Artificial intelligence technology has been widely used in many areas. One of the applicationsinthe education industry is automated scoring system for subjective questions.The text similarity calculation is a major difficulty in subjective question scoring. The word similarity calculation based onCiLin has achieved good results, but the long text will lead to the performance degradation of the traditional word similarity calculation method. In this paper, the extended named entity recognition method is used to extract some keywords from the candidate answers of the subjective questions. the improved CiLinword semantic similarity calculation method is used to calculate the similarity between the candidate keywords and the target keywords. The proposed method can effectively improve the word matching efficiency. On the basis of the original CiLinword semantic similarity algorithm, the performance is improved and the calculation time is effectively shortened.