MaartenGr/BERTopic

Project overview

BERTopic's modular architecture lets users freely swap embedding backends, dimensionality reduction, clustering, and vectorizer components while producing highly interpretable topics through c-TF-IDF.

Project type

RAG · Data Processing

Use cases

Data Analysis

Deployment

Refer to project documentation

License

MIT

Best for

• Data teams and researchers who need interpretable, customizable topic modeling with the ability to swap embedding, clustering, and vectorizer components.

Key capabilities

• Extracts topics from documents using clustering and class-based TF-IDF.
• Allows guiding the topic modeling process using a seed topic list.
• Visualizes topics, documents, hierarchies, term scores, and topic distributions over time.

Limitations and risks

• The large number of available functions can quickly become overwhelming for new users.

Getting started

• Install the library via pip (e.g., `pip install bertopic`), import and instantiate BERTopic, then fit the model using a sample dataset such as the 20 newsgroups corpus.

Evidence and sources

• README: [](https://pypi.org/project/bertopic/)
• README: BERTopic is a topic modeling technique that leverages 🤗 transformers and c-TF-IDF to create dense clusters allowing for easily interpretable topics whilst keeping important words…
• README: pip install bertopic
• README: pip install bertopic[flair,gensim,spacy,use]
• README: topics, probs = topic_model.fit_transform(docs)

Reviewed at: 2026-07-15T11:24:56Z