Open Source Packages

We build and maintain a suite of open source Python libraries and tools for computational text analysis. The packages are designed as a coherent ecosystem — each layer building on the one below.

chronowords

chronowords logo

Temporal word embedding analysis. Detect how word meanings shift across time in large text corpora using memory-efficient PPMI-based embeddings, NMF topic modeling, and Procrustes alignment.

Originally developed to study gender bias in Hungarian online media — the first version powered our 2019 analysis of how 102,240 news articles represent women, men, and minorities in the semantic space (read the analysis).

pip install chronowords


kenon

kenon logo

Semantic network construction from text corpora. Build and analyse word association graphs, find paths between concepts, and compare text-derived networks to human association norms (Nelson norms, Small World of Words).

Named after the Greek kenon (κενόν) — the void that enables connection.

pip install kenon


keyflux

keyflux logo

Corpus keyness, rank-turbulence divergence, and allotaxonographs — in pure Python. Derive keywords and lockwords from a focus-versus-reference comparison using proper corpus-linguistic measures (log-likelihood for significance, log ratio for effect size), compare the ranked lists with rank-turbulence divergence, and render the allotaxonograph — the rank-rank map plus the ranked list of which exact words drove the shift.

It replaces the usual "Jaccard overlap on the top-N keywords" summary with a transparent, pip-installable pipeline. Figures are matplotlib — no JavaScript runtime.

pip install keyflux


lexograph

lexograph logo

Spatialize linear text into pictures you can read. Through one segment → layout → encode → render pipeline, lexograph turns a text into a figure: punctuation spirals, 2-D and 3-D sentence walks, recurrence dotplots that plot a text against itself, and concordance plots for a term's dispersion.

Every preset returns a matplotlib figure and never calls show(), so it renders inline in Jupyter and saves cleanly. The visualisation member of the corpus-lx family, alongside chronowords, kenon, and keyflux.

Named after the Greek graphein (γράφειν) — to write, to draw.

pip install lexograph


corvus

corvus logo

A cookiecutter template for data science and text analysis projects. Pre-configured scaffold with uv, ruff, DVC, MLflow, Sphinx docs, and structured directories for raw/processed data, models, notebooks, and a Python package — eliminate manual setup and start analysing.

Originally developed as our internal project template for computational linguistics and NLP research, now publicly available.

uvx cookiecutter https://github.com/crow-intelligence/corvus.git


All packages are MIT licensed and open source.