Insight Extraction from Social Work Reflection
This project is designed to support adult social workers by analysing their reflective comments. These narratives often contain valuable signals about practice challenges, emotional wellbeing, and areas where additional support may be needed.
We are developing a lightweight, open-source Python package called themex, which leverages large language models (LLMs) to:
- Automatically extract key themes from reflective comments
- Identify the emotional tone of each theme (positive, neutral, or negative)
- Help managers or researchers spot patterns and prioritise who might need support
By automating this process, the project aims to promote early intervention, improve workforce wellbeing, and strengthen reflective practice across adult social care.
Currently, themex supports:
- Run any prompt that returns structured JSON
- Integrate with models via LangChain, Azure OpenAI, or HuggingFace Transformers
- Output reusable and interpretable results for downstream use
You can explore the tool here:
