線上個人化參考文獻管理系統

Abstract

參考文獻管理系統提供使用者儲存與管理文件的功能，文件儲存主要以記錄特徵(Feature)的方式，例如文件作者、出版年份、內容摘要等來代表某篇文件。而蒐集的文件可以依照使用者自定的類別來歸類，達到個人化管理的功能。 本論文提出的個人化文獻管理系統除了上述機制外，並且能夠做到個人化文獻推薦的功能，其概念為分析文件內容語意，亦即語意歧異解析(Word Sense Disambiguation, WSD)，將相同語意的文件分在同一群，若使用者有文件屬於某一群，則系統能將此群中的其他文件推薦給該使用者。在語意歧異解析方面，本論文提出一個新的語意歧異解析方法來決定文件內容中字詞的語意，這個方法是以建立字詞語彙鍵結(Lexical Chain)為基礎，並搭配WordNet得到詞彙概念(Concept)，建立的語彙鍵結依概念關係的強度而有不同的權重，利用這些權重來判斷該字詞可能的語意。實驗中，我們採用Semantic Concordance Corpus (SemCor)文件集來評估語意歧異解析方法的好壞。結果顯示，我們所提的方法有不錯的表現，平均來說，正確率可以達到65.26%，優於[Suarez99]所提出的59.11%。

A bibliographic management system facilitates the storage and management of the bibliographic information on documents that are important for user. Features of a bibliographic information like authors, published year, and abstracts are stored in the system to represent one document. A user can personalize his/her folders in terms of user-configurable categories. In addition to the above-mentioned functions, the personal bibliographic management system that we design also recommends documents relevant to an individual user. The idea is to analyze semantic meaning of documents. Documents with the same semantic meaning will be grouped into a cluster. If someone has documents belonging to one cluster, our system will recommend him other documents belonging to this cluster. Word Sense Disambiguation (WSD) is employed to discover the semantic meaning of a document. In this thesis, we propose a new method for Word Sense Disambiguation. The method is based on lexical chains and employs the taxonomy of WordNet. On the basis of the strength of the relationship between words, automatic disambiguating of word semantics can be accomplished by giving different weights. An evaluation of the method was done on Semantic Concordance Corpus (SemCor). The average percentage of correct resolutions achieved was 65.26%.