A Kind Of Network Model Building For Field Term Based On Expert Knowledge And Deep Learning

It has been a kind of mature approach to represent knowledge by using network model which consists of node and node relationships.How to use machine to learn knowledge,it is a long time task in the AI field,and also a need of experts in reality.In this paper,expert interaction based on Delphi method is used to get expert knowledge,and Technical system and technical list are generated by expert knowledge.In this case,more accurate text data set could be obtained by retrieving literature data.Technical terms in the data set were extracted and the co-occurrence network was constructed,then a kind of representation learning model Node2 vec was used to train the network and get the terms representation for further notability and correlation analysis.This paper found that the continuous iteration of expert knowledge and feedback after data analysis enables us to obtain more accurate and detailed domain knowledge.At the same time,Node2 vec based on the network model indicates that the learning model can also ignore the influence of redundant texts on experiments,and the representation obtained is limited to technical terms,making the analysis more targeted and efficient.