C R Hunt - Mar 16, 2013:

WordNet is great for determining parts of speech, synonyms, some hierarchal relationships, etc. I use it in my project for parsing purposes to build up a knowledge base, to perform morphological functions, etc. But WordNet doesn’t supply the type of “common sense” or relational knowledge you would need to put together sentences that make sense. That is, WordNet doesn’t associate the word “dog” any more closely with “bark” than with “hover”.

However, if you want to generate meaningful sentences, then it would be better to start from an ontology. (See, for example, OpenCyc.)  Your project would then be focused on converting from the internal database structure of the ontology into a natural language string. Fortunately, you would only need to concern yourself with the subset of grammar represented by the ontology relations.

What I would recommend is putting together a list of resources you would like to use. Then figure out which languages or wrappers you would need to put these resources together in a single program. (What programming languages do you know, btw? Do you have a preference?) Then it’ll be more straightforward to fill in the “missing” pieces with your own code.

That’s great !!!

http://swizec.com/blog/natural-language-generation-system-architectures/swizec/4535

Just found another article in my G+ stream that is pertinent to this thread. Short, but interesting and enjoyable read.

Andrew Smith - Apr 30, 2013:

http://swizec.com/blog/natural-language-generation-system-architectures/swizec/4535

Just found another article in my G+ stream that is pertinent to this thread. Short, but interesting and enjoyable read.

Thanks so much that’s great !! It is really interesting article