Computing the Story: Novel Narratives Decoded
- Thierry Poibeau (Lattice, CNRS,ENS-PSL)
Abstract
Seventy years of NLP research has led to today’s hype around large language models, but semantic and syntactic parsing have been integral from the start. Early techniques were inadequate, but LLMs (and improved computing power) have changed the situation. Now, we can parse and analyze massive corpora, like the web, with high accuracy. How can these advancements be applied in sociology? Do we still need traditional parsing, or should we rely on LLMs’ abstract representations, for the analysis of narratives for example? And why do sociologists seem fixated on SVO triplets, a concept linguists rarely use? These are the questions I will explore in my talk—though answers may be limited, it may open up discussions.