Automatic Recognition of Parallel Sentence Based on Sentences-Interaction CNN and Its Application
Guang‐Hui Liu, Lijun Fu, Bo Yu, Li Cui · 2022 7th International Conference on Computer and Communication Systems (ICCCS) · 2022
Automated essay evaluation is to make the computer simulating human evaluate students’ essay. The final grade of one essay is highly related to its literacy. The use of rhetorical devices can increase the literacy of one essay, and parallelism is one of the common rhetoric. Therefore, the identification of parallelism can contribute to automated essay evaluation. However, the research of automatic recognition of parallel sentence is so rare so far. In this paper, we design a deep learning method based on sentences interactive-convolutional neural network to recognize parallelism in the light of word co-occurrence, similar syntactic structure and tone of expression. Firstly, one or more interactive matrices whose elements represent the similarities in some ways between every token in two sentences are constructed. Then the convolutional neural network is employed to capture local features. The experimental results show that our feature selection is effective and the recognition effect is comparable with similar studies.