A technical talk on data annotation, quality control, and annotation automation with LLMs.
Data Fest is one of the largest professional events of Russia's Open Data Science community. This talk, held on the NLP track, compared three approaches to annotation: experts, crowdworkers, and LLMs.
The talk covered complex linguistic annotation schemes, quality-control tools, simpler annotator instructions, and automation using large language models. The recording received tens of thousands of views.
