VINIF.2020.DA14 – Developing a system to support fake news detection in online social networks

Principle Investigator
Dr. Nguyen Thi Minh Huyen
Host Organization
University of Science, Viet Nam National University

Fake news spread on social networks can cause serious economic, social, and health consequences for internet users. Detecting fake news, thereby limiting its spread and minimizing damage, is an urgent need and is attracting the attention of many research disciplines, including data science and social science and communication. This project aims to build a system that supports users in confirming the authenticity of news spread on social networks by automatically collecting and synthesizing supporting or denying evidence of the news from the Internet.

Objectives of the project

To build the above system, the project needs to research and develop machine learning methods for a number of core problems, including:

  • Inferring semantic relationships between multimedia texts
  • Inferring semantic relationships between texts in different languages
  • Summarize and synthesize multimedia and/or multilingual documents

Project impact

  • Allow the development of systems that suit the needs of diverse users to strengthen community efforts in fighting fake news.
  • Promote further research in data science and natural language processing, especially for Vietnamese; bring opportunities for training, research and development of advanced technologies in Vietnam.
  • Contribute to raising awareness of fake news and its damage in both the research community and society as a whole. Strengthen and encourage further development of platforms to educate users about responsibility in the Internet environment.
Principle Investigator
Dr. Nguyen Thi Minh Huyen
Host Organization
University of Science, Viet Nam National University

Tags

Expect Progress
01/11/2020
01/11/2021
Phase 1

– First version of toolkits for image caption labeling and semantic matching of sentence pairs
– Data sets with at least 10,000 records each
– The first experimental version of the system supports detecting fake news on social networks
– Submit 01 rank A conference paper for publication.

01/11/2022
Phase 2

– Add at least 20,000 records to each data set
– The second test version of the system supports detecting fake news on social networks
– Accepted publication of 01 grade A conference paper
– Submit 01 Q1 magazine article, and 01 rank A conference article
– 01 phase 2 report.

01/11/2023
Phase 3

– Continue adding at least 20,000 records for each data set obtained after phase 2
– Third version of the system
– 01 open source library for the problem of retrieving documents related to a post on social networks, using search engines on the internet; minimum accuracy of  90% on 100 post test dataset
– 01 open source library for the problem of inferring semantic constraints on two documents; Minimum 90% accuracy on a test dataset of 1000 sentence pairs; and at least 70% on a test dataset of 1000 text pairs
– 01 open source library for summarizing documents on the same topic; Save at least 90% time compared to manual text summarization
– 02 Q1 class journal articles and 01 A class conference article accepted for publication.

Tags