pyBTM

Python wrapper for Biterm Model algorithm (Yan et al., 2013)
BTM learning and inference algorithm source code can be found here: https://github.com/xiaohuiyan/BTM
 

Demo

See demo code in demo/ for more usage information

Data

Each corpus is expected to be in a single plain text file Words should be separated by spaces ' ' Documents should be separated by new lines '\n'

Initialization

Initialize a new model:

btm = pyBTM.BTM(K=k, input_path=path_to_data, alpha=alpha, beta=beta, niter=niter)

Encode data into bag-of-words format

btm.index_documents()

Learning

Learn topics:

btm.learn_topics()

Inference

Infer topics for each document:

btm.infer_documents()

Evaluate Performance

Evaluate performance of the model using CV and UMass coherence scores:

coherence_models = btm.build_coherence_model(measures=['c_v', 'u_mass'])
print([*map(lambda e : (e[0], e[1][1]), cm_dict.items())])

View Results

Show top 10 words for each topic:

topics = btm.get_topics(include_likelihood=False, use_words=True, L=10)

Extra model information

print(btm.info())

References

Yan, X., Guo, J., Lan, Y., & Cheng, X. (2013). A Biterm Topic Model for Short Texts.