frontmatter 
title: "Chatlang: A Two-Window Approach to Chatbots for Language Learning"
venue: EAIT 2023 - International Conference on Education and Artificial Intelligence Technologies
doi:
date: 2023-09-21
authors:
  - Sheng Yu
  - "[[Jared Coleman]]"
  - Bhaskar Krishnamachari
links:
  - https://anrg.usc.edu/www/papers/chatlang.pdf
  - https://github.com/ANRGUSC/chatlang
abstract: "Role-playing is an effective tool in language education, allowing learners to practice a target language in realistic scenarios and receive immediate feedback from teachers, native speakers, and fellow learners. This method is less viable, however, outside of the classroom, where access to speakers is more limited. Recent advancements in large language models (LLMs) have the potential to change this by enhancing the quality and accessibility of chatbot-based role-playing systems. This paper introduces \"Chatlang,\" a two-threaded LLM-powered chatbot-based role-playing system. Users interact with the system using two chat windows displayed side-by-side in a web interface. The first window is for role-playing, where the user and chatbot are expected to never break character and only use the target language. In the second window, the user can use their native language to ask questions about the conversation or target language and get context-relevant feedback on their mistakes. This design enables uninterrupted roleplaying alongside instant feedback and support. We present the design of Chatlang and discuss potential avenues for future research."
tags:
  - artificial-intelligence
  - large-language-models

Chatlang: A Two-Window Approach to Chatbots for Language Learning

Sheng Yu, Jared Coleman, Bhaskar Krishnamachari

EAIT 2023 - International Conference on Education and Artificial Intelligence Technologies · September 21, 2023 · PDF · GitHub

Abstract

Role-playing is an effective tool in language education, allowing learners to practice a target language in realistic scenarios and receive immediate feedback from teachers, native speakers, and fellow learners. This method is less viable, however, outside of the classroom, where access to speakers is more limited. Recent advancements in large language models (LLMs) have the potential to change this by enhancing the quality and accessibility of chatbot-based role-playing systems. This paper introduces "Chatlang," a two-threaded LLM-powered chatbot-based role-playing system. Users interact with the system using two chat windows displayed side-by-side in a web interface. The first window is for role-playing, where the user and chatbot are expected to never break character and only use the target language. In the second window, the user can use their native language to ask questions about the conversation or target language and get context-relevant feedback on their mistakes. This design enables uninterrupted roleplaying alongside instant feedback and support. We present the design of Chatlang and discuss potential avenues for future research.

Updated