Show HN: Cognitum Text Classifier for Social Science

3 points

a year ago

I'm building an AI text classifier python library targeting social scientists and academics who are unaware of the classification capabilities of SOTA language models.

A few things included:

- Intended for classifying free-text poll responses using LLMs - Support for both single and multi-label classification with confidence scores - Built-in evaluation against ground truth data and random sampling capabilities - Currently supports Llama and OpenAI models, with easy dataset handling - Works on Apple Silicon, planning to add support for more systems and models soon - Built on top of LMQL (https://lmql.ai/)

This is an early version with lots of rough edges. Hope this helps anyone working with survey analysis, and let me know if you have any feature requests or ideas for improvement.