A synthetic dataset for training small language models to translate TRIDENT intermediate language into Overpass QL queries for OpenStreetMap.
Given a AreaWithConcern instruction, generate a valid Overpass QL query:
Input:
AreaWithConcern: Shinjuku, Tokyo, Japan; Cafes
Output:
[out:json][timeout:30];
area["name:en"="Tokyo"]->.outer;
area["name:en"="Shinjuku"]->.inner;
(
nwr
"amenity"="cafe"(area.outer);
);
out geom;
Areas are listed… See the full description on the dataset page:
https://huggingface.co/datasets/yuiseki/text2geoql.
