Command Line Options

Common options

This article supports Phillip version of phil.3.10. <PHILLIP> in this article means the directory where Philip is installed.

• -m MODE :
  Set the execution mode.
  You must choose one among following modes:

  • -m {inference, infer} :
    A mode to perform inference.

  • -m {compile_kb, compile} :
    A mode to compile knowledge-base.

• -h :
  Print a simple help.

• -l PATH :
  Load a config file.
  A config file includes command options in each of lines. -l PATH options can be used also in a config file.
  You can see some examples in the directory <PHILLIP>/conf/

• -p NAME=VALUE :
  Set a parameter.
  To know available parameters, see Parameters and Flags.

• -f NAME :
  Set a flag.
  To know available flags, see Parameters and Flags.

• -t INT :
  Set the number of threads for parallelization.
  The number in default is 1. This option has been abolished from the version of phil.3.10.

• -v INT :
  Set verbosity.
  The value of verbosity must be from 0 to 5. The value in default is 1.

• -P INT :
  Set the number of threads for parallelization.
  The number in default is 1. This option is available from the version of phil.3.10.

Options for compile_kb

• -k PATH :
  Set the prefix of path of compiled knowledge-base.

• -c dist=NAME :
  Set the distance-provider for pre-estimation of reachability. You can currently use one of the following.

  • -c dist=basic :
    The most basic distance-provider, which defines distance of each axiom as fixed number 1.0. Moreover, adding an option formatted in :d<FLOAT> to an implication axiom, you can define its distance as the floating point number.
    For example, the distance of the following axiom is defined as 3.0.

        (B (name foo) (=> (^ (p x) (q y)) (r x) :d3.0))

  • -c dist=cost :
    The distance-provider which define distance of each axiom as its cost in cost-based abduction.
    In cost-based abduction, the optional parameter of an implication axiom is considered as the cost of the axiom.
    For example, the cost of the following axiom is defined as 10.0.

        (B (name foo) (=> (^ (p x) (q y)) (r x) :10.0))

• -c tab=NAME : Set the category table, which enables Phillip to do soft-backchaining and soft-unification. You can currently use one of the following.

  • -c tab=null : Disables the function of the category table, as well as the former versions. In default, this is selected.

  • -c tab=basic :
    Encodes axioms of categorical knowledge (e.g. (=> (dog-nn x) (animal-nn x))) into the category table.

Options for inference

• -c TYPE=NAME :
  Set a component into Phillip.
  Phillip needs following 3 type components to perform inference.

  • -c lhs=NAME :
    Components for making latent hypotheses sets.

    • -c lhs=depth :
      A component for generation based on Henry-like procedure.

    • -c lhs=a* :
      A component for generation based on A* search.

  • -c ilp=NAME :
    Components for conversion latent hypotheses sets into ILP problems.

    • -c ilp=weighted :
      A Component for conversion based on Weighted Abduction.

    • -c ilp=costed :
      A Component for conversion based on Cost-based Abduction.

  • -c sol=NAME :
    Components for optimizing ILP problems.

    • -c sol=gurobi :
      A component to optimize ILP problems with Gurobi optimizer 6.0.

    • -c sol=lpsolve :
      A component to optimize ILP problems with LP-Solve.5.5.

• -k PATH :
  Sets the prefix of path of compiled knowledge-base.

• -o PATH :
  Prints the solution hypothesis in XML format to the given file path. This option is available from phil.3.10.

  • -o lhs=PATH :
    Prints the latent hypotheses set in XML format to the given file path. (This option is mainly used for debug.)

  • -o ilp=PATH :
    Prints the ILP problem in XML format to the given file path. (This option is mainly used for debug.)

  • -o sol=PATH :
    Prints the solution of the ILP problem in XML format to the given file path. (This option is mainly used for debug.)

• -t NAME :
  Specifies the name of observations to solve.
  If NAME begins with !, Specifies the name of observations not to solve.
  For instance, using -t hoo, Phillip solves only the observation by the name of hoo. On the other hand, using -t !var, Phillip solves all observations excluding one by the name of 'var'.
  This option is available from phil.3.10.

• -T INT :
  Sets timeout of inference in seconds.

  • -T lhs=INT : Sets timeout of latent hypotheses set generation in seconds.

  • -T ilp=INT : Sets timeout of ILP problem conversion in seconds.

  • -T sol=INT : Sets timeout of ILP problem optimization in seconds.

• -G :
  Forces to satisfy requirements.
  This option is available from phil.3.10.

• -H :
  Adds the human readable hypothesis to output XML. For example:

      <hypothesis>
      (^ (banana-nn X3) (chimp-nn X2) (dobj E2 X3) (give-vb E2) (hungry-jj E1) (iobj E2 X2) (nsubj E1 E2) (nsubj E2 X1) (scientist-nn X1) (= _u1 _u2 _u3) (= _u0 _u4))
  </hypothesis>