ChatBot2023
Adaptive Chatbot for English Conversation Practice
| Alireza M. Kamelabad | |
|---|---|
| ali.mk | alimk@kth.se |
| +46 8 790 9269 | alimk@ieee.org |
If you got problems anywhere in the process, please contact me.
Publication
The paper describing this work is published in Frontiers in Education:
Personalized Language Learning with an LLM Chatbot: Effects of Immediate vs. Delayed Corrective Feedback
Alireza M. Kamelabad, Beatrice Turano, Mattias Lundin, Gabriel Skantze
Frontiers in Education, vol. 11, p. 1703664, Feb. 2026. Frontiers Media S.A.
🔗 DOI: 10.3389/feduc.2026.1703664
Abstract
The emergence of Large Language Models (LLMs) has opened new possibilities for language learning through conversational interaction with chatbots. Yet, little empirical evidence exists on how students experience such interactions and how corrective feedback should be provided. Research suggests that immediate corrective feedback is generally more effective than delayed feedback. Nevertheless, learners' perception of this effectiveness and their preferences for feedback timing, particularly in the domain of Computer-Assisted Language Learning (CALL), remain underexplored. This study investigates the feasibility of providing immediate feedback and examines the impact of feedback timing on user experience and grammar learning gains in English. An in-the-wild experiment was conducted with 66 L2 English learners, who integrated chatbot sessions into their English course as an extracurricular activity over one semester. Participants were randomly assigned to two groups receiving feedback either during or after the conversation. Findings reveal no significant difference in learning gains, but immediate feedback enhanced user experience, leading to overall positive perceptions of the chatbot. Additionally, we explore users' perceptions of the chatbot's social role and personality, offering a roadmap for future enhancements. These results provide valuable insights into the potential of LLMs and chatbots for language learning.Citation
If you use this work, please cite the paper using one of the formats below.
BibTeX
```bibtex @article{m.kamelabad2026-PersonalizedLanguageLearning, title = {Personalized Language Learning with an Chatbot: Effects of Immediate vs. Delayed Corrective Feedback}, shorttitle = {Personalized Language Learning with an Chatbot}, author = {M. Kamelabad, Alireza and Turano, Beatrice and Lundin, Mattias and Skantze, Gabriel}, year = 2026, month = feb, journal = {Frontiers in Education}, volume = {11}, pages = {1703664}, publisher = {Frontiers Media S.A.}, issn = {2504-284X}, doi = {10.3389/feduc.2026.1703664}, url = {https://www.frontiersin.org/articles/10.3389/feduc.2026.1703664/full}, abstract = {The emergence of Large Language Models (LLMs) has opened new possibilities for language learning through conversational interaction with chatbots. Yet, little empirical evidence exists on how students experience such interactions and how corrective feedback should be provided. Research suggests that immediate corrective feedback is generally more effective than delayed feedback. Nevertheless, learners' perception of this effectiveness and their preferences for feedback timing, particularly in the domain of Computer-Assisted Language Learning (CALL), remain underexplored. This study investigates the feasibility of providing immediate feedback and examines the impact of feedback timing on user experience and grammar learning gains in English. An in-the-wild experiment was conducted with 66 L2 English learners, who integrated chatbot sessions into their English course as an extracurricular activity over one semester. Participants were randomly assigned to two groups receiving feedback either during or after the conversation. Findings reveal no significant difference in learning gains, but immediate feedback enhanced user experience, leading to overall positive perceptions of the chatbot. Additionally, we explore users' perceptions of the chatbot's social role and personality, offering a roadmap for future enhancements. These results provide valuable insights into the potential of LLMs and chatbots for language learning.}, copyright = {Creative Commons Attribution 4.0 International}, langid = {english}, } ```APA 7
M. Kamelabad, A., Turano, B., Lundin, M., & Skantze, G. (2026). Personalized language learning with an LLM chatbot: Effects of immediate vs. delayed corrective feedback. *Frontiers in Education*, *11*, 1703664. https://doi.org/10.3389/feduc.2026.1703664IEEE
A. M. Kamelabad, B. Turano, M. Lundin, and G. Skantze, "Personalized Language Learning with an LLM Chatbot: Effects of Immediate vs. Delayed Corrective Feedback," *Frontiers in Education*, vol. 11, p. 1703664, Feb. 2026, doi: 10.3389/feduc.2026.1703664.ACM
Alireza M. Kamelabad, Beatrice Turano, Mattias Lundin, and Gabriel Skantze. 2026. Personalized Language Learning with an LLM Chatbot: Effects of Immediate vs. Delayed Corrective Feedback. *Frontiers in Education* 11 (Feb. 2026), 1703664. https://doi.org/10.3389/feduc.2026.1703664Repository
Access the code here: github.com/horotat/ChatBot2023
or
git clone https://github.com/horotat/ChatBot2023.git
Installation
To run the program properly you need an installation of conda (anaconda is suggested over miniconda). Once you have conda installed on your machine, follow the steps below.
Requirements
First, you need to create a conda virtual environment. The packages and their dependecies are checked and are functional in Ubuntu 22.04, Windows 11, and macOS ventura 13.0 arm64.
Create a conda virtual environment and install packages from the alice_env.yml file.
conda env create -f env.yml
After the installation
Spacy Language Model
We need to download the en language model for spacy. To do so, run the following command in the terminal.
python -m spacy download en
Set Environmental Variables
In order for the system to work properly, you need to set the following environemnt variables in the system.
$OPENAI_API_KEY: OpenAI API Key- Used in
webapp.py
- Used in
$FLASK_SECRET_KEY: A random Flask Secret Key to run the application.- Used in
webapp.py
- Used in
$ALLOWED_EMAILS_CSV: Link to a CSV file that contains list of allowed emails to signup coupled with the class and condition. (e.g.group, email, condition)- Used in
dataset.py
- Used in
Database
We used mongodb as our database. You can install it from here. After installation make sure that the mongod service is running.
Server Configurations
Deployment of the application is done using gunicorn and nginx.
We used these instructions to set a Password Authentication for the nginx server.
Running the Chatbot
To run the chatbot, first activate the conda virtual environment by:
conda activate alice_env
Then, run the webapp.py file with the following terminal command:
FLASK_APP=webapp.py FLASK_ENV=development flask run
Scenarios
The scenarios are feched from an online CSV. There is no automatic update for the scenarios. To update the scenarios, you need to run the scenarios_update.py file. To do so, run the following command in the terminal.
python scenarios_update.py