JSGF Grammar Tools
This set of tools can be used primarily to generate strings from a JSGF grammar, but it also provides an easy to use JSGFParser module which creates abstract syntax trees for JSGF grammars. Developers can use these ASTs to help create more tools for their purposes. For more detailed documentation, refer to the Sphinx documentation located in docs/_build/html/index.html
Dependencies
- Python 2.7
- PyParsing module (http://pyparsing.wikispaces.com/Download+and+Installation)
Instructions
The two main Python scripts are DeterministicGenerator.py and ProbabilisticGenerator.py. Both files require a grammar file as a command line argument, and the latter also requires a number, which refers to the number of sentences to generate. Importantly, DeterministicGenerator.py should not take grammars with recursive rules as an argument. A recursive rule is of the form:
<nonTerminal> = this (comes | goes) back to <nonTerminal>;
There are two example grammars included with the scripts: Ideas.gram and IdeasNonRecursive.gram. Ideas.gram is an example of a grammar with recursive rules, though the recursion is not as direct as the above example. It's a good idea to run these grammars with the generator scripts to see how the scripts work:
> python DeterministicGenerator.py IdeasNonRecursive.gram
> python ProbabilisticGenerator.py Ideas.gram 20
Notes
-
Larger grammars take a longer time to parse, so if nothing seems to be generating, wait a few seconds and the grammar should be parsed.
-
Most of JSGF as described in http://www.w3.org/TR/2000/NOTE-jsgf-20000605/ is supported, but there are a few things that have not been implemented by these tools yet:
- Kleene operators
- Imports and Grammar Names
- Tags