The Research Of Chinese Sentence Similarity Based On Layered

In computer science and technology disciplines, more and more research field involving the processing of text information. These fields keep text information processing as the central task, so text information processing quality restricts their further development. Text information is mainly expressed by natural language, processing of text information is reflected on the natural language processing. The similarity of sentences in Chinese text is extremely important and it’s a very basic work in the process of our daily Chinese text information processing. It was a hot and difficult point people have long been studied.During the research of Chinese sentence similarity computation, the similarity computation that we have studied focus on the characteristics in a certain respect of the sentence, this paper proposes a method of Chinese sentence similarity calculation based on the layered. The method is calculated from three levels:surface, middle and deep. The sentence surface information such as the sentence length, the distance of same key word etc; The sentence middle information such as the Sentence structure information; The sentence deep information such as the information of sentence emotional tendencies. The method reference a sentence similarity computing model based on the multi-features combination, using the weight to describe each layer of the sentence, then gets result of sentence similarity. The main research achievements of this paper are as follows:1)、In this paper,’we divide the sentences into three types: surface, middle and deep. We believe that the overall message of the sentence is composed of the sentence surface information, the sentence middle information and the sentence deep information. The surface information of sentence is determined primarily by the words information in sentence; The middle information of sentence is determined by sentence structure; The deep information of sentence is determined by sentence emotional tendencies. The surface layer and middle layer decided the theme of sentence. The sentence deep layer decided the tendency of emotion.2)、Aiming at the Common sentence similarity computation that we have studied is focus on the characteristics in a certain respect of the sentence, this article uses a layered structure to divide sentences feature, and fully integrated the advantages of this sentence feature similarity calculation, learn from each other. This structure is in favor of future expansion. Due to the low coupling characteristics of the hierarchy, we can choose the appropriate similarity calculation method to adapt to different application environments.3)、This article will add emotional tendencies into the sentence deep information. It makes the two sentence similarity is more consistent with human usage of language and the semantic understanding habits from the perspective of human thought and cognition.