About cartolex¶
cartolex maps a body of texts by the words of their authors. It finds the vocabulary they actually use (no list of keywords is given to it), places people, organisations and texts in a lexical space, groups that vocabulary into themes and draws it all as an atlas you can explore and share. In the app, the logo at the top left opens the same page.
Who talks the same¶
The heart of the method is a distance. Two authors are close in the lexical space when they write about their subject with the same words: cartolex measures how much people, teams or institutions talk the same, whether or not they work together; the co-authorship network, drawn on the same map, shows who does. The map shows what people write about, never how good their work is.
It was made for research fields, from a laboratory to a discipline. Any texts whose authors are known will do, though: the articles of journalists, for instance.
How a map is made¶
Texts, collected from open services (OpenAlex first) or imported.
Keywords: the noun phrases of the texts, found by spaCy’s language models and scored by how specific they are (TF-IDF); you, or an AI, keep the field’s own.
Space: a truncated singular value decomposition of the authors × keywords matrix (latent semantic analysis), the space where the distances are measured.
Themes: the keywords grouped into a tree by Ward’s clustering; you can edit it.
Map: the space drawn in two dimensions (UMAP or t-SNE), with the co-authorship network.
Everything runs on your computer; only the collection of texts and the optional AI clean-up use the network, and they say so first (Privacy and personal data).
Key references¶
Sparck Jones, K. (1972). A statistical interpretation of term specificity and its application in retrieval. Journal of Documentation, 28(1), 11–21. doi:10.1108/eb026526
Deerwester, S., Dumais, S. T., Furnas, G. W., Landauer, T. K., & Harshman, R. (1990). Indexing by latent semantic analysis. Journal of the American Society for Information Science, 41(6), 391–407. doi:10.1002/(SICI)1097-4571(199009)41:6<391::AID-ASI1>3.0.CO;2-9
Ward, J. H., Jr. (1963). Hierarchical grouping to optimize an objective function. Journal of the American Statistical Association, 58(301), 236–244. doi:10.1080/01621459.1963.10500845
Every method and library cartolex relies on, step by step: References.
Built on¶
spaCy’s language models find the noun phrases; scikit-learn, SciPy, NumPy, umap-learn and openTSNE build the space, the themes and the map; OpenAlex and other open services provide the texts.