Hierarchical Attention Network with Pairwise Loss for Chinese Zero Pronoun Resolution

Abstract

Recent neural network methods for Chinese zero pronoun resolution didn't take bidirectional attention between zero pronouns and candidate antecedents into consideration, and simply treated the task as a classification task, ignoring the relationship between different candidates of a zero pronoun. To solve these problems, we propose a Hierarchical Attention Network with Pairwise Loss (HAN-PL), for Chinese zero pronoun resolution. In the proposed HAN-PL, we design a two-layer attention model to generate more powerful representations for zero pronouns and candidate antecedents. Furthermore, we propose a novel pairwise loss by introducing the correct-antecedent similarity constraint and the pairwise-margin loss, making the learned model more discriminative. Extensive experiments have been conducted on OntoNotes 5.0 dataset, and our model achieves state-of-the-art performance in the task of Chinese zero pronoun resolution.

Tackling Zero Pronoun Resolution and Non-Zero Coreference Resolution Jointly

“…Yin et al (2018a) introduces reinforcement learning model into zero pronoun resolution to integrate local and global resolution information. Lin and Yang (2020) considers bidirectional attention between zero pronouns and candidate antecedents and proposes a pairwise-margin loss and a similarity constraint to optimize their model.…”