Grammar Control in Dialogue Response Generation for Language Learning Chatbots

TL;DR

This paper introduces a method to control grammatical features in chatbot responses for language learning by grounding response generation in a grammar repository, enhancing targeted grammar practice.

Contribution

It proposes a novel approach to integrate grammar control into dialogue generation models, improving language learning tools and supporting specific grammar acquisition.

Findings

Decoding strategies with Llama3 outperform GPT-3.5 in grammar control.

Grammar-controlled responses support targeted grammar learning.

The approach benefits language learning chatbots and second language acquisition research.

Abstract

Chatbots based on large language models offer cheap conversation practice opportunities for language learners. However, they are hard to control for linguistic forms that correspond to learners' current needs, such as grammar. We control grammar in chatbot conversation practice by grounding a dialogue response generation model in a pedagogical repository of grammar skills. We also explore how this control helps learners to produce specific grammar. We comprehensively evaluate prompting, fine-tuning, and decoding strategies for grammar-controlled dialogue response generation. Strategically decoding Llama3 outperforms GPT-3.5 when tolerating minor response quality losses. Our simulation predicts grammar-controlled responses to support grammar acquisition adapted to learner proficiency. Existing language learning chatbots and research on second language acquisition benefit from these affordances. Code available on GitHub.