Structural Characterization for Dialogue Disentanglement

Ma, Xinbei; Zhang, Zhuosheng; Zhang, Hai

doi:10.18653/v1/2022.acl-long.23

Cited by 12 publications

(16 citation statements)

References 26 publications

Supporting

Mentioning

Contrasting

Order By: Relevance

“…Each pointing operation is modeled as a multinomial distribution over the set of previous utterances. Ma et al (2022) propose characteristic features of speaker property and reference dependency for dialogue structure.…”

Section: Address Whommentioning

confidence: 99%

“…Reading List The presenters have survey papers for comprehensive references (Gu et al, 2022b;Zhang and Zhao, 2021). The following papers are also recommended: Ouchi and Tsuboi (2016); ; Kummerfeld et al (2019); Gu et al (2021); Ma et al (2022); Gu et al (2022a); Li and Zhao (2023); Gu et al (2023). Breadth While dozens of relevant papers over the tutorial are provided, we plan to cover around 10-15 research papers in detail.…”

Section: Prerequisitementioning

confidence: 99%

See 1 more Smart Citation

Learning WHO Saying WHAT to WHOM in Multi-Party Conversations

Gu,

Zhang,

Ling

2023

Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacifi

View full text Add to dashboard Cite

Multi-party conversations (MPC) are a more practical and challenging scenario involving more than two interlocutors. This research topic has drawn significant attention from both academia and industry, and it is nowadays counted as one of the most promising research areas in the field of dialogue systems. In general, MPC algorithms aim at addressing the issues of Who saying What to Whom, specifically, who speaks, say what, and address whom. The complicated interactions between interlocutors, between utterances, and between interlocutors and utterances develop many variant tasks of MPC worth investigation. In this tutorial, we present a comprehensive survey of recent advances in MPC. In particular, we summarize recent advances on the research of MPC modeling which is categorized by Who saying What to Whom. Finally, we highlight the challenges which are not yet well addressed in MPC and present future research directions.

show abstract

Section: Address Whommentioning

confidence: 99%

Section: Prerequisitementioning

confidence: 99%

Learning WHO Saying WHAT to WHOM in Multi-Party Conversations

Gu,

Zhang,

Ling

2023

Proceedings of the 13th International Joint Conference on Natural Language Processing and the 3rd Conference of the Asia-Pacifi

View full text Add to dashboard Cite

show abstract

“…denoted the decomposable attention model (Parikh et al, 2016), ESIM denoted the enhanced sequential inference model (Chen et al, 2017), and MHT denoted masked hierarchical Transformer . Numbers in bold denoted the best performance without comparing with Ptr-Net (Yu and Joty, 2020) and structural characterization (Ma et al, 2022), which are the latest proposed methods for dialogue disentanglement and are included for reference.…”

Section: Comparison Baselinesmentioning

confidence: 99%

Conversation- and Tree-Structure Losses for Dialogue Disentanglement

Li¹,

Gu²,

Ling³

et al. 2022

Proceedings of the Second DialDoc Workshop on Document-Grounded Dialogue and Conversational Question Answering

View full text Add to dashboard Cite

When multiple conversations occur simultaneously, a listener must decide which conversation each utterance is part of in order to interpret and respond to it appropriately. This task is referred as dialogue disentanglement. A significant drawback of previous studies on disentanglement lies in that they only focus on pair-wise relationships between utterances while neglecting the conversation structure which is important for conversation structure modeling. In this paper, we propose a hierarchical model, named Dialogue BERT (DIALBERT), which integrates the local and global semantics in the context range by using BERT to encode each message-pair and using BiLSTM to aggregate the chronological context information into the output of BERT. In order to integrate the conversation structure information into the model, two types of loss of conversation-structure loss and tree-structure loss are designed. In this way, our model can implicitly learn and leverage the conversation structures without being restricted to the lack of explicit access to such structures during the inference stage. Experimental results on two large datasets show that our method outperforms previous methods by substantial margins, achieving great performance on dialogue disentanglement.

show abstract

“…DGM (Ouyang et al, 2021) constructs two discourse graphs and uses R-GCNs to fully capture interactions among the elements. Ma et al (2022) employs R-GCNs to enhance reference dependencies for dialogue disentanglement. In contrast with previous works, our work proposes a sentence-level graph that is finely designed to mine the relationships between multiple elements in a sentence, extract rich structural semantics and facilitate information flow over the graph as well.…”

Section: Graph Modeling For Language Understandingmentioning

confidence: 99%

EventBERT: Incorporating Event-Based Semantics for Natural Language Understanding

Zou

Zhang

Zhao

2022

Lecture Notes in Computer Science

Self Cite

View full text Add to dashboard Cite

Natural language understanding tasks require a comprehensive understanding of natural language and further reasoning about it, on the basis of holistic information at different levels to gain comprehensive knowledge. In recent years, pre-trained language models (PrLMs) have shown impressive performance in natural language understanding. However, they rely mainly on extracting context-sensitive statistical patterns without explicitly modeling linguistic information, such as semantic relationships entailed in natural language. In this work, we propose EventBERT, an event-based semantic representation model that takes BERT as the backbone and refines with event-based structural semantics in terms of graph convolution networks. EventBERT benefits simultaneously from rich event-based structures embodied in the graph and contextual semantics learned in pre-trained model BERT. Experimental results on the GLUE benchmark show that the proposed model consistently outperforms the baseline model.

show abstract

Structural Characterization for Dialogue Disentanglement

Cited by 12 publications

References 26 publications

Learning WHO Saying WHAT to WHOM in Multi-Party Conversations

Learning WHO Saying WHAT to WHOM in Multi-Party Conversations

Conversation- and Tree-Structure Losses for Dialogue Disentanglement

EventBERT: Incorporating Event-Based Semantics for Natural Language Understanding

Contact Info

Product

Resources

About