Automating the Pyramid Method for Evaluating Text Summarization: Exploring Novel Approaches to Approximate Summary Content Units

Automating the Pyramid method for text summarization evaluation by exploring novel approaches to approximate Summary Content Units (SCUs), including semantic meaning units (SMUs) from Abstract Meaning Representation (AMR) and semantic GPT units (SGUs) from large language models. The study examines the intrinsic quality and downstream utility of these SCU approximations compared to existing methods.