Measuring Word Similarity Based on Pattern Vector Space Model

It is an important work in natural language processing to measure the semantic similarity between two words. This paper proposes a new method for computing word similarity based on pattern vector space model. Analogous to traditional vector space model in information retrieval, a word is represented as a vector in this paper. Each dimension corresponds to a contextual pattern. The similarity between two words is calculated by the cosine of the angle between their vectors. In the experiment, the proposed model is compared to other two baseline models on a Chinese version Miller-Charles data set. It shows that this method achieves competitive results.