#25 Towards the Development of a Fake News Detection...

Abstract:

Studies on automatic fake news detection has gain global interest as the spread of these non-factual, deceitful news items has led to confusion in various facets of our society-political, social, business and entertainment. From trivial celebrity news to influencing national elections, mis- or dis- information has brought disorder in large-scale proportion. Furthermore, as people get their news from social media, these platforms have been considered enabler of fake news as it was programmed to share posts to make them viral.

Since manual detection is problematic, humans cannot be trusted to detect fake news because we have our own biases. So, experts are relying on Artificial Intelligence to curtail its propagation. However, the human language has complexities like grammar, semantics, context, factors that influence the language like social, cultural, historical and political factors. Consequently, developing an AI detector is not a straight-forward simple algorithm.

Although no single autonomous system is enough to detect, for the time being, development of more automated detection tools and algorithms helps to block and warn the public.

This paper examines selected machine learning algorithms such as Naïve Bayeas and Support Vector Machine or SVM in combination with Natural Language Processing or NLP, a branch of AI, to develop a predictive model that checks whether a news tweet is fake or not. Based on studies, these algorithms have showed better performance in terms of accuracy and precision in detecting fake news. Datasets were developed from collected news-related tweets from the Philippines from the past six months. The machine learning model is tested through dataset splitting (60% training, 20% testing, and 20% validation). After training and fitting the data using Naïve Bayeas and SVM, performance of the predictive model is compared using classification Metrics: accuracy, precision, recall, F1-score, precision and recall. The developed model is to be implemented as a web agent or a browser extension notifying whether a requested news tweet is fake or not. 

nationalueduph-my.sharepoint.com/personal/acmwap2021_national-u_edu_ph/_layouts/15/onedrive.aspx?id=%2Fpersonal%2Facmwap2021%5Fnational%2Du%5Fedu%5Fph%2FDocuments%2FSubmitted%20Posters%2FPaperID25%5FSalazarArmida%2Emp4&parent=%2Fpersonal%2Facmwap2021%5Fnational%2Du%5Fedu%5Fph%2FDocuments%2FSubmitted%20Posters&originalPath=aHR0cHM6Ly9uYXRpb25hbHVlZHVwaC1teS5zaGFyZXBvaW50LmNvbS86djovZy9wZXJzb25hbC9hY213YXAyMDIxX25hdGlvbmFsLXVfZWR1X3BoL0VXT2FaVFJveVFWSWdUOEdzbTFsM29JQjFiRTJRUlRJZWU1a3ZpQlltOHhBVVE%5FcnRpbWU9aVdvdlQzcjUyRWc

What are the determinants to consider a tweet or post as fake news? Then, what are the aspects to be considered i.e. content, frequency of post, source or intent. Are we after the source or the spread of the fake news?

How can people use this application? Will this block fake news tweets?

Congrats Maam Armi <3 Hope this research really helps in eradicating misinformation and shape our society in a better place.