索引典之自動化建置與視覺化

Automatic Construction and Visualization of a Thesaurus

本論文描述自動化索引典建置與資訊視覺化的方法與結果。我們提出一個自動化方法，依據論文中的文字資訊，從無到有地建置索引典。這個方法利用統計訊息從論文的文字資訊中選取具有代表性的術語，偵測術語之間的概念關係，並利用資訊視覺化技術，將索引典資訊表示成直覺且資訊豐富的圖形。本論文以政大圖書與資訊學刊為對象，進行索引典建置與資訊視覺化的試驗，並且根據不同的用途，提出各種檢索與瀏覽的使用方式。試驗結果說明了這個方法的可行性與效果。

This paper describes automatic methods for thesaurus construction and information visualization and their testing results. A method is developed to construct a thesaurus from scratch, using only the textual materials in the examined domain of papers as the main information resource. All the terms relevant to the domain are selected and their mutual conceptual relationships are detected, based on the statistical processing of the input texts. The selected terms and their conceptual relationships are applied to construct a thesaurus dedicated to the examined domain. Another method for visualizing information in the constructed thesaurus is also proposed by generating a set of graphs, in which the mapped positions of related terms are displayed in juxtaposition. These graphs are useful in showing the knowledge structure of the examined domain and very suitable for the applications of retrieval and browsing. In the study, on the basis of our proposed methods, we have performed tests on the textual materials selected from papers published by the periodical Bulletin of Library and Information Science, NCCU. Various usages of the graphs in the constructed thesaurus have been experimented in terms of different sorts of applications. The final results confirm the feasibility as well as the effectiveness of the two methods.