Event-centric Query Suggestion for Online News
Query suggestion refers to the task of suggesting relevant and related queries to a search engine user to help in query formulation process and to expedite information retrieval with minimum amount of effort. It is highly useful in situations where the search requirements are not well understood and hence it has been widely adopted by search engines to guide users' search activity. For news websites, user queries have a time sensitive nature inherent in them. When some new event happens, there is a sudden burst in queries related to that event and such queries are sustained over a period of time before fading away with that event. In addition to this temporal aspect of search queries fired at news websites, they have an addition distinct quality, i.e., they are intended to get event related information majority of the times. Existing work on generating query suggestions involves analyzing query logs to suggest queries which are relevant and related to the search intent of the user. But in case of news websites, when there is a sudden burst in information related to a particular event, there are not enough search queries fired by other users which leads to lack of click data, and hence giving query suggestions related to some old event or even some irrelevant suggestions altogether. Another problem with query logs in the context of online news is that, they mostly contain queries related to popular events and hence fail to capture less popular events or events which got overshadowed by some other more sensational event. We propose a novel approach to generate event-centric query suggestions using metadata of news articles published by news media. We compared our proposed framework with existing state of the art query suggestion mechanisms provided by Google News, Bing News, Google Search and Bing Search on various parameters.