This public atlas is a reduced visualisation export for the working paper Measuring Global AI Cooperation, Competition and Innovation through News Text: A Graph-based Approach.
Each annual graph is an undirected ADM1 co-occurrence network. A link means that two first-order administrative regions were mentioned in the same news document after the study's AI-related filtering. The edge weight is the number of unique documents in that year that co-mentioned the pair.
Co-occurrence is an observational relationship in news text. It should not be interpreted, by itself, as evidence of direct collaboration, competition, innovation exchange, or causality.
The browser receives only simplified ADM1 geometry, region names and ISO3 country codes, annual aggregated region metrics, and annual source-target-weight network records required by the visualisation.
The public bundle excludes article text, headlines, document URLs, news-source names, document identifiers, raw location strings, original geometry identifiers, document-level records, local file paths, database paths, term-audit tables, source-audit tables and processing intermediates.
All ADM1 units in the supplied global boundary file are retained. A region with no observed activity in a selected year is represented as zero and remains visible, searchable and clickable. The supplied geometry contains 3,224 polygon features representing 3,223 displayed ADM1 units because two features share the same Mazandaran, IRN region key and are rendered as one unit.