These are chat archives for cltk/cltk

16th
Mar 2017
Kyle P. Johnson
@kylepjohnson
Mar 16 2017 16:46
@aakash5897 Give it a try!
I bet it could. If so, then make a pr to add it to the docs
@sagarkar10 and @/all Luke and I are reviewing applications as fast as we can. The best way for us to help is to send it as a Google doc that we can edit and add comments to
A new, really exciting ticket for someone to work on: cltk/cltk#505
It is for SyntaxNet, which is a deep learning system for doing NLP and is built on TensorFlow. I do not have time to help people with this, so you be responsible for putting together the precise steps necessary for installing on Mac and Linux (Ubuntu 16.10, preferably).
I am sorry that @priyaraistar is upset about something he thinks I did. For the record, I have not accepted any former student last year.
Kyle P. Johnson
@kylepjohnson
Mar 16 2017 16:52
If anyone wants to submit PRs, they are free to. However one should do so because they care about the mission of the project and want to learn the software, not because they expect something in return. Thanks
Samriddhi Sinha
@djokester
Mar 16 2017 16:59
@kylepjohnson is #505 up forGSoC?
Kyle P. Johnson
@kylepjohnson
Mar 16 2017 17:00
No not for GSoC, unfortunately
Samriddhi Sinha
@djokester
Mar 16 2017 17:00
@kylepjohnson mind if I take it up after the proposal submission? I am still fine tuning stuff?
#505 interests me a lot.
Kyle P. Johnson
@kylepjohnson
Mar 16 2017 17:02
I would love to, however model-building and algorithm development is very dependent on annotated data. And sadly, annotated data for classical languages is rare and very expensive to create. Thus, for this year, we are focusing on using available open source data and focusing on one or two languages
Samriddhi Sinha
@djokester
Mar 16 2017 17:05
@kylepjohnson doesn't SyntaxNet seem a bit conflicting given that it is Python 2.7 only. Whereas we at CLTK are 3.x+?
Kyle P. Johnson
@kylepjohnson
Mar 16 2017 17:05
@djokester Sure, at your leisure, however I will leave it open for others to work on. In fact, it might be best to have a few iterations of the docs. That is, someone does it, then others test it and fix further
@djokester you are right about the Python version conflict. We are not yet integrating it into the core, but only providing docs as a service to the community. Eventually, once it is ported to Py3, we can think about how to integrate it (and maybe other parts of TensortFlow) into the CLTK … but realistically this is at least a year away
Samriddhi Sinha
@djokester
Mar 16 2017 17:09
But it definitely is a wonderful add-on to the project