Open AccessAutomatic scoring method of English composition based on language depth perception
Author(s) -
Tang Dan,
Yu Sun
Publication year - 2020
Publication title -
journal of physics. conference series
Language(s) - English
Resource type - Journals
SCImago Journal Rank - 0.21
H-Index - 85
eISSN - 1742-6596
pISSN - 1742-6588
DOI - 10.1088/1742-6596/1486/4/042045
Subject(s) - computer science , natural language processing , composition (language) , artificial intelligence , bengali , similarity (geometry) , context (archaeology) , vocabulary , perception , grammar , natural language , language identification , linguistics , psychology , paleontology , philosophy , neuroscience , image (mathematics) , biology
AutomatedEssaySystem (AES) refers to a system for evaluating and rating compositions by using techniques in the fields of statistics, natural language processing and linguistics. The selection of composition features is one of the key issues in the study of automatic composition scoring. This paper abstracts the features of language sense from the depth of language perception, which makes up for the shortcomings of the automatic scoring system of shallow features (such as vocabulary difficulty, grammar, etc.). The features of language sense are mainly the similarity of keywords extracted from the first and last paragraphs of a composition by rake algorithm, the similarity of texts calculated by cosine similarity to the context content of a composition and the similarity detection of composition language materials by automata. Finally, from the three aspects of correlation, error and overall accuracy, it analyzes whether the deep-seated language features improved the scoring ability of AES system. The experimental results show that the added depth perception features enhance the language ability of automatic scoring and are closer to manual scoring.