| With the vigorous development of the Internet and the rapid growth of Chinese network size, Internet has increasingly become a new platform where people express their interest demands, vent their emotions on the happy, satisfied or dissatisfied and discuss with other people. However, due to the uneven quality of Internet users, hot issue is what it does or malicious speculation, it is need to analyze the netizen comments and identify the network public opinion in various roles, so as to make the correct response. When the emergency comes, it needs to analyze and response the discussion immediately. Analysis of the entire replies by manpower is very unrealistic, immediate recognition model is not feasible, so we need to select a model most similar from a prior to identify the role of the public opinion. That’s where the role identification model and transfer between different areas studied by this paper work.First, this paper describes the background and significance of transfer learning bound between heterogeneous relation data and the application in role identification, cites the related work and points out the reason of impractical. Then, based on existing research, put forward a new measure, set up the transfer learning bound between heterogeneous relation data, prove and extend it.Secondly, we present the process of establishing role identification model and a method of enhancing the adaptability and model transfer between areas. And then compare the role identification effect between different areas, as well as with the conventional method to make a comparison.Finally, we model different network carrier, judge whether suitable for transfer between areas according the similarity and give the result of character recognition as a sample for future practical.based on comparative analysis of the experimental results, we demonstrate the effectiveness of our proposed method, the feasibility of model transfer between relational data domains. |
PDF Full Text Request