Survey on User Intent Determination in Conversational Search and Question Answering Systems

Abstract

Conversational Search can be defined as an approach to findinginformation in multi-turn interactions. With the significant rise in the usage of conversational assistants (CAs), the need to easily understand user intent prediction and answer questions in conversational search systems such as Chatbots has been increasing.As the key aspect of Conversational search, user intent prediction can help providerelevantanswers to users. Typical intent predictions involve predicting user questions before users ask or predicting answers from limited or little information in questions from users. Researchers have since adopted different approaches to identify or predict user intent, such as feature engineering based on human queries to better understand the informational needs of the user. In many cases, the inter-relationship between questions is ignored and the existing methods often focus on optimizing immediate reward and overlook long-term rewards of capturing the entire decision-making process. More so, the complexity of conversations in Conversational Search often makes it challenging to determine user intent. This paper discusses the different approaches to User Intent determination in conversational search and Question Answering systems and how the systems can understand the needs of the user.