Using XAR

0. 'Extractor' is the main method that is invoked to start XAR. Under the XAR execution directory we have the "specs" directory which constaints specifications for various domains (schema, rules, constraints for each domain), the "databank" directory which contains databanks. In the XAR execution directory there is a file called config.txt which is a configuration file (we look at this shortly).

1. XAR can be run from the command line or from a GUI. For the command line enter 'n' when prompted whether to use the GUI or not.

2. There are various options at the command line menu, many of which we have gone through in the pre-assembled example. These include: specifying a domain, selecting a databank, creating a databank, generating attribute worlds, generating tuple worlds, and saving the attribute and tuple worlds as XML.

3. config.txt contains a number of flags, many of which are trace flags for various aspects and also options as to whether to use any deep analyzers as well (in extracting features). These flags, and what the system does when set to 1 are explained:.

FEATURESTRACE : trace the generation of features
RULESTRACE : trace the application of rules
ATTTRACE : trace generation of attribute worlds
TUPTRACE : trace generation of tuple worlds
SCHEMATRACE : schema loading trace
ALLTRACE : activates all above traces
DEEPFEATURES : use deep analyzer (i.e., stanford parser also for feature generation, NOTE: this now takes more time)
REXPLAIN : trace which rule generating which value for every attribute of every slot
EVALUATE : determine precison and recall ( measured against a set of answer keys provided) for each tuple and also aggregated over all tuples.