R$^2$-Net: Relation of Relation Learning Network for Sentence Semantic Matching
This paper addresses the problem of underestimating semantic information from output labels in sentence semantic matching for NLP researchers, especially for tasks with few labels. It offers an incremental improvement to existing deep neural network approaches.
This paper proposes R$^2$-Net, a model for sentence semantic matching that aims to better leverage the semantic information within output labels. It combines BERT for global encoding with a CNN for local keyword capture, and introduces a self-supervised relation of relation classification task along with triplet loss to improve label utilization and distinguish relations. The model demonstrates superior performance on two sentence semantic matching tasks.
Sentence semantic matching is one of the fundamental tasks in natural language processing, which requires an agent to determine the semantic relation among input sentences. Recently, deep neural networks have achieved impressive performance in this area, especially BERT. Despite the effectiveness of these models, most of them treat output labels as meaningless one-hot vectors, underestimating the semantic information and guidance of relations that these labels reveal, especially for tasks with a small number of labels. To address this problem, we propose a Relation of Relation Learning Network (R2-Net) for sentence semantic matching. Specifically, we first employ BERT to encode the input sentences from a global perspective. Then a CNN-based encoder is designed to capture keywords and phrase information from a local perspective. To fully leverage labels for better relation information extraction, we introduce a self-supervised relation of relation classification task for guiding R2-Net to consider more about labels. Meanwhile, a triplet loss is employed to distinguish the intra-class and inter-class relations in a finer granularity. Empirical experiments on two sentence semantic matching tasks demonstrate the superiority of our proposed model. As a byproduct, we have released the codes to facilitate other researches.