def __init__()

in ttw/data_loader.py [0:0]

• 51 lines of code
• 11 McCabe index (conditional complexity)

    def __init__(self, data_dir, set, last_turns=1, min_freq=3, min_sent_len=2, orientation_aware=False,
                 include_guide_utterances=True):
        self.dialogues = json.load(open(os.path.join(data_dir, 'talkthewalk.{}.json'.format(set))))
        self.dict = Dictionary(file=os.path.join(data_dir, 'dict.txt'), min_freq=min_freq)
        self.map = Map(data_dir, neighborhoods, include_empty_corners=True)
        self.act_dict = ActionAgnosticDictionary()
        self.act_aware_dict = ActionAwareDictionary()

        self.feature_loader = GoldstandardFeatures(self.map)

        self.data = dict()
        self.data['actions'] = list()
        self.data['goldstandard'] = list()
        self.data['landmarks'] = list()
        self.data['target'] = list()
        self.data['utterance'] = list()

        for config in self.dialogues:
            loc = config['start_location']
            neighborhood = config['neighborhood']
            boundaries = config['boundaries']
            act_memory = list()
            obs_memory = [self.feature_loader.get(neighborhood, loc)]

            dialogue_context = list()
            for msg in config['dialog']:
                if msg['id'] == 'Tourist':
                    act = msg['text']
                    act_id = self.act_aware_dict.encode(act)
                    if act_id >= 0:
                        new_loc = step_aware(act, loc, boundaries)
                        old_loc = loc
                        loc = new_loc

                        if orientation_aware:
                            act_memory.append(act_id)
                            obs_memory.append(self.feature_loader.get(neighborhood, new_loc))
                        else:
                            if act == 'ACTION:FORWARD':  # went forward
                                act_dir = self.act_dict.encode_from_location(old_loc, new_loc)
                                act_memory.append(act_dir)
                                obs_memory.append(self.feature_loader.get(neighborhood, loc))
                    elif len(msg['text'].split(' ')) > min_sent_len:
                        dialogue_context.append(self.dict.encode(msg['text']))
                        utt = self.dict.encode(START_TOKEN) + [y for x in dialogue_context[-last_turns:] for y in x] \
                              + self.dict.encode(END_TOKEN)
                        self.data['utterance'].append(utt)

                        landmarks, tgt = self.map.get_landmarks(config['neighborhood'], boundaries, loc)
                        self.data['landmarks'].append(landmarks)
                        self.data['target'].append(tgt)

                        self.data['actions'].append(act_memory)
                        self.data['goldstandard'].append(obs_memory)

                        act_memory = list()
                        obs_memory = [self.feature_loader.get(neighborhood, loc)]
                elif include_guide_utterances:
                    dialogue_context.append(self.dict.encode(msg['text']))