A treaty-to-treaty diffusion and genealogy network derived entirely from Dataset 1 (texts and article extracts) and Dataset 2 (clause-level annotations) of the Turkish BIT corpus, covering all 132 international investment agreements of the Republic of Türkiye (1962–2025; signed between 1962 and April 2024). It quantifies the textual similarity of every treaty pair, groups the treaties into textual families, and traces each treaty to its most-similar earlier predecessor, mapping how treaty drafting diffused across six decades of Turkish practice.
Files include the 132×132 TF-IDF cosine similarity matrix, all 8,646 pairwise comparisons (cosine similarity, Jaccard similarity, and shared clause-feature packages), the assignment of every treaty to one of 37 textual families, 86 parent–child genealogy edges, five figures, and an interactive offline network visualisation (vis-network, bundled for offline use). The full derivation pipeline (corpus construction, TF-IDF configuration, the nine clause-feature packages, family clustering, and the genealogy rule) is documented step by step in 00_README.txt.
Key findings
- Two dominant templates: 81 of 132 treaties fall into two mega-families, a liberalisation-era model (n=41, 1992–2009) and the post-2009 New Turkish model (n=40, 2009–2024); the remaining treaties form 10 small families (2–6 members) and 25 singletons.
- Directional diffusion: linking each treaty to its most-similar earlier predecessor (cosine ≥ 0.45) yields 86 descent edges that propagate forward in short steps, with dense clusters around the 1992 liberalisation wave and the post-2010 new-model wave.
- Text and legal content agree: pairwise textual similarity correlates with the number of shared clause-feature packages from Dataset 2 (Pearson r = 0.57 across all 8,646 pairs): treaties that read alike are also coded alike.
Related datasets
This dataset is part of a three-dataset Turkish BIT corpus, all deposited on DataverseNO at the University of Oslo:
Dataset 1: Turkish BIT Treaty Corpus
DataverseNO DOI: 10.18710/JX4WHH
(also on Zenodo: 10.5281/zenodo.20492364)
Dataset 2: Turkish BIT Clause-Level Annotation Dataset
(118 substantive variables coded per treaty)
DataverseNO DOI: 10.18710/FPNLKS
Dataset 3: Turkish BIT Treaty Diffusion and Genealogy Network (this dataset)
(textual similarity, treaty families, genealogy edges)
DataverseNO DOI: 10.18710/WA7HEO
(2026-07-07)
