Hacker News Reader

Comment by syllogism | Hacker News Reader

Looking forward to seeing the evaluation numbers!

I'm mostly curious about how their NER and parser compare against what I've implemented for https://spaCy.io . I've tried the architectures they're using, and I've found they need very wide (and therefore slow) hidden layers to get competitive accuracy.

I'm sure they have some evaluations, right? I mean you can't really develop these things without running experiments...

View on HN

Also it seems like their reported NER is just serving up the spaCy NER.

“spacy_ner service which provides Spacy NER annotations.” [0]http://nlp_architect.nervanasys.com/service.html

Well, the NLP Architect mentioned (https://github.com/NervanaSystems/nlp-architect) uses Spacy, so... :)

Well, that's what makes me curious! They have a class that runs the BIST parser (by Kilperwasser and Goldberg) using spaCy to pre-process for tokenization, sentence boundard detection and POS tags. Now, spaCy by default uses the parser to add sentence boundaries...So you've already got a parse right there.

So, is their BIST parser more accurate than spaCy's parser? I'm getting similar (but slightly higher) WSJ accuracies to what Eli and Yoav reported in their paper, so I'd be surprised if they're doing so much better?

Love Spacy, thanks for working on it.

This is a giant corporation getting academics to build software in a lab. I hope it's good, but I've usually seen it doesn't end up being much.

Reply on news.ycombinator.com