Modeling Transitions of Focal Entities for Conversational Knowledge Base Question Answering

Proceedings of the 59th Annual Meeting of the Association for Computational Linguistics and the 11th International Joint Conference on Natural Language Processing, pages 3288–3297 August 1–6, 2021. ©2021 Association for Computational Linguistics 3288 Modeling Transitions of Focal Entities for Conversational Knowledge Base Question Answering Yunshi Lan Singapore Management University yslan@smu.edu.sg Jing Jiang Singapore Management University jingjiang@smu.edu.sg Abstract Conversational KBQA is about answering a se- quence of questions related to a KB. Follow-up questions in conversational KBQA often have missing information referring to entities from the conversation history. In this paper, we pro- pose to model these implied entities, which we refer to as the focal entities of the conversation. We propose a novel graph-based model to cap- ture the transitions of focal entities and apply a graph neural network to derive a probability distribution of focal entities for each question, which is then combined with a standard KBQA module to perform answer ranking. Our exper- iments on two datasets demonstrate the effec- tiveness of our proposed method. 1 Introduction Recently, conversational Knowledge Base Ques- tion Answering (KBQA) has started to attract peo- ple’s attention (Saha et al., 2018; Christmann et al., 2019; Guo et al., 2018; Shen et al., 2019). Mo- tivated by real-world conversational applications, particularly personal assistants such as Apple Siri and Amazon Alexa, the task aims to answer ques- tions over KBs in a conversational manner. Figure 1 shows an example of conversational KBQA. As we can see, the conversation can be roughly divided into two parts: Q1, Q2 and Q3 revolve around the book “The Great Gatsby,” while Q4 and Q5 revolve around its author, “F. Scott Fitzgerald”. Although these entities are not explic- itly mentioned in the questions, they are implied by the conversation history, and they are critical for answering the questions. For example, Q3, when taken out of context, cannot be answered because Q3 itself does not state the title of the book being discussed. But since Q3 is a follow-up question of Q1, humans can easily infer that the book of interest here is “The Great Gatsby” and can hence answer the question correctly. We therefore can Conversation Based KBQA 𝐐𝟏: Which actor voiced the Unicorn in The last Unicorn? 𝐑𝟏: Mia Farrow 𝐐𝟐: And Alan Arkin was behind …? 𝐑𝟐: Schmendrick 𝐐𝟑: So, who sang for the film? R𝟑: America 𝐐𝟒: Genre of this band’s music? R𝟒: Folk rock, Soft rock 𝐐𝟓: By the way, who was the director? R𝟓: Jules Bass 𝐐𝟏: What novel has the character named Nick Carraway? 𝐑𝟏: The Great Gatsby Q2: Where is Jay Gatsby born? The Great Gatsby R2: North Dakota 𝐐𝟑: What is the name of the author? The Great Gatsby 𝐑𝟑: F. Scott Fitzgerald 𝐐𝟒: What’s his first novel? F. Scott Fitzgerald R𝟒: This Side of Paradise 𝐐𝟓: Who was his child? F. Scott Fitzgerald R𝟓: Frances Scott Fitzgerald Figure 1: An example conversation in conversational KBQA. The entities shown in blue are what we call the focal entities, which are implicit but …

Content truncated to 500 words