Online Conversation Disentanglement with Pointer Networks

Abstract

Huge amounts of textual conversations occur online every day, where multiple conversations take place concurrently. Interleaved conversations lead to difficulties in not only following the ongoing discussions but also extracting relevant information from simultaneous messages. Conversation disentanglement aims to separate intermingled messages into detached conversations. However, existing disentanglement methods rely mostly on handcrafted features that are dataset specific, which hinders generalization and adaptability. In this work, we propose an end-to-end online framework for conversation disentanglement that avoids time-consuming domain-specific feature engineering. We design a novel way to embed the whole utterance that comprises timestamp, speaker, and message text, and propose a custom attention mechanism that models disentanglement as a pointing problem while effectively capturing inter-utterance interactions in an end-to-end fashion. We also introduce a joint-learning objective to better capture contextual information. Our experiments on the Ubuntu IRC dataset show that our method achieves state-of-the-art performance in both link and conversation prediction tasks.

Disentangling Online Chats with DAG-structured LSTMs

“…This is particularly interesting especially when we compare with models employ-ing contextual embeddings like . For the cluster scores, the best model is the pointer network model of Yu and Joty (2020a), which is anyway within less than 0.5% of the best contextual model, and within 2.5% of our model. The difference mainly arises from a difference in recall and corresponds to an absolute difference of less than 10 true positive clusters on the test set.…”

Unsupervised Conversation Disentanglement through Co-Training

“…The two-step methods (Elsner and Charniak, 2008, 2010, 2011Chen et al, 2017;Jiang et al, 2018;Kummerfeld et al, 2019) firstly retrieve the relations between two messages, e.g., "reply-to" relations (Guo et al, 2018;, and then adopt a clustering algorithm to construct individual sessions. The end-to-end models (Tan et al, 2019;Yu and Joty, 2020), instead, perform the disentanglement operation in an end-to-end manner, where the context information of detached sessions will be exploited to classify a message to a session. End-to-end models tend to achieve better performance than two-step models, but both often need large annotated data to get fully trained , which is expensive to obtain and thus encourages the demand on unsupervised algorithms.…”

“…In the two-step methods Charniak, 2011, 2008;Jiang et al, 2018), a model first retrieves the "local" relations between two messages by utilizing either feature engineering approaches or deep learning methods, and then a clustering algorithm is employed to divide an entire conversation into separate sessions based on the message pair relations. In contrast, end-to-end methods (Tan et al, 2019;Yu and Joty, 2020) capture the "global" information contained in the context of detached sessions and calculate the matching degree between a session and a message in an end-to-end manner.…”

Zero-Shot Dialogue Disentanglement by Self-Supervised Entangled Response Selection

“…All previous work (Shen et al, 2006;Elsner and Charniak, 2008;Wang and Oard, 2009;Elsner and Charniak, 2011;Jiang et al, 2018;Kummerfeld et al, 2018;Yu and Joty, 2020) treat the task as a sequence of multiplechoice problems. Each of them consists of a sliding window of n utterances.…”

“…It would be ideal if we could design an algorithm to automatically organize an entangled conversation into its constituent threads. This is referred to as the task of dialogue disentanglement (Shen et al, 2006;Elsner and Charniak, 2008;Wang and Oard, 2009;Elsner and Charniak, 2011;Jiang et al, 2018;Kummerfeld et al, 2018;Yu and Joty, 2020 Training data for the dialogue disentanglement task is difficult to acquire due to the need for manual annotation. Typically, the data is annotated in the reply-to links format, i.e.…”