Artificial Intelligence And Big Data Technology Literature Information Mining

In recent years,artificial intelligence(AI)and big data have developed rapidly,and great attentions have been paid from both academia and industry.AI and big data are interdependent in a sense.As one of the most important carriers of research achievements,scientific literature has become an important tool of tracking the technological development.The scientific literature studied in this paper is mainly patent papers,journal papers and conference papers.Though there are some related analysis on scientific papers,the improvement is still one important open problem.In the aspect of patent analysis,most of the work is to retrieve relevant patent information from the patent database for analysis in the case of a given search keyword.As different authors have subjectivity to the setting of keywords,some authors deliberately abstract the description of their own patents.Therefore,searching for patents by the keyword matching way could lead to omission of query results easily.We extracted phrases related to "Big Data" from the patent abstracts by designing a filter.When compared the result with TF-IDF,SegPhrase,C-value,Word2 vec,we found that the method based on filtering rules would be more accurate.It could extract more high "Big Data" related phrases.In the analysis of journal and conference papers,many methods are based on the citation counting analysis of papers,however,most methods do not consider the change of paper citation weight over time.In order to make the experimental data representative,we refer to the CCF's recommendation list in 2015.After a detailed analysis of the changes in the citation weights of journals and conferences over time,related schemes are designed to classify journals and conferences,and the reasonable of the proposed schemes were verified.Finally,the journal and conference papers are mixed together for comparison and some interesting conclusions are found driven by the data.At the same time,the advisor-advisee relationship is also an important information hidden in the scientific literature,mainly included in the joint network.The related work is generally mainly from the perspective of the entire author's network,and the structure is easily affected by some special authors(nodes with large degrees).Considering of this problem,this paper divides the original author into some small independent associations firstly,and then mining the advisor-advisee relationship in this small community.This could ensure that the author's advisors belong to his collaborator and other authors do not affect the results.In the end,the advisors of a given author can be discovered and his academic inheritance tree can be drawn as well.