Concept Action Grammar Generator (CAGG)

This Java library implements an evaluator of grammars represented through the Backus-Naur Form syntax. It can compile the syntax into a tree and evaluate it with an input sentence as a String.

Publications

For more details, check the following publication and the related presentation

L. Buoncompagni and F. Mastrogiovanni, “An Open Framework to Develop and Validate Techniques for Speech Analysis”, published in the Proceedings of the 3rd Italian Workshop on Artificial Intelligence and Robotics, AI*IA, Genova, Italy, 2016.

The CAGG Language

// An hello-world example of a CAGG grammar.
#CAGv1;  
!start <MAIN>;
<MAIN>	: 	<R1> {@?;@~;} | <R2> {@T;@bb;};
<R1> 	: 	hi | hello;
<R2> 	: 	(!optional(good {@!T;}) bye) | !repeat(bye{@!!;}, 1, 2);

A CAGG grammar defines rules (identified by < and >) as a tree. In particular, the grammar divides rules into sub-rules, which represent a logic expression as a binary tree. Rules have an identifying name and an expression, separated by :. Each line in the grammar must end with ;, and the CAGG syntax supports C-like comments and annotations.

Some directives can be introduced in the rules, which are denoted by !. Specifically,

• !start defines the root rule of the syntax,
• !optional identifies that a specific sub-rule can be omitted,
• !repeat enables looping over a rule within a lower and an upper bound on the range.

Sub-rules or terms (i.e., words in a sentence, which are an expression’s leaves) can be aggregated using an empty space to denote an and logical operator, or with | to identify an x-or operator.

Each rule or term can be semantically tagged using {@...;...}, which specifies an array of symbols (i.e., tags) used to identify the rule path identified by the CAGG evaluator given an input string. Tags are used to implement different behaviours based on the evaluation of an input string. Special tags can be also used, such as:

• ? is replaced during compilation with the rule’s head or term (e.g., R1),
• ~ is replaced at compilation time with the leaf term evaluated by CAGG (e.g., hi or hello),
• ! removes a tag given its symbol (e.g., T),
• !! removes all tags for that specific rule.

Given the hello-world grammar above, examples of evaluated tags are:

• If the string "hi" is given, then the tag {R1\hi} is returned, while "hello" provides {R1/hello}.
• If the string "bye" is given, then {T\bb} is retrieved, if "good bye" is given then the tag would be {bb, T\bb}, and for "bye bye" no tags are returned during the evaluation.

Examples and Debugging

To know more about the code, check out the example grammars and runners. In addition, this software comes with a GUI to debug developed grammars (as shown below), please, check this running file to use it.

---

Author: Luca Buoncompagni ([email protected])