Modifying Online Text Clustering Algorithm using Inverted Index based Operation

摘要

This article concerns the experiments where two strategies of encoding documents are compared with each other for text clustering. We used two test beds for these experiments: NewsPage.com and 20NewsGroups. In order to compute an operation on string vectors, inverted indices of words are built from a corpus as the basis for doing that. In these experiments, single pass algorithm is adopted as the approaches to text clustering. The goal of these experiments is to observe whether modified versions of the two approaches are comparable to their traditional versions, when we use the inverted indices as the basis for performing the operation on string vectors, instead of a restricted sized similarity matrix.