The need for AI/ML text mining to identify negative narratives on social media for national security

A. Skumanich
Innov8ai Inc., California, United States

Keywords: AI, narratives, social media, misinformation

Social media can provide useful insights but there are challenges in extracting the key points. Standard top-down analyses often struggle to capture key signals of shifting events. With rapidly evolving global events, narratives on social media can provide insights, including instances of disinformation, which can emerge and develop swiftly. To address this need for an inductive approach, we examine a niche social media - GAB - to extract information and methodologies which can be more broadly applied. This paper examines narrative evolution on GAB and introduces quantitative methods from the field of corpus-based discourse analysis. The paper outlines the technical and methodological aspects of collecting and preprocessing GAB posts data for a keyness (Log Ratio) metric analysis, which identifies significant nouns and verbs for further investigation. Empirically, this method is applied to a case study involving a new GAB dataset from June 22 to June 26, 2023, focusing on the Wagner Militia mutiny attempt. The analysis highlights data features which indicate and capture the continuous shifting of narratives over time. We contrast some GAB features against a new Twitter dataset to indicate that the approach can assist in analyzing the larger social media channels. The need for this type of methodology is to better extract information from social media for security preparedness.