https://scholars.lib.ntu.edu.tw/handle/123456789/24371
標題: | 資訊檢索查詢之自然語言處理 | 作者: | 陳光華 | 關鍵字: | 資訊檢索;自然語言處理;剖析技術;Information Retrieval;Natural Language Processing;Parsing Technology | 公開日期: | 1996 | 出版社: | 臺北市:國立臺灣大學圖書資訊學系 | 期: | 57 | 起(迄)頁: | 141-153 | 來源出版物: | 中國圖書館學會會報 | 摘要: | 資訊檢索系統的發展已行之有年,然而使用者查詢的方式仍是以關鍵詞為基礎,並且結合布林邏輯運算。對於使用者而言,最自然的查詢方式是使用自然語言查詢所需要的資訊。尤其當語音輸入技術達到實用階段時,自然語言查詢的需求,將益形重要。少數具有自然語言查詢方式的系統,使用的僅是樣式比對(Pattern Matching)的技術,無法知道各詞彙之間的關係,當然也就無法知道各詞彙扮演的角色。本文提出一種自然語言剖析的技術,可以快速並且穩定地剖析自然語言,分析各詞彙的關係與彼此的角色,而建構這種剖析技術所需的資源僅僅是具有詞類標記的語料庫。在分析大規模的實驗結果之後,根據Crossing準則,平均準確率為81%;根據PARSEVAL準則,平均的召回率與精確率皆為33% Many researchers have devoted to developing information retrieval systems for a long time. However, the form which users use to retrieve information is still keyword-based queries with Boolean operators. From user’s point of view, the most natural way to issue queries is using they mother tongue. That is, users are more likely to apply natural language queries to retrieve useful information. Some systems accept natural language queries, but the technique they use is simple pattern-matching. Such kind of technique does not discover the relationship among the keywords, and then does not understand the roles which these keywords play in natural language sentences. This paper proposes a new parsing technique, which can quickly and stably parse natural language sentences. Moreover, this parsing technique can find out the relationship among keywords and the roles of each grammatical constituents. The resource needed to construct a parser based on the proposed parsing technique is tagged text corpora. After a large-scale experiment, the recall and the precision are 33% based on the PARSEVAL criterion. The accuracy is 81% based on the Crossing criterion. |
URI: | http://ntur.lib.ntu.edu.tw//handle/246246/29199 | Rights: | 國立臺灣大學圖書資訊學系 |
顯示於: | 圖書資訊學系 |
檔案 | 描述 | 大小 | 格式 | |
---|---|---|---|---|
blac1996.pdf | 465.89 kB | Adobe PDF | 檢視/開啟 |
在 IR 系統中的文件,除了特別指名其著作權條款之外,均受到著作權保護,並且保留所有的權利。