Grade Clustering and Seriation of Words Based on Their Co- Occurrences

We present the use of grade correspondence analysis (GCA) in text mining. A sample of words extracted from 20 newsgroups has been linearly arranged according to concordance between their co-occurrence distributions. Words’ co-occurrence matrix, obtained using HAL (Hyperspace Analogue to Language) system and normalized to deemphasize too frequent terms, has been reordered by the GCA algorithm, implemented in the GradeStat program. The aim of this reordering was to approach TP2 regularity of dependence, measured using Kendall’s tau. Deviations from regularity have been used to order words into series, cluster them and visualize using overrepresentation maps. Seriation provided a contextual scale between computerrelated terms and politicsand religion-related terms. Words appearing in various contexts occupy average positions on this scale. Word ordering based on similarity between their occurrence patterns can be used in thesauri building and query extension in information retrieval applications.