Using keywords/phrases in automatically generating hypertext links: An exploratory study.

摘要

The Traditional manual method of generating hypertext links is impractical for large and dynamic collections of documents. However, automatic link generation is a challenging problem in that it involves advanced techniques, and only limited success has been achieved. This study explores the applicability of using keywords and key phrases, and the semantic relationship between the contexts in which they occur, to automatically generate hyperlinks for large collection of reference materials. The types of hyperlinks that can be generated through keywords/phrases and their usefulness are major objects of this research.; As background for developing an automatic system, a content analysis of a representative Web site was performed to explore the semantic relationships expressed by links that are proximate to keywords/phrases in a human-generated set of hypertext documents. The results indicate that the keywords/phrases seem to be closely associated with these hyperlinks.; Techniques from the information retrieval field were used to automatically generate hyperlinks based on occurrences of keywords/phrases. Two approaches were developed, differing in whether the keywords/phrases were manually generated by a human author, or mechanically extracted by an automatic process. In the first prototype, based on keywords/phrases produced by the document authors, different types of intra-document and inter-document links were generated by using Salton et al.'s Vector Space Model. An analysis of the pattern of link types generated indicates this automatic approach may be more comprehensive than manual linking because it produces a wider variety of link types. A small user study shows that this prototype system may be sufficiently useful to help people with information searching and browsing tasks.; A second prototype was developed to explore the possibility of generating hyperlinks through automatically extracted keywords/phrases. This system used the Oracle Text gist/theme generation tool to extract keywords/phrases. Links were generated using the best linking method identified in the first prototype. Analysis of these links indicates that this approach seems to be practical and promising.; Through the development and evaluation of these prototypes for generating links, this study develops a new approach to automatically generating hyperlinks using keyword/phrase extraction, off-the-shelf database software, and proximity measuring techniques.