Skip to content

linker.md

Latest commit

 

History

History
31 lines (23 loc) · 1.72 KB

linker.md

File metadata and controls

31 lines (23 loc) · 1.72 KB

Description

EntityLinker links text labels to the corresponding geo-entities in external knowledge bases (e.g., OpenStreetMap) to enable advanced search queries on scanned maps. In the current version of the mapKurator, EntityLinker retrieves the candidate geo-entities in OpenStreetMap that satisfy two criteria: 1) the suggested word (i.e., output from PostOCR) is a substring of the candidate geo-entity's name and 2) the geocoordinates of text bounding polygon (i.e., output from Geocoordinate Converter) is in the buffer of OpenStreetMap geo-entities.

Commands

The inputs for this module are 1) metadata in CSV format and 2) PostOCR results in GeoJSON format.

1) Use run.py

python3 run.py --sample_map_csv_path='/home/maplord/maplist_csv/luna_omo_metadata_56628_20220724.csv' --expt_name='57k_maps' --module_entity_linking

where

  • --sample_map_csv_path stores the metadata of the input map, a sample file can be found here.
  • --output_folder: output directory
  • --expt_name: experiment name for running the pipeline
  • --module_entity_linking turns on the entity linker module in this run

2) Use entity_linking.py

If you wish to specify the input and output specifically, you can use entity_linking.py in m6_entity_linker folder.

Sample command:

python3 entity_linking.py --sample_map_path='/home/maplord/maplist_csv/luna_omo_metadata_56628_20220724.csv' --in_geojson_dir='{directory_path of PostOCR results}' --out_geojson_dir='{directory_path}'
  • --sample_map_path: stores the list of the map
  • --in_geojson_dir: input path of PostOCR output
  • --out_geojson_dir: output path to save the generated GeoJSON