Inducing Character-level Structure in Subword-based Language Models with Type-level Interchange Intervention Training

Jing Huang, Zhengxuan Wu, Kyle Mahowald, Christopher Potts

Findings: Interpretability and Analysis of Models for NLP Findings Paper

Session 1: Interpretability and Analysis of Models for NLP (Virtual Poster)
Conference Room: Pier 7&8
Conference Time: July 10, 11:00-12:30 (EDT) (America/Toronto)
Global Time: July 10, Session 1 (15:00-16:30 UTC)
Spotlight Session: Spotlight - Metropolitan West (Spotlight)
Conference Room: Metropolitan West
Conference Time: July 10, 19:00-21:00 (EDT) (America/Toronto)
Global Time: July 10, Spotlight Session (23:00-01:00 UTC)
Keywords: hierarchical & concept explanations, robustness
TLDR: Language tasks involving character-level manipulations (e.g., spelling corrections, arithmetic operations, word games) are challenging for models operating on subword units. To address this, we develop a causal intervention framework to learn robust and interpretable character representations inside...
You can open the #paper-P399 channel in a separate window.
Abstract: Language tasks involving character-level manipulations (e.g., spelling corrections, arithmetic operations, word games) are challenging for models operating on subword units. To address this, we develop a causal intervention framework to learn robust and interpretable character representations inside subword-based language models. Our method treats each character as a typed variable in a causal model and learns such causal structures by adapting the interchange intervention training method of Geiger et al. (2021). We additionally introduce a suite of character-level tasks that systematically vary in their dependence on meaning and sequence-level context. While character-level models still perform best on purely form-based tasks like string reversal, our method outperforms character-level models on more complex tasks that blend form, meaning, and context, such as spelling correction in context and word search games. Compared with standard subword-based models, our approach also significantly improves robustness on unseen token sequences and leads to human-interpretable internal representations of characters.