Similarity Comparison between Chinese Sentences

Identifying and extracting similar sentences from the example base is an essential procedure in machine-aided human translation (MAHT) and example-based machine translation (EBMT) system. A method for measuring the similarity between a pair of Chinese sentences has been proposed in this paper. Obviating from the common thesaurus-based strategy, a new principle based on word grammatical features is presented thereafter. Moreover, a dynamic mechanism is built into the method to increase the robustness and flexibility of the matching algorithm. From observations on the initial results, we've found that the expected most similar sentence in the example base for an input is listed among the first four candidate sentences in most cases, which is very helpful for both MAHT and EBMT.