Summarization Search

Abstract

Mobile networks in developing regions have been plagued by three basic problems: low bandwidth, high latency and high costs. Due to these factors, mobile web users in these regions are known to have a highly intermittent and relatively non-interactive user experience. In this paper, we consider the case of mobile search and propose Summarization search, a new search paradigm for mobile users where the basic goal is to provide a single summarized and relatively complete search response for every user search query. Summarization search is specifically designed as a relatively non-interactive, one-time quick shot search experience for mobile users where a mobile user can issue focused queries for specific information search needs. Our summarization search engine is designed as a meta-search service on top of conventional search services to generate a single summarized and condensed search response based on extracting the most useful information from the underlying search result pages. Our system yielded a high accuracy (82-92%) for three classes of focused queries: task oriented queries, category specific queries and focused issue focused queries for specific information search needs. queries sampled from the AOL log. We also performed a user study with 400 queries and 30 users. The user study results show that in a scale of 1 (best) to 5(worst), 45% of the cases, users rated the result as '1' and 40% of the cases were rated as '2', with an average rating of 2.09 across all users for all queries. Another task in the user study revealed that for around 55% of queries, users found their information in the summarized result. This demonstrates the effectiveness of the summarization search interface, where users' information requirement is fulfilled with just one operation of submitting the search query.