Research On The Method Of Measuring The Originality Of Innovation In Scientific Papers

At present,the evaluation of papers mostly uses the methods of bibliometrics indicators and peer review.However,bibliometrics focus on "quantity" rather than "quality",peer review is affected by subjective factors,and fairness is also controversial.Therefore,how to objectively measure the originality degree of papers according to the content of innovation has become a difficult problem in the field of scientific evaluation.It can not only provide reference for scientific evaluation,but also has important value in the application of scientific research project,sci-tech search,information retrieval.The research takes innovation of papers as the object of study and the goal is to provide the degree of originality.Using semantic relevance and contextual information,we can acquire new knowledge in innovation of papers.Based on the level of new knowledge in the existing knowledge hierarchy and its impact on following research,we design indicators and comprehensive measurement methods to measure originality degree.Main contributions are shown as follow:1)We extract the topics of the innovation,expand the form with the help of external knowledge and domain features.According to the citation network and the topic-time retrieval,we can track the source of knowledge and obtain relevant literature.The F-score of the method is 89.67%,which is superior to traditional retrieval methods and has usability.2)We combine the context to represent the text and enrich the semantic information.By improving the traditional model,the BILM + HATT semantic relevance binary classification model is designed.Thus,we can make relevant judgments comparing with existing research and realize the novelty recognition of topics based on semantic content.The accuracy of recognition method is 89.1%,and the confirmation results reach the practical level.3)The combination words in the innovation are mined by a co-occurrence network and screened according to the correlation degree and co-occurrence rate.We calculate document frequency and weight of combination words to realize novelty recognition.The recognition method has an accuracy rate of 86.57%,which can effectively obtain new combination words in the innovation.4)We design Knowledge Hierarchy Index(KHI)according to the level of new knowledge in the existing knowledge.The measurement formula is designed based on the existing Disruption(D)and the improved Diffusion Index(DI),so a comprehensive originality measurement method is proposed.According to the score of originality,quantification level is classified into three levels: major,medium and general originality.Finally,the empirical analysis is carried out to prove the effectiveness of the proposed method.The paper includes 28 pictures and 29 tables.