Study On Intelligent Retrieval And Recommendation System For Science And Technology Project Experts

As the number of the science and technology projects increases more and more rapidly,experts play a more important role in the phases of reviewing, establishment, intermediateinspection, conclusing and acceptance of the projects. At present, the experts that review theprojects are generally assigned by management department or chosen mechanically andrandomly. There exists a phenomenon that the reviewed results for the projects are unfair andunobjective. Therefore, it is extremely important to do some research on it. Putting the projectand expert information, which is provided by the science and technology project wholeprocess management system of Zhejiang Province, as the experimental data, this paper tries todo some research about the key technologies of the keywords extraction, knowledgerepresentation model and similarity calculation method firstly. And the intelligent retrievaland recommendation system for science and technology experts is implemented finally. Themain research contents and innovations are as follows:(1) Aiming at the expert’s and project’s information have the semi-structured character,we have presented the Matter-Element Multi-fields Knowledge Representation Models forexpert’s and project’s keywords, which are based on Extenics Matter-Element Model andVector Space Model. Firstly, we propose a novel keyword extraction method for the project’sand expert’s Chinese main text information obtained from database, which includes wordsegmentation and stop words filtering by stop word lists. Secondly, after establishing thewords network of a project’s information, we extract the project information’s keywords bymeasuring the word’s statistical feature and gathering feature. The expert’s information isstreamlined, so we extract the expert inforamation’s keyword directly by filtering stop words.Finally, we establish the knowledge representation model of expert and project on the basis ofkeywords, and create corresponding index.(2) The traditional scientific expert retrieval matches terms mechanically and lacks ofsemantic analysis, and experts are not listed by the retrieval relevance in descending order.Relying on the Matter-Element Multi-fields Knowledge Representation Model, we willpropose a method for expert information full-text retrieval based on semantic analysis. We putforward the retrieval relevance computing method between search term and expertinformation based on word semantic similarity. This method lists experts by the retrievalrelevance in descending order. The practical application and the experiment of full-textretrieval for expert information show that the proposed method has achieved the anticipated effect, which can retrieval experts for science project more accurately.(3) Relying on the Matter-Element Multi-fields Knowledge Representation Model, wealso have proposed an expert recommend method for science projects. It automaticallygenerates and recommends candidate experts for one or a group of projects. Firstly, weevaluate the relevance computing method between project/projects and expert informationbased on two parts graph maximum semantic matching analysis. Secondly, the methodgenerates an initial list of recommended experts by the threshold truncation. Finally, usingregression analysis，a scoring model for scientific research ability based on AnalyticHierarchy Process (AHP) is put forward to optimize the order of recommended experts.Aiming at the case of recommending experts for a group of projects, it needs to merge thefeatures of Matter-Element Multi-fields Knowledge Representation Model. Thus, severalmodels of different projects are transformed into one multi-fields knowledge representationmodel.Furthermore, the paper’s research can actually be applied into the development for expertintelligent retrieval and recommendation. It automatically retrieves and recommends expertsfor science and technology projects. The preliminary application shows that, it alleviates theproblems of huge workload and lacking scientificity and unreason in the process of expertselection intelligently. The research promotes science decision-making and management to ahigher level effectively.