Source code for nltk.inference.api

# Natural Language Toolkit: Classifier Interface
#
# Author: Ewan Klein <ewan@inf.ed.ac.uk>
#         Dan Garrette <dhgarrette@gmail.com>
#
# URL: <https://www.nltk.org/>
# For license information, see LICENSE.TXT

"""
Interfaces and base classes for theorem provers and model builders.

``Prover`` is a standard interface for a theorem prover which tries to prove a goal from a
list of assumptions.

``ModelBuilder`` is a standard interface for a model builder. Given just a set of assumptions.
the model builder tries to build a model for the assumptions. Given a set of assumptions and a
goal *G*, the model builder tries to find a counter-model, in the sense of a model that will satisfy
the assumptions plus the negation of *G*.
"""

import threading
import time
from abc import ABCMeta, abstractmethod


[docs]class Prover(metaclass=ABCMeta): """ Interface for trying to prove a goal from assumptions. Both the goal and the assumptions are constrained to be formulas of ``logic.Expression``. """
[docs] def prove(self, goal=None, assumptions=None, verbose=False): """ :return: Whether the proof was successful or not. :rtype: bool """ return self._prove(goal, assumptions, verbose)[0]
@abstractmethod def _prove(self, goal=None, assumptions=None, verbose=False): """ :return: Whether the proof was successful or not, along with the proof :rtype: tuple: (bool, str) """
[docs]class ModelBuilder(metaclass=ABCMeta): """ Interface for trying to build a model of set of formulas. Open formulas are assumed to be universally quantified. Both the goal and the assumptions are constrained to be formulas of ``logic.Expression``. """
[docs] def build_model(self, goal=None, assumptions=None, verbose=False): """ Perform the actual model building. :return: Whether a model was generated :rtype: bool """ return self._build_model(goal, assumptions, verbose)[0]
@abstractmethod def _build_model(self, goal=None, assumptions=None, verbose=False): """ Perform the actual model building. :return: Whether a model was generated, and the model itself :rtype: tuple(bool, sem.Valuation) """
[docs]class TheoremToolCommand(metaclass=ABCMeta): """ This class holds a goal and a list of assumptions to be used in proving or model building. """
[docs] @abstractmethod def add_assumptions(self, new_assumptions): """ Add new assumptions to the assumption list. :param new_assumptions: new assumptions :type new_assumptions: list(sem.Expression) """
[docs] @abstractmethod def retract_assumptions(self, retracted, debug=False): """ Retract assumptions from the assumption list. :param debug: If True, give warning when ``retracted`` is not present on assumptions list. :type debug: bool :param retracted: assumptions to be retracted :type retracted: list(sem.Expression) """
[docs] @abstractmethod def assumptions(self): """ List the current assumptions. :return: list of ``Expression`` """
[docs] @abstractmethod def goal(self): """ Return the goal :return: ``Expression`` """
[docs] @abstractmethod def print_assumptions(self): """ Print the list of the current assumptions. """
[docs]class ProverCommand(TheoremToolCommand): """ This class holds a ``Prover``, a goal, and a list of assumptions. When prove() is called, the ``Prover`` is executed with the goal and assumptions. """
[docs] @abstractmethod def prove(self, verbose=False): """ Perform the actual proof. """
[docs] @abstractmethod def proof(self, simplify=True): """ Return the proof string :param simplify: bool simplify the proof? :return: str """
[docs] @abstractmethod def get_prover(self): """ Return the prover object :return: ``Prover`` """
[docs]class ModelBuilderCommand(TheoremToolCommand): """ This class holds a ``ModelBuilder``, a goal, and a list of assumptions. When build_model() is called, the ``ModelBuilder`` is executed with the goal and assumptions. """
[docs] @abstractmethod def build_model(self, verbose=False): """ Perform the actual model building. :return: A model if one is generated; None otherwise. :rtype: sem.Valuation """
[docs] @abstractmethod def model(self, format=None): """ Return a string representation of the model :param simplify: bool simplify the proof? :return: str """
[docs] @abstractmethod def get_model_builder(self): """ Return the model builder object :return: ``ModelBuilder`` """
[docs]class BaseTheoremToolCommand(TheoremToolCommand): """ This class holds a goal and a list of assumptions to be used in proving or model building. """
[docs] def __init__(self, goal=None, assumptions=None): """ :param goal: Input expression to prove :type goal: sem.Expression :param assumptions: Input expressions to use as assumptions in the proof. :type assumptions: list(sem.Expression) """ self._goal = goal if not assumptions: self._assumptions = [] else: self._assumptions = list(assumptions) self._result = None """A holder for the result, to prevent unnecessary re-proving"""
[docs] def add_assumptions(self, new_assumptions): """ Add new assumptions to the assumption list. :param new_assumptions: new assumptions :type new_assumptions: list(sem.Expression) """ self._assumptions.extend(new_assumptions) self._result = None
[docs] def retract_assumptions(self, retracted, debug=False): """ Retract assumptions from the assumption list. :param debug: If True, give warning when ``retracted`` is not present on assumptions list. :type debug: bool :param retracted: assumptions to be retracted :type retracted: list(sem.Expression) """ retracted = set(retracted) result_list = list(filter(lambda a: a not in retracted, self._assumptions)) if debug and result_list == self._assumptions: print(Warning("Assumptions list has not been changed:")) self.print_assumptions() self._assumptions = result_list self._result = None
[docs] def assumptions(self): """ List the current assumptions. :return: list of ``Expression`` """ return self._assumptions
[docs] def goal(self): """ Return the goal :return: ``Expression`` """ return self._goal
[docs] def print_assumptions(self): """ Print the list of the current assumptions. """ for a in self.assumptions(): print(a)
[docs]class BaseProverCommand(BaseTheoremToolCommand, ProverCommand): """ This class holds a ``Prover``, a goal, and a list of assumptions. When prove() is called, the ``Prover`` is executed with the goal and assumptions. """
[docs] def __init__(self, prover, goal=None, assumptions=None): """ :param prover: The theorem tool to execute with the assumptions :type prover: Prover :see: ``BaseTheoremToolCommand`` """ self._prover = prover """The theorem tool to execute with the assumptions""" BaseTheoremToolCommand.__init__(self, goal, assumptions) self._proof = None
[docs] def prove(self, verbose=False): """ Perform the actual proof. Store the result to prevent unnecessary re-proving. """ if self._result is None: self._result, self._proof = self._prover._prove( self.goal(), self.assumptions(), verbose ) return self._result
[docs] def proof(self, simplify=True): """ Return the proof string :param simplify: bool simplify the proof? :return: str """ if self._result is None: raise LookupError("You have to call prove() first to get a proof!") else: return self.decorate_proof(self._proof, simplify)
[docs] def decorate_proof(self, proof_string, simplify=True): """ Modify and return the proof string :param proof_string: str the proof to decorate :param simplify: bool simplify the proof? :return: str """ return proof_string
[docs] def get_prover(self): return self._prover
[docs]class BaseModelBuilderCommand(BaseTheoremToolCommand, ModelBuilderCommand): """ This class holds a ``ModelBuilder``, a goal, and a list of assumptions. When build_model() is called, the ``ModelBuilder`` is executed with the goal and assumptions. """
[docs] def __init__(self, modelbuilder, goal=None, assumptions=None): """ :param modelbuilder: The theorem tool to execute with the assumptions :type modelbuilder: ModelBuilder :see: ``BaseTheoremToolCommand`` """ self._modelbuilder = modelbuilder """The theorem tool to execute with the assumptions""" BaseTheoremToolCommand.__init__(self, goal, assumptions) self._model = None
[docs] def build_model(self, verbose=False): """ Attempt to build a model. Store the result to prevent unnecessary re-building. """ if self._result is None: self._result, self._model = self._modelbuilder._build_model( self.goal(), self.assumptions(), verbose ) return self._result
[docs] def model(self, format=None): """ Return a string representation of the model :param simplify: bool simplify the proof? :return: str """ if self._result is None: raise LookupError("You have to call build_model() first to " "get a model!") else: return self._decorate_model(self._model, format)
def _decorate_model(self, valuation_str, format=None): """ :param valuation_str: str with the model builder's output :param format: str indicating the format for displaying :return: str """ return valuation_str
[docs] def get_model_builder(self): return self._modelbuilder
[docs]class TheoremToolCommandDecorator(TheoremToolCommand): """ A base decorator for the ``ProverCommandDecorator`` and ``ModelBuilderCommandDecorator`` classes from which decorators can extend. """
[docs] def __init__(self, command): """ :param command: ``TheoremToolCommand`` to decorate """ self._command = command # The decorator has its own versions of 'result' different from the # underlying command self._result = None
[docs] def assumptions(self): return self._command.assumptions()
[docs] def goal(self): return self._command.goal()
[docs] def add_assumptions(self, new_assumptions): self._command.add_assumptions(new_assumptions) self._result = None
[docs] def retract_assumptions(self, retracted, debug=False): self._command.retract_assumptions(retracted, debug) self._result = None
[docs] def print_assumptions(self): self._command.print_assumptions()
[docs]class ProverCommandDecorator(TheoremToolCommandDecorator, ProverCommand): """ A base decorator for the ``ProverCommand`` class from which other prover command decorators can extend. """
[docs] def __init__(self, proverCommand): """ :param proverCommand: ``ProverCommand`` to decorate """ TheoremToolCommandDecorator.__init__(self, proverCommand) # The decorator has its own versions of 'result' and 'proof' # because they may be different from the underlying command self._proof = None
[docs] def prove(self, verbose=False): if self._result is None: prover = self.get_prover() self._result, self._proof = prover._prove( self.goal(), self.assumptions(), verbose ) return self._result
[docs] def proof(self, simplify=True): """ Return the proof string :param simplify: bool simplify the proof? :return: str """ if self._result is None: raise LookupError("You have to call prove() first to get a proof!") else: return self.decorate_proof(self._proof, simplify)
[docs] def decorate_proof(self, proof_string, simplify=True): """ Modify and return the proof string :param proof_string: str the proof to decorate :param simplify: bool simplify the proof? :return: str """ return self._command.decorate_proof(proof_string, simplify)
[docs] def get_prover(self): return self._command.get_prover()
[docs]class ModelBuilderCommandDecorator(TheoremToolCommandDecorator, ModelBuilderCommand): """ A base decorator for the ``ModelBuilderCommand`` class from which other prover command decorators can extend. """
[docs] def __init__(self, modelBuilderCommand): """ :param modelBuilderCommand: ``ModelBuilderCommand`` to decorate """ TheoremToolCommandDecorator.__init__(self, modelBuilderCommand) # The decorator has its own versions of 'result' and 'valuation' # because they may be different from the underlying command self._model = None
[docs] def build_model(self, verbose=False): """ Attempt to build a model. Store the result to prevent unnecessary re-building. """ if self._result is None: modelbuilder = self.get_model_builder() self._result, self._model = modelbuilder._build_model( self.goal(), self.assumptions(), verbose ) return self._result
[docs] def model(self, format=None): """ Return a string representation of the model :param simplify: bool simplify the proof? :return: str """ if self._result is None: raise LookupError("You have to call build_model() first to " "get a model!") else: return self._decorate_model(self._model, format)
def _decorate_model(self, valuation_str, format=None): """ Modify and return the proof string :param valuation_str: str with the model builder's output :param format: str indicating the format for displaying :return: str """ return self._command._decorate_model(valuation_str, format)
[docs] def get_model_builder(self): return self._command.get_prover()
[docs]class ParallelProverBuilder(Prover, ModelBuilder): """ This class stores both a prover and a model builder and when either prove() or build_model() is called, then both theorem tools are run in parallel. Whichever finishes first, the prover or the model builder, is the result that will be used. """
[docs] def __init__(self, prover, modelbuilder): self._prover = prover self._modelbuilder = modelbuilder
def _prove(self, goal=None, assumptions=None, verbose=False): return self._run(goal, assumptions, verbose), "" def _build_model(self, goal=None, assumptions=None, verbose=False): return not self._run(goal, assumptions, verbose), "" def _run(self, goal, assumptions, verbose): # Set up two thread, Prover and ModelBuilder to run in parallel tp_thread = TheoremToolThread( lambda: self._prover.prove(goal, assumptions, verbose), verbose, "TP" ) mb_thread = TheoremToolThread( lambda: self._modelbuilder.build_model(goal, assumptions, verbose), verbose, "MB", ) tp_thread.start() mb_thread.start() while tp_thread.is_alive() and mb_thread.is_alive(): # wait until either the prover or the model builder is done pass if tp_thread.result is not None: return tp_thread.result elif mb_thread.result is not None: return not mb_thread.result else: return None
[docs]class ParallelProverBuilderCommand(BaseProverCommand, BaseModelBuilderCommand): """ This command stores both a prover and a model builder and when either prove() or build_model() is called, then both theorem tools are run in parallel. Whichever finishes first, the prover or the model builder, is the result that will be used. Because the theorem prover result is the opposite of the model builder result, we will treat self._result as meaning "proof found/no model found". """
[docs] def __init__(self, prover, modelbuilder, goal=None, assumptions=None): BaseProverCommand.__init__(self, prover, goal, assumptions) BaseModelBuilderCommand.__init__(self, modelbuilder, goal, assumptions)
[docs] def prove(self, verbose=False): return self._run(verbose)
[docs] def build_model(self, verbose=False): return not self._run(verbose)
def _run(self, verbose): # Set up two thread, Prover and ModelBuilder to run in parallel tp_thread = TheoremToolThread( lambda: BaseProverCommand.prove(self, verbose), verbose, "TP" ) mb_thread = TheoremToolThread( lambda: BaseModelBuilderCommand.build_model(self, verbose), verbose, "MB" ) tp_thread.start() mb_thread.start() while tp_thread.is_alive() and mb_thread.is_alive(): # wait until either the prover or the model builder is done pass if tp_thread.result is not None: self._result = tp_thread.result elif mb_thread.result is not None: self._result = not mb_thread.result return self._result
[docs]class TheoremToolThread(threading.Thread):
[docs] def __init__(self, command, verbose, name=None): threading.Thread.__init__(self) self._command = command self._result = None self._verbose = verbose self._name = name
[docs] def run(self): try: self._result = self._command() if self._verbose: print( "Thread %s finished with result %s at %s" % (self._name, self._result, time.localtime(time.time())) ) except Exception as e: print(e) print("Thread %s completed abnormally" % (self._name))
@property def result(self): return self._result