| Summary: | 碩士 === 國立成功大學 === 資訊工程學系 === 102 === Conventional search engines usually consider a search query corresponding only to a simple task. Nevertheless, due to the explosive growth of web usage in recent years, more and more queries are driven by complex tasks consisting of multiple sub-tasks. In order to accomplish a complex task, users usually have to issue a series of queries. For example, the complex task “travel to Beijing” may involve several sub-task goals, including “book flights,” “reserve hotel,” and “survey map”. Understanding complex tasks can allow a search engine to predict a variety of sub-task goals to be efficiently accomplished simultaneously.
In this work, we propose a topic-event-based complex task model (TECTM) to deal with the above problem. Our TECTM contains three main stages. The first is task-coherence clustering which groups queries into the same complex task. The second is sub-task goal identification which identifies some sub-task goals for a complex task based on queries from the same task. The third is task name generation which utilizes the identified sub-task goals to generate the complex task name. For improving the performance of TECTM, we exploit multiple web resources including query log, clicked pages, community question answering (CQA), search engine results page (SERP), and microblogs. In addition, we develop an application, complex-task-based search engine (CTSE) which provides integrated search results for sub-task goals based on TECTM. Experimental results show that our TECTM is effective in generating complex task names with corresponding identified sub-task goals for a complex task. Furthermore, CTSE also provides more suitable ranking of search results to help users accomplish their complex tasks with less effort.
|
|---|
