Ejemplo n.º 1
0
    def __init__(self):
        self.builder = StatementBuilder()
        self.adder = StatementSafeAdder()
        self.question_handler = QuestionHandler()
        self.sfactory = SentenceFactory()

        self.output_sentence = []
Ejemplo n.º 2
0
    def __init__(self):
        self.builder = StatementBuilder()
        self.adder = StatementSafeAdder()
        self.question_handler = QuestionHandler()
        self.sfactory = SentenceFactory()

        self.output_sentence = []
Ejemplo n.º 3
0
class ContentAnalyser(object):
    def __init__(self):
        self.builder = StatementBuilder()
        self.adder = StatementSafeAdder()
        self.question_handler = QuestionHandler()
        self.sfactory = SentenceFactory()

        self.output_sentence = []

    def analyse(self, sentence, current_speaker):
        """Analyse a sentence intent and perform the corresponding behaviour.

        :returns: a pair with
         - a list of symbolic statements produced
        during the analyse. Note that these statements have 
        already been sent to the knowledge base
         - a situation ID that identifies a situation the speaker is
         desiring (for an imperative sentence or an order) or 
         experiencing (like an 'InterrogativeState' for a question).

         These two values can be None.

        Note also that an answer to the speaker is stored in self.output_sentence
        after analyse() completes.

        It can be used to tell the human to acknowledge an order, answer a 
        gratulation, the answer a question (or the fact the answer is not known),
        etc.
        """

        self.builder.clear_all()
        self.output_sentence = []

        sentence = self.pre_analyse_content(sentence)

        if sentence.data_type == [INTERJECTION, EXCLAMATION]:
            pass

        if sentence.data_type in [START, END]:
            self.output_sentence.append(sentence)

        if sentence.data_type == GRATULATION:
            self.output_sentence.extend(
                self.sfactory.create_gratulation_reply())

        if sentence.data_type in [AGREEMENT, DISAGREEMENT]:
            self.output_sentence.extend(self.sfactory.create_agree_reply())

        if sentence.data_type in [IMPERATIVE, STATEMENT]:
            logger.debug(
                colored_print(
                    "Processing the content of " +
                    ("an imperative sentence" if sentence.data_type
                     == IMPERATIVE else "a statement "), "magenta"))
            return self.process_sentence(sentence, current_speaker)

        if sentence.data_type in [W_QUESTION, YES_NO_QUESTION]:
            logger.debug(
                colored_print(
                    "Processing the content of " +
                    ("a W question " if sentence.data_type == W_QUESTION else
                     "a YES/NO question"), "magenta"))
            return self.process_question(sentence, current_speaker)

        return None, None  # default: no statement generated, no situation ID

    def process_sentence(self, sentence, current_speaker):
        self.builder.set_current_speaker(current_speaker)

        stmts, situation_id = self.builder.process_sentence(sentence)

        if stmts:

            self.add_stmts(stmts)

            emotions.satisfied()

            if situation_id:
                # If a new situation has been created, mark it as
                # active.
                self.output_sentence.extend(self.sfactory.create_agree_reply())
                ResourcePool().mark_active(situation_id)
        else:
            logger.info("No statements produced")

        # Class grounding
        if self.builder.lear_more_concept:
            self.output_sentence.extend(
                self.sfactory.create_what_is_a_reference(
                    sentence, self.builder.lear_more_concept))

        return stmts, situation_id

    def process_question(self, sentence, current_speaker):
        self.question_handler.set_current_speaker(current_speaker)

        # 'stmts' contains a list of statement describing the current
        # 'interrogative state' of the interactor.
        answer, stmts, situation_id = self.question_handler.process_sentence(
            sentence)

        if stmts:
            self.add_stmts(stmts)
        else:
            logger.info("No statements produced")

        if answer:
            emotions.satisfied()
            logger.info(
                level_marker(level=2, color="yellow") + "Found: \n" +
                colored_print(str(answer), None, "magenta"))
        else:
            emotions.sorry()
            logger.info(
                level_marker(level=2, color="yellow") +
                "Couldn't find anything!")

        if sentence.data_type == W_QUESTION:
            self.output_sentence.extend(
                self.sfactory.create_w_question_answer(
                    sentence, answer, self.question_handler._current_speaker,
                    self.question_handler.get_query_on_field()))

        if sentence.data_type == YES_NO_QUESTION:
            self.output_sentence.extend(
                self.sfactory.create_yes_no_answer(sentence, answer))

        return stmts, situation_id

    def add_stmts(self, stmts):
        logger.info("Generated statements: ")
        for s in stmts:
            logger.info(">> " + colored_print(s, None, 'magenta'))

        self.adder._current_speaker = self.builder._current_speaker
        self.adder._unclarified_ids = self.builder._unclarified_ids
        self.adder._statements = stmts
        self.adder._statements_to_remove = self.builder._statements_to_remove
        stmts = self.adder.process()

        logger.debug("...added to the ontology")

    def analyse_output(self):
        return self.output_sentence

    def pre_analyse_content(self, sentence):
        """ this method analyse the content of a sentence and possibly changes its purpose.

        E.g: Can you give me the bottle?
        The sentence above is of YES_NO_QUESTION type but should actually
        be processed as an order in which the current speaker desires 'the
        bottle'. Therefore, we turn it into 'give me the bottle'.
        """
        # Case of :
        #   -INPUT:  Yes_no_question + can + action verb
        #   -OUTPUT: Imperative + action verb
        #

        if sentence.data_type == YES_NO_QUESTION:
            for sv in sentence.sv:
                for verb in sv.vrb_main:
                    if 'can+' in verb:

                        vrb_main = verb.lstrip('can+')

                        if not vrb_main in ResourcePool().state + ResourcePool(
                        ).action_verb_with_passive_behaviour.keys(
                        ) + ResourcePool().goal_verbs:
                            logger.debug(
                                colored_print(
                                    "Interpreting the <can + action verb> sequence as a desire.\nApplying transformation:",
                                    "magenta"))

                            sv.vrb_main[sv.vrb_main.index(verb)] = verb.lstrip(
                                'can+')
                            sentence.data_type = IMPERATIVE

                            logger.debug(str(sentence))

                            return sentence

        return sentence
Ejemplo n.º 4
0
class ContentAnalyser(object):
    def __init__(self):
        self.builder = StatementBuilder()
        self.adder = StatementSafeAdder()
        self.question_handler = QuestionHandler()
        self.sfactory = SentenceFactory()

        self.output_sentence = []

    def analyse(self, sentence, current_speaker):
        """Analyse a sentence intent and perform the corresponding behaviour.

        :returns: a pair with
         - a list of symbolic statements produced
        during the analyse. Note that these statements have 
        already been sent to the knowledge base
         - a situation ID that identifies a situation the speaker is
         desiring (for an imperative sentence or an order) or 
         experiencing (like an 'InterrogativeState' for a question).

         These two values can be None.

        Note also that an answer to the speaker is stored in self.output_sentence
        after analyse() completes.

        It can be used to tell the human to acknowledge an order, answer a 
        gratulation, the answer a question (or the fact the answer is not known),
        etc.
        """

        self.builder.clear_all()
        self.output_sentence = []

        sentence = self.pre_analyse_content(sentence)

        if sentence.data_type == [INTERJECTION, EXCLAMATION]:
            pass

        if sentence.data_type in [START, END]:
            self.output_sentence.append(sentence)

        if sentence.data_type == GRATULATION:
            self.output_sentence.extend(self.sfactory.create_gratulation_reply())

        if sentence.data_type in [AGREEMENT, DISAGREEMENT]:
            self.output_sentence.extend(self.sfactory.create_agree_reply())

        if sentence.data_type in [IMPERATIVE, STATEMENT]:
            logger.debug(colored_print("Processing the content of " + (
                "an imperative sentence" if sentence.data_type == IMPERATIVE else "a statement "), "magenta"))
            return self.process_sentence(sentence, current_speaker)

        if sentence.data_type in [W_QUESTION, YES_NO_QUESTION]:
            logger.debug(colored_print("Processing the content of " + (
                "a W question " if sentence.data_type == W_QUESTION else "a YES/NO question"), "magenta"))
            return self.process_question(sentence, current_speaker)

        return None, None # default: no statement generated, no situation ID

    def process_sentence(self, sentence, current_speaker):
        self.builder.set_current_speaker(current_speaker)

        stmts, situation_id = self.builder.process_sentence(sentence)

        if stmts:

            self.add_stmts(stmts)

            emotions.satisfied()

            if situation_id:
                # If a new situation has been created, mark it as
                # active.
                self.output_sentence.extend(self.sfactory.create_agree_reply())
                ResourcePool().mark_active(situation_id)
        else:
            logger.info("No statements produced")

        # Class grounding
        if self.builder.lear_more_concept:
            self.output_sentence.extend(
                self.sfactory.create_what_is_a_reference(sentence, self.builder.lear_more_concept))

        return stmts, situation_id


    def process_question(self, sentence, current_speaker):
        self.question_handler.set_current_speaker(current_speaker)

        # 'stmts' contains a list of statement describing the current 
        # 'interrogative state' of the interactor.
        answer, stmts, situation_id = self.question_handler.process_sentence(sentence)

        if stmts:
            self.add_stmts(stmts)
        else:
            logger.info("No statements produced")

        if answer:
            emotions.satisfied()
            logger.info(
                level_marker(level=2, color="yellow") + "Found: \n" + colored_print(str(answer), None, "magenta"))
        else:
            emotions.sorry()
            logger.info(level_marker(level=2, color="yellow") + "Couldn't find anything!")

        if sentence.data_type == W_QUESTION:
            self.output_sentence.extend(self.sfactory.create_w_question_answer(sentence,
                                                                               answer,
                                                                               self.question_handler._current_speaker,
                                                                               self.question_handler.get_query_on_field()))

        if sentence.data_type == YES_NO_QUESTION:
            self.output_sentence.extend(self.sfactory.create_yes_no_answer(sentence, answer))

        return stmts, situation_id


    def add_stmts(self, stmts):
        logger.info("Generated statements: ")
        for s in stmts:
            logger.info(">> " + colored_print(s, None, 'magenta'))

        self.adder._current_speaker = self.builder._current_speaker
        self.adder._unclarified_ids = self.builder._unclarified_ids
        self.adder._statements = stmts
        self.adder._statements_to_remove = self.builder._statements_to_remove
        stmts = self.adder.process()

        logger.debug("...added to the ontology")

    def analyse_output(self):
        return self.output_sentence

    def pre_analyse_content(self, sentence):
        """ this method analyse the content of a sentence and possibly changes its purpose.

        E.g: Can you give me the bottle?
        The sentence above is of YES_NO_QUESTION type but should actually
        be processed as an order in which the current speaker desires 'the
        bottle'. Therefore, we turn it into 'give me the bottle'.
        """
        # Case of : 
        #   -INPUT:  Yes_no_question + can + action verb
        #   -OUTPUT: Imperative + action verb
        #   

        if sentence.data_type == YES_NO_QUESTION:
            for sv in sentence.sv:
                for verb in sv.vrb_main:
                    if 'can+' in verb:

                        vrb_main = verb.lstrip('can+')

                        if not vrb_main in ResourcePool().state + ResourcePool().action_verb_with_passive_behaviour.keys() + ResourcePool().goal_verbs:
                            logger.debug(colored_print(
                                "Interpreting the <can + action verb> sequence as a desire.\nApplying transformation:",
                                "magenta"))

                            sv.vrb_main[sv.vrb_main.index(verb)] = verb.lstrip('can+')
                            sentence.data_type = IMPERATIVE

                            logger.debug(str(sentence))

                            return sentence

        return sentence