Publication

Abstract : The researchers need to access a large number of articles so as to keep abreast with recent developments in their respective research areas. Finding a relevant article from a huge collection of information, therefore, becomes a huge challenge. Recommender systems play a vital role in assisting researchers to filter correlated information from large volume articles. In this work, we have used the topic modeling approach, latent Dirichlet allocation (LDA), for recommending research articles based on user queries. We have collected reference articles of a query given by the user from the Wikipedia site. The proposed algorithm comprises four steps: 1. LDA first iteration to find the top N words. 2. LDA second iteration elicits weighted topic distribution. 3. Pointwise mutual information (PMI) obtains the correlation between the top N words and the weighted topic distribution words. 4. Reference articles are collected for top correlated words, and the top-ranked articles are recommended to the user by applying collaborative filtering algorithm. The two-pass LDA helps to avoid the ambiguities. The proposed algorithm performance is evaluated using the parameters of precision and recall. As compared to the results obtained from the existing recommender systems, our work demonstrates that the LDA-based algorithm gives a powerful recommender framework. © 2020, Springer Nature Singapore Pte Ltd.