現今過量資訊常讓使用者無法適時獲得需要資訊。以研究論文而言，當研究新鮮人欲踏入某一新領域時，須花費許多時間並過濾大量研究資訊後，方能找到有興趣的研究方向。現行搜尋系統係透過內控的Rank機制進行運算，呈現方式亦均為條列式（View by list）的資訊展現，大量資訊閱讀常花費許多時間。為使搜尋結果能讓使用者更易於接受，本論文針對電腦科學論文開發一套搜尋系統，整合現行熱門且具權威性的學術網站資訊，配合資訊擷取技術，及視覺化與OLAP互動分析概念，提出一套新穎的論文搜尋系統。由於論文研究方法、技術及應用領域、環境的異同之處，常常是該篇論文有所貢獻或獨到見解之所在，為讓使用者能掌握研究論文的精髓與精要，本篇論文針對論文標題，透過機器學習方式建立學習模組進行該篇論文可能的運用層面/環境（Application/Environment, AE）及使用方法/技術（Method/Technique, MT）的擷取，再藉由與使用者互動方式探尋其有興趣或欲研讀的論文集。Information overload is an increasing problem which often causes inefficiency. Take junior graduates as an example, searching good research papers can take a lot of time. Existing retrieval systems often display search results based on some internal ranking score. In this paper, we propose the integration of information extraction with OLAP (online analytical processing) and visualization technique to achieve interactive search experience. We define two extraction targets: application environment (AE) and involved method/technique (MT) as the basic elements that define the research problem and the contribution of a paper. Both manually constructed rules and machine learning based approaches are applied to compared the performances. The experimental results on 500 titles show that AE has an F-measure of 0.73 while MT has an F-measure of 0.6. With the extracted application environment and involved method/technology, users can then use visualization and OLAP tool to get a better view of the search results.
