The RelEx processor is architected to allow a number of different output modules to co-exist. Each module walks over the internal graphical representation of the parsed, processed text, and prints output. There are 7 output modules at this time:
The default output, designed for overview readability, not for post-processing. General aspects of this format are described on the main RelEx page. Users who need to post-process RelEx output should NOT use this format; they should use one of the others below, or create a new output module.
An XML-based batch processing/post-processing/archival format. It captures all of the information generated by RelEx in a format that is very easy to machine process, and is also quite human-readable. The idea here is that processing stages that come after RelEx could just read this input, instead of paying the cost to re-parse the text every time.
Ideal for piping the output of RelEx into the OpenCog NLP subsystem. This format is discussed in greater detail in a README file in the OpenCog source code tree -- this format is tailored to the needs of opencog processing. Note also that there is a perl script that will convert the compact format to the OpenCog format.
The relations and features are encoded into OpenCog via a series of InheritanceLinks. The detailed encoding can be found in the file "src/nlp/scm/type-definitions.scm" of the OpenCog source distribution.
Generates semantic web/web-ontology type output. Experimental.
Debugging output. Dumps full graph, useful for debugging processing rules.
All of the different output formats are generated by code that can be found in the src/java/relex/output directory.