# The design and implementation of Language Learning Chatbot with XAI using Ontology and Transfer Learning

**Authors:** Nuobei Shi, Qin Zeng, Raymond Lee  
**arXiv:** [2009.13984](https://arxiv.org/abs/2009.13984)  
**Date:** 2020-09-29  
**Summary page:** https://www.summarizepaper.com/en/paper/2009.13984v1/

## Key points

- Authors propose a transfer learning-based English language learning chatbot
- Chatbot utilizes GPT-2 to generate output explained by an ontology graph
- Three levels of systematic English learning: phonetics, semantic, and free-style conversation
- Implementation includes an ontology graph and Explainable AI (XAI) techniques
- Language Learning agent integrates with WeChat as front-end and GPT-2 model as back-end
- Paper published in International Conference on NLP & Big Data (NLPD 2020)
- 19 pages long with 20 figures
- Primary categories: cs.AI, cs.CL, cs.LG

## Summary

In this paper titled "The design and implementation of Language Learning Chatbot with XAI using Ontology and Transfer Learning," authors Nuobei Shi, Qin Zeng, and Raymond Lee propose a transfer learning-based English language learning chatbot. The chatbot utilizes GPT-2 to generate output that can be explained by an ontology graph rooted in a fine-tuning dataset. The authors introduce three levels for systematic English learning: the phonetics level for speech recognition and pronunciation correction, the semantic level for specific domain conversation, and the simulation of free-style conversation in English - representing the highest level of language chatbot communication as a free-style conversation agent. To contribute academically, the authors implement an ontology graph to explain the performance of free-style conversation. They follow the concept of Explainable Artificial Intelligence (XAI) to visualize the connections of neural networks in bionics and explain the output sentence from the language model. From an implementation perspective, their Language Learning agent integrates a mini-program in WeChat as a front-end, while utilizing a fine-tuned GPT-2 model of transfer learning as a back-end to interpret responses through the ontology graph. This paper was published in the International Conference on NLP & Big Data (NLPD 2020) and is 19 pages long with 20 figures. It appears in CS & IT - CSCP 2020, pp. 305–323. The primary categories associated with this paper are cs.AI, cs.CL, and cs.LG. Overall, this research presents an innovative approach to English language learning through a transfer learning based chatbot that incorporates XAI techniques and an ontology graph for improved performance and explanation capabilities.