Multi-expert Opinion Short Text Integration Research And Practice

The integration of short texts of multi-expert opinions is a common method used by the government to make group decisions for specific projects by adopting the "Expert Group Review Method".The realization of machine integration of multi-expert opinion short texts not only has theoretical research value,but also has practical significance.Aiming at the integration of expert opinions under the same goal and fixed knowledge domain,this paper proposes a research method of short text integration of multi-expert opinions based on knowledge fusion.Realize comprehensive opinion generation dominated by knowledge fusion at the cognitive level.Provide efficient,objective,comprehensive,and clear group decision-making results,which are suitable for the expert group decision-making mode usually adopted by various government projects.This paper transforms the research on the integration of expert opinion short texts from the textual form at the physical level to the knowledge expression and organization at the cognitive level.Content mining of expert opinions through opinion mining technology completes knowledge acquisition,use the target knowledge concept tree and ontology to express and supplement knowledge,on the basis of knowledge specification,the lack of semantics is supplemented,more explicit knowledge and its association are identified,and knowledge fusion is carried out based on this to form a new complete knowledge resource to support the expression of expert opinion short text integration.Relying on the pre-approval of e-government projects as the application scenario,this paper realizes the integration method to complete the generation of the comprehensive approval opinions of the expert group,and replaces the manual integration work,scientifically and efficiently "intelligent" assists government project approval and guarantees the quality of project construction.The research contents are as follows:(1)Based on cognitive science theory and analysis of knowledge organization process in opinion generation process,this paper proposes a theoretical method based on knowledge norms and using knowledge fusion at the cognitive level to achieve multi-expert opinion short text integration.Specifically,it includes three parts: knowledge representation and knowledge supplementation,knowledge acquisition using opinion mining technology,and short text integration of opinions based on knowledge fusion.(2)In this paper,a fine-grained opinion mining method based on knowledge specification is designed,which aims to obtain deep characteristic knowledge such as experts’ adjustment suggestions for project elements and emotional attitudes.First,the word features and the bert word embedding model are integrated into the Bi LSTM-CRF attention mechanism model for entity extraction of government information system projects;Secondly,according to the characteristics of the text,the expert opinion SAO structured knowledge is obtained on the basis of grammatical analysis.Then,the constructed knowledge concept tree and project approval knowledge ontology support the evaluation feature hierarchical semantic recognition and tacit knowledge discovery.At the same time,the sentiment polarity is obtained by using a deep learning model.(3)In this paper,the integration strategy of expert opinion short text based on knowledge fusion is carried out in detail.It is essentially a multi-source knowledge fusion process based on knowledge extraction.It includes the micro-level knowledge fusion supported by the concept tree and the macro-level knowledge fusion supported by the knowledge ontology.On this basis,the subjective and objective information fusion method is further designed.It has practical(4)significance for the generation of the comprehensive opinion structure model.Finally,the effectiveness of the strategy is verified by the method of textual information and knowledge meta-measurement.(5)Based on the multi-expert opinion short text integration strategy based on knowledge fusion proposed in this paper,use the 2017-2018 e-government expert approval data of T City Network Information Office for instant verification and analysis.Explain the applicability of the research in this paper in the examination and approval of auxiliary government affairs projects.