Research On Key Technologies Of Electronic Journal Literature Recommendation In Digital Library

With the constant improvement of Internet and information technology, electronic literature has become the indispensable important academic resource in academic digital library. As an important component of electronic literature, Electronic journal literature represents the developing tendency of research areas and scholars’ research achievements, which is an important source for research users to acquire scientific information. Because the storage of journal literature in digital library is huge and continues increasing, users often spend a lot of time and effort retrieving the literature they need. Therfore, as a college information service platform, it is urgent for digital library to explore a kind of technology to find the literature users really need. So the research on electronic journal literature recommendation in digital library has very important practical significance. Electronic journal literature and users as the research object, recommending users needful literature for the purpose, this paper analyzed and studied the realization and key technologies of electronic journal literature recommendation.Based on insight into the information recommendation patterns in digital libraries and recommendation methods at home and abroad, this paper mainly introduced and analyzed the key technologies which included recommendation algorithms, information retrieval model, Chinese word segmentation and text feature extraction. Then, the paper did emphatical discussion and improvement on the defects of the trational TF-IDF text feature extraction. Through the analysis of causes of the defects and bad influence on recommended results, the paper took literature words distribution between classes and in class into consideration. It improved the IDF calculation formula and introduced words dispersion in class, TD. Finally, the improved TF-IDF’-TD algorithm was got. For the specific characteristics of electronic journal literature, this paper has designed a recommendation model by adopting content-filtering algorithm and given the realization methods of main function modules. The paper introduced literature value on the basis of literature similarity, considering three indicators, journal influence factor, cited frequency and published time, which have great influence on journal literature quality. Then, the paper has put forward comprehensive recommended degree which combines literature similarity and value to recommend users the good literature which is similar to their preferences.Using Java language and Eclipse development platform to program and experiment, the results show that, improved TF-IDF feature extraction algorithm has increased the similarity of users’ relevant literature and improved the performance of the recommendation model effectively. Literature comprehensive recommended degree has given consideration to users’ requirements and literature quality, and improved the quality of recommended literature. The recommendation model designed in this paper could recommend users electronic journal literature, and the effect is good.