Automated learning of decision rules for text categorization

Abstract

We describe the results of extensive experiments using optimized rule-based induction methods on large document collections. The goal of these methods is to discover automatically classification patterns that can be used for general document categorization or personalized filtering of free text. Previous reports indicate that human-engineered rule-based systems, requiring many man-years of developmental efforts, have been successfully built to “read” documents and assign topics to them. We show that machine-generated decision rules appear comparable to human performance, while using the identical rule-based representation. In comparison with other machine-learning techniques, results on a key benchmark from the Reuters collection show a large gain in performance, from a previously reported 67% recall/precision breakeven point to 80.5%. In the context of a very high-dimensional feature space, several methodological alternatives are examined, including universal versus local dictionaries, and binary versus frequency-related features.

The use of bigrams to enhance text categorization

“…In a number of experiments (Lewis, 1992a(Lewis, , 1992b(Lewis, , 1992cApté et al, 1994), it was found that the use of phrases 2 actually caused text categorization performance to degrade. Despite of these discouraging results, investigations of using phrases have been actively pursued (Mladeni) and Grobelnik 1998;Fürnkranz, 1998;Schütze et al, 1995;Schapire et al, 1998).…”

Learning-Free Text Categorization

“…A non exhaustive list of machines learning approaches to text categorization includes naive Bayes [5]), k-nearest neighbors [4], SVM [6], boosting [7], and rule-learning algorithms [8]. However, most of these studies apply text classification to a small set of classes (usually a few hundreds, as in the paradigmatic Reuters' collection [9]).…”

“…For each triplet provided in table 1, the first letter refers to the term frequency, the second refers to the inverse document frequency and the third letter refers to a normalization factor. 8 Available on the first author's homepage.…”

User Behavior Analysis of the Open-Ended Document Classification System

“…In the 1980's a common approach to document classification was rule-based, which involved a human in the construction of classifier. Though such a method provides accurate rules and has the additional benefit of being human understandable, the construction of such rules requires significant human input and the human needs some knowledge concerning the details of rule construction as well as domain knowledge, which become a bottleneck of this approach [1]. As an alternative, the machine learning (ML) approach has become the dominant one since 1990's.…”