A Text Representation and Retrieval Method Based on Concept Algebra

This paper introduces the concept algebra (CA) theory as a basis for the conceptual representation and the derivation of text processing to realize a semantic based retrieval system. We also take advantage of Hownet to create the concept attributes space for concept algebra. With the help of LTP, we get the key words and their dependent relations of every sentence to build the CA concept representation of the content with a five-tuple. Concepts make it possible to express both the keyword itself and the semantic relation with its context. According to the demands of text retrieval, some CA operations are optimized to calculate the relations and similarity between concepts. Besides, a text retrieval system framework which processes information based on the concept relations at a concept level is also proposed to verify the advantages of our method.