“Relation extraction (RE) is the core NLP task of inferring semantic relationships between entities from text. Standard supervised RE techniques entail training modules to tag tokens comprising entity spans and then predict the relationship between them. Recent work has instead treated the problem as a sequence-to-sequence task, linearizing relations between entities as target strings to be generated conditioned on the input. Here we … [use] larger language models (GPT-3 and Flan-T5 large) than considered in prior work and evaluat[e] their performance on standard RE tasks under varying levels of supervision.”

Find the paper and full list of authors at ArXiv.