Performance Evaluation of the Recommender System in Scientific Databases

Scientific article recommender system assists and advance information retrieval process by proposing and offering articles tailored to the researchers needs. The main purpose of this study is to evaluate the performance of the recommender System in three scientific databases.  

This applied study is directed by the valuation method. Sample consisted of three scientific databases: Elsevier, Taylor & Francis, and Google Scholar, which share recommendation tools. "Information storage and retrieval" was selected as the search subject. Ten specialized keywords related to the topic of information storage and retrieval were selected. After searching each key words, the first retrieved article was reviewed. Then, for each first article, the first 5 recommended articles were mined in each of the three mentioned databases. Data was collected through direct observation using a researcher-made checklist. To evaluate subject relevance, bibliographic information of the first article retrieved in each subject and database along with the bibliographic information of 5 recommended articles was provided to two groups of librarians and IT professionals. Sample was selected by snowball method. Descriptive and inferential statistics were used to analyze the data.

Findings showed that among the databases, Elsevier recommends more relevant results from the perspective of IT professionals and librarians in the field of information storage and retrieval, with Google Scholar and Taylor & Francis in the next ranks. In total, the most relevant articles in terms of subject experts were the articles that ranked fifth.

To sum up, Elsevier performed better than the other two databases in terms of recommending related articles. Also, there is a significant difference between the views of librarians and IT professionals regarding the relevance of recommended articles in the field of information storage and retrieval. Thus, from the point of view of IT professionals, the significance of the recommended articles is greater.