Quantum-inspired Representation for Long-tail Senses of Word Sense Disambiguation

Abstract

Data imbalance, also known as the long-tailed distribution of data, is an important challenge for data-driven models. Due to the long tail phenomenon of word sense distribution in linguistics, it is difficult to learn accurate representations for Long-Tail Senses (LTSs) in Word Sense Disambiguation (WSD) tasks. Without data augmentation, exploring representation methods that do not rely on large sample sizes is an important means to combat the long tail. In this paper, inspired by the superposition state in quantum mechanics, a representation method in Hilbert space is proposed to reduce the dependence on large sample sizes. We theoretically prove the correctness of the method, verify its effectiveness on the WSD task, and achieve state-of-the-art performance.

Tasks

Reproductions