Boxer Options

Boxer supports various command-line options. If you run boxer without any arguments you will get a brief overview of all options and range of values. Here follows a brief description of the (more or less stable) options:

--help

Shows all currently supported options.

--version

Shows the version number of Boxer.

--stdin

Allows one to use Boxer reading a file from standard input.

--input <file>

This specifies the input file containing a CCG derivation. Usually this is the file output by the C&C parser using the options --models models/boxer and --candc-printer boxer. The input file can also be an unresolved DRS (see below under ---inputtype).

--output <file>

This specifies the output file. If you omit this option, output will be written to standard output.

--box <arg>

When set to true this prints an additional pretty-printed DRSs in box format, as we know it from Kamp's Discourse Representation Theory. Because this output format looses some information, it is not meant for further processing, but only for looking at! Default value is false.

--copula <arg>

When set to true the copula will introduce an equality condition (this is the default value). When set to false it will be analysed as a transitive verb in a neo-Davidsonian fashion.

--elimeq <arg>

When set to true all equalities are removed from the DRSs and discourse referents standing in the equality relation are unified, but only if this can be done in a meaning-preserving manner. Default value is false.

--flat <arg>

When set to true the DRSs are output in a flat list rather than a recursive structure. This is done by introducing labels that point to DRSs or DRS-conditions. Default: false.

--format <arg>

Currently boxer supports two output formats: prolog (default) and xml. The DTDs for the XML output are in directory src/data/boxer. You can also give it the value no in which case it doesn't output anything (but it could still print boxes if --box is set to true).

--instantiate <arg>

When set to true will generate Prolog atoms for all discourse referents. When set to false, the boxer will use Prolog variables for discourse referents. The latter is the default value.

--resolve <arg>

When set to true, boxer will resolve all alpha-DRSs and perform merge-reduction. Resolution follows Van der Sandt's theory of binding and accommodation. Default: false.

--vpe <arg>

When set to true (and --resolve is set to true as well) boxer will attempt to detect verb phrase ellipses and resolve them. (This is under construction.) Default: false.

--robust <arg>

When set to true, boxer applies some recovery rules in case a purely compositional analysis on the CCG derivation fails. This generally leads to a higher coverage, but also to lower quality of the analysis.

--semantics <arg>

The basic (and default) formalism of semantics is drs, but other formats are also possible: fol (first-order formula syntax); triples (semantics in form of triples); sdrs (segmented DRSs); ccg (CCG derivation, nicely printed).

--tense <arg>

When set to true tense is represented following Kamp & Reyle. When set to false the semantic analysis of tense is ignored.

--verbnet <arg>

Uses the semantics roles of Verbnet in the neo-Davidsonian semantics.

--warnings <arg>

If you want to see all the error and warning messages produced by boxer, set this option to true. These are then written to standard error. Default: false.

--window <arg>

This sets the size of the window (in number of sentences) to create a DRS. The idea is that when set to 1 (the default), boxer generates a DRS for each sentence. When set to 2, boxer will generate a DRS spanning each two sentences. This option is under development and it is unclear whether it will be in future releases.