The Application Of Text Mining Technology In The Analysis Of Academic Figures

Nowadays,the Internet technology is highly developed,and a large amount of valuable research literature,such as papers,periodicals,patent specification,conference documents,technical documents and other information,are in the form of text on the web.These kinds of information are generally unstructured,their contents are mostly natural languages,and their semantics are difficult to be processed by computers.In view of this,this thesis uses text mining technology to mine the literature information,extracting the necessary knowledge,including the co-author relations,research content,and visualizes the knowledge obtained.The main contents of this thesis include:(1)Documents preprocessing.The preprocessing includes cleaning,denoising,drying,word and clause partition and so on.Feature extraction and feature representation are completed on the preprocessed documents which are used to extract keywords in the documents.(2)Keywords extraction.In the first part,we use some rules to identify candidate phrases,then calculate the characteristic values of each candidate phrase,and then use the machine learning algorithm to predict the key phrases from each candidate phrase.(3)Visualized representation of knowledge.Interactive graphics software echarts is used to visualize the extracted knowledge in a way that can be understood by others.In the experimental stage,the effect of the KEA algorithm is evaluated by comparing with the labeled keywords and using a large corpus.According to the experimental results,the original algorithms are improved,and the structural features and grammatical features are added in the new algorithms.The confidence level has been significantly improved in the new algorithm for p<0.05.