A Deep Sequential Model for Discourse Parsing on Multi-Party Dialogues

Abstract

Discourse structures are beneficial for various NLP tasks such as dialogue understanding, question answering, sentiment analysis, and so on. This paper presents a deep sequential model for parsing discourse dependency structures of multi-party dialogues. The proposed model aims to construct a discourse dependency tree by predicting dependency relations and constructing the discourse structure jointly and alternately. It makes a sequential scan of the Elementary Discourse Units (EDUs) 1 in a dialogue. For each EDU, the model decides to which previous EDU the current one should link and what the corresponding relation type is. The predicted link and relation type are then used to build the discourse structure incrementally with a structured encoder. During link prediction and relation classification, the model utilizes not only local information that represents the concerned EDUs, but also global information that encodes the EDU sequence and the discourse structure that is already built at the current step. Experiments show that the proposed model outperforms all the state-of-the-art baselines.

Multi-turn Response Selection using Dialogue Dependency Relations

“…Asher and Lascarides (2005) proposed the SDRT theory with the STAC Corpus (Asher et al, 2016) which made a great contribution to the discourse parsing on multiparty dialogues. Shi and Huang (2019) proposed a sequential neural network and achieved the stateof-the-art results on this dataset. Another similar task is dialogue disentanglement (Du et al, 2017).…”

“…Motivated by the above, we aim to analyze the discourse structures in dialogue history at first. We utilize the discourse dependency parsing model for dialogues proposed by Shi and Huang (2019). It is a deep sequential model that achieves the state-ofthe-art performance on the STAC corpus.…”

“…Due to the scarcity of annotated dialogues across domains under SDRT theory, the predicted dependency relations had never been used for down-streaming tasks, such as response selection and dialogue summarization. In this paper, we take advantage of both the simplicity of the "reply-to" relation and the sequential parsing methods (Shi and Huang, 2019) to do dialogue dependency parsing. Developing general discourse parsing with relations types and take relation types into consideration may be future work.…”

Molweni: A Challenge Multiparty Dialogues-based Machine Reading Comprehension Dataset with Discourse Structure

“…In most cases, every node (utterance) in the discourse dependency graph only has one parent node. Shi and Huang (2019) proposed the deep sequential model for discourse parsing on multiparty chat dialogs which adopted an iterative algorithm to learn the structured representation and highlight the speaker information in the dialog. The model jointly and alternately learns the dependency structure and discourse relations.…”

“…Research into multiparty dialog has recently grown considerably, partially due to the growing ubiquity of dialog agents. Multiparty dialog applications such as discourse parsing and meeting summarization are now mainstream research (Shi and Huang, 2019;Hu et al, 2019;Sun et al, 2019;Perret et al, 2016;Afantenos et al, 2015). Such applications must consider the more complex, graphical nature of discourse structure: coherence between adjacent utterances is not a given, unlike standard prose where sequential guarantees hold.…”