP tasks.
- An Attentive Fine-Grained Entity Typing Model with Latent Type Representation48
A fine-grained entity typing model with a novel attention mechanism and a hybrid type classifier, which predicts a low-dimensional vector that encodes latent type features and reconstructs the type vector from this latent representation.
- LEMMA: Towards LVLM-Enhanced Multimodal Misinformation Detection with External Knowledge Augmentation46
This paper proposes LEMMA: LVLM-Enhanced Multimodal Misinformation Detection with External Knowledge Augmentation, which leverages LVLM intuition and reasoning capabilities while augmenting them with external knowledge to enhance the accuracy of misinformation detection.
- Analysis and Enhancement of Wikification for Microblogs with Context Expansion44
This work evaluates an existing Twitter dataset as well as the D2W task in general and test the effects of two tweet context expansion methods, based on tweet authorship and topic-based clustering, on a state-of-the-art D1W system and evaluates the results.
- Understanding Emergent In-Context Learning from a Kernel Regression Perspective43
This paper investigates the reason why a transformer-based language model can accomplish in- Context learning after pre-training on a general language corpus by proposing one hypothesis that LLMs can simulate kernel regression algorithms when faced with in-context examples and empirically investigates the in- context behaviors of language models.
- ADEPT: A DEbiasing PrompT Framework43
ADEPT is proposed, a method to debias PLMs using prompt tuning while maintaining the delicate balance between removing biases and ensuring representation ability, and achieves competitive results while maintaining (and in some cases even improving) the PLM's representation ability.
- Improving name tagging by reference resolution and relation detection43
This work uses an N-best approach to generate multiple hypotheses and have them re-ranked by subsequent stages of processing to reduce the errors produced by a Chinese name tagger.
- KompaRe: A Knowledge Graph Comparative Reasoning System42
This paper develops KompaRe, the first of its kind prototype system that provides comparative reasoning capability over large knowledge graphs, and presents both the system architecture and its core algorithms, including knowledge segment extraction, pairwise reasoning and collective reasoning.
- Cross-document Misinformation Detection based on Event Graph Reasoning41
A graph-based detector that constructs a cross- document knowledge graph using cross-document event coreference resolution and employs a heterogeneous graph neural network to conduct detection at two levels of misinformation detection, which significantly outperforms existing methods on this new task.
- Separating Skills and Concepts for Novel Visual Question Answering40
This work presents a novel method for learning to compose skills and concepts that separates these two factors implicitly within a model by learning grounded concept representations and disentangling the encoding of skills from that of concepts.
- 40
- Toward Generalizable Evaluation in the LLM Era: A Survey Beyond Benchmarks39
This survey probes the core challenges that the rise of LLMs poses for evaluation, and identifies and analyzes two pivotal transitions: from task-specific to capability-based evaluation and from manual to automated evaluation.
- 38
- Graph-Based Clustering for Computational Linguistics: A Survey38
This survey overviews graph-based clustering and its applications in computational linguistics as a five-part story: hypothesis, modeling, measure, algorithm and evaluation, and envision that graph- based clustering is a promising solution for some emerging NLP problems.
- 37
- MuMuQA: Multimedia Multi-Hop News Question Answering via Cross-Media Knowledge Extraction and Grounding37
A new QA evaluation benchmark with 1,384 questions over news articles that require cross- media grounding of objects in images onto text, and introduces a novel multimedia data augmentation framework, based on cross-media knowledge extraction and synthetic question-answer generation, to automatically augment data that can provide weak supervision for this task.
- NewsClaims: A New Benchmark for Claim Detection from News with Attribute Knowledge37
This work extends the claim detection problem to include extraction of additional attributes related to each claim and releases 889 claims annotated over 143 news articles, seeing that zero-shot and prompt-based baselines show promising performance on this benchmark, while still considerably behind human performance.
- Fine-grained Information Extraction from Biomedical Literature based on Knowledge-enriched Abstract Meaning Representation37
A novel biomedical Information Extraction (IE) model to tackle two challenges and extract scientific entities and events from English research papers, and performs Abstract Meaning Representation (AMR) to compress the wide context to uncover a clear semantic structure for each complex sentence.
- Syntax-aware Multi-task Graph Convolutional Networks for Biomedical Relation Extraction37
A novel graph convolutional networks model is proposed that incorporates dependency parsing and contextualized embedding to effectively capture comprehensive contextual information in biomedical relation extraction.
- Open-Domain Multi-Document Summarization via Information Extraction: Challenges and Prospects37
Novel approaches to taking advantage of cross-document IE for multi-document summarization are explored and one of them, re-ranking the output of a high performing summarization system with IE-informed metrics, leads to improvements in both manually-evaluated content quality and readability.
- Who, what, when, where, why?: comparing multiple approaches to the cross-lingual 5W task37
An error analysis of a new cross-lingual task: the 5W task, a sentence-level understanding task which seeks to return the English 5W's corresponding to a Chinese sentence, which shows that MT significantly degrades sentence- level understanding.
- Acting Less is Reasoning More! Teaching Model to Act Efficiently36
This work proposes Optimal Tool Call-controlled Policy Optimization (OTC-PO), a simple yet effective RL-based framework that encourages models to produce accurate answers with minimal tool calls and is the first RL-based framework that explicitly optimizes tool-use efficiency in TIR 1.
- BERT might be Overkill: A Tiny but Effective Biomedical Entity Linker based on Residual Convolutional Neural Networks36
This work proposes an efficient convolutional neural network with residual connections for biomedical entity linking that achieves comparable or even better linking accuracy than the state-of-the-art BERT-based models while having about 60 times fewer parameters.
- Tweet Ranking Based on Heterogeneous Networks36
This paper introduces Tri-HITS, a model to iteratively propagate ranking scores across heterogeneous networks and shows that integrating both formal genre and inferred social networks with tweet networks produces a higher-quality ranking than the tweet networks alone.
- CUNY-UIUC-SRI TAC-KBP2011 Entity Linking System Description36
This paper describes a joint effort by the City University of New York (CUNY), University of Illinois at Urbana-Champaign (UIUC) and SRI International at participating in the mono-lingual entity linking (MLEL) and cross-lingUAL entities linking (CLEL) tasks for the NIST Text Analysis Conference (TAC) Knowledge Base Population (KBP2011) track.
- 35
- Corpus-based Open-Domain Event Type Induction35
This work presents a corpus-based open-domain event type induction method that automatically discovers a set of event types from a given corpus by selecting salient predicates and object heads, and obtains event types by jointly embedding and clustering pairs in a latent spherical space.
- Joint Multimedia Event Extraction from Video and Article35
This work introduces the new task of Video MultiMedia Event Extraction (Video M2E2) and proposes the first self-supervised multimodal event coreference model that can determine coreference between video events and text events without any manually annotated pairs and introduces the first multimodals transformer which extracts structured event information jointly from both videos and text documents.
- Controllable and Diverse Text Generation in E-commerce35
A fine-grained controllable generative model that uses an algorithm borrowed from automatic control to precisely manipulate the diversity/accuracy trade-off of generated text and outperforms existing generative models in terms of diversity and relevance.
- The Gun Violence Database: A new task and data set for NLP35
A new and growing dataset, the Gun Violence Database, is introduced in order to facilitate the adaptation of current NLP technologies to the domain of gun violence, thus enabling better social science research on this important and under-resourced problem.
- ChemReasoner: Heuristic Search over a Large Language Model's Knowledge Space using Quantum-Chemical Feedback34
This work introduces an AI-guided computational screening framework unifying linguistic reasoning with quantum-chemistry based feedback from 3D atomistic representations, and pioneers AI-accelerated, trustworthy catalyst discovery.
- Genre Separation Network with Adversarial Training for Cross-genre Relation Extraction34
A genre-separation network is designed, which applies two encoders, one genre-independent and onegenre-shared, to explicitly extract genre-specific and genre-agnostic features, and which outperforms the state-of-the-art by 1.7% absolute F1 gain.
- Mining Name Translations from Comparable Corpora by Creating Bilingual Information Networks.32
A novel approach to acquire name translation pairs in a bootstrapping framework is designed and the experimental results show this approach can generate highly accurate nametranslation pairs for persons, geopolitical and organization entities.
- KEBLM: Knowledge-Enhanced Biomedical Language Models30
This work proposes a general framework for incorporating various types of domain knowledge from multiple sources into biomedical PLMs and shows that the proposed framework consistently improves the performance of the underlying PLMs on various downstream tasks such as natural language inference, question answering, and entity linking.
- A Context-Dependent Gated Module for Incorporating Symbolic Semantics into Event Coreference Resolution30
A novel context-dependent gated module to adaptively control the information flows from the input symbolic features and achieve state-of-the-art results on two datasets: ACE 2005 and KBP 2016.
- 30
- Overview of TAC-KBP2016 Tri-lingual EDL and Its Impact on End-to-End KBP.30
An overview of the Tri-lingual Entity Discovery and Linking task at the Knowledge Base Population (KBP) track at TAC2016 is given and several new and effective research directions including cross-lingUAL knowledge transfer and exploiting language-specific resources are summarized.
- What should I Ask: A Knowledge-driven Approach for Follow-up Questions Generation in Conversational Surveys29
This paper constructed a new human-annotated dataset of human-written follow-up questions with dialogue history and labeled knowledge in the context of conversational surveys and proposes a two-staged knowledge-driven model for the task, which generates informative and coherent follow- up questions by using knowledge to steer the generation process.
- 29
- Towards Fast Adaptation of Pretrained Contrastive Models for Multi-channel Video-Language Retrieval28
It is found that discrete text tokens coupled with a pretrained contrastive text model yields the best performance, which can even outperform state-of-the-art on the iVQA and How2QA datasets without additional training on millions of video-text data.
- Event Schema Induction with Double Graph Autoencoders28
This work proposes a new event schema induction framework using double graph autoencoders, which captures the global dependencies among nodes in event graphs and induces high-quality event schemas with global consistency.
- Improving Slot Filling Performance with Attentive Neural Networks on Dependency Structures28
This paper proposes an effective DNN architecture for SF with the following new strategies: take a regularized dependency graph instead of a raw sentence as input to DNN, to compress the wide contexts between query and candidate filler.
- Harnessing web page directories for large-scale classification of tweets28
This paper introduces an inexpensive way of labeling large sets of tweets, which can be easily regenerated or updated when needed, and uses human-edited web page directories to infer categories from URLs contained in tweets.
- Combining Social Cognitive Theories with Linguistic Features for Multi-genre Sentiment Analysis28
This paper combines local linguistic features and global social evidence in a propagation scheme to improve sentiment analysis results and obtains significant improvement for various genres in the domain of presidential election.
- 27
- Two-Stage Hashing for Fast Document Retrieval27
A two-stage unsupervised hashing framework which harmoniously integrates two state-of-theart hashing algorithms Locality Sensitive Hashing and Iterative Quantization is proposed, which capitalizes on both term and topic similarity among documents, leading to precise document retrieval.
- Unsupervised Language-Independent Name Translation Mining from Wikipedia Infoboxes26
This paper describes an unsupervised and language-independent approach to mine name translation pairs from entity profiles, using Wikipedia Infoboxes as a stand-in for high quality entity profile extraction.
- Re-ranking algorithms for name tagging26
It is demonstrated how re-ranking can improve name tagging in a Chinese information extraction system by incorporating information from relation extraction, event extraction, and coreference.
- SynerGPT: In-Context Learning for Personalized Drug Synergy Prediction and Drug Design25
A novel setting and models for in-context drug synergy learning are proposed, inspired by recent work that pre-trains a GPT language model (LM) to “in-context learn” common function classes, and novel pre-training schemes that enable a G PT model to in- context learn “drug synergy functions” are devised.
- Improving Candidate Retrieval with Entity Profile Generation for Wikidata Entity Linking25
This work proposes a novel candidate retrieval paradigm based on entity profiling that complements the traditional approach of using a Wikipedia anchor-text dictionary, enabling it to further design a highly effective hybrid method for candidate retrieval.
- EventCube25
This proposed EventCube demo will show the power of the system not only on the originally designed ASRS (Aviation Safety Report System) data sets, but also on news datasets collected from multiple news agencies, and academic datasets constructed from the DBLP and web data.
- Enhancing Multi-Document Summarization with Cross-Document Graph-based Information Extraction24
This work focuses on improving Multi-Document Summarization (MDS) performance by using cross-document IE output, incorporating two novel components: the use of auxiliary entity and event recognition systems to focus the summary generation model and incorporating an alignment loss between IE nodes and their text spans to reduce inconsistencies between the IE graphs and text representations.
- The Battlefront of Combating Misinformation and Coping with Media Bias24
This tutorial dives into the important research questions of how to develop a robust fake news detection system, which not only fact-check information pieces provable by background knowledge but also reason about the consistency and the reliability of subtle details for emerging events.
- Logical Entity Representation in Knowledge-Graphs for Differentiable Rule Learning23
Empirical results demonstrate that the proposed Logical Entity RePresentation (LERP) model outperforms other rule learning methods in knowledge graph completion and is comparable or even superior to state-of-the-art black-box methods.
- CONCRETE: Improving Cross-lingual Fact-checking with Cross-lingual Retrieval23
The proposed Cross-lingual Inverse Cloze Task (X-ICT), a self-supervised algorithm that creates training instances by translating the title of a passage, is presented, a first fact-checking framework augmented with cross-lingually retrieval that aggregates evidence retrieved from multiple languages through a cross-lingsual retriever.
- 23
- Event Time Extraction and Propagation via Graph Attention Networks23
This paper first formulates this problem based on a 4-tuple temporal representation used in entity slot filling, which allows us to represent fuzzy time spans more conveniently, and proposes a graph attention network-based approach to propagate temporal information over document-level event graphs constructed by shared entity arguments and temporal relations.
- Cross-lingual predicate cluster acquisition to improve bilingual event extraction by inductive learning23
A new inductive learning framework is proposed to automatically augment background data for low-confidence events and then conduct global inference and obtained significant improvement over a state-of-the-art bilingual (English and Chinese) event extraction system.
- Enhanced Knowledge Selection for Grounded Dialogues via Document Semantic Graphs22
This work proposes to automatically convert the background knowledge documents into document semantic graphs and then perform knowledge selection over such graphs to preserve sentence-level information through the use of sentence nodes and provide concept connections between sentences.
- Identifying and Tracking Sentiments and Topics from Social Media Texts during Natural Disasters22
The experimental results demonstrate that LDST performs very well at discovering topics and sentiments from social media and tracking their shifts in different geographical regions during emergencies and disasters.
- Can Language Models Follow Multiple Turns of Entangled Instructions?21
This work presents a systematic investigation of LLMs' capabilities in handling multiple turns of instructions, covering three levels of difficulty: retrieving information from instructions, tracking and reasoning across turns, and (3) resolving conflicts among instructions.
- 21
- Be Appropriate and Funny: Automatic Entity Morph Encoding21
A wide variety of novel approaches to automatically encode proper and interesting morphs, which can effectively pass decoding tests are developed.
- Open-Vocabulary Argument Role Prediction for Event Extraction20
This paper formulate the role prediction problem as an in-filling task and construct prompts for a pre-trained language model to generate candidate roles to infer a set of argument roles for a given event type, resulting in a novel unsupervised framework, RolePred.
- Analysis and repair of name tagger errors20
A joint inference model to improve Chinese name tagging by incorporating feedback from subsequent stages in an information extraction pipeline: name structure parsing, cross-document coreference, semantic relation extraction and event extraction is presented.
- Creative Language Encoding under Censorship19
This position paper systematically categorize human-created obfuscated language on various levels, investigates their basic mechanisms, gives an overview on automated techniques needed to simulate human encoding, and summarizes remaining challenges for future research on the interaction between Natural Language Processing and encryption.
- 19
- Must Read: A Comprehensive Survey of Computational Persuasion18
A taxonomy for persuasion research is introduced and key challenges for future research are discussed to enhance the safety, fairness, and effectiveness of AI-powered persuasion while addressing the risks posed by increasingly capable language models.
- CALM: Unleashing the Cross-Lingual Self-Aligning Ability of Language Model Question Answering18
Evaluations on the MEDQA and X-CSQA datasets demonstrate CALM's effectiveness in enhancing cross-lingual knowledge question answering, both in zero-shot and retrieval-augmented settings and offer a qualitative analysis of how cross-lingual consistency can enhance knowledge alignment.
- Community Moderation and the New Epistemology of Fact Checking on Social Media18
The current approaches to misinformation detection across major platforms are examined, the emerging role of community-driven moderation is explored, and both the promises and challenges of crowd-checking at scale are critically evaluated.
- Geometry Informed Tokenization of Molecules for Language Model Generation18
The Geo2Seq is proposed, which converts molecular geometries into $SE(3)$-invariant 1D discrete sequences and consists of canonical labeling and invariant spherical representation steps, which together maintain geometric and atomic fidelity in a format conducive to LMs.
- 18
- Context-aware Entity Morph Decoding18
This paper presents the first end-to-end context-aware entity morph decoding system that can automatically identify, disambiguate, verify morph mentions based on specific contexts, and resolve them to target entities.
- RPI BLENDER TAC-KBP2015 system description18
To extract English name mentions, a linear-chain CRFs model trained from ACE 20032005 corpora is applied and several regular expression based rules are encoded to extract poster name mentions in discussion forum posts.
- 17
- 17
- Global Attention for Name Tagging17
This work proposes a model that learns to incorporate document-level and corpus-level contextual information alongside local contextual information via document- level and Corpus-level attentions, which dynamically weight their respective contextual information and determines the influence of this information through gating mechanisms.
- Challenges from Information Extraction to Information Fusion17
The requirements and possible solutions to perform information fusion are specified and it is believed this is a critical step to advance IE to a higher level of performance and portability.
- Scientific Opinion Summarization: Paper Meta-review Generation Dataset, Methods, and Evaluation16
The experiments show that human-written summaries do not always satisfy all necessary criteria such as depth of discussion, and identifying consensus and controversy for the specific domain, and the combination of task decomposition and iterative self-refinement shows strong potential for enhancing the opinions and can be applied to other complex text generation using black-box LLMs.
- 16
- 16
- 15
- Name-aware Machine Translation15
A Name-aware Machine Translation approach which can tightly integrate name processing into MT model is proposed, by jointly annotating parallel corpora, extracting name-aware translation grammar and rules, adding name phrase table and name translation driven decoding.
- 15
- 14
- Multimedia Generative Script Learning for Task Planning14
This work proposes a new task, Multimedia Generative Script Learning, to generate subsequent steps by tracking historical states in both text and vision modalities, as well as presenting the first benchmark containing 5,652 tasks and 79,089 multimedia steps.
- Domain-Independent Novel Event Discovery and Semi-Automatic Event Annotation14
This work exploits both monolingual texts and bilingual sentence-aligned parallel texts to cluster event triggers and discover novel event types, and generates event argument annotations semiautomatically, framed as a sentence ranking and semantic role labeling task.
- Collaborative entity extraction and translation14
The use of a new source of data is proposed: the information gleaned from large bitexts and captured by a statistical, phrase-based machine translation system, which can significantly improve Chinese entity extraction and machine translation systems.
- Information Association for Language Model Updating by Mitigating LM-Logical Discrepancy13
This work proposes a new task formulation of the information updating task that only requires the provision of an unstructured updating corpus and evaluates the performance of information updating on the generalizability to question-answer pairs pertaining to the updating information.
- Semi-supervised New Event Type Induction and Description via Contrastive Loss-Enforced Batch Attention13
This work presents a novel approach to semi-supervised new event type induction using a masked contrastive loss, which learns similarities between event mentions by enforcing an attention mechanism over the data minibatch.
- Learning to Learn Words from Visual Scenes13
A meta-learning framework that learns how to learn word representations from unconstrained scenes using the natural compositional structure of language to create training episodes that cause a meta-learner to learn strong policies for language acquisition is introduced.
- Leveraging Entity Linking and Related Language Projection to Improve Name Transliteration13
A novel approach to leverage state-of-the-art Entity Linking techniques to automatically correct name transliteration results, us-ing collective inference from source contexts and additional evidence from knowledge base is proposed.
- Towards LifeSpan Cognitive Systems12
This paper proposes a potential instantiation for LSCS that can integrate all four classes of technologies based on a conceptual metric called Storage Complexity, which measures the relative space required to store past experiences.
- CoRNStack: High-Quality Contrastive Data for Better Code Ranking12
It is demonstrated that contrastive training of embedding models using C O RNS TACK leads to state-of-the-art performance across a variety of code retrieval tasks, and significant improvements in function localization for GitHub issues, an important component of real-world software development.
- 12
- COVID-19 Claim Radar: A Structured Claim Extraction and Tracking System12
A COVID-19 Claim Radar is presented to automatically extract supporting and refuting claims on a daily basis, enabling users to explore equivalent, refuting, or supporting claims with structural evidence, such as shared claimers, similar centroid events and arguments.
- 12
- Joint Detection and Coreference Resolution of Entities and Events with Document-level Context Aggregation12
By extending the jointly trained model to document-level, this work can improve the results by incorporating cross-sentence dependencies and additional contextual information that might not be available at the sentence level, which allows for more globally optimized predictions.
- A study of using an out-of-box commercial MT system for query translation in CLIR12
The experiments show that MT is an excellent tool for the query translation task, and with the help of relevance feedback, it can achieve significant improvement over the monolingual baseline.
- 11
- 11
- Non-Sequential Graph Script Induction via Multimedia Grounding11
This paper designs a multimodal framework to ground procedural videos to WikiHow textual steps and thus transform each video into an observed step path on the latent ground truth graph script, enabling a script knowledge model capable of both generating explicit graph scripts for learnt tasks and predicting future steps given a partial step sequence.
- External Knowledge Acquisition for End-to-End Document-Oriented Dialog Systems11
EKo-Doc is presented, an architecture for document-oriented conversations with access to external knowledge: it is assumed that a conversation is centered around a topic document and that external knowledge is needed to produce responses.
- New Frontiers of Information Extraction11
This tutorial will provide audience with a systematic introduction to recent advances of IE, by answering several important research questions, including how to develop an robust IE system from noisy, insufficient training data, while ensuring the reliability of its prediction.
- Personalized Entity Resolution with Dynamic Heterogeneous KnowledgeGraph Representations11
This work proposes a new framework that leverages personalized features to improve the accuracy of product ranking, and incorporates product, customer, and history representations into a neural reranking model to predict which candidate is most likely to be purchased by a specific customer.
- Learning Hierarchical Relationships among Partially Ordered Objects with Heterogeneous Attributes and Links11
This paper studies a generic form of relationship along which objects can form a treelike structure, a pervasive structure in various domains, and proposes a discriminative undirected graphical model that integrates a wide range of features and rules by defining potential functions with simple forms.
- Analysis and Refinement of Cross-Lingual Entity Linking11
A strong baseline system derived from a combination of state-of-the-art machine translation and monolingual entity linking is enhanced to achieve 11.2% improvement in B-Cubed+ F-measure.
- 11
- SyncMind: Measuring Agent Out-of-Sync Recovery in Collaborative Software Engineering10
This work introduces SyncMind, a framework that systematically defines the out-of-sync problem faced by large language model (LLM) agents in collaborative software engineering (CSE) and creates SyncBench, a benchmark featuring 24,332 instances of agent out-of-sync scenarios in real-world CSE derived from 21 popular GitHub repositories with executable verification tests.
- 10
- Sketching as a Tool for Understanding and Accelerating Self-attention for Long Sequences10
This work proposes Skeinformer to accelerate self-attention and further improve the accuracy of matrix approximation to self-Attention with column sampling, adaptive row normalization and pilot sampling reutilization.
- 10
- Linguistic Resources for Entity Linking Evaluation: from Monolingual to Cross-lingual10
The linguistic resource creation efforts at the Linguistic Data Consortium (LDC) in support of Named Entity Linking evaluation of KBP are described, focusing on annotation methodologies, process, and features of corpora from 2009 to 2011, with a highlighted analysis of the cross-lingual NEL data.
- Finer: Investigating and Enhancing Fine-Grained Visual Concept Recognition in Large Vision Language Models9
A multiple granularity attribute-centric benchmark and training mixture, Finer, is proposed, which aims to establish a ground to evaluate LVLMs’ fine-grained visual comprehension ability and provide significantly improved explainability.
- Relabeling Distantly Supervised Training Data for Temporal Knowledge Base Population9
A temporal knowledge base population system is enhanced to improve the quality of distantly supervised training data and identify a minimal feature set for classification and results in notably less computational cost than a parallel system trained on the full feature set.
- One sense per context cluster: Improving word sense disambiguation using web-scale phrase clustering9
- Automatic recognition of logical relations for English, Chinese and Japanese in the GLARF framework9
GLARF, a framework for representing three linguistic levels and systems for generating this representation, focuses on a logical level, like LFG's F-structure, but compatible with Penn Treebanks.
- Atomic Reasoning for Scientific Table Claim Verification8
This work proposes that enhancing a model's ability to interpret table-based claims involves reducing cognitive load by developing modular, reusable reasoning components (i.e., atomic skills), and introduces a skill-chaining schema that dynamically composes these skills to facilitate more accurate and generalizable reasoning with a reduced cognitive load.
- 8
- 8
- 8
- Cross-lingual Joint Entity and Word Embedding to Improve Entity Linking and Parallel Sentence Mining8
A novel method is proposed, CLEW, to generate cross-lingual data that is a mix of entities and contextual words based on Wikipedia that provides reliable alignment on both the word/entity level and the sentence level, and thus it is used to mine parallel sentences for all (302, 2) language pairs in Wikipedia.
- Applying Coreference to Improve Name Recognition8
A novel method of applying the results of coreference resolution to improve Name Recognition for Chinese by considering first some methods for gauging the confidence of individual tags assigned by a statistical name tagger and presenting rules which use coreference information to correct some name tagging errors.
- 7
- 7
- 7
- 7
- How to Speak a Language without Knowing It7
This work develops a system that lets people overcome language barriers by letting them speak a language they do not know by converts the translation into a phonetic spelling in the user’s own orthography.
- Cross-lingual Slot Filling from Comparable Corpora7
A novel supervised rescoring based validation approach to incorporate global evidence from very large bilingual comparable corpora to overcome information extraction and machine translation errors in this new task of crosslingual slot filling.
- ARMADA: Attribute-Based Multimodal Data Augmentation6
ARMADA is a novel multimodal data generation framework that extracts knowledge-grounded attributes from symbolic KBs for semantically consistent yet distinctive image-text pair generation, and uses the commonsense knowledge of LLMs to modulate auxiliary visual attributes such as backgrounds for more robust representation of original entities.
- Schema-Guided Culture-Aware Complex Event Simulation with Multi-Agent Role-Play6
This work develops a controllable complex news event simulator guided by both the event schema representing domain knowledge about the scenario and user-provided assumptions representing case-specific conditions and introduces a geo-diverse commonsense and cultural norm-aware knowledge enhancement component.
- 6
- Cross-lingual Structure Transfer for Zero-resource Event Extraction6
A novel share-and-transfer framework to reach the goal of transferring graph structures across languages with comparable performance to a state-of-the-art supervised model trained from more than 1,500 manually annotated event mentions.
- 6
- Adding smarter systems instead of human annotators: re-ranking for system combination6
The impact of the learning-to-rank framework is demonstrated to combine output from multiple slot filling systems to populate entity-attribute facts in a knowledge base and it is shown that the approach can be used to create answer keys more efficiently and at a lower cost than laborious human annotation.
- 6
- 5
- 5
- Named Entity Recognition Under Domain Shift via Metric Learning for Life Sciences5
This work investigates the applicability of transfer learning for enhancing a named entity recognition model trained in the biomedical domain to be used in the chemical domain, and proposes a model to transfer the knowledge from the source domain to the target domain, but, at the same time, to project the source entities and target entities into separate regions of the feature space.
- 5
- MIMIR: A Customizable Agent Tuning Platform for Enhanced Scientific Applications5
MIMIR is a streamlined platform that leverages large LLMs to generate agent-tuning data for fine-tuning smaller, specialized models, and enables larger models to simulate various roles and create interaction data, which can be used to fine-tune open-source models like LLaMA2.
- Knowledge-Enriched Natural Language Generation5
This tutorial presents a roadmap to line up the state-of-the-art methods to tackle knowledge-enriched text generation challenges, and dives deep into various technical components: how to represent knowledge, how to feed knowledge into a generation model, and how to evaluate generation results.
- Learning to Learn Words from Narrated Video.5
Experiments show that the proposed framework significantly outperforms the state-of-the-art in visual language modeling for acquiring new words and predicting new compositions and suggest that the visual modality helps the approach more robustly generalize at these tasks.
- Joint Learning of Chinese Words, Terms and Keywords5
A four-level Dirichlet Process based model (DP-4) to jointly learn the word distributions from the corpus, domain and document levels simultaneously simultaneously is proposed and a sentence-wise Gibbs sampler is adopted to obtain proper segmentation results.
- Combining Flat and Structured Approaches for Temporal Slot Filling or: How Much to Compress?5
Experiments on the KBP2011 temporal training data set show that both surface and structured approaches outperform a baseline bag-of-word based classifier and the proposed hybrid method can further improve the performance significantly.
- Enhancing Multi-lingual Information Extraction via Cross-Media Inference and Fusion5
A new information fusion approach to integrate facts extracted from cross-media objects (videos and texts) into a coherent common representation including multi-level knowledge (concepts, relations and events) is described.
- 5
- The effects of high quality translations of named entities in cross-language information exploration5
- 5
- 4
- MolCap-Arena: A Comprehensive Captioning Benchmark on Language-Enhanced Molecular Property Prediction4
This study presents Molecule Caption Arena: the first comprehensive benchmark of LLM-augmented molecular property prediction, and introduces a novel, battle-based rating system to confirm the ability of LLM-extracted knowledge to enhance state-of-the-art molecular representations.
- Understanding the Effect of Data Augmentation on Knowledge Distillation4
This work observes that KD prefers as much data as possible, which is different from fine-tuning that too much data will not gain more performance, and sheds light on the preference difference in data augmentation between fine- tuning and knowledge distillation.
- 4
- ManiTweet: A New Benchmark for Identifying Manipulation of News on Social Media4
An exploratory analysis of human-written tweets reveals intriguing connections between manipulation and the domain and factuality of news articles, as well as revealing that manipulated sentences are more likely to encapsulate the main story or consequences of a news outlet.
- 4
- 4
- Re-ranking Summaries Based on Cross-Document Information Extraction4
A method to automatically incorporate IE results into sentence ranking based on cross-document information extraction results is described, which can significantly improve a high-performing multi-document summarization system.
- Alice: Proactive Learning with Teacher's Demonstrations for Weak-to-Strong Generalization3
Alice is introduced, a framework that leverages complementary knowledge between teacher and student to enhance the learning process and significantly enhances the W2SG performance, yielding substantial improvements in three key tasks compared to the original W2SG.
- MIRACLE: An Online, Explainable Multimodal Interactive Concept Learning System3
With its ability to learn new concepts in an online, few-shot manner and its support of dynamic changes to its knowledge base, MIRACLE represents a step forward in interpretable multimodal learning systems.
- 3
- GLaD: Synergizing Molecular Graphs and Language Descriptors for Enhanced Power Conversion Efficiency Prediction in Organic Photovoltaic Devices3
GLaD achieves precise predictions of PCE, thereby facilitating the synthesis of new OPV molecules with improved efficiency and proving valuable for tasks in low-data regimes within the chemical space, as it enriches molecular representations by incorporating molecular property descriptions learned from large-scale pretraining.
- 3
- 3
- VAULT: VAriable Unified Long Text Representation for Machine Reading Comprehension3
VAULT is proposed: a light-weight and parallel-efficient paragraph representation for MRC based on contextualized representation from long document input, trained using a new Gaussian distribution-based objective that pays close attention to the partially correct instances that are close to the ground-truth.
- MUSE: Textual Attributes Guided Portrait Painting Generation3
A novel approach to automatically generate portrait paintings guided by textual attributes takes a set of attributes written in text, in addition to facial features extracted from a photo of the subject as input, and designs a novel stacked neural network architecture by extending an image-to-image generative model to accept textual attributes.
- 3
- 3
- 3
- Free-form text summarization in social sensing3
A system which can automatically summarize real-time textual data common to social sensing applications, focusing on text messages that describe events in the environment is demonstrated.
- A toolkit for knowledge base population3
This work intends to release an open source KBP toolkit that is publicly available for research purposes and aims to distill entity information from multiple unstructured and semi-structured data sources and incorporate the information into a knowledge base (KB).
- Transducing logical relations from automatic and manual GLARF3
GLARF relations are generated from treebank and parses for English, Chinese and Japanese and evaluation of system output for these input types requires consideration of multiple correct answers.
- Establishing Knowledge Preference in Language Models2
It is demonstrated that a 7B model, fine-tuned on only a few thousand examples automatically generated by the proposed method, effectively achieves superior performance in adhering to the desired knowledge preference hierarchy.
- Automating Intervention Discovery from Scientific Literature: A Progressive Ontology Prompting and Dual-LLM Framework2
A novel framework leveraging large language models (LLMs), which integrates a progressive ontology prompting (POP) algorithm with a dual-agent system, named LLM-Duo, culminating in the creation of a publicly accessible intervention knowledge base with great potential to benefit the speech-language pathology community.
- ReFIT: Relevance Feedback from a Reranker during Inference2
This work leverages the re-ranker to also improve retrieval by providing inference-time relevance feedback to the retriever by updating the retriever’s query representation using a lightweight inference-time distillation of the re-ranker’s prediction for that instance.
- RESIN-EDITOR: A Schema-guided Hierarchical Event Graph Visualizer and Editor2
An interactive event graph visualizer and editor designed for analyzing complex events that allows users to render and freely edit hierarchical event graphs extracted from multimedia and multi-document news clusters with guidance from human-curated event schemas.
- 2
- NLUBot101 at SemEval-2023 Task 3: An Augmented Multilingual NLI Approach Towards Online News Persuasion Techniques Detection2
This work tackled a novel task of classifying persuasion techniques in online news articles at a paragraph level with a cross-lingual data augmentation approach and leveraged a recently proposed multilingual natural language inference model to address these challenges.
- 2
- New Frontiers of Scientific Text Mining2
This tutorial presents a comprehensive overview of recent research and development on scientific text mining, focusing on the biomedical and chemistry domains and introduces a set of methods that perform effective scientific information extraction, such as named entity recognition, relation extraction, and event extraction.
- 2
- 2
- 2
- 2
- Annotating Event Chains for Carbon Sequestration Literature2
This project of annotating event chains for an important scientific domain ― carbon sequestration ― aims to reduce carbon emissions and has been identified by the U.S. National Academy of Engineering as a grand challenge problem for the 21st century.
- EMCompress: Video-LLMs with Endomorphic Multimodal Compression1
This work proposes a novel, cognitively-inspired task -- Endomorphic Multimodal Compression (EMC) -- as a structurally-constrained sufficient-statistic problem for VideoQA, and forms it as an endomorphic transformation F_EMC that compresses the multimodal input while preserving answer invariance across reasonable downstream models.
- 1
- 1
- 1
- 1
- Gene-Metabolite Association Prediction with Interactive Knowledge Transfer Enhanced Graph for Metabolite Production1
An Interactive Knowledge Transfer mechanism based on Metabolism Graphs (IKT4Meta) is introduced to enhance prediction accuracy by integrating cross-metabolism knowledge and using Pretrained Language Models (PLMs) to generate inter-graph links mitigates heterogeneity issues, while intra-graph links are propagated via these anchors.
- 1
- 1
- 1
- 1
- Improving Consistency for Text Summarization with Energy Functions1
To improve the general consistency of abstractive summarization models, EnergySum is introduced, where the Residual Energy-based Model is applied by designing energy scorers that reflect each type of consistency.
- 1
- 1
- Zero-Shot Cross-lingual Name Retrieval for Low-Resource Languages1
A novel method is presented which is able to perform name retrieval on a new language without any additional training data by leveraging freely available, cross-lingual resources and a small amount of training data from another language.
- Collaborative ranking and collaborative clustering1
Comparing to the tradition non-collaborative schemes, collaborative ranking leverages strengths from multiple query collaborators and ranker collaborators while collaborative clustering leverages strength from multiple instance collaborators and clusterer collaborators.
- 1
- Improving information extraction and translation using component interactions1
A new framework based on component interactions to approach the creation of a high-performance NLP pipeline that applies all stages in a suitable order, with each stage generating multiple hypotheses and propagating them through the whole pipeline.
- –
- –
- –
- –
- –
- –
- That's Deprecated! Understanding, Detecting, and Steering Knowledge Conflicts in Language Models for Code Generation–
A domain-agnostic framework for constructing and interpreting such conflicts, along with a novel evaluation method and dataset tailored to code conflict scenarios, are proposed.
- –
- –
- –
- –
- –
- –
- Language + Molecules–
This tutorial is designed to provide an introduction to this area of research and highlight the key roles language can fill for 1) abstract, compositional control of generative models, 2) bridging different biochemical modalities, 3) planning experimental procedures, and 4) broadening access to computational approaches.
- Enriching Conceptual Knowledge in Language Models through Metaphorical Reference Explanation–
Experimental results show that the approach significantly enhances the model’s ability to understand and apply enriched concepts, demonstrating the potential of metaphorical reference identification in improving conceptual knowledge of LMs.
- –
- InfoPattern: Unveiling Information Propagation Patterns in Social Media–
The demo InfoPattern centers on the interplay between language and human ideology, capable of red teaming to simulate adversary responses from opposite ideology communities, and information propagation graph discovery to reveal the evolution of claims across various communities over time.
- –
- –
- –
- –
- MUSE: Illustrating Textual Attributes by Portrait Generation.–
A novel approach, MUSE, to illustrate textual attributes visually via portrait generation by extending an image-to-image generative model to accept textual attributes and proposes a new attribute reconstruction metric to evaluate whether the generated portraits preserve the subject's attributes.
- –
- –
- Research on the Coordinative Development of Interactive Design and Local Traditional Culture Communication–
The transformation of local traditional culture from “one-way communication” to multi-dimensional “interactive communication” under the mode of media development is discussed, and the relationship between interactive design and the communication of local traditional culture is explained.
- Narrating a Knowledge Base–
This work builds a generation framework based on a pointer network which can copy facts from the input KB, and adds two attention mechanisms: (i) slot-aware attention to capture the association between a slot type and its corresponding slot value; and (ii) a new table position self-attention to captured the inter-dependencies among related slots.
- –
- Information Network Construction and Alignment from Automatically Acquired Comparable Corpora–
A novel approach to discover cross-lingual comparable corpora based on video comparison, and a weakly-supervised and language-independent approach to mine name translation pairs that can produce highly reliable name pairs are described.
- Speech and Spoken Document–
Different levels of speech segmentation, approaches to automatically recovering segment bound-ary locations, and experimental results demonstrating impact on several language pro-cessing tasks are described.
- –
- –
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.