Core Framework code for NLU system, can be extended for different tasks.
- General description of original system: https://www.icsi.berkeley.edu/pubs/ai/naturallanguage15.pdf
- Extension/Implementation for multi-agent collaboration: ftp://ftp.icsi.berkeley.edu/pub/feldman/AI-HRI.pdf
System requirements:
- PyEnchant: http://pythonhosted.org/pyenchant/
- Pyre, port of Lyre: uncompress ecg-package.tar.gaz and follow the instructions in INSTALL
- Jython: http://www.jython.org/
- ecg-grammars: https://github.com/icsi-berkeley/ecg-grammars
Running the system:
- analyzer.sh: This will run the grammars/Analyzer on an RPC server. If you’d like this to run as a background process, you can modify the “jython” command to add a “&” at the end. By default, it runs the “research” grammar from ecg-grammars.
- ntuples.sh: This will run the Specializer (as a separate process/Terminal tab). This will open up a prompt, in which you can enter sentences and view the resulting n-tuples, printed in a cleaner visualization.
- ui.sh: This will run both the grammars/Analyzer and the Agent-UI in one tab.
- setup.sh: this will run the Problem Solver as a background process. In the “core” framework system, the problem solver simply prints out n-tuples.
- tests.sh: This runs a suite of sentences and makes sure they match ‘gold standard’ n-tuples. Note that analyzer.sh must already be running.
NOTE: You’ll also need to set your PYTHONPATH in your bash_profile to point to {dir_path}/framework-code/src/main.