Context-Aware In-Page Search

1 Context-Aware In-Page Search ��� Yu-Hao Lin, ��� Yu-Lan Liu, ��� Tzu-Xi Yen, ��� Jason S. Chang ������������ Department of Computer Science National Tsing Hua University { catonmars.lin, ikulan12, joseph.yen, jason.jschang}@gmail.com Abstract In this paper we introduce a method for searching appropriate articles from knowledge bases (e.g. Wikipedia) for a given query and its context. In our approach, this problem is transformed into a multi-class classification of candidate articles. The method involves automatically augmenting smaller knowledge bases using larger ones and learning to choose adequate articles based on hyperlink similarity between article and context. At run-time, keyphrases in given context are extracted and the sense ambiguity of query term is resolved by computing similarity of keyphrases between context and candidate articles. Evaluation shows that the method significantly outperforms the strong baseline of assigning most frequent articles to the query terms. Our method effectively determines adequate articles for given query-context pairs, suggesting the possibility of using our methods in context-aware search engines. Keywords: entity linking, word sense disambiguation, Wikipedia, support vector machine, search engine 1 Introduction Today we surf the Internet through search engines most of the time. With the explosive growth of web pages, the accuracy and relevancy of search results have become ever more important. Traditional search engines accept keywords, and return a page full of possible relevant results. Then users can click one of the results to visit the sites they are interested in. We call this type of search “keyword-search”. Today, almost all search engines are keyword-based. However, various classes of results mixed in the search results. For example, when a user query the search engine with the keyword “apple”, the search results comprise of two major class, “Apple Inc.”, the computer company, and “apple”, a kind of fruit. With only one keyword, even state-of-the-art keyword-based search engines could not distinguish between different search intents. Unlike keyword search, context-aware search assume each query is Proceedings of the Twenty-Fourth Conference on Computational Linguistics and Speech Processing (ROCLING 2012) 292 2 associated with a context. Figure 1. An example of context-aware search Figure 2. The mention “John McCarthy” and its context. In this paper, we present a prototypical system, In-Page Search, that automatically extract context information and use them to disambiguate ambiguous queries. Users could select the terms they are interested in, and then with a click of the mouse, the In-Page Search system shows a pop-up window with the most relevant results for the given context.(See Figure 1.) In-Page Search is similar to the “entity-linking problem”, which has long been an active research topic in IR and Database community. Entity-linking problem could be informally described as follows: given a knowledge base, in which every entry is an entity and its associated information. Given a mention and the context with the mention, determine the correct entity that the given mention really links to. For example, Figure 2 shows the mention “John McCarthy” and it’s context, in a knowledge base, there are more than 10 entities which may be linked …

Content truncated to 500 words