Using Question-Answer Pairs in Extractive Summarization of Email Conversations

Abstract

While sentence extraction as an approach to summarization has been shown to work in documents of certain genres, because of the conversational nature of email communication, sentence extraction may not result in a coherent summary. In this paper, we present our work on augmenting extractive summaries of threads of email conversations with automatically detected question-answer pairs. We compare various approaches to integrating questionanswer pairs in the extractive summaries, and show that their use improves the quality of email summaries.

A Joint Model of Conversational Discourse and Latent Topics on Microblogs

“…Zhang et al (2013) and Li et al (2015a) leverage dialogue acts to indicate summary-worthy messages. In the fields conversation summarization from other domains, e.g., meetings, forums, and emails, it is also popular to leverage the pre-detected discourse structure for summarization (Murray et al 2006;Wang and Cardie 2013;Bhatia, Biyani, and Mitra 2014;McKeown, Shrestha, and Rambow 2007;Bokaei, Sameti, and Liu 2016). Oya and Carenini (2014) and Qin, Wang, and Kim (2017) address discourse tagging together with salient content discovery on emails and meetings, and show the usefulness of their relations in summarization.…”

Automatic Summarization

“…As a consequence, questionanswer exchanges figure as one of the dominant uses of email conversations. These observations led to research on identification of question and answer pairs in email [140,187] and the integration of such pairs in extractive summaries of email [132]. McKeown et al experiment with different methods for integrating question and answer pairs into email summarization and show that all achieve significant improvement over an extractive approach alone, which may end up including either a question without its answer or an answer without its question.…”

Speech Act Modeling of Written Asynchronous Conversations with Task-Specific Embeddings and Conditional Structured Models

“…The two-part structures across posts like 'question-answer' and 'request-grant' are called adjacency pairs (Schegloff, 1968). Identification of speech acts is an important step towards deep conversation analysis in these media (Bangalore et al, 2006), and has been shown to be useful in many downstream applications including summarization (McKeown et al, 2007) and question answering (Hong and Davison, 2009).…”