So far, named entity recognition (NER) has been involved with three major types, including flat, overlapped (aka. nested), and discontinuous NER, which have mostly studied individually. Recently, a growing interest built for unified tackling the above jobs concurrently one single model. Current best-performing methods mainly include span-based sequence-to-sequence models, where unfortunately former merely focus on boundary identification latter may suffer from exposure bias. In this work, we...

It has been a hot research topic to enable machines understand human emotions in multimodal contexts under dialogue scenarios, which is tasked with emotion analysis conversation (MM-ERC). MM-ERC received consistent attention recent years, where diverse range of methods proposed for securing better task performance. Most existing works treat as standard classification problem and perform feature disentanglement fusion maximizing utility. Yet after revisiting the characteristic MM-ERC, we...

Unified opinion role labeling (ORL) aims to detect all possible structures of 'opinion-holder-target' in one shot, given a text. The existing transition-based unified method, unfortunately, is subject longer terms and fails solve the term overlap issue. Current top performance has been achieved by employing span-based graph model, which however still suffers from both high model complexity insufficient interaction among opinions roles. In this work, we investigate novel solution revisiting...

With the continuous emergence of various social media platforms frequently used in daily life, multimodal meme understanding (MMU) task has been garnering increasing attention. MMU aims to explore and comprehend meanings memes from perspectives by performing tasks such as metaphor recognition, sentiment analysis, intention detection, offensiveness detection. Despite making progress, limitations persist due loss fine-grained metaphorical visual clue neglect text-image weak correlation. To...

So far, discontinuous named entity recognition (NER) has received increasing research attention and many related methods have surged such as hypergraph-based methods, span-based sequence-to-sequence (Seq2Seq) etc. However, these more or less suffer from some problems decoding ambiguity efficiency, which limit their performance. Recently, grid-tagging benefit the flexible design of tagging systems model architectures, shown superiority to adapt for various information extraction tasks. In...

Few-shot named entity recognition (NER) exploits limited annotated instances to identify mentions. Effectively transferring the internal or external resources thus becomes key few-shot NER. While existing prompt tuning methods have shown remarkable performances, they still fail make full use of knowledge. In this work, we investigate integration rich knowledge for stronger We propose incorporating deep framework with threefold (namely <italic xmlns:mml="http://www.w3.org/1998/Math/MathML"...

Despite significant advancements in multi-label text classification, the ability of existing models to generalize novel and seldom-encountered complex concepts, which are compositions elementary ones, remains underexplored. This research addresses this gap. By creating unique data splits across three benchmarks, we assess compositional generalization classification models. Our results show that these often fail concepts encountered infrequently during training, leading inferior performance...

Unified opinion role labeling (ORL) aims to detect all possible structures of 'opinion-holder-target' in one shot, given a text. The existing transition-based unified method, unfortunately, is subject longer terms and fails solve the term overlap issue. Current top performance has been achieved by employing span-based graph model, which however still suffers from both high model complexity insufficient interaction among opinions roles. In this work, we investigate novel solution revisiting...

With the proliferation of dialogic data across Internet, Dialogue Commonsense Multi-choice Question Answering (DC-MCQ) task has emerged as a response to challenge comprehending user queries and intentions. Although prevailing methodologies exhibit effectiveness in addressing single-choice questions, they encounter difficulties handling multi-choice due heightened intricacy informational density. In this paper, inspired by human cognitive process progressively excluding options, we propose...

The multi-document summarizer using genetic algorithm-based sentence extraction (MSBGA) regards summarization process as an optimization problem where the optimal summary is chosen among a set of summaries formed by conjunction original articles sentences. To solve NP hard problem, MSBGA adopts algorithm, which can choose on global aspect. evaluation function employs four features according to criteria good summary: satisfied length, high coverage, informativeness and low redundancy. improve...

Few-shot named entity recognition (NER) exploits limited annotated instances to identify mentions. Effectively transferring the internal or external resources thus becomes key few-shot NER. While existing prompt tuning methods have shown remarkable performances, they still fail make full use of knowledge. In this work, we investigate integration rich knowledge for stronger We propose incorporating deep framework with threefold (namely TKDP), including 1) context and 2) label & 3) sememe TKDP...

Multimodal Named Entity Recognition (MNER) is a pivotal task designed to extract named entities from text with the support of pertinent images. Nonetheless, notable paucity data for Chinese MNER has considerably impeded progress this natural language processing within domain. Consequently, in study, we compile NER dataset (CMNER) utilizing sourced Weibo, China's largest social media platform. Our encompasses 5,000 Weibo posts paired 18,326 corresponding The are classified into four distinct...

Financial sentiment analysis is a fine-grained task that needs to predict the value toward given target entity. Recently, dependency-based graph neural networks have been introduced for target-based analysis. However, financial with implicit expression more challenging than explicit analysis, requiring deep understanding of complex association between clue in context and In previous work related most methods focused on learning simple word-to-word relations contextual words entity based...

The dependency syntactic structure is widely used in event extraction. However, the reflecting features essentially different from that reflects semantic features, leading to performance degradation. In this article, we propose use Event Trigger Structure for Extraction (ETSEE), which can compensate inconsistency between two structures. First, leverage ACE2005 dataset as case study, and annotate three kinds of ETSs, is, “light verb + trigger”, “preposition structures” “tense trigger”. Then...