ScalaNLP/Scalala renamed to Breeze

[David Hall]

Hi everyone,

Given the facts on the ground, I'm going to merge the scalanlp/scalala
projects into a new project, called Breeze. Breeze is going to have a
modular structure like ScalaNLP did, and it will have a linear algebra
core, based on a (yet again) rewritten version of Scalala.

https://github.com/dlwh/breeze

The components will be:

* breeze-math: Linear algebra and numerics routines (née Scalala)
* breeze-process: Libraries for managing datasets and job pipelines.
(née scalanlp-data)
* breeze-learn: Machine Learning and Optimization.

Possible future releases:
* breeze-viz: Vizualization and plotting (The plotting parts of scalala)
* breeze-fst: Finite state toolkit
* breeze-graphs: I toyed around with making a graph library, but then
it looked like someone else was working on it, so I stopped.

It's still very much work in progress. I'll send a followup email to
scalala@googlegroups about the -math library. I might think about
breaking out learn into a stats component and a learning component, or
more likely an optimize, learning, and stats component. But I'm not
excited about that at the moment.

Versioning restarts at 0.1 across the board, and I'm going to say it's
Apache licensed, taking DanR's comment that he wanted to move in that
direction as good-as-we're-going-to-get.

Comments about the name are welcome, though I'm pretty happy with it.
(The name has personal significance to me.) Getting a domain name will
be a little tough, but there's only one project on github called
breeze, and it's some unmaintained php web framework.

Since we're in merger mode, I'm happy to absorb other people's code. :-)

-- David

[Ivan Brusic]

Exciting to read this news. I was a previous user of Scalala and was
looking at potentionally using ScalaNLP for a new personal project. My
main usage for NLP is named entity extraction, so I am still on the
fence between Scala and Python (NLTK).

Cheers,

Ivan

> --
> You received this message because you are subscribed to the Google Groups "ScalaNLP" group.
> To post to this group, send email to scal...@googlegroups.com.
> To unsubscribe from this group, send email to scalanlp+u...@googlegroups.com.
> For more options, visit this group at http://groups.google.com/group/scalanlp?hl=en.
>

[David Hall]

There's a basic CRF-based NER system in the learn component. (You'll
want to probably go back a few commits before the scalala merger.)

https://github.com/dlwh/breeze/blob/master/learn/src/main/scala/breeze/sequences/CRFTrain.scala

It expects data in the CONLL format (see
http://www.cnts.ua.ac.be/conll2003/ner/) and you can specify feature
templates, though the default is pretty good for English.