Tags all substrings (or tokens) existing as place names in the geonames gazetteer in O(len(text)) time.
Download the allCountries.zip file from the geonames Web site and
extract the allCountries.txt file (e.g., into the geonames_tagger/data folder).
Needs ~1 hour on MacBook Pro and 4.9GB of RAM:
$ python gnt_index.py data/allCountries.txt idx001
Start the http server:
$ python gnt_server.py -h "localhost" -p 55555 idx001
Make tagging POST request via a python script (check source of gnt_request.py). The submitted text must always be utf-8 encoded:
$ python gnt_request -h "localhost" -p 55555 txtfile
Or, make http POST request directly (returns the result as json):
http://localhost:55555/service
POST parameter: text=<utf-8 encoded text>
Using curl (POST):
$ curl --data "text=the text to parse" http://localhost:55555
Using curl (GET):
$ curl http://localhost:55555?text=aplacename
The resulting json looks like:
{
"success": 1,
"n_matches": N,
"matches": [
{
"token": TOKEN,
"idx": [START,END],
"gnrecs": [
[COLS OF REC IN allCountries.txt FILE],
...
]
},
...
]
}
In case of an error:
{
"success": 0,
"msg": MESSAGE
}