Recognition Study On Research Fronts Of Artificial Intelligence Based On LDA Model

Artificial intelligence has increasingly become an important development strategy in science and technology.Identifying the research fronts of artificial intelligence timely and accurately is a prior work.This paper studies the reaserch fronts identification of artifiicial intelligence fields using topic model analysis,social network analysis,expert evaluation,and comparative research.The main work is as follows.Firstly,based on the literature research methodology,the research methods of identifying research fronts,fronts characteristic indicators,LDA model related theories,research fronts related theories,and the research tools are sorted out and analyzed.Then,the source of the literature data is analyzed,the search formula is constructed,the search results are counted and stored.The word segmentation,the stop word removal,the stemming and the lemmatization are implemented using the Python language.Afterwards,the construction,parameter setting and solution of the LDA model are completed.The probability-based document-topic matrix and topic-vocabulary matrix are obtained,which provides data support for the calculation of research fronts.Consequently,the topic association network is constructed with the help of social network analysis and visualization technology so that the research fronts of artificial intelligence are indentified quantitatively in terms of the three indicators of topic intensity,novelty and betweenness centrality Finally,the effectiveness of the identification is validated via the comparative research and expert evaluation.Based on LDA model,16 research fronts in artificial intelligence field were obtained from 2013 to 2017.According to Small's fronts life cycle theory,the research fronts was classified and interpreted,and divided into growth type(10)and stable type(6).Finally,the accuracy of topic recognition based on LDA model was verified from two aspects of expert evaluation and comparative research.The results show that LDA model can simulate clustering topics well and give the research fronts in the fields accurately,according to the selected indicators related to research fronts.