Improving Sentence-Level Relation Extraction through Curriculum Learning
This work improves relation extraction for natural language processing applications by incrementally enhancing performance on standard datasets.
The paper tackled the problem of sentence-level relation extraction by addressing difficult or noisy data through a curriculum learning-based model that splits data by difficulty, achieving state-of-the-art F1-scores of 75.0% on TACRED and 91.4% on Re-TACRED.
Sentence-level relation extraction mainly aims to classify the relation between two entities in a sentence. The sentence-level relation extraction corpus often contains data that are difficult for the model to infer or noise data. In this paper, we propose a curriculum learning-based relation extraction model that splits data by difficulty and utilizes them for learning. In the experiments with the representative sentence-level relation extraction datasets, TACRED and Re-TACRED, the proposed method obtained an F1-score of 75.0% and 91.4% respectively, which are the state-of-the-art performance.