---
title: CHB BERTopic V3
emoji: 🔬
colorFrom: blue
colorTo: indigo
sdk: gradio
sdk_version: "5.31.0"
app_file: app.py
pinned: false
python_version: "3.11"
short_description: BERTopic analysis app for Computers in Human Behavior
---

# CHB BERTopic V3

This Space runs the final CHB BERTopic app for MDM Assignment-4.

## What is included

- Gradio app entrypoint: `app.py`
- Pipeline code: `tools.py`
- Bundled corpus: `data/ComputersinHumanBehavior_TopicModelling_Export_7553_for_app.csv`
- Local Space output folder: `outputs`

## Runtime

The pipeline uses SPECTER2 embeddings, UMAP, HDBSCAN, BERTopic, and PAJAIS taxonomy mapping.

LLM-assisted labeling, theme grouping, taxonomy reasoning, and narrative generation use Hugging Face inference. Add `HF_TOKEN` as a Space secret before running those phases. If the token is missing or inference fails, the app uses deterministic fallback logic where the pipeline supports it.

## Run Order

Use the default input path shown in the app, then run the phases in order:

1. Load corpus
2. Discover abstract and title topics
3. Label topics
4. Consolidate themes
5. Map themes to PAJAIS
6. Generate comparison
7. Generate narrative

## Author

Prasad Vijay Gade (2024301006) | MDM Assignment-4