Skip to content


Version 0.14.1

Release date: 2 March, 2023


  • Use ChatGPT to create topic representations!:
  • Added delay_in_seconds parameter to OpenAI and Cohere representation models for throttling the API
    • Setting this between 5 and 10 allows for trial users now to use more easily without hitting RateLimitErrors
  • Fixed missing title param to visualization methods
  • Fixed probabilities not correctly aligning (#1024)
  • Fix typo in textgenerator @dkopljar27 in #1002


Within OpenAI's API, the ChatGPT models use a different API structure compared to the GPT-3 models. In order to use ChatGPT with BERTopic, we need to define the model and make sure to set chat=True:

import openai
from bertopic import BERTopic
from bertopic.representation import OpenAI

# Create your representation model
openai.api_key = MY_API_KEY
representation_model = OpenAI(model="gpt-3.5-turbo", delay_in_seconds=10, chat=True)

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)

Prompting with ChatGPT is very satisfying and can be customized in BERTopic by using certain tags. There are currently two tags, namely "[KEYWORDS]" and "[DOCUMENTS]". These tags indicate where in the prompt they are to be replaced with a topics keywords and top 4 most representative documents respectively. For example, if we have the following prompt:

prompt = """
I have topic that contains the following documents: \n[DOCUMENTS]
The topic is described by the following keywords: [KEYWORDS]

Based on the information above, extract a short topic label in the following format:
topic: <topic label>

then that will be rendered as follows and passed to OpenAI's API:

I have a topic that contains the following documents: 
- Our videos are also made possible by your support on
- If you want to help us make more videos, you can do so on or get one of our posters from our shop.
- If you want to help us make more videos, you can do so there.
- And if you want to support us in our endeavor to survive in the world of online video, and make more videos, you can do so on

The topic is described by the following keywords: videos video you our support want this us channel patreon make on we if facebook to patreoncom can for and more watch 

Based on the information above, extract a short topic label in the following format:
topic: <topic label>


Whenever you create a custom prompt, it is important to add

Based on the information above, extract a short topic label in the following format:
topic: <topic label>
at the end of your prompt as BERTopic extracts everything that comes after topic:. Having said that, if topic: is not in the output, then it will simply extract the entire response, so feel free to experiment with the prompts.

Version 0.14.0

Release date: 14 February, 2023


  • Fine-tune topic representations with bertopic.representation
    • Diverse range of models, including KeyBERT, MMR, POS, Transformers, OpenAI, and more!'
    • Create your own prompts for text generation models, like GPT3:
      • Use "[KEYWORDS]" and "[DOCUMENTS]" in the prompt to decide where the keywords and and set of representative documents need to be inserted.
    • Chain models to perform fine-grained fine-tuning
    • Create and customize your represention model
  • Improved the topic reduction technique when using nr_topics=int
  • Added title parameters for all graphs (#800)


  • Improve documentation (#837, #769, #954, #912, #911)
  • Bump pyyaml (#903)
  • Fix large number of representative docs (#965)
  • Prevent stochastisch behavior in .visualize_topics (#952)
  • Add custom labels parameter to .visualize_topics (#976)
  • Fix cuML HDBSCAN type checks by @FelSiq in #981

API Changes:

  • The diversity parameter was removed in favor of bertopic.representation.MaximalMarginalRelevance
  • The representation_model parameter was added to bertopic.BERTopic

Representation Models

Fine-tune the c-TF-IDF representation with a variety of models. Whether that is through a KeyBERT-Inspired model or GPT-3, the choice is up to you!


The algorithm follows some principles of KeyBERT but does some optimization in order to speed up inference. Usage is straightforward:


from bertopic.representation import KeyBERTInspired
from bertopic import BERTopic

# Create your representation model
representation_model = KeyBERTInspired()

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)



Our candidate topics, as extracted with c-TF-IDF, do not take into account a keyword's part of speech as extracting noun-phrases from all documents can be computationally quite expensive. Instead, we can leverage c-TF-IDF to perform part of speech on a subset of keywords and documents that best represent a topic.


from bertopic.representation import PartOfSpeech
from bertopic import BERTopic

# Create your representation model
representation_model = PartOfSpeech("en_core_web_sm")

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)



When we calculate the weights of keywords, we typically do not consider whether we already have similar keywords in our topic. Words like "car" and "cars" essentially represent the same information and often redundant. We can use MaximalMarginalRelevance to improve diversity of our candidate topics:


from bertopic.representation import MaximalMarginalRelevance
from bertopic import BERTopic

# Create your representation model
representation_model = MaximalMarginalRelevance(diversity=0.3)

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)

mmr (1)

Zero-Shot Classification

To perform zero-shot classification, we feed the model with the keywords as generated through c-TF-IDF and a set of candidate labels. If, for a certain topic, we find a similar enough label, then it is assigned. If not, then we keep the original c-TF-IDF keywords.

We use it in BERTopic as follows:

from bertopic.representation import ZeroShotClassification
from bertopic import BERTopic

# Create your representation model
candidate_topics = ["space and nasa", "bicycles", "sports"]
representation_model = ZeroShotClassification(candidate_topics, model="facebook/bart-large-mnli")

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)


Text Generation: 🤗 Transformers

Nearly every week, there are new and improved models released on the 🤗 Model Hub that, with some creativity, allow for further fine-tuning of our c-TF-IDF based topics. These models range from text generation to zero-classification. In BERTopic, wrappers around these methods are created as a way to support whatever might be released in the future.

Using a GPT-like model from the huggingface hub is rather straightforward:

from bertopic.representation import TextGeneration
from bertopic import BERTopic

# Create your representation model
representation_model = TextGeneration('gpt2')

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)


Text Generation: Cohere

Instead of using a language model from 🤗 transformers, we can use external APIs instead that do the work for you. Here, we can use Cohere to extract our topic labels from the candidate documents and keywords. To use this, you will need to install cohere first:

pip install cohere

Then, get yourself an API key and use Cohere's API as follows:

import cohere
from bertopic.representation import Cohere
from bertopic import BERTopic

# Create your representation model
co = cohere.Client(my_api_key)
representation_model = Cohere(co)

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)


Text Generation: OpenAI

Instead of using a language model from 🤗 transformers, we can use external APIs instead that do the work for you. Here, we can use OpenAI to extract our topic labels from the candidate documents and keywords. To use this, you will need to install openai first:

pip install openai

Then, get yourself an API key and use OpenAI's API as follows:

import openai
from bertopic.representation import OpenAI
from bertopic import BERTopic

# Create your representation model
openai.api_key = MY_API_KEY
representation_model = OpenAI()

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)


Text Generation: LangChain

Langchain is a package that helps users with chaining large language models. In BERTopic, we can leverage this package in order to more efficiently combine external knowledge. Here, this external knowledge are the most representative documents in each topic.

To use langchain, you will need to install the langchain package first. Additionally, you will need an underlying LLM to support langchain, like openai:

pip install langchain, openai

Then, you can create your chain as follows:

from langchain.chains.question_answering import load_qa_chain
from langchain.llms import OpenAI
chain = load_qa_chain(OpenAI(temperature=0, openai_api_key=MY_API_KEY), chain_type="stuff")

Finally, you can pass the chain to BERTopic as follows:

from bertopic.representation import LangChain

# Create your representation model
representation_model = LangChain(chain)

# Use the representation model in BERTopic on top of the default pipeline
topic_model = BERTopic(representation_model=representation_model)

Version 0.13.0

Release date: 4 January, 2023


  • Calculate topic distributions with .approximate_distribution regardless of the cluster model used
    • Generates topic distributions on a document- and token-levels
    • Can be used for any document regardless of its size!
  • Fully supervised BERTopic
    • You can now use a classification model for the clustering step instead to create a fully supervised topic model
  • Manual topic modeling
    • Generate topic representations from labels directly
    • Allows for skipping the embedding and clustering steps in order to go directly to the topic representation step
  • Reduce outliers with 4 different strategies using .reduce_outliers
  • Install BERTopic without SentenceTransformers for a lightweight package:
    • pip install --no-deps bertopic
    • pip install --upgrade numpy hdbscan umap-learn pandas scikit-learn tqdm plotly pyyaml
  • Get meta data of trained documents such as topics and probabilities using .get_document_info(docs)
  • Added more support for cuML's HDBSCAN
    • Calculate and predict probabilities during fit_transform and transform respectively
    • This should give a major speed-up when setting calculate_probabilities=True
  • More images to the documentation and a lot of changes/updates/clarifications
  • Get representative documents for non-HDBSCAN models by comparing document and topic c-TF-IDF representations
  • Sklearn Pipeline Embedder by @koaning in #791



Personally, I believe that documentation can be seen as a feature and is an often underestimated aspect of open-source. So I went a bit overboard😅... and created an animation about the three pillars of BERTopic using Manim. There are many other visualizations added, one of each variation of BERTopic, and many smaller changes.

Topic Distributions

The difficulty with a cluster-based topic modeling technique is that it does not directly consider that documents may contain multiple topics. With the new release, we can now model the distributions of topics! We even consider that a single word might be related to multiple topics. If a document is a mixture of topics, what is preventing a single word to be the same?

To do so, we approximate the distribution of topics in a document by calculating and summing the similarities of tokensets (achieved by applying a sliding window) with the topics:

# After fitting your model run the following for either your trained documents or even unseen documents
topic_distr, _ = topic_model.approximate_distribution(docs)

To calculate and visualize the topic distributions in a document on a token-level, we can run the following:

# We need to calculate the topic distributions on a token level
topic_distr, topic_token_distr = topic_model.approximate_distribution(docs, calculate_tokens=True)

# Create a visualization using a styled dataframe if Jinja2 is installed
df = topic_model.visualize_approximate_distribution(docs[0], topic_token_distr[0]); df

Supervised Topic Modeling

BERTopic now supports fully-supervised classification! Instead of using a clustering algorithm, like HDBSCAN, we can replace it with a classifier, like Logistic Regression:

from bertopic import BERTopic
from bertopic.dimensionality import BaseDimensionalityReduction
from sklearn.datasets import fetch_20newsgroups
from sklearn.linear_model import LogisticRegression

# Get labeled data
data= fetch_20newsgroups(subset='all',  remove=('headers', 'footers', 'quotes'))
docs = data['data']
y = data['target']

# Allows us to skip over the dimensionality reduction step
empty_dimensionality_model = BaseDimensionalityReduction()

# Create a classifier to be used instead of the cluster model
clf= LogisticRegression()

# Create a fully supervised BERTopic instance
topic_model= BERTopic(
topics, probs = topic_model.fit_transform(docs, y=y)

Manual Topic Modeling

When you already have a bunch of labels and simply want to extract topic representations from them, you might not need to actually learn how those can predicted. We can bypass the embeddings -> dimensionality reduction -> clustering steps and go straight to the c-TF-IDF representation of our labels:

from bertopic import BERTopic
from bertopic.backend import BaseEmbedder
from bertopic.cluster import BaseCluster
from bertopic.dimensionality import BaseDimensionalityReduction

# Prepare our empty sub-models and reduce frequent words while we are at it.
empty_embedding_model = BaseEmbedder()
empty_dimensionality_model = BaseDimensionalityReduction()
empty_cluster_model = BaseCluster()

# Fit BERTopic without actually performing any clustering
topic_model= BERTopic(
topics, probs = topic_model.fit_transform(docs, y=y)

Outlier Reduction

Outlier reduction is an frequently-discussed topic in BERTopic as its default cluster model, HDBSCAN, has a tendency to generate many outliers. This often helps in the topic representation steps, as we do not consider documents that are less relevant, but you might want to still assign those outliers to actual topics. In the modular philosophy of BERTopic, keeping training times in mind, it is now possible to perform outlier reduction after having trained your topic model. This allows for ease of iteration and prevents having to train BERTopic many times to find the parameters you are searching for. There are 4 different strategies that you can use, so make sure to check out the documentation!

Using it is rather straightforward:

new_topics = topic_model.reduce_outliers(docs, topics)

Lightweight BERTopic

The default embedding model in BERTopic is one of the amazing sentence-transformers models, namely "all-MiniLM-L6-v2". Although this model performs well out of the box, it typically needs a GPU to transform the documents into embeddings in a reasonable time. Moreover, the installation requires pytorch which often results in a rather large environment, memory-wise.

Fortunately, it is possible to install BERTopic without sentence-transformers and use it as a lightweight solution instead. The installation can be done as follows:

pip install --no-deps bertopic
pip install --upgrade numpy hdbscan umap-learn pandas scikit-learn tqdm plotly pyyaml

Then, we can use BERTopic without sentence-transformers as follows using a CPU-based embedding technique:

from sklearn.pipeline import make_pipeline
from sklearn.decomposition import TruncatedSVD
from sklearn.feature_extraction.text import TfidfVectorizer

pipe = make_pipeline(

topic_model = BERTopic(embedding_model=pipe)

As a result, the entire package and resulting model can be run quickly on the CPU and no GPU is necessary!

Document Information

Get information about the documents on which the topic was trained including the documents themselves, their respective topics, the name of each topic, the top n words of each topic, whether it is a representative document, and the probability of the clustering if the cluster model supports it. There are also options to include other metadata, such as the topic distributions or the x and y coordinates of the reduced embeddings that you can learn more about here.

To get the document info, you will only need to pass the documents on which the topic model was trained:

>>> topic_model.get_document_info(docs)

Document                               Topic    Name                        Top_n_words                     Probability    ...
I am sure some bashers of Pens...       0       0_game_team_games_season    game - team - games...          0.200010       ...
My brother is in the market for...      -1     -1_can_your_will_any         can - your - will...            0.420668       ...
Finally you said what you dream...      -1     -1_can_your_will_any         can - your - will...            0.807259       ...
Think! It is the SCSI card doing...     49     49_windows_drive_dos_file    windows - drive - docs...       0.071746       ...
1) I have an old Jasmine drive...       49     49_windows_drive_dos_file    windows - drive - docs...       0.038983       ...

Version 0.12.0

Release date: 5 September, 2022


  • Perform online/incremental topic modeling with .partial_fit
  • Expose c-TF-IDF model for customization with bertopic.vectorizers.ClassTfidfTransformer
    • The parameters bm25_weighting and reduce_frequent_words were added to potentially improve representations:
  • Expose attributes for easier access to internal data
  • Major changes to the Algorithm page of the documentation, which now contains three overviews of the algorithm:
  • Added an example of combining BERTopic with KeyBERT
  • Added many tests with the intention of making development a bit more stable


  • Fixed iteratively merging topics (#632 and (#648)
  • Fixed 0th topic not showing up in visualizations (#667)
  • Fixed lowercasing not being optional (#682)
  • Fixed spelling (#664 and (#673)
  • Fixed 0th topic not shown in .get_topic_info by @oxymor0n in #660
  • Fixed spelling by @domenicrosati in #674
  • Add custom labels and title options to barchart @leloykun in #694

Online/incremental topic modeling:

Online topic modeling (sometimes called "incremental topic modeling") is the ability to learn incrementally from a mini-batch of instances. Essentially, it is a way to update your topic model with data on which it was not trained on before. In Scikit-Learn, this technique is often modeled through a .partial_fit function, which is also used in BERTopic.

At a minimum, the cluster model needs to support a .partial_fit function in order to use this feature. The default HDBSCAN model will not work as it does not support online updating.

from sklearn.datasets import fetch_20newsgroups
from sklearn.cluster import MiniBatchKMeans
from sklearn.decomposition import IncrementalPCA
from bertopic.vectorizers import OnlineCountVectorizer
from bertopic import BERTopic

# Prepare documents
all_docs = fetch_20newsgroups(subset=subset,  remove=('headers', 'footers', 'quotes'))["data"]
doc_chunks = [all_docs[i:i+1000] for i in range(0, len(all_docs), 1000)]

# Prepare sub-models that support online learning
umap_model = IncrementalPCA(n_components=5)
cluster_model = MiniBatchKMeans(n_clusters=50, random_state=0)
vectorizer_model = OnlineCountVectorizer(stop_words="english", decay=.01)

topic_model = BERTopic(umap_model=umap_model,

# Incrementally fit the topic model by training on 1000 documents at a time
for docs in doc_chunks:

Only the topics for the most recent batch of documents are tracked. If you want to be using online topic modeling, not for a streaming setting but merely for low-memory use cases, then it is advised to also update the .topics_ attribute as variations such as hierarchical topic modeling will not work afterward:

# Incrementally fit the topic model by training on 1000 documents at a time and track the topics in each iteration
topics = []
for docs in doc_chunks:

topic_model.topics_ = topics


Explicitly define, use, and adjust the ClassTfidfTransformer with new parameters, bm25_weighting and reduce_frequent_words, to potentially improve the topic representation:

from bertopic import BERTopic
from bertopic.vectorizers import ClassTfidfTransformer

ctfidf_model = ClassTfidfTransformer(bm25_weighting=True)
topic_model = BERTopic(ctfidf_model=ctfidf_model)


After having fitted your BERTopic instance, you can use the following attributes to have quick access to certain information, such as the topic assignment for each document in topic_model.topics_.

Attribute Type Description
topics_ List[int] The topics that are generated for each document after training or updating the topic model. The most recent topics are tracked.
probabilities_ List[float] The probability of the assigned topic per document. These are only calculated if a HDBSCAN model is used for the clustering step. When calculate_probabilities=True, then it is the probabilities of all topics per document.
topic_sizes_ Mapping[int, int] The size of each topic.
topic_mapper_ TopicMapper A class for tracking topics and their mappings anytime they are merged, reduced, added, or removed.
topic_representations_ Mapping[int, Tuple[int, float]] The top n terms per topic and their respective c-TF-IDF values.
c_tf_idf_ csr_matrix The topic-term matrix as calculated through c-TF-IDF. To access its respective words, run .vectorizer_model.get_feature_names() or .vectorizer_model.get_feature_names_out()
topic_labels_ Mapping[int, str] The default labels for each topic.
custom_labels_ List[str] Custom labels for each topic as generated through .set_topic_labels.
topic_embeddings_ np.ndarray The embeddings for each topic. It is calculated by taking the weighted average of word embeddings in a topic based on their c-TF-IDF values.
representative_docs_ Mapping[int, str] The representative documents for each topic if HDBSCAN is used.

Version 0.11.0

Release date: 11 July, 2022


hierarchical_topics = topic_model.hierarchical_topics(docs, topics) 
tree = topic_model.get_topic_tree(hierarchical_topics)
# Use input embeddings
topic_model.visualize_documents(docs, embeddings=embeddings)

# or use 2D reduced embeddings through a method of your own (e.g., PCA, t-SNE, UMAP, etc.)
reduced_embeddings = UMAP(n_neighbors=10, n_components=2, min_dist=0.0, metric='cosine').fit_transform(embeddings)
topic_model.visualize_documents(docs, reduced_embeddings=reduced_embeddings)
# Run the visualization with the original embeddings
topic_model.visualize_hierarchical_documents(docs, hierarchical_topics, embeddings=embeddings)

# Or, if you have reduced the original embeddings already which speed things up quite a bit:
reduced_embeddings = UMAP(n_neighbors=10, n_components=2, min_dist=0.0, metric='cosine').fit_transform(embeddings)
topic_model.visualize_hierarchical_documents(docs, hierarchical_topics, reduced_embeddings=reduced_embeddings)
  • Create custom labels to the topics throughout most visualizations
# Generate topic labels
topic_labels = topic_model.generate_topic_labels(nr_words=3, topic_prefix=False, word_length=10, separator=", ")

# Set them internally in BERTopic
# Merge topics 1, 2, and 3
topics_to_merge = [1, 2, 3]
topic_model.merge_topics(docs, topics, topics_to_merge)

# Merge topics 1 and 2, and separately merge topics 3 and 4
topics_to_merge = [[1, 2], [3, 4]]
topic_model.merge_topics(docs, topics, topics_to_merge)


  • Fix support for k-Means in .visualize_heatmap (#532)
  • Fix missing topic 0 in .visualize_topics (#533)
  • Fix inconsistencies in .get_topic_info (#572) and (#581)
  • Add optimal_ordering parameter to .visualize_hierarchy by @rafaelvalero in #390
  • Fix RuntimeError when used as sklearn estimator by @simonfelding in #448
  • Fix typo in visualization documentation by @dwhdai in #475
  • Fix typo in docstrings by @xwwwwww in #549
  • Support higher Flair versions

Version 0.10.0

Release date: 30 April, 2022


  • Use any dimensionality reduction technique instead of UMAP:
from bertopic import BERTopic
from sklearn.decomposition import PCA

dim_model = PCA(n_components=5)
topic_model = BERTopic(umap_model=dim_model)
  • Use any clustering technique instead of HDBSCAN:
from bertopic import BERTopic
from sklearn.cluster import KMeans

cluster_model = KMeans(n_clusters=50)
topic_model = BERTopic(hdbscan_model=cluster_model)


  • Add a CountVectorizer page with tips and tricks on how to create topic representations that fit your use case
  • Added pages on how to use other dimensionality reduction and clustering algorithms
  • Additional instructions on how to reduce outliers in the FAQ:
import numpy as np
probability_threshold = 0.01
new_topics = [np.argmax(prob) if max(prob) >= probability_threshold else -1 for prob in probs] 


  • Fixed None being returned for probabilities when transforming unseen documents
  • Replaced all instances of arg: with Arguments: for consistency
  • Before saving a fitted BERTopic instance, we remove the stopwords in the fitted CountVectorizer model as it can get quite large due to the number of words that end in stopwords if min_df is set to a value larger than 1
  • Set "hdbscan>=0.8.28" to prevent numpy issues
  • Although this was already fixed by the new release of HDBSCAN, it is technically still possible to install 0.8.27 with BERTopic which leads to these numpy issues
  • Update gensim dependency to >=4.0.0 (#371)
  • Fix topic 0 not appearing in visualizations (#472)
  • Fix (#506)
  • Fix (#429)
  • Fix typo in DTM documentation by @hp0404 in #386

Version 0.9.4

Release date: 14 December, 2021

A number of fixes, documentation updates, and small features:

  • Expose diversity parameter
    • Use BERTopic(diversity=0.1) to change how diverse the words in a topic representation are (ranges from 0 to 1)
  • Improve stability of topic reduction by only computing the cosine similarity within c-TF-IDF and not the topic embeddings
  • Added property to c-TF-IDF that all IDF values should be positive (#351)
  • Improve stability of .visualize_barchart() and .visualize_hierarchy()
  • Major documentation overhaul (mkdocs, tutorials, FAQ, images, etc. ) (#330)
  • Drop python 3.6 (#333)
  • Relax plotly dependency (#88)
  • Additional logging for .transform (#356)

Version 0.9.3

Release date: 17 October, 2021

  • Fix #282
    • As it turns out the old implementation of topic mapping was still found in the transform function
  • Fix #285
    • Fix getting all representative docs
  • Fix #288
    • A recent issue with the package pyyaml that can be found in Google Colab

Version 0.9.2

Release date: 12 October, 2021

A release focused on algorithmic optimization and fixing several issues:


  • Update the non-multilingual paraphrase- models to the all- models due to improved performance
  • Reduce necessary RAM in c-TF-IDF top 30 word extraction


  • Fix topic mapping
    • When reducing the number of topics, these need to be mapped to the correct input/output which had some issues in the previous version
    • A new class was created as a way to track these mappings regardless of how many times they were executed
    • In other words, you can iteratively reduce the number of topics after training the model without the need to continuously train the model
  • Fix typo in embeddings page (#200)
  • Fix link in README (#233)
  • Fix documentation .visualize_term_rank() (#253)
  • Fix getting correct representative docs (#258)
  • Update memory FAQ with HDBSCAN pr

Version 0.9.1

Release date: 1 September, 2021

A release focused on fixing several issues:


  • Fix TypeError when auto-reducing topics (#210)
  • Fix mapping representative docs when reducing topics (#208)
  • Fix visualization issues with probabilities (#205)
  • Fix missing normalize_frequency param in plots (#213)

Version 0.9.0

Release date: 9 August, 2021


  • Implemented a Guided BERTopic -> Use seeds to steer the Topic Modeling
  • Get the most representative documents per topic: topic_model.get_representative_docs(topic=1)
    • This allows users to see which documents are good representations of a topic and better understand the topics that were created
  • Added normalize_frequency parameter to visualize_topics_per_class and visualize_topics_over_time in order to better compare the relative topic frequencies between topics
  • Return flat probabilities as default, only calculate the probabilities of all topics per document if calculate_probabilities is True
  • Added several FAQs


  • Fix loading pre-trained BERTopic model
  • Fix mapping of probabilities
  • Fix #190

Guided BERTopic:

Guided BERTopic works in two ways:

First, we create embeddings for each seeded topics by joining them and passing them through the document embedder. These embeddings will be compared with the existing document embeddings through cosine similarity and assigned a label. If the document is most similar to a seeded topic, then it will get that topic's label. If it is most similar to the average document embedding, it will get the -1 label. These labels are then passed through UMAP to create a semi-supervised approach that should nudge the topic creation to the seeded topics.

Second, we take all words in seed_topic_list and assign them a multiplier larger than 1. Those multipliers will be used to increase the IDF values of the words across all topics thereby increasing the likelihood that a seeded topic word will appear in a topic. This does, however, also increase the chance of an irrelevant topic having unrelated words. In practice, this should not be an issue since the IDF value is likely to remain low regardless of the multiplier. The multiplier is now a fixed value but may change to something more elegant, like taking the distribution of IDF values and its position into account when defining the multiplier.

seed_topic_list = [["company", "billion", "quarter", "shrs", "earnings"],
                   ["acquisition", "procurement", "merge"],
                   ["exchange", "currency", "trading", "rate", "euro"],
                   ["grain", "wheat", "corn"],
                   ["coffee", "cocoa"],
                   ["natural", "gas", "oil", "fuel", "products", "petrol"]]

topic_model = BERTopic(seed_topic_list=seed_topic_list)
topics, probs = topic_model.fit_transform(docs)

Version 0.8.1

Release date: 8 June, 2021


  • Improved models:
    • For English documents the default is now: "paraphrase-MiniLM-L6-v2"
    • For Non-English or multi-lingual documents the default is now: "paraphrase-multilingual-MiniLM-L12-v2"
    • Both models show not only great performance but are much faster!
  • Add interactive visualizations to the plotting API documentation

For better performance, please use the following models:

  • English: "paraphrase-mpnet-base-v2"
  • Non-English or multi-lingual: "paraphrase-multilingual-mpnet-base-v2"


  • Improved unit testing for more stability
  • Set transformers version for Flair

Version 0.8.0

Release date: 31 May, 2021


  • Additional visualizations:
    • Topic Hierarchy: topic_model.visualize_hierarchy()
    • Topic Similarity Heatmap: topic_model.visualize_heatmap()
    • Topic Representation Barchart: topic_model.visualize_barchart()
    • Term Score Decline: topic_model.visualize_term_rank()
  • Created bertopic.plotting library to easily extend visualizations
  • Improved automatic topic reduction by using HDBSCAN to detect similar topics
  • Sort topic ids by their frequency. -1 is the outlier class and contains typically the most documents. After that 0 is the largest topic, 1 the second largest, etc.


  • Fix typo #113, #117
  • Fix #121 by removing these two lines
  • Fix mapping of topics after reduction (it now excludes 0) (#103)

Version 0.7.0

Release date: 26 April, 2021

The two main features are (semi-)supervised topic modeling and several backends to use instead of Flair and SentenceTransformers!


  • (semi-)supervised topic modeling by leveraging supervised options in UMAP
    •, y=target_classes)
  • Backends:
    • Added Spacy, Gensim, USE (TFHub)
    • Use a different backend for document embeddings and word embeddings
    • Create your own backends with bertopic.backend.BaseEmbedder
    • Click here for an overview of all new backends
  • Calculate and visualize topics per class
    • Calculate: topics_per_class = topic_model.topics_per_class(docs, topics, classes)
    • Visualize: topic_model.visualize_topics_per_class(topics_per_class)
  • Several tutorials were updated and added:
Name Link
Topic Modeling with BERTopic Open In Colab
(Custom) Embedding Models in BERTopic Open In Colab
Advanced Customization in BERTopic Open In Colab
(semi-)Supervised Topic Modeling with BERTopic Open In Colab
Dynamic Topic Modeling with Trump's Tweets Open In Colab


  • Fixed issues with Torch req
  • Prevent saving term frequency matrix in CTFIDF class
  • Fixed DTM not working when reducing topics (#96)
  • Moved visualization dependencies to base BERTopic
    • pip install bertopic[visualization] becomes pip install bertopic
  • Allow precomputed embeddings in bertopic.find_topics() (#79):
model = BERTopic(embedding_model=my_embedding_model), my_precomputed_embeddings)

Version 0.6.0

Release date: 1 March, 2021


  • DTM: Added a basic dynamic topic modeling technique based on the global c-TF-IDF representation
    • model.topics_over_time(docs, timestamps, global_tuning=True)
  • DTM: Option to evolve topics based on t-1 c-TF-IDF representation which results in evolving topics over time
    • Only uses topics at t-1 and skips evolution if there is a gap
    • model.topics_over_time(docs, timestamps, evolution_tuning=True)
  • DTM: Function to visualize topics over time
    • model.visualize_topics_over_time(topics_over_time)
  • DTM: Add binning of timestamps
    • model.topics_over_time(docs, timestamps, nr_bins=10)
  • Add function get general information about topics (id, frequency, name, etc.)
    • get_topic_info()
  • Improved stability of c-TF-IDF by taking the average number of words across all topics instead of the number of documents


  • _map_probabilities() does not take into account that there is no probability of the outlier class and the probabilities are mutated instead of copied (#63, #64)

Version 0.5.0

Release date: 8 Februari, 2021


  • Add Flair to allow for more (custom) token/document embeddings, including 🤗 transformers
  • Option to use custom UMAP, HDBSCAN, and CountVectorizer
  • Added low_memory parameter to reduce memory during computation
  • Improved verbosity (shows progress bar)
  • Return the figure of visualize_topics()
  • Expose all parameters with a single function: get_params()


  • To simplify the API, the parameters stop_words and n_neighbors were removed. These can still be used when a custom UMAP or CountVectorizer is used.
  • Set calculate_probabilities to False as a default. Calculating probabilities with HDBSCAN significantly increases computation time and memory usage. Better to remove calculating probabilities or only allow it by manually turning this on.
  • Use the newest version of sentence-transformers as it speeds ups encoding significantly

Version 0.4.2

Release date: 10 Januari, 2021


  • Selecting embedding_model did not work when language was also used. This led to the user needing to set language to None before being able to use embedding_model. Fixed by using embedding_model when language is used (as a default parameter).

Version 0.4.1

Release date: 07 Januari, 2021


  • Simple fix by lowering the languages variable to match the lowered input language.

Version 0.4.0

Release date: 21 December, 2020


  • Visualize Topics similar to LDAvis
  • Added option to reduce topics after training
  • Added option to update topic representation after training
  • Added option to search topics using a search term
  • Significantly improved the stability of generating clusters
  • Finetune the topic words by selecting the most coherent words with the highest c-TF-IDF values
  • More extensive tutorials in the documentation

Notable Changes:

  • Option to select language instead of sentence-transformers models to minimize the complexity of using BERTopic
  • Improved logging (remove duplicates)
  • Check if BERTopic is fitted
  • Added TF-IDF as an embedder instead of transformer models (see tutorial)
  • Numpy for Python 3.6 will be dropped and was therefore removed from the workflow.
  • Preprocess text before passing it through c-TF-IDF
  • Merged get_topics_freq() with get_topic_freq()


  • Fix error handling topic probabilities

Version 0.3.2

Release date: 16 November, 2020


  • Fixed a bug with the topic reduction method that seems to reduce the number of topics but not to the nr_topics as defined in the class. Since this was, to a certain extend, breaking the topic reduction method a new release was necessary.

Version 0.3.1

Release date: 4 November, 2020


  • Adding the option to use custom embeddings or embeddings that you generated beforehand with whatever package you'd like to use. This allows users to further customize BERTopic to their liking.

Version 0.3.0

Release date: 29 October, 2020


  • transform() and fit_transform() now also return the topic probability distributions
  • Added visualize_distribution() which visualizes the topic probability distribution for a single document

Version 0.2.2

Release date: 17 October, 2020


  • Fixed n_gram_range not being used
  • Added option for using stopwords

Version 0.2.1

Release date: 11 October, 2020


  • Improved the calculation of the class-based TF-IDF procedure by limiting the calculation to sparse matrices. This prevents out-of-memory problems when faced with large datasets.

Version 0.2.0

Release date: 11 October, 2020


  • Changed c-TF-IDF procedure such that it implements a version of scikit-learns procedure. This should also speed up the calculation of the sparse matrix and prevent memory errors.
  • Added automated unit tests

Version 0.1.2

Release date: 1 October, 2020


  • When transforming new documents, self.mapped_topics seemed to be missing. Added to the init.

Version 0.1.1

Release date: 24 September, 2020


  • Fixed requirements --> Issue with pytorch
  • Update documentation

Version 0.1.0

Release date: 24 September, 2020


  • First release of BERTopic
  • Added parameters for UMAP and HDBSCAN
  • Option to choose sentence-transformer model
  • Method for transforming unseen documents
  • Save and load trained models (UMAP and HDBSCAN)
  • Extract topics and their sizes

Notable Changes:

  • Optimized c-TF-IDF
  • Improved documentation
  • Improved topic reduction