Research on Automatic Task Evaluation System of English-Assisted Teaching Based on Text Abstract Extraction

With the continuous improvement and maturity of keyword extraction technology, its application scope continues to expand and has now penetrated into multiple fields. This study innovatively introduces the concept of word similarity and optimizes the TF-IDF algorithm. By comprehensively considering word similarity and inverse frequency, this study has developed a novel TF-IDF optimization algorithm that significantly improves the accuracy and relevance of keyword extraction, thereby more accurately reflecting students’ learning status and needs. On this basis, this study further optimized the construction of the sentence matrix and proposed an improved initial point selection strategy to address the problem of traditional K-means algorithm being prone to local optima, thereby improving the global optimality and stability of clustering results. In the stage of text summarization extraction, this study fully considered the potential impact of Chinese writing habits and keywords on sentence importance, and customized and optimized the TextRank algorithm. With the help of machine learning technology, optimization algorithms can gain a deeper understanding of text structure and semantic associations, thereby more accurately identifying and extracting topic sentences that represent the main ideas of the article, and generating higher-quality text summaries.