Search Clarification Selection via Query-Intent-Clarification Graph Attention

Chang Gao, Wai Lam

Abstract: Proactively asking clarifications in response to search queries is a useful technique for revealing the intent of the query. Search clarification is important for both web and conversational search. This paper focuses on the clarification selection task. Inspired by the fact that a good clarification should clarify the query’s different intents, we propose a graph attention-based clarification selection model that can exploit the relations among a given query, its intents, and its clarifications via constructing a query-intent-clarification attention graph. The comparison with competitive baselines on large-scale search clarification data demonstrates the effectiveness of our model.