I'm sure that REST services have their place in the world, but I sure
hope that NLP will not move more and more to the cloud.
Consider running a business processing confidential or sensitive data
with NLP. I'd not at all feel comfortable contracting third-party
web-services to do that processing for me. I may consider running a
private cloud at some cloud hosting service, but actually, I may most
probably tend to set up a private in-house cloud or cluster.
Within research, there have been studies that web services are
detrimental to reproducibility [1,2]. Services change or become
unavailable, which can make it difficult to impossible to reproduce
experimental results.
I sure hope that NLP tools will continue to be distributed as portable
software. Of course, REST in *addition* to underlying portable
software and resources are a fine thing.
Cheers,
-- Richard
[1] How Reliable is Your workflow: Monitoring Decay in Scholarly Publications
J. M. Gómez-Pérez and E. Garcia-Cuesta and J. Zhao and A. Garrido and J. E.
Ruiz and G. Klyne
75 (2013)
[2] Best practices for workflow design: how to prevent workflow decay
K. M. Hettne and K. Wolstencroft and K. Belhajjame and C. A. Goble and E. Mina
and H. Dharuri and D. De Roure and L. Verdes-Montenegro and J. Garrido and M.
Roos
(2012)
On 31.01.2014, at 10:24, andy mcmurry <mcmurry.a...@gmail.com> wrote:
> I completely agree.
>
> Pure functional programming has the benefit of being immutable which is
> nice for parallel computing. Clojure/Haskell are both purely functional,
> whereas Scala can be written functionally if the author chooses. They are
> all good languages.
>
> Groovy also has its place. Groovy is wonderfully simple, especially
> for "*getting
> started*" examples for new users familiar with Java.
>
> REST would really demystify things (I hope).
> I'm not a big proponent of "cloud" services, but I imagine more and more
> NLP processing will up on Amazon with REST calls.
>
> --AndyMC
>
>
> On Wed, Jan 29, 2014 at 9:43 AM, Steven Bethard
> <steven.beth...@gmail.com>wrote:
>
>> On Wed, Jan 29, 2014 at 5:24 AM, andy mcmurry <mcmurry.a...@gmail.com>
>> wrote:
>>> Clojure, having its origins in LISP, is a better fit for serious NLP
>> work than Groovy
>>
>> Sorry, I have to call this one out. I don't think having origins in
>> LISP makes anything a better fit for serious NLP work. Not that I'm
>> against Clojure or that I'm recommending Groovy. But there's nothing
>> inherent about LISP that makes it a better fit for NLP.
>>
>> If you want to argue that functional paradigms (e.g. LISP, Haskell,
>> Scala, Map-Reduce) are better for serious NLP work, I might believe
>> that argument. But I don't think there's anything special about LISP
>> that makes it better for NLP than other functional languages.
>>
>> Steve
