Predicate Argument Annotations for Natural Language text.
This software annotates predicate-argument structures of natural-language texts, by utilizing information from the text's syntactic-analysis.
For example, given the sentence "I have a computer that can process natural languages.", the software extracts the following:
have
(SUBJECT) I
(OBJECT) computer
process
(SUBJECT) computer
(OBJECT) natural languages
Predicate-argument structures can be thought of as a shallower approach than Semantic Role Labeling, yet more semantically expressive than syntactic parse-trees.
The annotation tool is named PASTA, which stands for Predicate Argument Structure Annotator.
Assuming a Maven project, the following dependency should be added to the POM file
<dependency>
<groupId>com.github.asher-stern</groupId>
<artifactId>text-understanding</artifactId>
<version>1.0.2</version>
</dependency>The project includes a demo program: com.as.text_understanding.pasta.DemoPasta. This is a command-line program which takes no arguments. It reads sentences interactively from the standard input, and prints their predicate-argument structures.
PASTA can be used in two ways: as a stand-alone tool, or as a UIMA annotator.
As a stand-alone tool, the user has to provide a constituency parse-tree as the input. The parse-tree should be in the format of com.as.text_understanding.representation.tree.TreeNode, and be converted to TreeTravelNode (using the static method TreeTravelNode.createFromTree(), which is the input-argument of com.as.text_understanding.pasta.Pasta
An easy way to obtain such a parse-tree is to use UIMA DKPro libarary just for the syntactic analysis, and convert it to TreeNode using TreeBuilderFromDkpro. This is easy since all the required libararies are included in the project's POM, and no models or parameters are needed (they are all provided seemlessly by DKPro). An example is provided in com.as.text_understanding.pasta.DemoPasta.
As a UIMA annotator: the annotator is com.as.text_understanding.uima_annotators.pasta.FromDkproPastaAnnotator. It assumes the document is already annotated by some DKPro annotations, as specified in FromDkproPastaAnnotator's TypeCapability annotation. To get this annotations, the precondition annotators, specified in DemoPasta.PRECONDITION_ANNOTATORS suffice. (Note that these precondition annotators can be replaced by other DKpro annotators, like, for example, Stanford-Parser).
As for the results, please note that the Predicate annotation mostly covers also some arguments. The actual predicate can be retrieved by the field verb of Predicate.
An example program which runs PASTA as a UIMA annotator, from within UIMA's Document-Analyzer, is provided at com.as.text_understanding.uima_annotators.pasta.DemoPastaAnnotator.