Our Research
Rigorous, peer-reviewed research is at the core of Reliable AI.
DANTE: Deductive Content Analysis Using Text Embeddings
Traditional qualitative content analysis of text is labor-intensive. We introduce DANTE, a reliable and transparent few-shot tool that makes text embedding-based qualitative analysis more accessible. It enables researchers to perform large-scale deductive content analysis without extensive programming skills.
Scalable and consistent few-shot classification of survey responses using text embeddings
We introduce a text embedding-based classification framework for efficient qualitative analysis of open-ended survey responses. Requiring only a handful of examples, this method integrates seamlessly into existing workflows and achieves performance comparable to expert human coders. It enables scalable, consistent, and audit-friendly analysis of thousands of responses.
Using Text Embeddings for Qualitative Analysis at Scale
We propose a novel technique for deductive qualitative data analysis using text embeddings. By representing text in a high-dimensional meaning space, we can quantify differences and model topics more flexibly than traditional methods. Validate against established datasets, our approach recovers key trends and offers a scalable solution for large text corpora.