Dear INLP students,
I have improved the tutor available for the iNLP introduction lecture and created one for the evaluation lecture.
They are available at:
- Introduction lecture:
https://chatgpt.com/g/g-AyzsbUz4B-inlp-tutor-intro
- Evaluation lecture:
https://chatgpt.com/g/g-tTpxFDl7k-inlp-tutor-eval
Both have been build on the automated transcripts of the lecture videos and then tailored with specific instructions.
You can use then in many different ways, for example:
1. Ask them to generate questions to check your understanding of the lecture content
--> "Give me a question on evaluation metrics"
2. Ask them for clarification on specific topics
--> "Tell me more about the kappa score?"
3. Ask them to show the parts ("chunks") of the transcripts they rely on to state something
--> "Show me chunk [23]" or simply "Show me [23]"
4. Ask them to solve exercises
--> "Assume that the 2x2 confusion between annotator 1 and annotator 2 has 2 and 7 on its diagonal and 1 and 0 on its anti-diagonal, what is the corresponding Kappa score?"
5. Ask them what are the main topics covered in the lecture
--> "Which are the topics covered by the lecture? Just give me the topics, no explanation."
--> "Could you order these topics by decreasing order of importance for an NLP course?"
and probably many more!...
With each of the tutors, simply click on the "Start" conversational starter and go...
I will be illustrating the tutor on Evaluation this morning and your feedback is more than welcome so that I can continue to improve these tools.
Also remember the caveats that Jean-Cedric already stressed:- we do not by any mean want to promote ChatGPT nor any OpenAI products; the choice of that particular LLM was only for practical purposes and convenience for these first experiments;
- a free access to ChatGPT is enough to test this LLM -- there is absolutely no need for any kind of paid account;
- we do NOT guarantee the quality of the
questions/answers produced by the system, which should therefore be
considered as a simple prototype, for which your feedback is very useful.