CLLD – Cross-Linguistic Linked Data

Mapping Glottocodes to ISO 639-3

posted Friday, November 13, 2015 by Robert Forkel

A lot of data about languages marks the associated language either using its Glottocode (i.e. its identifier according to Glottolog) or its ISO 639-3 code. So often, when merging data from various sources, the issue of mapping between the two code systems comes up.

From the use case of merging data, a one-to-one correspondence would be ideal. Conceptionally, though, the mapping may be many-to-many, and in addition, a ISO 639-3 code may map to a Glottolog (sub)family or dialect, not necessarily a Glottolog language. So a table listing only the one-to-one correspondences may be deceptive.

But since such a table is clearly useful, if only to get done with the simple cases, here is a recipe to create such a mapping dynamically. We use the general information about languages in Glottolog in JSON format. This JSON object contains a member resources listing all Glottolog languages. The information about one language looks like

            "id": "aari1239", 
            "identifiers": [
                    "identifier": "aiw", 
                    "type": "iso639-3"
                    "identifier": "aar", 
                    "type": "WALS"
                    "identifier": "aiw", 
                    "type": "multitree"
            "latitude": 5.95034, 
            "longitude": 36.5721, 
            "name": "Aari"

Simplifying such an object to

            "glottocode": "aari1239", 
            "isocode": "aiw"

is a typical task for the excellent jq tool. Turning JSON into csv can be done using csvkit’s in2csv command.

So wrapping up, we can turn Glottolog’s information for languages into a csv file of the form


with a single command line:

curl ""\
| jq '[.resources[] | {glottocode: .id, isocode: [.identifiers[]] | map(select(.type == "iso639-3"))[0].identifier}] | map(select(.isocode!=null))'\
| in2csv -f json > glottocode2iso.csv