International Journal of Science and Research (IJSR)

International Journal of Science and Research (IJSR)
Call for Papers | Fully Refereed | Open Access | Double Blind Peer Reviewed

ISSN: 2319-7064

Downloads: 1 | Views: 57 | Weekly Hits: ⮙1 | Monthly Hits: ⮙1

Masters Thesis | Computer Science & Engineering | India | Volume 12 Issue 4, April 2023

A Thesis on News Recommendation

Abhik Naskar | Sudeshna Sarkar

Abstract: An integral part of any individualised news service is the news recommendation technique. Research on news recommendation is much more limited than that on product and movie recommendations, primarily due to the absence of a benchmark dataset of sufficiently high quality. In this paper, we introduce a massive dataset for news recommendation called MIND. MIND is built from Microsoft News user click logs and contains over 160k English news articles with rich textual content like titles, abstracts, and body paragraphs. We show that MIND is a useful testbed for news recommendation by comparing several cutting-edge methods that were developed on various private datasets. Our findings demonstrate that the success of news recommendation is heavily dependent on the accuracy with which content is understood and user interests are modelled. Several NLP techniques, including efficient text representation methods and pre-trained language models, have been shown to significantly boost news recommendation performance. Because the internet allows access to news stories from millions of sources throughout the world, online news reading has grown in popularity. The ability of news websites to assist readers in finding stories that are interesting to read is a major difficulty. We share our work on constructing a personalised news recommendation engine in the Microsoft News Dataset in this thesis (MIND). The recommendation engine creates profiles of users' news interests based on their prior click activity for users who are signed in and have expressly enabled web history. To better understand how users' news interests vary over time, we employ semantic analysis to anticipate users' current news interests based on their behaviours and the news trend seen across all users' activities. Moreover to capture the knowledge aware concept of different news, their clicked behaviours etc. we also construct knowledge graphs to capture that information and generate personalised news recommendations. To access the MIND dataset, visit

Keywords: Content-Based Filtering, Collaborative Filtering , News Recommendation Section, Pre-trained Language Model-empowered News Recommendation, News encoder, Microsoft News Dataset, MIND

Edition: Volume 12 Issue 4, April 2023,

Pages: 1324 - 1330

How to Download this Article?

Type Your Valid Email Address below to Receive the Article PDF Link

Verification Code will appear in 2 Seconds ... Wait