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
- Donghong JiSuggested from co-authorship
Is this you? Claim this profile to confirm or dismiss it.
Works75 from public data
- 110
- Encoder-Decoder Based Unified Semantic Role Labeling with Label-Aware Syntax89
A novel unified SRL framework based on the sequence-to-sequence architecture with double enhancement in both the encoder and decoder sides is investigated, and in-depth analysis demonstrates that the model can effectively capture the correlations between syntactic and SRL structures.
- Rethinking Boundaries: End-To-End Recognition of Discontinuous Mentions with Pointer Networks84
An innovative model for discontinuous NER based on pointer networks, where the pointer simultaneously decides whether a token at each decoding frame constitutes an entity mention and where the next constituent token is, which is the kernel source to NER.
- Revisiting Disentanglement and Fusion on Modality and Context in Conversational Multimodal Emotion Recognition76
This work argues that both the feature multimodality and conversational contextualization should be properly modeled simultaneously during the feature disentanglement and fusion steps of MM-ERC, and proposes a Contribution-aware Fusion Mechanism and a Context Refusion Mechanism for multimodal and context integration, respectively.
- 51
- OneEE: A One-Stage Framework for Fast Overlapping and Nested Event Extraction49
The model is equipped with an adaptive event fusion module to generate event-aware representations and a distance-aware predictor to integrate relative distance information for word-word relation recognition, which are empirically demonstrated to be effective mechanisms.
- End-to-end Semantic Role Labeling with Neural Transition-based Model45
This paper presents the first work of transition-based neural models for end-to-end SRL, and suggests high-order compositions to extract non-local features, which can enhance the proposed transition model further.
- 44
- Fine-grained depression analysis based on Chinese micro-blog reviews43
A neural model for joint depression degree and cause prediction is designed and compared with several widely-used neural models such as TextCNN, BiLSTM and BERT, which shows that neural models achieve promising results, but there is still room for improvement.
- Latent Target-Opinion as Prior for Document-Level Sentiment Classification: A Variational Approach from Fine-Grained Perspective37
This paper proposes to first learn the latent target-opinion distribution behind the documents, and then leverage such fine-grained prior knowledge into the classification process, using a hierarchical model based on Transformer encoders.
- PanoSent: A Panoptic Sextuple Extraction Benchmark for Multimodal Conversational Aspect-based Sentiment Analysis36
A multimodal conversational ABSA is introduced, where two novel subtasks are proposed: 1) Panoptic Sentiment Sextuple Extraction and 2) Sentiment Flipping Analysis, detecting the dynamic sentiment transformation throughout the conversation with the causal reasons.
- 34
- Multimodal Emotion-Cause Pair Extraction with Holistic Interaction and Label Constraint32
This work proposes a novel model that captures holistic interaction and label constraint (HiLo) features for the MECPE task and innovatively leverages emotion transition features as pivotal cues to enhance causal inference within conversations.
- Constructing Holistic Spatio-Temporal Scene Graph for Video Semantic Role Labeling31
A novel holistic spatio-temporal scene graph (namely HostSG) representation based on the existing dynamic scene graph structures, which well model both the fine-grained spatial semantics and temporal dynamics of videos for VidSRL is explored.
- Actively Learn from LLMs with Uncertainty Propagation for Generalized Category Discovery27
This work innovatively employs uncertainty propagation to select data samples from high-uncertainty regions, which are then labeled using LLMs through a comparison-based prompting scheme, which enhances accuracy in identifying new categories and introduces a soft feedback propagation mechanism to minimize the spread of inaccurate feedback.
- Harnessing Holistic Discourse Features and Triadic Interaction for Sentiment Quadruple Extraction in Dialogues27
A novel framework that not only capitalizes on comprehensive discourse feature modeling, but also captures the intrinsic interaction for optimal quadruple extraction is introduced, and a novel triadic scorer is proposed, strengthening weak token relations within a quadruple, thereby enhancing the cohesion of the quadruple extraction.
- NUS-Emo at SemEval-2024 Task 3: Instruction-Tuning LLM for Multimodal Emotion-Cause Analysis in Conversations25
The architecture of the system developed for participation in Task 3 of SemEval-2024: Multimodal Emotion-Cause Analysis in Conversations is described, which designs an approach of emotion-cause-aware instruction-tuning for LLMs, to enhance the perception of the emotions with their corresponding causal rationales.
- Refining and Synthesis: A Simple yet Effective Data Augmentation Framework for Cross-Domain Aspect-based Sentiment Analysis17
A simple yet effective framework to address shortages in ABSA data augmentation, called Refining and Synthesis Data Augmentation (RSDA), which refines generated labeled data using a natu-ral language inference (NLI) filter to control data quality.
- TKDP: Threefold Knowledge-Enriched Deep Prompt Tuning for Few-Shot Named Entity Recognition17
This work proposes incorporating the deep prompt tuning framework with threefold knowledge (namely TKDP), including the internal 1) context knowledge and the external 2) label knowledge & 3) sememe knowledge, and presents a knowledge-enriched model that significantly outperforms 9 strong-performing baseline systems in 5-/10-/20-shot settings.
- Electronic word-of-mouth effects on studio performance leveraging attention-based model17
This study proves that deep learning with the attention mechanism has the best accuracy in both eWOM and stock price movement and presents both a hierarchical two-layer attention network and hierarchical convoluted attention network (HCAN), which quantify the importance of crucial eWom features in capturing valuable information from audience members’ reviews.
- 14
- Unveiling the Cognitive Compass: Theory-of-Mind-Guided Multimodal Emotion Reasoning12
This work introduces HitEmotion, a ToM-grounded hierarchical benchmark that diagnoses capability breakpoints across increasing levels of cognitive depth, and proposes a ToM-guided reasoning chain that tracks mental states and calibrates cross-modal evidence to achieve faithful emotional reasoning.
- MMLSCU: A Dataset for Multi-modal Multi-domain Live Streaming Comment Understanding12
This work introduces the MMLSCU dataset, containing 50,129 intention-annotated comments across multiple modalities (text, images, vi-deos, audio) from eight streaming domains, and implements an end-to-end model to sequentially perform the following tasks: viewer comment intent detection, intent cause mining, viewer comment explanation and streamer policy suggestion.
- FormFactory: An Interactive Benchmarking Suite for Multimodal Form-Filling Agents11
A comprehensive evaluation of state-of-the-art MLLMs is conducted and it is observed that no model surpasses 5% accuracy, underscoring the inherent difficulty of the task and revealing significant limitations in current models' visual layout reasoning and field-value alignment abilities.
- 11
- Revisiting Conversation Discourse for Dialogue Disentanglement10
A structure-aware framework is developed to integrate the rich structural features for better modeling the conversational semantic context and an easy-first decoding algorithm is devised, which performs utterance pairing under the easy-to-hard manner with a global context, breaking the constraint of traditional sequential decoding order.
- 10
- 9
- 9
- SpeechEE: A Novel Benchmark for Speech Event Extraction8
This paper pioneers a SpeechEE, defined as detecting the event predicates and arguments from a given audio speech, and tailor an E2E SpeechEE system based on the encoder-decoder architecture, where a novel Shrinking Unit module and a retrieval-aided decoding mechanism are devised.
- Enhancing Cross-Document Event Coreference Resolution by Discourse Structure and Semantic Information8
This work proposes the construction of document-level Rhetorical Structure Theory (RST) trees and cross-document Lexical Chains to model the structural and semantic information of documents and develops a large-scale Chinese cross-document event coreference dataset to fill this gap.
- 7
- What Factors Influence LLMs’ Judgments? A Case Study on Question Answering7
This study reveals that LLMs’ judging abilities are susceptible to the influence of these four factors, and analyzing from the newly proposed dimension of question difficulty is highly necessary, providing valuable insights into optimizing LLMs’ performance as judges.
- Outer synchronization for two discrete-time complex dynamic networks via the synthesized coupling connections7
A new outer synchronization scheme for the two networks is proposed by synthesizing the coupling connections between nodes, including the inner and outer connections, such that the outer synchronization error of two networks converges asymptotically to the origin.
- Adversarial shared-private model for cross-domain clinical text entailment recognition7
This work proposes a domain adaptation framework for the cross-domain clinical RTE, which achieves significantly enhanced performances against baseline domain adaptation methods, on the few-shot and zero-shot transferring settings.
- Neural transition model for aspect-based sentiment triplet extraction with triplet memory7
A neural transition model for end-to-end ASTE, which model the triplet prediction as a graph structure, based on which a transition system with neural design is implemented, and proposes a triplet memory mechanism to fully leverage the underlying interactions from the previously recognized triplets relevant to the current parse.
- M$^{3}$D: A Multimodal, Multilingual and Multitask Dataset for Grounded Document-level Information Extraction6
A Missing Modality Construction Module (MMCM) is designed to alleviate the issues caused by missing modalities and set a reasonable standard for subsequent research on multimodal IE, and an innovative hierarchical multimodal IE model is proposed.
- A joint triple extraction method by entity role attribute recognition6
A new model for entity role attribute recognition based on triple holistic fusion features, which can extract triples (including overlapping triples) under a limited number of relationships, and its prediction process is simple and easy explain is proposed.
- Event Extraction in Large Language Model5
It is argued that EE should be viewed as a system component that provides a cognitive scaffold for LLM centered solutions, aiming to evolve EE from static extraction into a structurally reliable, agent ready perception and memory layer for open world systems.
- ECQED: Emotion-Cause Quadruple Extraction in Dialogs5
This paper presents a new task, Emotion-Cause Quadruple Extraction in Dialogs (ECQED), which requires detecting emotion-cause utterance pairs and emotion and cause types, and presents an ECQED model based on a structural and semantic heterogeneous graph as well as a parallel grid tagging scheme.
- 5
- 5
- Nominal Compound Chain Extraction: A New Task for Semantic-Enriched Lexical Chain5
This paper introduces a novel task, Nominal Compound Chain Extraction (NCCE), extracting and clustering all the nominal compounds that share identical semantic topics and model the task as a two-stage prediction, which is handled via a proposed joint framework.
- CMNER: A Chinese Multimodal NER Dataset based on Social Media4
A Chinese Multimodal NER dataset is compiled utilizing data sourced from Weibo, China's largest social media platform, and the results substantiate the hypothesis that Chinese and English multimodal NER data can mutually enhance the performance of the NER model.
- 3
- 3
- STPar: A Structure-Aware Triaffine Parser for Screenplay Character Coreference Resolution3
A Structure-aware Triaffine Parser (STPar) is introduced for the MovieCoref task, which combines discourse and syntactic structures in the feature encoding process, enabling comprehensive analysis of ternary relationships and complex interactions.
- 3
- Double tracking control for the complex dynamic network with an unavailable link state3
To realise the double tracking control (DT-Control) of CDN, the presented DT-Control scheme includes the synthesis of controller for NDS and the coupling term in LDS, which can ensure that the two subsystems track the given reference targets.
- 3
- ROLL-SENSITIVE ONLINE CAMERA ORIENTATION DETERMINATION ON THE STRUCTURED ROAD2
A novel online camera orientation determination method based on a longitudinal vanishing point without the zero- roll hypothesis that can reconstruct the ground space accurately compared with previous methods with zero-roll hypothesis is proposed.
- 2
- 1
- Synergizing Understanding and Generation with Interleaved Analyzing-Drafting Thinking1
The interleaved Analyzing-Drafting problem-solving loop (AD-Loop), a new think paradigm that dynamically alternates between analytic and drafting operations, is introduced, highlighting AD-Loop as a principled and broadly applicable strategy for synergizing comprehension and creation.
- Tracking control for discrete-time complex dynamic network via link dynamics1
This paper investigates the tracking control problem for discrete-time complex dynamic network (DCDN) via the link dynamics through the difference vector equation instead of matrix difference equation, which provides the more intuitive geometric features and convenient mathematical method due to the flexible mathematical operational properties of vector.
- 1
- 1
- Chinese Event Discourse Deixis Resolution: Design of the Dataset and Model1
A deep neural network model is proposed for EDDR, which formulates the task into two subtasks, namely event anaphor recognition and event antecedent recognition, and outperforms all the baselines and achieves about 53%, 44%, 53%, and 63% F1s using standard anaphora resolution metrics.
- AN IMPROVED ENERGY SEGMENTATION BASED STEREO MATCHING ALGORITHM1
An improved energy segmentation based stereo matching algorithm that incorporates gradient information of each segmented region into the energy function according to the facts that gradients inside the same segmentation should be closer than those from different segmentation.
- –
- –
- Global Commander and Local Operative: A Dual-Agent Framework for Scene Navigation–
By disentangling global reasoning from local action, DACo alleviates cognitive overload and improves long-horizon stability, and provides a principled and extensible paradigm for robust long-horizon navigation.
- Orthogonal Spatial-temporal Distributional Transfer for 4D Generation–
A spatial-temporal-disentangled 4D (STD-4D) Diffusion model, which synthesizes 4D-aware videos through disentangled spatial and temporal latents and a spatial-temporal-aware HexPlane to integrate the transferred spatiotemporal features for better 4D deformation and 4D Gaussian feature modeling.
- CogMAEC'25: The 1st Workshop on Cognition-oriented Multimodal Affective and Empathetic Computing–
The 1st Workshop on Cognition-oriented Multimodal Affective and Empathetic Computing was held at ACM Multimedia 2025, exploring how machines can not only recognize emotions but also explain their causes and simulate human-like affective reasoning.
- The ACM Multimedia 2025 Grand Challenge of Multimodal Conversational Aspect-based Sentiment Analysis–
The Multimodal Conversational Aspect-based Sentiment Analysis (MCABSA) challenge is organized, the task, dataset, and evaluation settings are introduced, the systems of the top teams are summarized, and the findings of the participants are discussed.
- Polyp segmentation method via cross-level correlation attention–
A novel deep learning architecture for automated polyp segmentation in colonoscopy images, aiming to improve early colorectal cancer detection, that integrates multi-scale feature fusion with attention mechanisms to address challenges in polyp boundary delineation and size variation.
- –
- –
- An Effective Multimodal Rumor Detection Model via Image Semantic Enhancement and Hierarchical Fusion–
- The ACM Multimedia 2024 Viual Spatial Description Grand Challenge–
The VSD challenge provides two benchmark datasets for three subtasks, i.e., visual spatial relationship classification, single spatial description generation, and open-ended spatial description generation.
- XFashion: Character Animation Generation via Facial-enhanced and Granularly Controlling–
The XFashion, a diffusion-based method to generate vivid fashion videos, where the user could control the generation with portrait, emotion and poses, is proposed, which effectively achieves stronger character animation generation results than state-of-the-art baselines.
- –
- –
- –
- –
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.