Context-Aware Intent Identification in Email Conversations

Proceedings of the 42nd Annual International ACM SIGIR Conference on Research and Development in Information Retrieval (SIGIR 2019) |

Published by ACM

Email continues to be one of the most important means of online communication. People spend a significant amount of time sending, reading, searching and responding to email in order to manage tasks, exchange information, etc. In this paper, we study intent identification in workplace email. We use a large scale publicly available email dataset to characterize intents in enterprise email and propose methods for improving intent identification in email conversations. Previous work has studied identifying intent in email conversations and discussed its implications on designing email clients and intelligent agents that support users with taking actions triggered by emails (e.g. creating a to-do item, setting up a meeting, etc.). To accomplish this, previous work focused on classifying email messages into broad topical category or detecting sentences that contain action items or follow certain speech acts. In this work, we focus on sentence-level intent identification and study how incorporating more context (such as the full message body and other metadata) could improve the performance of the intent identification models. We experiment with several models for leveraging context including both classical machine learning and deep learning approaches. We show that modeling the interaction between sentence and context can significantly improve performance.