Extraction of person names, places, etc.

[michaelnmmeyer]

This is the kind of tool I had in mind for extracting keywords from documents. GPE means "Geopolitical entity"; NORP means "Nationality or religious or political group". I used one of Manu's inscriptions as example.

I ran the same thing against all our corpus. There are a lot of tagging errors. This can be improved by reworking the underlying AI model, or by using another programming library which provides better results but is extremely slow.

Even so, the data is already useful as is. We have around 4000 person names which could be used for faceted search or in an autocomplete interface, for instance.