Navigating Semantically Annotated Queries for Task Understanding

Abstract

As search systems gradually turn into intelligent personal assistants, users increasingly resort to a search engine to accomplish a complex task, such as planning a trip, renting an apartment, or investing in stocks. A key challenge for the search engine is to understand the users underlying task given a sample query like ``tickets to panama'', ``studios in los angeles'', or ``spotify stocks'', and to recommend other queries to help the user complete the task. In this dissertation, we propose three strategies for task understanding by navigating a semantically annotated query log using a mixture of explicit and latent representations of entire queries and of query parts. We thoroughly evaluate our proposed strategies in the context of the TREC 2016 Tasks track and via crowdsourcing. Our results demonstrate the effectiveness of the proposed strategies in terms of diversity and novelty, as well as their complementarity, with significant improvements compared to multiple state-of-the-art query suggestion baselines adapted for this task. Moreover, we show that our proposal is particularly effective for long-tail queries as well as for hard queries, which encompass a large number of subtasks.