Generative AI and Language Teaching

Anthropomorphism

You can anthropomorphize AI in various ways and create interactive learning activities.

• AI as a Japanese literature teacher
• AI as a Japanese language teacher
• AI as a Japanese language learner

Prompt templates from Google Docs 

• Translation
• Quiz Generation
• Create a vocabulary list
• Translate into easy English
• Mollick and Mollick: Use LLMs as a language consultant
• AI as a job interviewer: Japanese language instructor position

How to generate Canvas quiz based on the output file(s) from LLMs (Thanks to Laurence Totino)

Quick Summary

• Both the SDSU tool and NYIT tool require far less formatting and less effort to copy text over from ChatGPT compared to the Kansas State tool, which Sam Medeiros (IS&T & Canvas team) had tested in February 2024 (see very end of doc for this email from Sheryl/Sam).
  • These two tools were recommended by a Canvas R1 Peer from UVA who had done research on text > quiz converters. The peer also mentioned Respondus and GetMarked, which MIT does not have and is not pursuing at this time. We were not able to contact a developer/maintainer of the Kansas State tool.
• Both tools successfully support non-Roman characters throughout the process of converting a text file to QTI and when that QTI is imported to Canvas/the text becomes Quiz questions.
• The SDSU tool offers a bit more flexibility in formatting the questions, but the QTI conversion takes place on a site that is technically not “secure” (“use at own risk” is what UVA peer told me).
• Caveat for both of these tools: my testing may not capture the complexity of what Takako is trying to do. I also only tested with MC questions – I did not yet do a test of every type of supported questions type, like true/false and multiple answers (though I am happy to do further testing!).

Tips:

1. You can use this prompt to create a reading comprehension quiz:https://aikawa.mit.edu/files/2025/01/読解クイズ-sample.prompt.docx
2. The output from GPT4-o looks like this: https://aikawa.mit.edu/files/2025/01/readingL6-1.txt
3. You can use the above output file to create a QTI.zip file using one of the above tools.

Prompt samples

• Classification task: This ACTFL-rubric prompt classifies students’ essays based on the ACTFL rubrics.
• Easy Japanese: This easy-Japanese prompt translates difficult text(s) into easy ones. (Prompt is in English and the text is Japanese)
• Easy Chinese: This easy-Chinese prompt translates difficut Chinese text(s) into easy ones.  
• Emulate Human-like Feedback: This emulate-human-like feedback prompt tries to do an error analysis of student’s writing.  It provides: corrected sentencence and feedback from GPT. Prompt is in English and the error sentences are Japanese.)
• Multiple-choice Reading Comprehension Quiz Prompt: This reading comprehension quiz prompt generates a multiple-choice reading comprehension quiz from a given text.
• Dialogue Generation Prompt: This dialogue generation prompt generates dialogues using given vocabularies with specific styles.
• Vocabulary List Creator Prompt: This voc-list prompt creats a vocabulary list from a newspaper article (Yomiuri-newspaper).
• Kanji Reading Quiz Prompt: This kanji reading quiz prompt generates a multiple-choice kanji reading quiz based on a given text.

Spring 2022 (UROP students: Stephen Wilson, Victor Luo, and Kenny Chen, “Development of a Multi-lingual Chatbot as a Pedagogical Approach to Language Education.” 

AI Chatbot: User inputs are translated to English with DeepL, OpenAI’s GPT-3 produces an AI response, and then the response is translated from English with DeepL. This model speaks very naturally, but some information may be lost in translation, and outputs are often too difficult for language learners.