Applications

• Sentence validation at semantic level

  Taking as example the following sentence: "The cat shot to the wall", each word is defined in every English dictionary (so it's valid at lexical level), and even the position of each word, forming an accepted structure (valid at grammar level). But it's not a realistic scenario, due to for shoot you need a gun, to use a gun, a hand is required, but cats doesn't have hands. Therefore it's an invalid sentence (failed at semantic level) at least in the real world.

  
  

• Disambiguation

  If you are trying to convert automatically a listened language (speech to text), you have to take in mind that at acoustic level some words sound quite similar.

  E.g.: a computer that understands you like your mother VS a computer that understand you lie cured mother

  The underlined words, sound pretty similar, therefore only with a semantic knowledge about "computers", "understanding" and "mothers" could be correctly interpreted.

  
  

• Object guessing

  Imagine you have an application be able to analyze an image that identifies "black legs", "2 eyes", ...

  If you can integrate this information with other sources, you will be able to identify (or at least reduce the options) which object are you seeing.

  
  

• Translation helping

  Translation is not a conversion of words, it requires to know what a concrete concept means, for choosing correctly their corresponding words in the other language.

What's next?

For the next version/release, it will try to add the following enhances:

• Solve the list described in limitations, specially the instantiation.

  
  

• Add mechanisms to catch the interaction between the objects (for example taking advantage of the <dependencies> provided by Stanford NLP Core)

  and even other information that could be extracted from a sentence as "who", "when", "where", etc.; this could be very useful to help other mental processes as logic, pragmatic analysis or making decisions.

  This does not mean add or implement higher mental processes into the model, the scope of this project is focus only on the semantic memory.

  
  

• Provide an online demo of the application for being tested without the need of installation.

Contributions

The application is an open source software, you are allowed to do everything you want with any part of the project (but remind the Stanford software is under the GNU General Public License).

If you use it, please mention this page it or at least don't forget contact to share the enhances and to know what uses people is doing.

This is a non-profit project that pretends to be open for everyone, not only for developers.

Then if you want to contribute, you can help with a lot of tasks as:

• Providing sentences to the model (information extraction)
• Testing the application
• Reviewing the model
• Checking the documentation
• Web page maintenance and enhances
• Proposing anything you consider
• Spreading the project on the web and into the interested people
• Hardware and resources support
• And of course, development tasking

Everyone is invited to join, and any help would be more than appreciated.