I'm trying to extract API integration parameters like Authorization headers, query params, and request body fields from API documentation. This is essentially a custom NER task.
I’ve experimented with LLaMA 3.1 and Mistral using Ollama, but both run too slowly on my local machine and even on Google Colab.
The challenge is that I don’t have a labeled dataset for this task. So I'm looking for a lighter or faster model (LLM or otherwise) that’s compatible with Ollama or suggestions for classic NLP/ML techniques (e.g. rule-based methods, CRFs, etc.) that can work without a labeled dataset
Any tips on using fewshot prompts effectively for this kind of extraction task
Has anyone faced a similar use case? I’d really appreciate any ideas, tools, or models that could help.
Thanks!
<script>tag within the<head>of the page at demo.360alumni.com/api/doc