Can LLMs Effectively Detect Poetry? Identifying Quality Differences Among Local, Free, and Commercial Models

Information extraction from large text corpora poses particular challenges for social media corpora due to their high diversity in languages, content, and genres. At the same time, social media posts, as relatively manageable and self-contained text fragments, are more advantageous for automated analysis and classification than longer texts. This poster presents an approach for testing the detection of formal parameters such as text type (poetry or not), text language (Russian, Ukrainian, Belarusian, and others), as well as several content features, using a sample of Facebook posts containing published poetry.