A Two-Agent Game for Zero-shot Relation Triplet Extraction

Relation triplet extraction is a fundamental task in natural language processing that aims to identify semantic relationships between entities in text. It is particularly challenging in the zeroshot setting, i.e., zero-shot relation triplet extraction (ZeroRTE), where the relation sets between training and test are disjoint. Existing methods deal with this task by integrating relations into prompts, which may lack sufficient understanding of the unseen relations. To address these limitations, this paper presents a novel Two-Agent Game (TAG) approach to deliberate and debate the semantics of unseen relations. TAG consists of two agents, a generator and an extractor. They iteratively interact in three key steps: attempting, criticizing, and rectifying. This enables the agents to fully debate and understand the unseen relations. Experimental results demonstrate consistent improvement over ALBERT-Large, BART, and GPT3.5 1 , without incurring additional inference costs in all cases. Remarkably, our method outperforms strong baselines by a significant margin, achieving an impressive 6%-16% increase in F1 scores, particularly when dealing with FewRel with five unseen relations 2 . 039 To reduce the overreliance on labeled data, Chia 040 et al. (2022) introduce a challenging task, zero-shot 041 relation triplet extraction (ZeroRTE), where rela-042 tion sets during the training and test stages are dis-043 joint. For example, in Fig. 1, training samples may 044 belong to the seen relation set sibling, performer, 045 while test samples may belong to the unseen rela-046 tion set mother, writer. Generalizing knowledge 047 from the training set to the test set is critical. 048 Existing methods deal with ZeroRTE by formu-049 lating it into prompts to leverage the power of lan-050 guage models. The main idea is to fine-tune the 051 model on seen relations from the training dataset.