Two Methods of Evaluation of Semantic Similarity of Nouns Based on Their Modifier Sets
- Cite this paper as:
- Bolshakov I.A., Gelbukh A. (2007) Two Methods of Evaluation of Semantic Similarity of Nouns Based on Their Modifier Sets. In: Kedad Z., Lammari N., Métais E., Meziane F., Rezgui Y. (eds) Natural Language Processing and Information Systems. NLDB 2007. Lecture Notes in Computer Science, vol 4592. Springer, Berlin, Heidelberg
Two methods of evaluation of semantic similarity/dissimilarity of English nouns are proposed based on their modifier sets taken from Oxford Collocation Dictionary for Student of English. The first method measures similarity by the portion of modifiers commonly applicable to both nouns under evaluation. The second method measures dissimilarity by the change of the mean value of cohesion between a noun and modifiers, its own or those of the contrasted noun. Cohesion between words is measured by Stable Connection Index (SCI) based of raw Web statistics for occurrences and co-occurrences of words. It is shown that the two proposed measures are approximately in inverse monotonic dependency, while the Web evaluations confer a higher resolution.
Unable to display preview. Download preview PDF.