Topic Extraction And Contrast Analysis Of Document Quality Evaluation

With the improvement of information technology,The deliverables of document types in the company is increasing,which will consume large amounts of resources to review.Computer automatic evaluation of deliverables will be an important issue in delivery management.The quality evaluation of the deliverables involves understanding the contents of the deliverables,and it is difficult to evaluate the quality based on the comparison of content and standards.Based on this,the method of subject extraction and subjective analysis is put forward,which can be used to evaluate the quality of deliverables and play a supporting role for manual evaluation.In order to improve the topic extraction effect,TextRank_CW algorithm introduce the influence factors of the influence,frequency,location and part of speech on the basis of the TextRank algorithm,Using the G1 weighting method gives different Weighting.The subject contrast analysis method calculates the similarity degree of the keywords according to the upper and lower structure tree of "HowNet ",and then obtains the Subject similarity.If the similarity is within the specified threshold,it can be judged that the topic of the document delivery discussion focuses on the standard theme,and then the fuzzy quality evaluation of the document class deliverables is realized.Experiments show that the TextRank_CW algorithm can obtain better topic extraction effect,and it is superior to the TextRank algorithm in precision and recall rate.And in the actual project,the subject contrast analysis method can evaluate the deliverables fuzzyly.