We need to allow GADM IDs (e.g., ETH.9.7.3_1) in addition to GeoNames IDs (e.g., 6783700) for normalizing locations. The geonorm library already supports this. (No changes were needed.) I have constructed a new geonorm index that includes all of GeoNames, plus some Ethiopian woredas from GADM that weren't in GeoNames:
Those Ints will need to be changed to Strings to work with the new index.
As far as I know, there was no reason to use Ints in the first place, so while it's a pain to change a bunch of types in Eidos, I think a String is a better representation for an ID anyway.
@EgoLaparra: can you look into this today? I believe @kwalcock wants all changes in today so that he can merge them, run Eidos on the documents over the weekend, and send Ben the new results on Monday.
We need to allow GADM IDs (e.g., ETH.9.7.3_1) in addition to GeoNames IDs (e.g., 6783700) for normalizing locations. The geonorm library already supports this. (No changes were needed.) I have constructed a new geonorm index that includes all of GeoNames, plus some Ethiopian woredas from GADM that weren't in GeoNames:
http://clulab.cs.arizona.edu/models/geonames+woredas.zip
To load this new index, some changes to Eidos are needed. Eidos assumes that geo IDs will be Ints:
https://github.com/clulab/eidos/blob/master/src/main/scala/org/clulab/wm/eidos/context/GeoNormFinder.scala#L23 https://github.com/clulab/eidos/blob/master/src/main/scala/org/clulab/wm/eidos/context/GeoNormFinder.scala#L137 etc.
Those Ints will need to be changed to Strings to work with the new index.
As far as I know, there was no reason to use Ints in the first place, so while it's a pain to change a bunch of types in Eidos, I think a String is a better representation for an ID anyway.
@EgoLaparra: can you look into this today? I believe @kwalcock wants all changes in today so that he can merge them, run Eidos on the documents over the weekend, and send Ben the new results on Monday.