A library for topic modeling and browsing
Switch branches/tags
Nothing to show
Clone or download
Fetching latest commit…
Cannot retrieve the latest commit at this time.



TOM (TOpic Modeling) is a Python 3 library for topic modeling and browsing, licensed under the MIT license. Its objective is to allow for an efficient analysis of a text corpus from start to finish, via the discovery of latent topics. To this end, TOM features functions for preparing and vectorizing a text corpus. It also offers a common interface for two topic models (namely LDA using either variational inference or Gibbs sampling, and NMF using alternating least-square with a projected gradient method), and implements three state-of-the-art methods for estimating the optimal number of topics to model a corpus. What is more, TOM constructs an interactive Web-based browser that makes it easy to explore a topic model and the related corpus.


Vector space modeling

  • Feature selection based on word frequency
  • Weighting
    • tf
    • tf-idf

Topic modeling

  • Latent Dirichlet Allocation
    • Standard variational Bayesian inference (Latent Dirichlet Allocation. Blei et al, 2003)
    • Online variational Bayesian inference (Online learning for Latent Dirichlet Allocation. Hoffman et al, 2010)
    • Collapsed Gibbs sampling (Finding scientific topics. Griffiths & Steyvers, 2004)
  • Non-negative Matrix Factorization (NMF)
    • Alternating least-square with a projected gradient method (Projected gradient methods for non-negative matrix factorization. Lin, 2007)

Estimating the optimal number of topics

  • Stability analysis (How many topics? Stability analysis for topic models. Greene et al, 2014)
  • Spectral analysis (On finding the natural number of topics with latent dirichlet allocation: Some observations. Arun et al, 2010)
  • Consensus-based analysis (Metagenes and molecular pattern discovery using matrix factorization. Brunet et al, 2004)


We recommend you to install Anaconda (https://www.continuum.io) which will automatically install most of the required dependencies (i.e. pandas, numpy, scipy, scikit-learn, matplotlib, flask). You should then install the lda module (pip install lda). Eventually, clone or download this repo and run the following command:

python setup.py install

Or install it directly either from PyPi:

pip install tom_lib

or from Conda Cloud:

conda install -c psoriano tom_lib


We provide two sample programs, topic_model.py (which shows you how to load and prepare a corpus, estimate the optimal number of topics, infer the topic model and then manipulate it) and topic_model_browser.py (which shows you how to generate a topic model browser to explore a corpus), to help you get started using TOM.

Load and prepare a textual corpus

A corpus is a TSV (tab separated values) file describing documents, formatted as follows: a document per line, with at least three columns, namely id (a number), title (a short text) and text (the full content of the document), e.g.:

id	title	text
1	Document 1's title	This is the full content of document 1.
2	Document 2's title	This is the full content of document 2.

The following code snippet shows how to load a corpus of French documents and vectorize them using tf-idf with unigrams.

corpus = Corpus(source_file_path='input/raw_corpus.csv',
print('corpus size:', corpus.size)
print('vocabulary size:', len(corpus.vocabulary))
print('Vector representation of document 0:\n', corpus.vector_for_document(0))

Instantiate a topic model and infer topics

It is possible to instantiate a NMF or LDA object then infer topics.


topic_model = NonNegativeMatrixFactorization(corpus)

LDA (using either the standard variational Bayesian inference or Gibbs sampling):

topic_model = LatentDirichletAllocation(corpus)
topic_model.infer_topics(num_topics=15, algorithm='variational')
topic_model = LatentDirichletAllocation(corpus)
topic_model.infer_topics(num_topics=15, algorithm='gibbs')

Instantiate a topic model and estimate the optimal number of topics

Here we instantiate a NMF object, then generate plots with the three metrics for estimating the optimal number of topics.

topic_model = NonNegativeMatrixFactorization(corpus)
viz = Visualization(topic_model)
                       tao=10, step=1, 

Save/load a topic model

To allow reusing previously learned topics models, TOM can save them on disk, as shown below.

utils.save_topic_model(topic_model, 'output/NMF_15topics.tom')
topic_model = utils.load_topic_model('output/NMF_15topics.tom')

Print information about a topic model

This code excerpt illustrates how one can manipulate a topic model, e.g. get the topic distribution for a document or the word distribution for a topic.

print('\nTopic distribution for document 0:',
print('\nMost likely topic for document 0:',
print('\nFrequency of topics:',
print('\nTop 10 most relevant words for topic 2:',
      topic_model.top_words(2, 10))

Topic model browser: screenshots

Topic cloud

Topic details

Document details