An Approach for Word Categorization Based on Semantic Similarity Measure Obtained from Search Engines

Word categorization based on semantic similarity is a problem need to be solved for several natural language applications. A similarity measure is need for word categorization. In this study it is proposed that the semantic similarity between two Turkish words is in direct proportion to the number of pages which the words are located next to each other. Google and Yahoo search engines were used to find the number of pages. In the first attempt to verify the proposal, the experiments were done with small datasets. The average success ratio is 87%.