Place tagger

Text in.
Ranked OpenStreetMap tags out.

Describe a place without naming what it is — “a quiet spot in F-7 with good cold coffee” — and this ranks all 227 tags in the vocabulary, showing the adjectives behind each one and how sure the model is of them.

Why is it invitation-only?

It runs on one laptop, not a server. Every description is parsed by a transformer and ranked in the same process, so a handful of people at once is the honest capacity. The invitation is a queue, not a judgement — and it keeps a research demonstrator from being read as a finished product.

Request an invitation
Stored only in a local file on the author’s laptop, only so a code can be sent back to you.
Thesis
Geospatial Tagging of Volunteered Place Descriptions: Bayesian Elastic Net and Extended FP-Growth Approach
Author
Faizan Shah Ahmad · MS Artificial Intelligence, Air University, Islamabad
Supervisor
Dr. Madiha Yousaf · co-supervisor Dr. Faridoon Khan

The tagger reads text. It does not geocode, query OpenStreetMap, or locate anything on Earth, and it makes no outbound network request of any kind.