| Summary: | 碩士 === 國立臺灣大學 === 資訊管理研究所 === 91 === Searching and browsing are the two primary activities of a document retrieval process.
Searching is convenient for locating documents that are relevant to a specific query
without much human guidance. Browsing is necessary when the user’s requirement cannot
be precisely stated. If the collection of documents is adequately organized, the user
might be able to find specific information in a reasonable amount of time by browsing the
document space. However, both retrieval methods have their disadvantages when used
alone. The goal of this research is to design and implement an integrated tool for research
papers retrieval that offers both the benefits of searching and browsing. It is hoped that
with such a tool the user will be able to quickly find the interesting papers by shuffing
between searching and browsing activities.
Our tool, SABer (Search-And-Browser), preprocesses and clusters the paper collection
into a hierarchy by a clustering method based on minimum spanning trees and novel
similarity measures. Given a set of keywords as input, SABer will locate a cluster that
contains the relevant publications. SABer will then guide the user to the targeted clusters.
The user may glance at any publication in the cluster if the publication meets her
requirement. The user may repeat searching and browsing activities until she finds the
needed publications. We believe that SABer will realize the possibility of fast access to
research papers and greatly alleviate a researchers’ burden of literature search.
|
|---|
