Evaluation of Two Systems on Multi-class Multi-label Document Classification

Abstract

In the world of text document classification, the most general case is that in which a document can be classified into more than one category, the multi-label problem. This paper investigates the performance of two document classification systems applied to the task of multi-class multi-label document classification. Both systems consider the pattern of co-occurrences in documents of multiple categories. One system is based on a novel sequential data representation combined with a kNN classifier designed to make use of sequence information. The other is based on the “Latent Semantic Indexing” analysis combined with the traditional kNN classifier. The experimental results show that the first system performs better than the second on multi-labeled documents, while the second performs better on uni-labeled documents. Performance therefore depends on the dataset applied and the objective of the application.