How to use the nlu.messages.msgutils.extract_close_keywords function in nlu

if subjects:
            self.frame['subject'] = [get_node_string(subject)
                                     for subject in subjects]
        words_temporary_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_pos,
                                       tokenized_string,
                                       2)
        words_temporary_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_neg,
                                       tokenized_string,
                                       2)
        words_permanent_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_pos,
                                       tokenized_string,
                                       2)
        words_permanent_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_neg,
                                       tokenized_string,
                                       2)
        words_temporary = words_temporary_pos + words_temporary_neg
        words_permanent = words_permanent_pos + words_permanent_neg
        if words_temporary and words_permanent:
            # Confused
            # self.frame['temporal'] = None
            # self.frame['word'] = None
            # This check is skipped due to an error in not using the POS
            # when looking up synsets.
            # TODO: Fix (example: fish)
            pass
        if words_temporary:
            self.frame['temporal'] = 'temporary'
            self.frame['word'] = words_temporary[0]

tokenized_string = g.generate_tokenized_string(raw_input_string)
        parseTree = g.generate_stanford_parse_tree(raw_input_string)
        
        subjects = extract_subject_nodes(parseTree)
        if subjects:
            self.frame['subject'] = [get_node_string(subject)
                                     for subject in subjects]
        words_temporary_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_pos,
                                       tokenized_string,
                                       2)
        words_temporary_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_neg,
                                       tokenized_string,
                                       2)
        words_permanent_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_pos,
                                       tokenized_string,
                                       2)
        words_permanent_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_neg,
                                       tokenized_string,
                                       2)
        words_temporary = words_temporary_pos + words_temporary_neg
        words_permanent = words_permanent_pos + words_permanent_neg
        if words_temporary and words_permanent:
            # Confused
            # self.frame['temporal'] = None
            # self.frame['word'] = None
            # This check is skipped due to an error in not using the POS
            # when looking up synsets.
            # TODO: Fix (example: fish)

def _parse(self, raw_input_string, g):
        """
        Fills out message meta and frame attributes.
        """
        tokenized_string = g.generate_tokenized_string(raw_input_string)
        parseTree = g.generate_stanford_parse_tree(raw_input_string)
        
        subjects = extract_subject_nodes(parseTree)
        if subjects:
            self.frame['subject'] = [get_node_string(subject)
                                     for subject in subjects]
        words_temporary_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_pos,
                                       tokenized_string,
                                       2)
        words_temporary_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_neg,
                                       tokenized_string,
                                       2)
        words_permanent_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_pos,
                                       tokenized_string,
                                       2)
        words_permanent_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_neg,
                                       tokenized_string,
                                       2)
        words_temporary = words_temporary_pos + words_temporary_neg

def _parse(self, raw_input_string):
        """
        Fills out message meta and frame attributes
        """
       
        tokenizer = nltk.WordPunctTokenizer()
        tokenized_string = tokenizer.tokenize(raw_input_string)
        tagger = utils.combined_taggers
        tagged_string = tagger.tag(tokenized_string)

        wordActionMap = {'exit':SystemMessage.exit_keywords, 'restart':SystemMessage.restart_keywords}
        for action, keywords in wordActionMap.items():
            matches = extract_close_keywords(keywords, tokenized_string, 3)
            if matches: # synset of keyword was found in the sentence
                self.frame['action'] = action

def _parse(self, raw_input_string, g):
        """
        Fills out message meta and frame attributes.
        """
        tokenized_string = g.generate_tokenized_string(raw_input_string)
        parseTree = g.generate_stanford_parse_tree(raw_input_string)
        
        subjects = extract_subject_nodes(parseTree)
        if subjects:
            self.frame['subject'] = [get_node_string(subject)
                                     for subject in subjects]
        words_temporary_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_pos,
                                       tokenized_string,
                                       2)
        words_temporary_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_temporary_neg,
                                       tokenized_string,
                                       2)
        words_permanent_pos = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_pos,
                                       tokenized_string,
                                       2)
        words_permanent_neg = extract_close_keywords(
                                       PreferenceMessage.keywords_permanent_neg,
                                       tokenized_string,
                                       2)
        words_temporary = words_temporary_pos + words_temporary_neg
        words_permanent = words_permanent_pos + words_permanent_neg
        if words_temporary and words_permanent:
            # Confused
            # self.frame['temporal'] = None