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.
Works236 from public data
- Joint Event Extraction via Structured Prediction with Global Features590
This work proposes a joint framework based on structured prediction which extracts triggers and arguments together so that the local predictions can be mutually improved, and proposes to incorporate global features which explicitly capture the dependencies of multiple triggers and argued.
- Incremental Joint Extraction of Entity Mentions and Relations535
An incremental joint framework to simultaneously extract entity mentions and relations using structured perceptron with efficient beam-search is presented, which significantly outperforms a strong pipelined baseline, which attains better performance than the best-reported end-to-end system.
- Refining Event Extraction through Cross-Document Inference515
This work designs a simple scheme to conduct cross-document inference for improving the ACE event extraction task, and applies the hypothesis of “One Sense Per Discourse” to information extraction.
- A Joint Neural Model for Information Extraction with Global Features481
A joint neural framework that aims to extract the globally optimal IE result as a graph from an input sentence and can be easily applied to new languages or trained in a multilingual manner, as OneIE does not use any language-specific feature.
- Knowledge Base Population: Successful Approaches and Challenges402
The techniques which can serve as a basis for a good KBP system are provided, the remaining challenges by comparison with traditional Information Extraction (IE) and Question Answering (QA) tasks are laid out, and some suggestions to address these challenges are provided.
- Towards a Unified Multi-Dimensional Evaluator for Text Generation397
This paper re-frame NLG evaluation as a Boolean Question Answering (QA) task, and by guiding the model with different questions, they can use one evaluator to evaluate from multiple dimensions, and introduces an intermediate learning phase that enables UniEval to incorporate external knowledge from multiple related tasks and gain further improvement.
- Document-Level Event Argument Extraction by Conditional Generation391
A document-level neural event argument extraction model is proposed by formulating the task as conditional generation following event templates by creating the first end-to-end zero-shot event extraction framework.
- MINT: Evaluating LLMs in Multi-turn Interaction with Tools and Language Feedback386
MINT is introduced, a benchmark that evaluates LLMs' ability to solve tasks with multi-turn interactions by (1) using tools and (2) leveraging natural language feedback, and repurpose a diverse set of established evaluation datasets focusing on reasoning, coding, and decision-making into a compact subset for efficient evaluation.
- Unleashing the Emergent Cognitive Synergy in Large Language Models: A Task-Solving Agent through Multi-Persona Self-Collaboration330
This work proposes Solo Performance Prompting (SPP), which transforms a single LLM into a cognitive synergist by engaging in multi-turn self-collaboration with multiple personas and shows that assigning multiple fine-grained personas in LLMs improves problem-solving abilities compared to using a single or fixed number of personas.
- CoType320
A novel domain-independent framework that jointly embeds entity mentions, relation mentions, text features and type labels into two low-dimensional spaces, and adopts a novel partial-label loss function for noisy labeled data and introduces an object "translation" function to capture the cross-constraints of entities and relations on each other.
- Visual Attention Model for Name Tagging in Multimodal Social Media314
This paper creates two new multimodal datasets and proposes a novel model architecture based on Visual Attention that not only provides deeper visual understanding on the decisions of the model, but also significantly outperforms other state-of-the-art baseline methods for this task.
- Moralization in social networks and the emergence of violence during protests228
By analysing the language of tweets around protests in Baltimore in 2015 and through behavioural laboratory experiments, Dehghani and colleagues find that moralization of protest issues leads to greater support for violence and increased incidence of violent protest.
- Zero-Shot Transfer Learning for Event Extraction220
A transferable architecture of structural and compositional neural networks is designed to jointly represent and map event mentions and types into a shared semantic space and can select, for each event mention, the event type which is semantically closest in this space as its type.
- FaitCrowd199
Experimental results on two real-world datasets show that FaitCrowd can significantly reduce the error rate of aggregation compared with the state-of-the-art multi-source aggregation approaches due to its ability of learning topical expertise from question content and collected answers.
- A Novel Neural Topic Model and Its Supervised Extension173
A novel neural topic model (NTM) is proposed where the representation of words and documents are efficiently and naturally combined into a uniform framework and is competitive in both topic discovery and classification/regression tasks.
- Revisiting Out-of-distribution Robustness in NLP: Benchmark, Analysis, and LLMs Evaluations163
This paper reexamines the research on out-of-distribution (OOD) robustness in the field of NLP and proposes a benchmark construction protocol that ensures clear differentiation and challenging distribution shifts.
- CRAFT: Customizing LLMs by Creating and Retrieving from Specialized Toolsets154
CRAFT is designed to be flexible and offers a plug-and-play approach to adapt off-the-shelf LLMs to unseen domains and modalities, without any finetuning, and achieves substantial improvements compared to strong baselines.
- LM-Infinite: Zero-Shot Extreme Length Generalization for Large Language Models151
LM-Infinite is a simple and effective method for enhancing LLMs’ capabilities of handling long contexts, which allows LLMs pre-trained with 2K or 4K-long segments to generalize to up to 200M length inputs while retaining perplexity.
- Keep Meeting Summaries on Topic: Abstractive Multi-Modal Meeting Summarization149
An abstractive meeting summarizer from both videos and audios of meeting recordings is developed, which significantly outperforms the state-of-the-art with both BLEU and ROUGE measures.
- Label Noise Reduction in Entity Typing by Heterogeneous Partial-Label Embedding149
A global objective is formulated for learning the embeddings from text corpora and knowledge bases, which adopts a novel margin-based loss that is robust to noisy labels and faithfully models type correlation derived from knowledge bases.
- 147
- AFET: Automatic Fine-Grained Entity Typing by Hierarchical Partial-Label Embedding146
This paper proposes a novel embedding method to separately model “clean” and “noisy” mentions, and incorporates the given type hierarchy to induce loss functions.
- 141
- R-Tuning: Instructing Large Language Models to Say ‘I Don’t Know’137
Refusal-Aware Instruction Tuning is presented, formalized by first identifying the disparity in knowledge encompassed by pre-trained parameters compared to that of instruction tuning data, and constructing the refusal-aware data based on the knowledge intersection to tune LLMs to refrain from responding to questions beyond its parametric knowledge.
- Cross-media Structured Common Space for Multimedia Event Extraction133
A novel method is proposed, Weakly Aligned Structured Embedding (WASE), that encodes structured representations of semantic information from textual and visual data into a common embedding space that enables exploiting available resources without explicit cross-media annotation.
- Code4Struct: Code Generation for Few-Shot Event Structure Prediction130
It is shown that, with sufficient in-context examples, formulating EAE as a code generation problem is advantageous over using variants of text-based prompts, and Code4Struct is comparable to supervised models trained on 4,202 instances and outperforms current state-of-the-art (SOTA) trained on 20-shot data by 29.5% absolute F1.
- Tweet, but verify: epistemic study of information verification on Twitter129
This study surveys users on credibility perceptions regarding witness pictures posted on Twitter related to Hurricane Sandy, and unveils insight about tweet presentation, as well as features that users should look at when assessing the veracity of tweets in the context of fast-paced events.
- Constructing Information Networks Using One Single Model122
A new framework is proposed that unifies the output of three information extraction tasks - entity mentions, relations and events as an information network representation, and extracts all of them using one single joint model based on structured prediction.
- Joint Entity and Event Extraction with Generative Adversarial Imitation Learning119
This work proposes a new framework for entity and event extraction based on generative adversarial imitation learning—an inverse reinforcement learning method using a generatives adversarial network (GAN).
- Bridge Text and Knowledge by Learning Multi-Prototype Entity Mention Embedding99
A novel Multi-Prototype Mention Embedding model is proposed, which learns multiple sense embeddings for each mention by jointly modeling words from textual contexts and entities derived from a knowledge base, and an efficient language model based approach to disambiguate each mention to a specific sense.
- Entity-aware Image Caption Generation86
A convolutional neural networks - long short term memory networks (CNN-LSTM) model is trained to generate a template caption based on the input image and a knowledge graph based collective inference algorithm is used to fill in the template with specific named entities retrieved via the hashtags.
- Biomedical Event Extraction based on Knowledge-driven Tree-LSTM84
A novel knowledge base (KB)-driven tree-structured long short-term memory networks (Tree-LSTM) framework is proposed, incorporating two new types of features: dependency structures to capture wide contexts and entity properties from external ontologies via entity linking.
- Low-Resource Name Tagging Learned with Weakly Labeled Data83
A novel neural model for name tagging solely based on weakly labeled (WL) data is proposed, so that it can be applied in any low-resource settings and demonstrate superior performance as well as efficiency.
- Heterogeneous Supervision for Relation Extraction: A Representation Learning Approach83
This work proposes a novel framework, REHession, to conduct relation extractor learning using annotations from heterogeneous information source, e.g., knowledge base and domain heuristics, and adopts embedding techniques to learn the distributed representations of context.
- Collaborative Ranking: A Case Study on Entity Linking81
In contrast to traditional non-collaborative ranking scheme which solely relies on the strengths of isolated queries and one stand-alone ranking algorithm, the new scheme integrates the strengths from multiple collaborators of a query and the strength from multiple ranking algorithms.
- Predicting unknown time arguments based on cross-event propagation81
Two methods are described, one based on rules and the other based on statistical learning, to predict the unknown time argument for an event by the propagation from its related events.
- Entity linking for biomedical literature79
A novel unsupervised collective inference approach is proposed to address the EL problem in a new domain and is able to outperform a current state-of-the-art supervised approach that has been trained with a large amount of manually labeled data.
- Comparison of the Impact of Word Segmentation on Name Tagging for Chinese and Japanese77
This paper presents several new observations and analysis on the impact of word segmentation on name tagging in Chinese and Japanese, and suggests it is crucial to keep segmentation settings consistent between training and testing for name tagging.
- The Wisdom of Minority: Unsupervised Slot Filling Validation based on Multi-dimensional Truth-Finding77
A novel unsupervised multi-dimensional truth finding framework which incorporates signals from multiple sources, multiple systems and multiple pieces of evidence by knowledge graph construction through multi-layer deep linguistic analysis is presented.
- Acquiring Background Knowledge to Improve Moral Value Prediction69
A novel approach to automatically acquire background knowledge from an external knowledge base to enrich input texts and thus improve moral value prediction by combining basic textual features with background knowledge achieves performance comparable to a single human annotator.
- Improving Subject-Area Question Answering with External Knowledge68
This work explores simple yet effective methods for exploiting two sources of externalknowledge for exploiting unstructured external knowledge for subject-area QA on multiple-choice question answering tasks in subject areas such as science.
- RESIN: A Dockerized Schema-Guided Cross-document Cross-lingual Cross-media Information Extraction and Event Tracking System65
A new information extraction system that can automatically construct temporal event graphs from a collection of news documents from multiple sources, multiple languages, and multiple data modalities is presented.
- 65
- Measuring and Improving Chain-of-Thought Reasoning in Vision-Language Models63
This work evaluates existing state-of-the-art VLMs, and finds that even the best-performing model is unable to demonstrate strong visual reasoning capabilities and consistency, indicating that substantial efforts are required to enable VLMs to perform visual reasoning as systematically and consistently as humans.
- RCOT: Detecting and Rectifying Factual Inconsistency in Reasoning by Reversing Chain-of-Thought61
This work proposes RCoT (Reversing Chain-of-Thought), a novel method to improve LLMs' reasoning abilities by automatically detecting and rectifying factual inconsistency in LLMs, generated solutions and finds that manually written fine-grained feedback can dramatically improve LLM's reasoning abilities.
- Can one language bootstrap the other: a case study on event extraction60
This paper proposes a new bootstrapping framework using cross-lingual information projection that is particularly effective for a challenging NLP task which is situated at the end of a pipeline and thus suffers from the errors propagated from upstream processing and has low-performance baseline.
- Describing a Knowledge Base59
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.
- Modeling Truth Existence in Truth Discovery57
This work proposes a probabilistic graphical model, which simultaneously infers truth as well as source quality without any a priori training involving ground truth answers, and proposes an initialization scheme based upon a quantity named truth existence score, which synthesizes two indicators, namely, participation rate and consistency rate.
- Data selection in semi-supervised learning for name tagging57
Two semi-supervised learning techniques are presented to improve a state-of-the-art multi-lingual name tagger and it is concluded that simply relying upon large corpora is not in itself sufficient: one must pay attention to unlabeled data selection too.
- Weakly-Supervised Aspect-Based Sentiment Analysis via Joint Aspect-Sentiment Topic Embedding55
A weakly-supervised approach for aspect-based sentiment analysis, which uses only a few keywords describing each aspect/sentiment without using any labeled examples, which generates quality joint topics and outperforms the baselines significantly on benchmark datasets.
- Improving Event Extraction via Multimodal Integration55
This paper first discovers visual patterns from large-scale text-image pairs in a weakly-supervised manner and then proposes a multimodal event extraction algorithm where the event extractor is jointly trained with textual features and visual patterns.
- 55
- Minimize Exposure Bias of Seq2Seq Models in Joint Entity and Relation Extraction52
A novel Sequence-to-Unordered-Multi-Tree (Seq2UMTree) model is proposed to minimize the effects of exposure bias by limiting the decoding length to three within a triplet and removing the order among triplets.
- CAMR at SemEval-2016 Task 8: An Extended Transition-based AMR Parser52
CAMR, the transition-based parser that is used in the SemEval-2016 Meaning Representation Parsing task, is described and three sets of new features are introduced: 1) rich named entities, 2) a verbalization list, 3) semantic role labels.
- 52
- Dynamic Global Memory for Document-level Argument Extraction48
A new global neural generation-based framework for document-level event argument extraction is introduced by constructing a document memory store to record the contextual event information and leveraging it to implicitly and explicitly help with decoding of arguments for later events.
- Curating and contextualizing Twitter stories to assist with social newsgathering48
TweetGathering is a prototype tool that provides curated and contextualized access to news stories on Twitter, built with the aim of assisting journalists both with gathering and with researching news stories as users comment on them.
- Cross-document Event Extraction and Tracking: Task, Evaluation, Techniques and Challenges48
It is demonstrated that global inference from background knowledge and cross-document event aggregation are crucial to enhance the performance of this new task of cross- document event extraction and tracking and its evaluation metrics.
- Open-Schema Event Profiling for Massive News Corpora47
This work develops a Bayesian non-parametric model to detect events and event types by exploiting the slot expressions of the entities mentioned in news articles and proposes an unsupervised embedding model for schema induction that encodes the insight.
- Seed-Based Event Trigger Labeling: How far can event descriptions get us?47
This work proposes an alternative approach, which takes the example trigger terms mentioned in the guidelines as seeds, and then applies an eventindependent similarity-based classifier for trigger labeling, which can skip manual annotation for new event types, while requiring only minimal annotated training data for few example events at system setup.
- 47
- SOLO: A Single Transformer for Scalable Vision-Language Modeling46
This paper introduces the first open-source training recipe for developing SOLO, an open-source 7B LVLM using moderate academic resources, and demonstrates performance comparable to LLaVA-v1.5-7B, particularly excelling in visual mathematical reasoning.
- Timeline Summarization based on Event Graph Compression via Time-Aware Optimal Transport45
A time-aware optimal transport distance is introduced for learning the compression model in an unsupervised manner and it is shown that the approach significantly improves on the state of the art on three real-world datasets, including two public standard benchmarks and the newly collected Timeline100 dataset.
- 43
- Incorporating Background Knowledge into Video Description Generation40
This work develops an approach that uses video meta-data to retrieve topically related news documents for a video and extracts the events and named entities from these documents and generates a description using a Knowledge-aware Video Description network.
- Language and Domain Independent Entity Linking with Quantified Collective Validation40
A novel unsupervised algorithm named Quantified Collective Validation is presented that avoids excessive linguistic analysis on the source documents and fully leverages the knowledge base structure for the entity linking task.
- RESIN-11: Schema-guided Event Prediction for 11 Newsworthy Scenarios39
This work introduces RESIN-11, a new schema-guided event extraction&prediction framework that can be applied to a large variety of newsworthy scenarios and builds a demo website based on the dockerized system and schema library.
- Utilizing Relative Event Time to Enhance Event-Event Temporal Relation Extraction38
This paper proposes a joint model for event-event temporal relation classification and an auxiliary task, relative event time prediction, which predicts the event time as real numbers, and adopts the Stack-Propagation framework to incorporate predicted relative eventTime prediction and keep the differentiability.
- Lifelong Event Detection with Knowledge Transfer37
A new lifelong learning framework that focuses on lifelong event detection as an exemplar case and proposes a new problem formulation that is also generalizable to other IE tasks, showing that it helps consolidate old knowledge and improve novel knowledge acquisition.
- Event-Centric Natural Language Processing37
This tutorial will provide audience with a systematic introduction of knowledge representations of events, various methods for automated extraction, conceptualization and prediction of events and their relations, and a wide range of NLU and commonsense understanding tasks that benefit from aforementioned techniques.
- Learning to Decompose Visual Features with Latent Textual Prompts36
Decomposed Feature Prompting (DeFo) leverages a flexible number of learnable embeddings as textual input while maintaining the vision-language dual-model architecture, which enables the model to learn decomposed visual features with the help of feature-level textual prompts.
- Building a Cross-document Event-Event Relation Corpus35
This work designs an annotation schema and builds a corpus for this new task of extracting event-event relations across documents, and lays out the main challenges based on detailed inter-annotator disagreement and error analysis.
- 34
- Knowledge Overshadowing Causes Amalgamated Hallucination in Large Language Models33
It is shown that the hallucination rate grows with both the imbalance ratio (between the popular and unpopular condition) and the length of dominant condition description, consistent with the derived generalization bound.
- 32
- 32
- Name Tagging for Low-resource Incident Languages based on Expectation-driven Learning32
This paper tackles a challenging name tagging problem in an emergent setting the tagger needs to be complete within a few hours for a new incident language (IL) using very few resources and proposes a new expectation-driven learning framework that rapidly acquire, categorize, structure and zoom in on ILspecific expectations.
- Cross-lingual Multi-Level Adversarial Transfer to Enhance Low-Resource Name Tagging31
A new neural architecture is developed that leverages multi-level adversarial transfer that projects source language words into the same semantic space as those of the target language without using any parallel corpora or bilingual gazetteers and yields language-agnostic sequential features.
- A Multimodal-Sensor-Enabled Room for Unobtrusive Group Meeting Analysis31
A smart meeting room that combines microphones and unobtrusive ceiling-mounted Time-of-Flight (ToF) sensors to understand group dynamics in team meetings and derives metrics from these automatic estimates and correlates them with user-reported rankings of emergent group leaders and major contributors to produce accurate predictors.
- Paper Abstract Writing through Editing Mechanism31
A paper abstract writing system based on an attentive neural sequence-to-sequence model that can take a title as input and automatically generate an abstract and a novel Writing-editing Network that can attend to both the title and the previously generated abstract drafts and then iteratively revise and polish the abstract.
- Harnessing the Reasoning Economy: A Survey of Efficient Reasoning for Large Language Models30
This survey provides a comprehensive analysis of reasoning economy in both the post-training and test-time inference stages of LLMs, encompassing the cause of reasoning inefficiency, behavior analysis of different reasoning patterns, and potential solutions to achieve reasoning economy.
- Joint inference for cross-document information extraction30
A joint inference approach based on information network structure to conduct cross-fact reasoning with an integer linear programming framework is described, which obtained 13.7%-24.4% user browsing cost reduction over a state-of-the-art IE system.
- Self-Improving LLM Agents at Test-Time29
This work explores a new test-time self-improvement method to create more effective and generalizable agentic LMs on-the-fly, and highlights the promise of TT-SI, demonstrating the potential of self-improvement algorithms at test-time as a new paradigm for building more capable agents toward self-evolution.
- RPI BLENDER TAC-KBP2016 System Description.29
To extract name mentions from Chinese and Spanish documents, bi-directional LSTMs (Long Short Term Memory) networks which can leverage long distance features are used which acquired and exploited non-traditional language-specific resources, and incorporate them into a novel neural network framework.
- 28
- Wikification and Beyond: The Challenges of Entity and Concept Grounding28
This work focuses on Wikification, a task that has received increased attention in recent years from the NLP and Data Mining communities, partly fostered by the U.S. NIST Text Analysis Conference Knowledge Base Population track.
- 27
- Stage-wise Fine-tuning for Graph-to-Text Generation27
This paper proposes a structured graph-to-text model with a two-step fine-tuning mechanism which first fine-tunes model on Wikipedia before adapting to the graph- to-text generation, and proposes a novel tree-level embedding method to capture the inter-dependency structures of the input graph.
- UserRL: Training Interactive User-Centric Agent via Reinforcement Learning25
This work systematically vary turn-level reward assignment and trajectory-level score calculation to analyze how different formulations affect learning under the GRPO algorithm, and establishes UserRL as a practical pathway for developing robust user-centric agentic models.
- Open Relation Extraction and Grounding25
This work proposes a novel importance-based open RE approach by exploiting the global structure of a dependency tree to extract salient triples from large-scale corpora by leveraging KB triples and weighted context words associated with relational triples.
- Bring you to the past: Automatic Generation of Topically Relevant Event Chronicles25
Experimental results demonstrate the first novel approach to automatically generate a topically relevant event chronicle during a certain period given a reference chronicles during another period is promising to tackle this new problem.
- Exploring and inferring user–user pseudo‐friendship for sentiment analysis with heterogeneous networks25
A novel information network‐based framework which can infer hidden similarity and dissimilarity between users by exploring similar and opposite opinions is proposed, so as to improve post‐level and user‐level sentiment classification at the same time.
- Resolving Entity Morphs in Censored Data25
This paper exploits temporal constraints to collect crosssource comparable corpora relevant to any given morph query and identify target candidates and proposes various novel similarity measurements including surface features, meta-path based semantic features and social correlation features and combine them in a learning-to-rank framework.
- Human-in-the-loop Schema Induction24
It is shown that the human-in-the-loop schema induction system not only transfers to new domains more easily than previous approaches, but also reduces efforts of human curation thanks to the interactive interface.
- Reliability-aware Dynamic Feature Composition for Name Tagging24
A novel reliability-aware name tagging model that is able to dynamically select and compose features such as word embedding and character-level representation using gating mechanisms and improves the performance for most genre pairs and obtains up to 5% absolute F-score gain.
- Liberal Entity Extraction: Rapid Construction of Fine-Grained Entity Typing Systems24
This article proposes a novel unsupervised entity-typing framework by combining symbolic and distributional semantics, and develops a novel joint hierarchical clustering and linking algorithm to type all mentions using these representations.
- Unsupervised Person Slot Filling based on Graph Mining24
A graph-based algorithm to automatically identify triggers based on personalized PageRank and Affinity Propagation for a given query and slot type and can be quickly adapted to any language and new slot types.
- Tackling representation, annotation and classification challenges for temporal knowledge base population22
Experimental results demonstrated that the annotation enhancement approach dramatically increased the speed of the training procedure, and that the flat and structured classification approaches were complementary, together yielding a state-of-the-art TSF system.
- EVEDIT: Event-based Knowledge Editing with Deductive Editing Boundaries21
This work introduces a theoretical framework for KE that highlights an overlooked set of knowledge that remains unchanged and aids in knowledge deduction during editing and names as the deduction anchor, and proposes a novel task of event-based knowledge editing that pairs facts with event descriptions.
- Enhanced Chart Understanding in Vision and Language Task via Cross-modal Pre-training on Plot Table Pairs21
ChartT5 is introduced, a V+L model that learns how to interpret table information from chart images via cross-modal pre-training on plot table pairs via Masked Header Prediction and Masked Value Prediction to facilitate the model with different skills to interpret the table information.
- Event Specific Multimodal Pattern Mining for Knowledge Base Construction21
This paper proposes a novel multimodal pattern mining approach towards constructing a high-level "event" schema semi-automatically, which has the capability to extend text only methods for schema construction.
- 20
- Democratizing LLMs: An Exploration of Cost-Performance Trade-offs in Self-Refined Open-Source Models19
A domain-agnostic self-refinement process coupled with a novel ranking metric facilitates informed decision-making in model selection, thereby reducing costs and democratizing access to high-performing language models, as evidenced by case studies.
- 19
- Multilingual Entity, Relation, Event and Human Value Extraction19
A state-of-the-art end-to-end multilingual (English, Russian, and Ukrainian) knowledge extraction system that can perform entity discovery and linking, relation extraction, event extraction, and coreference is demonstrated.
- A Grounded Unsupervised Universal Part-of-Speech Tagger for Low-Resource Languages19
This work describes an approach for low-resource unsupervised POS tagging that yields fully grounded output and requires no labeled training data and demonstrates the tagger’s utility by incorporating it into a true ‘zero-resource’ variant of the MALOPA dependency parser model.
- List-only Entity Linking19
This work selects most linkable mentions as seed mentions and disambiguate other mentions by comparing them with the seed mentions rather than directly with the entities.
- Cross-media Event Extraction and Recommendation19
A comprehensive system that searches, identifies, organizes and summarizes complex events from multiple data modalities, and recommends events related to the user’s ongoing search based on previously selected attribute values and dimensions of events being viewed is developed.
- Cross-document Event Coreference Resolution based on Cross-media Features19
A novel approach to jointly encode effective features from both closed captions and video key frames is developed, demonstrating that visual features provided 7.2% absolute F-score gain on stateof-the-art text based event extraction and coreference resolution.
- Constrained information-theoretic tripartite graph clustering to identify semantically similar relations19
This paper formulates relation clustering as a constrained tripartite graph clustering problem, presents an efficient clustering algorithm and exhibits the advantage of the constrained framework.
- 19
- RPI-BLENDER TAC-KBP2013 Knowledge Base Population System19
This year the RPI-BLENDER team participated in the following four tasks: English Entity Linking, Regular slot Filling, Temporal Slot Filling and Slot F filling validation.
- 18
- A Comparison of Event Representations in DEFT18
Comparisons of event representations are presented, along with a comparison of data annotated according to each event representation, with the purpose of being able to compare actual annotation across all of these approaches as directly as possible.
- 17
- Detecting Deceptive Groups Using Conversations and Network Analysis17
Experimental results demonstrate that the approach significantly outperforms human voting and state-of-theart subgroup detection methods at dynamically differentiating the deceptive groups from truth-tellers.
- 16
- Expertise in Social Networks: How Do Experts Differ from Other Users?16
A heuristic-based approach for automated identification of expertis eon Twitter is proposed and it is shown that these heuristics provide interesting insights regarding how experts differ from other user groups which can help guide future studies in this areas and algorithms for expertise location.
- The unobtrusive group interaction (UGI) corpus16
A new, publicly available multi-modal dataset for group dynamics study that differs from previous datasets in its use of ceiling-mounted, unobtrusive depth sensors, which can be used for fine-grained analysis of head and body pose and gestures.
- Multi-lingual Common Semantic Space Construction via Cluster-consistent Word Embedding16
This work introduces a new cluster-consistent correlational neural network to construct the common semantic space by aligning words as well as clusters, and introduces multiple cluster-level alignments and enforce the word clusters to be consistently distributed across multiple languages.
- ELISA-EDL: A Cross-lingual Entity Extraction, Linking and Localization System16
ELISA-EDL, a state-of-the-art re-trainable system to extract entity mentions from low-resource languages, link them to external English knowledge bases, and visualize locations related to disaster topics on a world heatmap is demonstrated.
- 16
- NewsNetExplorer16
The proposed NewsNetExplorer system combines a set of news information network exploration and mining mechanisms that explore news in multi-dimensional space, and helps to gain insight on how to perform effective information extraction, integration and mining in large unstructured datasets.
- 16
- Top-Down and Bottom-Up: A Combined Approach to Slot Filling16
This work describes two bottom-up Information Extraction style pipelines and a top-down Question Answering style pipeline to address the Slot Filling task, and proposes several novel approaches to enhance these pipelines, including statistical answer re-ranking and Markov Logic Networks based cross-slot reasoning.
- HySPA: Hybrid Span Generation for Scalable Text-to-Graph Extraction15
A Hybrid Span Generator (HySPA) is proposed that invertibly maps the information graph to an alternating sequence of nodes and edge types, and directly generates such sequences via a hybrid span decoder which can decode both the spans and the types recurrently in linear time and space complexities.
- Chengyu Cloze Test15
A neural recommendation model for Chengyu, which is a special type of Chinese idiom, is presented, which achieves 89.5% accuracy on cloze test and outperforms human subjects who attended competitive universities in China.
- Platforms for Non-speakers Annotating Names in Any Language15
Two annotation platforms that allow an English speaker to annotate names for any language without knowing the language are demonstrated and other methods of creating silver- and gold-standard annotations using native speakers are compared.
- Automatic Entity Recognition and Typing in Massive Text Data15
Data-driven methods to recognize typed entities of interest in massive, domain-specific text corpora are introduced and can automatically identify token spans as entity mentions in documents and label their fine-grained types in a scalable way.
- CUNY-BLENDER TAC-KBP2012 Entity Linking System and Slot Filling Validation System15
This year the CUNY-BLENDER team participated in the English Entity Linking and Slot Filling Validation tracks and applies two new techniques, collaborative clustering and query reformulation to entity linking and answer validation.
- 14
- EscapeBench: Towards Advancing Creative Intelligence of Language Model Agents14
EscapeBench is introduced —a benchmark suite of room escape game environments designed to challenge agents with creative reasoning, unconventional tool use, and iterative problem-solving to uncover implicit goals.
- 14
- 14
- 14
- Name List Only? Target Entity Disambiguation in Short Texts14
This paper proposes a graphbased model called TremenRank to collectively identify target entities in short texts given a name list only that propagates trust within the graph, allowing for an arbitrary number of target entities and texts using inverted index technology.
- CUNY BLENDER TAC-KBP2011 Temporal Slot Filling System Description14
The participation of the CUNY-BLENDER team in the Temporal Slot Filling (TSF) pilot task organized as part of the TAC-KBP2010 evaluation is described.
- Text2DB: Integration-Aware Information Extraction with Large Language Model Agents13
An LLM agent framework OPAL (Observe-PlanAnalyze LLM) which includes an Observer component that interacts with the database, the Planner component that generates a code-based plan with calls to IE models, and the Analyzer component that provides feedback regarding code quality before execution is proposed.
- A Language-First Approach for Procedure Planning13
This work verifies the planning ability of LMs and demonstrates how LMs can serve as a powerful “reasoning engine” even when the input is provided in another modal029 ity.
- GAIA - A Multi-media Multi-lingual Knowledge Extraction and Hypothesis Generation System.13
This research highlights the need to understand more fully the role of emotion in human interaction and the role that language plays in the development of language and culture.
- 12
- 12
- 12
- Identifying News from Tweets12
This work proposes a novel newsworthiness classifier trained through active learning and investigates human assessment and automatic methods to encode it on both the tweet and trending topic levels and shows that content and journalism based features proved to be effective for ranking and classifying content on Twitter.
- Knowledge Graph and Semantic Computing: Semantic, Knowledge, and Linked Big Data12
A model which jointly learns the embeddings of multiple knowledge bases (KBs) in a uniform vector space to align entities in KBs and shows that the proposed approach which only utilize the structure information of KBs also works well.
- 11
- A Weibo Dataset for the 2022 Russo-Ukrainian Crisis11
The Russia-Ukraine Crisis Weibo (RUW) dataset, with over 3.5M user posts and comments in the first release, is presented.
- 11
- 11
- Bitext Name Tagging for Cross-lingual Entity Annotation Projection11
This paper focuses on the named entity recognition (NER) task and proposes a weakly-supervised framework to project entity annotations from English to IL through bitexts, which combines advantages of rule-based methods and deep learning methods.
- 10
- GENE: Global Event Network Embedding10
A novel framework is proposed, Global Event Network Embedding (GENE), that encodes the event network with a multi-view graph encoder while preserving the graph topology and node semantics.
- A Multi-media Approach to Cross-lingual Entity Knowledge Transfer10
A novel approach to discover comparable documents in high-resource languages (HLs) and project Entity Discovery and Linking results from HLs documents back to LLs, which leverage a wide variety of language-independent forms from multiple data modalities.
- 9
- Embracing Non-Traditional Linguistic Resources for Low-resource Language Name Tagging9
Experiments show that feeding linguistic knowledge can make DNN significantly more robust to noise, achieving 8%-22% absolute F-score gains on name tagging without using any human annotation.
- Incident-Driven Machine Translation and Name Tagging for Low-resource Languages9
Novel approaches to tackling the problem of natural language processing for low-resource languages are described, including universal tools, rapid resource and knowledge acquisition, rapid language projection, and joint methods for MT and name tagging.
- ActionIE: Action Extraction from Scientific Literature with Programming Languages8
A CTION IE is introduced, a method that leverages Large Language Models (LLMs) to bridge this divide by converting actions written in natural language into executable Python code, which enables us to capture the entities of interest, and the relationship between each action, given the features of Programming Languages.
- Parameter-Efficient Tuning Helps Language Model Alignment8
The approach, alignMEnt with parameter-Efficient Tuning (MEET), improves the quality of control tokens, thus improving controllable generation quality consistently by an apparent margin on two well-recognized datasets compared with prior works.
- 8
- Multi-lingual Entity Discovery and Linking8
This tutorial discusses and compares multiple methods that make use of multi-lingual word embeddings and presents EL methods that work for both name tagging and linking in very low resource languages.
- 8
- An Iterative Link-based Method for Parallel Web Page Mining8
The basic idea of the method is that the translation similarity of two pages can be inferred from their neighbor pages, which can be adopted as an important source of external similarity.
- 7
- 7
- Entity-Conditioned Question Generation for Robust Attention Distribution in Neural Information Retrieval7
This work uses a novel targeted synthetic data generation method that identifies poorly attended entities and conditions the generation episodes on those to teach neural IR to attend more uniformly and robustly to all entities in a given passage.
- Multi-level cross-lingual attentive neural architecture for low resource name tagging7
This paper combines multi-level cross-lingual knowledge as attention into a neural architecture, which guides low resource name tagging to achieve a better performance and is able to yield improvements in name tagging over all previous baselines.
- NYU's Chinese ACE 2005 EDR System Description Can Global Re-Ranking and Semantic Role Labeling help Chinese EDR?7
The overall architecture of the NYU Chinese ACE 2005 EDR (Entity Detection and Recognition) system is presented, focussing mainly on mention detection, using an N-Best approach to generate multiple name hypotheses and apply parsing, chunking and list matching to generatemultiple nominal hypotheses.
- Visually Descriptive Language Model for Vector Graphics Reasoning6
The Visually Descriptive Language Model (VDLM) is proposed, which performs text-based reasoning about vector graphics and achieves stronger zero-shot performance compared to state-of-the-art LMMs, such as GPT-4V, in various low-level multimodal perception and reasoning tasks on vector graphics.
- Augmentation with Projection: Towards an Effective and Efficient Data Augmentation Paradigm for Distillation6
This work proposes AugPro (Augmentation with Projection), an effective and efficient data augmentation method for distillation that builds on top of representation interpolation augmentation methods to maintain the diversity of expressions and converts the augmented data to tokens to avoid shifting decision boundaries.
- 6
- From Image to Translation6
This paper developed an end-to-end system to read a scanned hand-written Nyushu article, segment it into characters, link them to standard characters, and then translate the article into Mandarin Chinese.
- 6
- Why Read if You Can Scan? Trigger Scoping Strategy for Biographical Fact Extraction6
This paper mimics a scanning process to extract biographical facts by using event and relation triggers as triggers to find specific facts without having to read all the words.
- A Large-Scale Study of Reranker Relevance Feedback at Inference5
This paper studies inference-time reranker relevance feedback extensively across multiple retrieval domains, languages, and modalities, while also investigating aspects such as the performance and latency implications of the number of distillation updates and feedback candidates.
- Shorten the Long Tail for Rare Entity and Event Extraction5
A new transformation module is designed that transforms infrequent candidate mention representation during evaluation with the average mention representation in the training dataset to handle the generalization challenge by making the evaluation instances closer to the frequent training cases.
- 5
- 5
- Toward Consistent and Informative Event-Event Temporal Relation Extraction5
This model combines a pretrained language model and a graph neural network to output event embeddings, which captures the contextual information of event graphs and achieves state-of-the-art performance and generates high-quality event graphs.
- Schema-Guided Event Graph Completion5
This work uses event schema, a template that describes the stereotypical structure of event graphs, to address the above issues and proposes a self-supervised strategy to construct training samples, as well as an inference algorithm that is specifically designed to complete event graphs.
- EventKE: Event-Enhanced Knowledge Graph Embedding5
This paper proposes an event-enhanced KG embedding model EventKE, which incorporates event nodes by building a heterogeneous network, where entity nodes and event nodes are distributed on the two sides of the network inter-connected by event argument links, and designs a novel and effective attention-based message passing method.
- Error Analysis of Uyghur Name Tagging: Language-specific Techniques and Remaining Challenges5
It is concluded that simply adopting a machine learning model which is proven successful for high-resource languages along with language-independent superficial features is unlikely to be effective for Uyghur, or low- resource languages in general.
- Cross-genre Event Extraction with Knowledge Enrichment5
This paper incorporates cross-genre knowledge to boost the event extractor performance and demonstrates that without any additional annotations, the proposed approach is able to provide 15% absolute F-score improvement over the state-of-the-art.
- Successful Data Mining Methods for NLP5
This tutorial aims to bridge the knowledge gap between these two fields and speed up the transition process by introducing two types of DM methods: those state-of-the-art DM methods that have already been proven effective for NLP; and some newly developedDM methods that the author believes will fit into some specific NLP problems.
- Incremental learning from news events5
This paper formally defines the problem of incremental knowledge learning from similar news events on the same subject, where each event consists of a set of news articles reporting about it, and proposes a scalable two-phase framework to incrementally learn a topic hierarchy for a subject from Events on the subject as the events occur.
- 4
- Training-free Deep Concept Injection Enables Language Models for Video Question Answering4
This paper makes the first attempt to demonstrate that the PLM is able to perform zero-shot crossmodal tasks without any crossmodal pretraining, when the observed visual concepts are injected as both additional input text tokens and augmentation in the intermediate features within each feed-forward network for the PLM.
- User Simulator Assisted Open-ended Conversational Recommendation System4
A User Simulator (US) that communicates with the CRS using natural language based on given user preferences, enabling long-term reinforcement learning and a framework that uses reinforcement learning with two novel rewards, i.e., recommendation and conversation rewards, to train the C RS.
- Bridging the Gap between Native Text and Translated Text through Adversarial Learning: A Case Study on Cross-Lingual Event Extraction4
Results demonstrate that the proposed adversarial training framework can effectively incorporate machine translation to improve event extraction, while simply adding machine-translated data yields unstable performance due to the representational gap.
- 4
- 4
- Discovering Concept-Level Event Associations from a Text Stream4
This work investigates the importance and challenge of this task and proposes a novel solution by using event sequential patterns, which can discover important event associations implicitly expressed.
- 3
- Rethinking Task Sampling for Few-shot Vision-Language Transfer Learning3
This work aims to further improve the generalization of existing vision-language models on unseen tasks via a simple yet efficient model-agnostic Multi-task Fine-tuning strategy based on uniform task sampling and discovers that the effectiveness of MAML is highly dependent on the zero-shot performance of the pre-trained model, and this simple algorithm can out-perform MAML on more challenging datasets with low zero-shot performance.
- 3
- RPI BLENDER TAC-KBP2017 13 Languages EDL System.3
CRFs model is adopted, which is particularly good at jointly modeling sequential tagging decisions, on tp of the Bi-LSTM networks for name tagging and nominal and pronominal mentions are extracted.
- The Emergence of Communities and Their Leaders on Twitter Following an Extreme Event3
The results show that the government agencies had limited participation on Twitter during 2011 Japan Tsunami compared to an extensive participation during 2012 Hurricane Sandy, and the behavior of Twitter users during both events was consistent with the issuance of actionable information.
- Unsupervised Sentiment-Bearing Feature Selection for Document-Level Sentiment Classification3
An unsupervised sentiment-bearing feature selection method (USFS) is proposed, which incorporates sentiment discriminant analysis (SDA) into sentiment strength calculation (SSC).
- Comment-guided Learning: Bridging the Knowledge Gap between Expert Assessor and Feature Engineer3
This work presents a simple, low-cost but effective comment- guided learning approach to exploit some hidden privileged knowledge declaratively in an automatic assessor, which outperformed previous methods such as Recognizing Textual Entailment (RTE) based fact validation.
- 2
- 2
- 2
- 2
- 2
- Coreference by Appearance: Visually Grounded Event Coreference Resolution2
A careful analysis reveals that the performance gain of the multimodal model especially under unsupervised settings comes from better learning of visually salient events.
- Fine-grained Coordinated Cross-lingual Text Stream Alignment for Endless Language Knowledge Acquisition2
Experiments show this approach not only outperforms previous approaches on bilingual lexicon extraction from coordinated text streams but also can harvest high-quality alignments from large amounts of streaming data for endless language knowledge mining, which makes it promising to be a new paradigm for automatic language knowledge acquisition.
- Visualizing Group Dynamics based on Multiparty Meeting Understanding2
This paper uses multi-party meeting opinion mining based on bipartite graphs to extract opinions and calculate mutual influential factors, using the Lunar Survival Task as a study case.
- 2
- One Tense per Scene: Predicting Tense in Chinese Conversations2
A set of novel sentence-level features are proposed using rich linguistic resources and a new hypothesis of “One tense per scene” is proposed to incorporate scene-level (global) evidence to enhance the performance.
- Utility Evaluation of Cross-document Information Extraction2
The results show that, in comparison to using source documents only, the quality of summary reports assembled using IE results, especially from cross-document IE, was significantly better and user satisfaction was higher.
- Cross-document temporal and spatial person tracking system demonstration2
A new cross-document information extraction task is defined and a system which can extract, rank and track events in two dimensions: temporal and spatial is demonstrated.
- 1
- 1
- Prioritizing Image-Related Tokens Enhances Vision-Language Pre-Training1
PRIOR is a simple vision-language pre-training approach that addresses image-related tokens through differential weighting in the NTP loss, drawing from the importance sampling framework, and exhibits superior scaling properties, as demonstrated by significantly higher scaling coefficients.
- 1
- 1
- Image-Image Search for Comparable Corpora Construction1
This work uses an image-image search engine to obtain similar images, together with the cap-tions in source language and target language, and utilizes captions of similar imag-es to construct sentence-level bilingual corpora.
- 1
- 1
- Cross-media Cross-genre Information Ranking Multi-media Information Networks1
The facts from MiNets are exploited in a novel random walk-based algorithm to iteratively propagate ranking scores across multiple data modalities and the power of cross-media cross-genre inference is demonstrated.
- Cross-media Cross-genre Information Ranking based on Multi-media Information Networks1
The facts from MiNets are exploited in a novel random walk-based algorithm to iteratively propagate ranking scores across multiple data modalities and the power of cross-media cross-genre inference is demonstrated.
- –
- –
- –
- –
- –
- –
- –
- –
- –
- –
- Biography-Dependent Collaborative Entity Archiving for Slot Filling–
This work introduces topic modeling methodologies into entity biography profiling, so as to build a bridge between fuzzy and exact matching, and leverage the archiving method to enhance slot filling systems.
- Analysis and Refinement of Temporal Relation Aggregation–
Three methods to modify temporal relation information prior to aggregation to address the challenge of imprecise and inconsistent temporal information in relation instances in text are motivated and evaluated.
- –
- –
- The Application and Prospect Analysis of Cell Broadcast in the Weather Warning Information Release–
The principle and characteristics of Cell Broadcast are introduced, which has a lot of features compared with SMS, and its characteristics of strong regionality, high timeliness and covering a wide range have great prospects in the sudden and local weather warning information release.
- –
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 2026-10-14. 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.