fgregg
commented
9 years ago Hi @mbatchkarov, we are trying to figure out the best way to make a 'parser maker toolkit.' As our next step, we are going to make a new name parser (Given Name, Surname, Generational Title, Honorific, etc). If you are willing to put up with the messiness, it would be great if you went ahead with the UK address parser. Once we have made a few new parsers, we should have a much better idea of how to separate concerns.
To your immediate questions.
- @tomsteinberg can help get us some messy UK addresses
- For the tag set, if you can't find an authoritative standard, just start with a list of tags that make sense. As you will find addresses that can't be sufficiently covered, add tags as you go.
- Happy to help some with the manual labelling.
ofergold
mbatchkarov
cathydeng
stevevance
jeancochrane
Hi,
I'd like to label some data and train a model for British addresses. The following would need to be done:
usaddresswill apply. A decent manual will be needed too (usaddressneeds one too!)If anyone would like to help with this, or has some data lying around, please let me know.
PS This isn't really an issue with
usaddressper se, but I couldn't think of a better place to post.