0

So far, I have evaluated few well known Java parsers, but none comes actually free for commercial use, in one way or other. Like OpenNLP, which is free, but uses training data which is not available for commercial purpose. Morphadorner is there, but it seems last updated in 2009. Stanford NLP is full GPL, while LingPipe provides a paid license.

Being an non-funded startup at the moment, it is really not possible for me to pay high license fee.I had to implement a rule based parser for my requirement, which incomparable to industry standard solutions.

Though, I am not a free-freak and it is not like I need something that only few people in world are using. I firmly believe, things like NLP are getting popularity, and implementing a POS Tagger again is like reinventing a reinvented wheel. Is there really no industry standard solution for this?

Update : The mindset seems industry-standard means a paid thing. Then how come several libraries of HTML parsers(among other stuffs) come free. Html parsers ar far more used than NLP, but still, there is fairly large requirement for NLP tasks.

Update : Question is closed. I shall soon be pleased to release my tiny small library for keyword extraction that makes use of Wordnet and various pretty rules. Thanks AB

akshayb
  • 1,219
  • 2
  • 18
  • 44

1 Answers1

4

You have already identified a solution:

Stanford NLP is full GPL ...

There is no ban on using GPL for commercial projects.

The catch is that if you distribute your code in binary form, you also have to provide source code as well. That means that your business model needs to be based on providing paid support and/or services to your customers (rather than selling software licenses) ... but there are many companies earning good money based on similar models.

Here's another Q & A on this topic with some other potential leads:

The mindset seems industry-standard means a paid thing. Then how come several libraries of HTML parsers(among other stuffs) come free. Html parsers ar far more used than NLP, but still, there is fairly large requirement for NLP tasks.

It is up to the team (or organization or company) that develops something like this to decide how they license it. It appears that none of the existing NLP offerings are "free as in beer". That's their choice. And you've got no right to complain ... given that you are also choosing NOT to make your application available for "free as in beer".

And as I pointed out Stanford NLP is "Free as in Libre". It is just that you are not prepared to live with the obligations of the GPL.

Nobody owes you a duty to do things for you for free. As a would-be entrepreneur, you should not expect it.

(FWIW, the reason that there are free-as-in-beer HTML parsers, and lots of other things is typically that either the problem is orders of magnitude easier than NLP, or that there are companies / organizations out there who have donated lots of resources to get the code developed.)

Community
  • 1
  • 1
Stephen C
  • 698,415
  • 94
  • 811
  • 1,216
  • Thanks. I would love to, but sentence parsing is only a part of my application. I would not favor the idea of releasing complete code at the moment. – akshayb May 27 '13 at 03:40
  • That is your choice. But if you make that choice, you may find that you have to pay out hard cash up-front. – Stephen C May 27 '13 at 04:51
  • Don't forget about the ASP loophole in GPL (incl v2). You can use GPL v2 software in your website, but not release the source code. http://www.funambol.com/blog/capo/2006/08/honest-public-license.html – Neil McGuigan May 27 '13 at 06:35