What BERT is not: Lessons from a new suite of psycholinguistic diagnostics for language models
Allyson Ettinger Department of Linguistics University of Chicago
Abstract
Pre-training by language modeling has become a popular and successful approach to NLP tasks, but we have yet to understand exactly what linguistic capacities these pre-training processes confer upon models. In this paper we introduce a suite of diagnostics drawn from human language experiments, which allow us to ask targeted questions about information used by language models for generating predictions in context. As a case study, we apply these diagnostics to the popular BERT model, finding that it can generally distinguish good from bad completions involving shared category or role reversal, albeit with less sensitivity than humans, and it robustly retrieves noun hypernyms, but it struggles with challenging inference and role-based event prediction—and in particular, it shows clear insensitivity to the contextual impacts of negation.
中文速览
预训练语言模型（如BERT）在各种NLP任务上表现出色，但人们并不清楚它们究竟学到了哪些语言知识。这篇论文借鉴认知心理语言学中针对人类语言加工设计的实验范式，构建了一套诊断测试集，专门考察语言模型在常识推理、语义角色、类别归属和否定理解等方面的能力，测试方式是直接检验模型对上下文中词语的预测概率，无需任何任务微调。研究者将这套工具应用于BERT，发现它在类别区分和名词上下位关系检索方面表现较好，但在挑战性的常识推理和基于事件角色的预测上明显吃力，尤其对"否定"的语义影响几乎完全不敏感。这一发现揭示了当前预训练语言模型的根本局限，提醒我们不能仅凭下游任务的高分就假设模型已掌握深层语言理解能力。
原文 arXiv:1907.13528;中英对照 + 大白话阅读 https://aha.fim.ai/paper/1907.13528v2