Summary: | 碩士 === 國立成功大學 === 資訊工程學系碩博士班 === 101 === Natural language search is to use human language questions as query to search and extract suitable webpages for users. Compare with short query, natural language query users can directly submit their query intents. For example, the question, “Which restaurants in Tainan have delicious hot pot?”. According to our observation, conventional search engines can’t efficiently process this queries and the search result is too messy. Therefore users need to spend lots of time on browsing and filter those noise information in the result pages . Rose et al. proposed a conceptual framework for classifying user’s goals. Such that, search engines can associate user goals with queries and exploit the goal information. In our paper, we focus in his proposed list-informational search goal.
We proposed a website-base entity set expansion method to expand our entities. We try to recommend evidence pages of each entity based on the intent behind users by analyzing question structures and answer structures. The question structure can be divided into three parts, they are question focus, question context and question opinion pair. We used the algorithm of question analysis to identify question features. As to the answer structure, it can be divided into entity homepage, context evidence page and opinion summary. We combine the relationship between the question structure and the answer structure with the real comments on the Internet to construct Opinion-based Entity Ranking Model (OBERM) to improve List-Informational search.
Experiment result shows that our proposed method OBERM outperforms Google Search. And it shows OBERM really can enhance performance in List-informational search.
|