Beyond Words: Analyzing Emotions and Linguistic Characteristics to Detect Hoax-Related Tweets During Spanish Regional Elections

Detecting misinformation on social media, especially Twitter/X, is crucial during electoral periods. This study presents a comprehensive methodology using a Random Forest model to identify hoax-related tweets in the Spanish political landscape by analyzing their syntactic, semantic, lexical and emotional characteristics. The results reveal consistent emotional patterns where, the emotions most closely related to misinformation tend to be anger, disgust, fear and negativity. As a tangible outcome, we propose a tool designed to enhance the efficiency of fact-checkers in the detection of hoaxes. Also considered are the limitations of generic and multilingual approaches, supporting the use of context-specific strategies. The study demonstrates the effectiveness and generalizability, of the screening tool proposed as a means of combating misinformation in tweets during Spanish electoral periods.