Automatic Scoring of Chinese Language Composition Combined with the Perception of Expression Recognition

Artificial intelligence is an important tool for current language education, however, the traditional automatic scoring model for composition has poor feature recognition performance and poor scoring result accuracy. Therefore, the study proposes an automatic scoring model for Chinese language compositions combined with utterance recognition perception. The experimental results show that compared with the teacher’s judgement, the judgement errors of the proposed model on context relevance indexes are -0.02 and -0.03 respectively, while those of the traditional bag-of-words model are 0.14 and -0.23 respectively, and the judgement errors of the model on thematic relevance and the number of speech defects are 0.1 and 1 respectively, while those of the traditional bag-ofwords model on the two indexes are -0.14 and 6 respectively. In addition, compared with the teacher’s final scoring of the composition, the absolute error of the model’s final scoring ranges from $1-7$, while the error range of the traditional bag-of-words model is $3-11$. In summary, the proposed automatic scoring model for Chinese language composition combined with utterance recognition perception is able to accurately determine the scoring of the composition, and to promote the application and development of artificial intelligence in the field of Chinese language learning.