Word Sense Disambiguation for Exploiting Hierarchical Thesauri in Text Classification

Abstract

Abstract. The introduction of hierarchical thesauri (HT) that contain significant semantic information, has led researchers to investigate their potential for improving performance of the text classification task, extending the traditional "bag of words" representation, incorporating syntactic and semantic relationships among words. In this paper we address this problem by proposing a Word Sense Disambiguation (WSD) approach based on the intuition that word proximity in the document implies proximity also in the HT graph. We argue that the high precision exhibited by our WSD algorithm in various humanly-disambiguated benchmark datasets, is appropriate for the classification task. Moreover, we define a semantic kernel, based on the general concept of GVSM kernels, that captures the semantic relations contained in the hierarchical thesaurus. Finally, we conduct experiments using various corpora achieving a systematic improvement in classification accuracy using the SVM algorithm, especially when the training set is small.

A Knowledge-Based Semantic Kernel for Text Classification

“…Semantic-aware kernels have been proposed by Mavroeidis et al [4] who propose a generalized vector space model with WordNet senses and their hypernyms to improve text classification performance. Bloehdorn at al.…”

“…This latter definition of SR for a pair of terms is the definition of the Omiotis measure that we are using in our case. 4 4 Omiotis-based Semantic Kernel…”

“…The application of Word Sense Disambiguation (WSD) techniques [2] during document preprocessing can be helpful; however, this is usually computationally expensive, and the performance of the unsupervised techniques is poor while use of supervised techniques requires large amounts of hand-annotated text documents. The use of external semantic knowledge provided by word thesauri or ontologies to adjust or "smooth" the BOW representation has shown much promise [3,4]. However, the embedding of semantic information is usually computationally expensive.…”

Convolutional Sentence Kernel from Word Embeddings for Short Text Categorization

“…Siolas and d'Alché Buc (2000) pioneered the idea of semantic kernels for text categorization, capitalizing on WordNet (Miller, 1995) to propose continuous word kernels based on the inverse of the path lengths in the tree rather than the common delta word kernel used so far, i. e. exact matching between unigrams. Bloehdorn et al (2006) extended it later to other tree-based similarity measures from WordNet while Mavroeidis et al (2005) exploited its hierarchical structure to define a Generalized Vector Space Model kernel.…”

Using wavelet analysis for text categorization in digital libraries: a first experiment with Strathprints

“…Whereas we regard document categorization by SVM [30,50,49,4,38,6] a particular implementation of machine learning, an increasingly successful solution to the classical problem of automatic classification, we also envisage information representation by vectors, a standard point of departure for TC by SVM, a limitation of the above attempt, and combine the former with semantic content representation in Hilbert space instead of Euclidean space. In this new approach, instead of term and document vectors, term and document functions are used to represent the semantic content of digital objects, with the advantage that functions, having more parameters than vectors, can host more semantic content in a comprehensive description than vector space based methods.…”