Adam Fallon

Monitoring

List of things that look promising, but I reserve judgement on

I will periodically update these to mention things like whether or not I still think they are useful.

Labelled Matches in R#

Don’t get me wrong - Spacy is a good NER library. Very extensible, well documented, flexible.

However - from experience, a lot of applications of entity labelling aren’t unique or domain specific - you are just trying to do something reasonbly straightfoward with dates, or places, etc.

Entity Recognition Example

A common use case is just marking up things like dates, links, percentages, money, names in text - and all of those follow simple, codified forms.

Yeah - you know where this is going…Regex.

This idea has been implemented by R# - a regex processing library that boasts 4,500x speed up over Spacy for simple labelling.

While I don’t think this is a silver bullet, I think there are numerous times with something this simple will work well.