Topic Modeling in Python for Social Sciences

Handy Jupyter Notebooks, python scripts, mindmaps and scientific literature that I use in for Topic Modeling. Including text mining from PDF files, text preprocessing, Latent Dirichlet Allocation (LDA), hyperparameters grid search and Topic Modeling visualiation.

List of Notebooks

• textract-PDF: batch text mining from PDFs
• text-pre-process: tokenization, stopwords and stemming from spaCy
• gensim-pre-process: tokenization, stopwords and stemming from gensim
• gensim-topics: LDA in gensim
• gensim-LDA-Mallet: LDA in gensim using a MALLET wrapper
• gensim-optimal-topics: choose the number of topics to give the highest coherence and perplexity values in LDA models in gensim
• gensim-topic-analysis: exploratory analysis of topics from LDA models in gensim
• scikit-learn-best-topic-model: LDA in scikit-learn and optimal hyperparameter grid search; it also includes a pyLDAvis wrapper for scikit-learn for interactive LDA results visualization
• gensim-topic-modeling-visualization: multiple strategies to visualize the results of a gensim LDA model
• gensim-pyLDAvis: pyLDAvis wrapper for gensim for interactive LDA results visualization
• Stop-words Comparison: english and portuguese stopwords comparison from NLTK, spaCy and gensim

List of python function in utils

• clean_up: Clean up you text and generate list of words for each document using spaCy.
• compute_performance: Generate a model list for number of topics and compute c_v coherence and perplexity (if applicable) using either gensim.models.ldamodel.LdaModel, gensim.models.ldamulticore.LdaMulticore or gensim.models.wrappers.LdaMallet. It enables visualization of the optimal topic number for your model.
• graph_performance: Graphics for visualizing the output of a compute performance for a LDA model list.

Scientific Literature

A BibTeX file with the relevant collection of scientific literature (mainly articles published in peer-reviewed journals). It is comprised of computer science literature about topic modeling algorithms and procedures; and social science literature about best practices and uses of topic modeling along with interesting applications of topic modeling to textual data.

Mind Maps

I tend to aggregate and organize my knowledge about a subject in mind maps. There are some mind maps about topic modeling as PDF files with some content already referenced with the relevant literature.

Stopwords Comparison

As of June 15h 2019

Author

Jose Eduardo Storopoli