Scripts that creates training and testing data for geo-parsing from Wikipedia
Genesis includes two parts, (1) a genesis.tsv file and (2) a folder of raw text with maps. The genesis.tsv file is useful for machine learning and easily imported into databases. Secondly, the folder contains a folder for each page on the English Wikipedia with enough text with places to make a map. Each of these subfolders includes two files. The first is a .txt file that includes the text of the page. The second is a geojson file that is a map of the places mentioned in the article.
- Machine Learning Tabular Training Data: https://s3.amazonaws.com/firstdraftgis/genesis.tsv.zip
- Folder: https://s3.amazonaws.com/firstdraftgis/wikidata-gazetteer.tsv
The training data is released under Creative Commons Attribution-ShareAlike 3.0 license, which is the same license as much of Wikipedia's content: https://en.wikipedia.org/wiki/Wikipedia:Text_of_Creative_Commons_Attribution-ShareAlike_3.0_Unported_License