GlobalNamesArchitecture/GlobalNames

Alexander Myltsev

@alexander-myltsev

ok, I understand

Dmitry Mozzherin

@dimus

We got our publication at Biodiversity Next https://biss.pensoft.net/article/35353/

Jonathan A Rees

@jar398

Is there a chat channel for the workshop?

Dmitry Mozzherin

@dimus

How about we use this channel

I think the workshop is totally legit for this room

https://github.com/SpeciesFileGroup/cs_492_spring

Matt

@mjy

Why the CS link? Just here, right?

Link to workshop org - https://github.com/bhl-gn-workshop

Dmitry Mozzherin

@dimus

we want to steal some of the code the guys made

Matt

@mjy

I see.

We need the BHL table of serials as CSV, easiest way to get it?

Dmitry Mozzherin

@dimus

https://about.biodiversitylibrary.org/tools-and-services/developer-and-data-tools/ may be

Matt

@mjy

Didn't we have that data for the CS course?

OK, never minds, we poke at that starting point.

Dmitry Mozzherin

@dimus

@gdower this one: http://bhlrpc.globalnames.org:80

Matt

@mjy

https://github.com/bhl-gn-workshop/bhl_gnames_tw_interfaces

Geoff Ower

@gdower

python -m grpc_tools.protoc -I. --python_out=. --grpc_python_out=. my_service.proto

Jocelyn Pender

@jocelynpender

I put what Dima helped me do in this new repo: https://github.com/bhl-gn-workshop/bhl-ner

David Shorthouse

@dshorthouse

Is http://bhlrpc.globalnames.org:80 up? FWIW, doesn't work from Holiday Inn wifi.

My bad; does work.

Dmitry Mozzherin

@dimus

https://files.gitter.im/SpeciesFileGroup/CS492/WQom/Verification_Heatmap2.html

https://files.gitter.im/SpeciesFileGroup/CS492/kico/Verification_Heatmap.html

Is http://bhlrpc.globalnames.org:80 up? FWIW, doesn't work from Holiday Inn wifi.

@dshorthouse I wonder if it is also setting on the network, it works from here fine :-/

Dmitry Mozzherin

@dimus

these are visualizations of 190K BHL volumes, and names index

Dmitry Mozzherin

@dimus

https://github.com/gnames/gnfinder/releases/tag/v0.8.8

https://gitlab.com/gogna/gnparser/-/releases

Jocelyn Pender

@jocelynpender

@dshorthouse Another approach to the geo parsing problem is an ensemble approach, combining results from several NER systems by a voting system: https://www.frontiersin.org/articles/10.3389/fdigh.2018.00002/full

Geoff Ower

@gdower

@jocelynpender https://pythonhosted.org/geoparsepy/readme.html

Joel Richard

@cajunjoel

mobot31753002212634_djvu.txt

@jocelynpender @gdower Falco OCR text (above)

Geoff Ower

@gdower

The Falco berigora was confusingly located in Falco.

Dmitry Mozzherin

@dimus

mitie-ner.txt

MITIE location person tagging sometimes it is right sometimes it is funny

David Shorthouse

@dshorthouse

Same in watson

Geoff Ower

@gdower

Here's output from the mordecai python package:

 >>> geo.geoparse("The Falco berigora was confusingly located in Falco.")
[{'word': 'Falco', 'spans': [{'start': 46, 'end': 51}], 'country_predicted': 'ITA', 'country_conf': 0.83302397, 'geo': {'admin1': 'Sicily', 'lat': '38.11706', 'lon': '15.07908', 'country_code3': 'ITA', 'geonameid': '6540054', 'place_name': 'Falcone', 'feature_class': 'A', 'feature_code': 'ADM3'}}]

David Shorthouse

@dshorthouse

watson_people.csv

watson_places.csv

Dmitry Mozzherin

@dimus

I think i like watson results a bit better than mitie results

David Shorthouse

@dshorthouse

They appear to be cleaner. But... somehow we have to get this locally. Apparently possible with Watson, but will require some high level conversations. Also need some actual entities like wikidata QIDs, which are not provided in Watson.

Jocelyn Pender

@jocelynpender

Let's add our experiences with the geoparsing tools here : https://docs.google.com/document/d/1YH1NRG1-zusAkZZWnmD0tScx7dSE8znVXnOtbQo_Ppc/edit?disco=AAAADLdNjJc

@dshorthouse @gdower @dimus @gdower

Joel Richard

@cajunjoel

Hi @debpaul