Studying on extraction method for semantic relationship among topic words

A concept lattice based method to extract the semantic relationship among topic words proposed. By counting the topic words frequency in the text, topic words are classified as low-frequency, middle-frequency, high-frequency topic words. It´s the base to forms the semantic space of topic words of semantic network. Using concept lattice, semantic relationships in the topic level are extracted. The relationships include similar relationship and imply relationship. A good result has been obtained in the vertical search engine object for clothing area. The extract method can also be directly applied to related search and query expansion of information retrieval etc.