Small HTTP server that allows you to access Relex – the Natural Language Processing capabilities of the OpenCog project. This project loads Relex and all it’s Java dependencies into a small Ruby/Sinatra server and exposes it’s functionality through JSON calls.

The Relex tools are very powerful and they have many Java and C++ dependencies. In order to communicate with it in a performant way, we need to make sure that the Relex process + JVM keeps on running. Relex comes with it’s own TCP-based RPC server which is mainly focussed on OpenCog integration. This project attempts to make the functionality available through HTTP JSON RPC calls.

• Developers that need access to some of the most powerful NLP tools
• Developers that quickly need JSON output of a parse
• Developers that use web-flavoured frameworks: jQuery, NodeJS, Rails, Sinatra, etc…

To learn more about Relex & it’s dependencies, please visit the OpenCog website.

• Java VM
• Linkgrammar and it’s Java bindings
• Wordnet
• Most (cross-platform) JAR files are in dependencies/*, including Relex.
• Ruby and these gems:

  sudo gem install json
  sudo gem install sinatra
  sudo gem install rjb

$ ruby relex.rb 
  Java VM, loading: commons-logging.jar
  Java VM, loading: jwnl.jar
  Java VM, loading: linkgrammar-4.6.5.jar
  Java VM, loading: relex-1.2.1.jar
  == Sinatra/0.9.4 has taken the stage on 9090 for development with backup from Mongrel

$ wget http://localhost:9090/frames?text=I am eating a pear
  {
      "Transitive_action": [{
          "Patient": ["eat", "pear"]
      },
      {
          "Agent": ["eat", "I"]
      }],
      "Process_continue": [{
          "Time": ["present", "present"]
      },
      {
          "Event": ["eat", "eat"]
      }],
      "Ingestion": [{
          "Ingestibles": ["eat", "pear"]
      },
      {
          "Ingestor": ["eat", "I"]
      }]
  }

$ wget http://localhost:9090/phrase_tree?text=I am eating a pear
  {
      "string": "(S (NP I) (VP am (VP eating (NP a pear))) .)",
      "tree_array": [[{
          "phrase": "I",
          "pos": "NP"
      }], [[[{
          "phrase": "a pear",
          "pos": "NP"
      }], {
          "phrase": "eating ",
          "pos": "VP"
      }], {
          "phrase": "am ",
          "pos": "VP"
      }], {
          "phrase": "",
          "pos": "S"
      }],
      "flat_array": [{
          "phrase": "I",
          "pos": "NP"
      },
      {
          "phrase": "a pear",
          "pos": "NP"
      },
      {
          "phrase": "eating ",
          "pos": "VP"
      },
      {
          "phrase": "am ",
          "pos": "VP"
      },
      {
          "phrase": "",
          "pos": "S"
      }]
  }

• Expose more functionality
• Integration tests
• Caching mechanisms
• Stress testing

MIT, but who cares?