Make spacy.en.English() 10× faster about spacy HOT 5 CLOSED

You'll need to run spacy as some sort of long-running server process, which your client communicates with.

I've tried to keep start up times low by using efficient binary file formats, but less than a second is unrealistic --- spacy needs to load its model files.

I assume you don't want to program in Python? If you do, you should just load spacy in your top-level module, and pass it around, as you iterate over your document collection.

from spacy.

Python is perfectly fine, preferred actually. The trouble is that I only really have one document. I'll be running my analysis on the file as it changes over time (at most once every few seconds), but I want the analysis to complete quickly, ideally in under 300 ms. It sounds like I'll need to follow your suggestion of using a long-running background process that the command line utility communicates with. Do you know of an existing Python utility that uses that architecture, for reference?

from spacy.

Why can't you just watch the file for changes from a Python process?

I mean...ultimately your problem is to feed text into Python at regular
intervals. I'm sure you can find some way to do this.

If you like writing in Python, I would suggest that having spaCy output to
the command line will be much worse than using its API. Presumably you also
need to use the annotations for something, right? Writing to file, reading
in, and then rewriting your own API should be unnecessary.

On Fri, Mar 27, 2015 at 3:14 AM, Jordan Suchow [email protected]
wrote:

Python is perfectly fine, preferred actually. The trouble is that I only
really have one document. I'll be running my analysis on the file as it
changes over time (at most once every few seconds), but I want the analysis
to complete quickly, ideally in under 300 ms. It sounds like I'll need to
follow your suggestion of using a long-running background process that the
command line utility communicates with. Do you know of an existing Python
utility that uses that architecture, for reference?

—
Reply to this email directly or view it on GitHub
#40 (comment).

from spacy.

Thanks, this was helpful!

from spacy.

This thread has been automatically locked since there has not been any recent activity after it was closed. Please open a new issue for related bugs.

from spacy.