基於大眾意見分析所建置之擬人化剪報系統

A Humanlike Clipping System Based on Public Opinion Analysis

網路使用者通常在搜尋引擎輸入相關查詢以便進行搜尋，而搜尋引擎通常根據所輸入的查詢回傳相對應的網頁列表，這個列表的內容不外乎包含查詢或相關的內容。然而根據文獻，使用者輸入的查詢平均長度為2.3個字。在這麼短的查詢裡，要尋找使用者的真正的搜尋需求是一件困難的工作，特別是針對語意不清的查詢。在本論文之中，我們提出一個搜尋引擎剪報系統，此系統模擬人的行為，針對網頁內容擷錄使用者真正的需求內容，使之成為一個全新的剪報頁面。本系統與傳統新聞剪報的概念是相同的，亦即我們只會擷錄相關文章並將所有相關文章整理成使用者所需結果。根據實驗結果，我們的系統效能明顯的優於現行分群搜尋引擎。

Internet users often use search engines to search for content based on queries submitted to the search engine. The search engines generally return a list of web pages that include at least one of the submitted queries or are somehow related to the submitted queries. However, according to the literature, the average length of Web queries is about 2.3 words. In such short queries, it is a difficult task to find users’ search needs, especially for ambiguous queries. In this paper, we present a search engine clipping system, which simulates humanlike behavior to clip the user information need into a page showing to the user at the Web page. The perspective of our system is the same as newspaper clipping that only clips the relative articles and converge them into a user need document. According to the results of experiments, we conclude that our system is significantly better than current clustering search engines.