Topic Discovery About Economy During COVID-19 Pandemic from Spanish Tweets

Automatic topic discovery from natural language texts has been a challenging and widely studied problem. The ability to discover the topics present in a collection of text documents is essential for information systems. Topic discovery has been used to obtain a compact representation of documents for grouping, classification, and retrieval. Some tasks that can benefit from topic discovery: recommendation systems, tracking misinformation, writing summaries, and text clustering. However, topic discovery from Spanish texts has been somewhat neglected. For this reason, this work proposes analyzing the behavior of topic discovery tasks in texts in Spanish, specifically in tweets about the Mexican economy during the COVID-19 pandemic, under three different approaches. A comparison was conducted, achieving promising results because the topic coherence metric indicates coherent topics. The highest score of 1.22 was obtained using PLSA with 50 topics, concluding that the topics encompassed the study domain.