关键词: 作文自动评分, 语言深度感知, 语感, 流畅度, AC自动机

Abstract:

The paper designs an automatic scoring algorithm for Chinese composition from the perspective of language deep perception. The computational modeling on the abstract features of language intuition mimics the evaluation criteria of human for natural language. It makes up for the insufficiency of mechanical statistical natural language processing technology in early automatic scoring of compositions. Aho-Corasiek automata is built in to quickly analyze the elements supporting language intuition, i.e. the indexes of personal language corpus of the writer. In evaluating the language ability on a sentence level, such as sentence fluency, technologies like text segmentation and structure recognition are embedded in the system, together with the similarity comparison between the composition to be treated and the standard modes of discourse, and therefore a comprehensive evaluating algorithm is created to author’s language competency in the composition automatic scoring system. Experimental results show that the algorithm improves rationality of automatic language ability scoring and gets closer to fit the expertized scoring samples.

Key words: composition automatic scoring, language deep perception, language sense, fluency, Aho-Corasiek automata