Method decode

extraction/predict_parser/target_predict_parser.py:226–309

:param gold_list: :param pred_list: :param text_list: :param raw_list: :return: dict: pred_event -> [(type1, trigger1), (type2, trigger2), ...] gold_event -> [(type1, trigger1), (type2, trigger2), ...]

(self, gold_list, pred_list, text_list=None, raw_list=None)

class ETPredictParser(PredictParser):
    
    def decode(self, gold_list, pred_list, text_list=None, raw_list=None) -> Tuple[List[Dict], Counter]:
        """

        :param gold_list:
        :param pred_list:
        :param text_list:
        :param raw_list:
        :return:
            dict:
                pred_event -> [(type1, trigger1), (type2, trigger2), ...]
                gold_event -> [(type1, trigger1), (type2, trigger2), ...]
                pred_role -> [(type1, role1, argument1), (type2, role2, argument2), ...]
                gold_role -> [(type1, role1, argument1), (type2, role2, argument2), ...]
            Counter:
        """
        counter = Counter()
        well_formed_list = []

        def convert_bracket(_text):
            _text = add_space(_text)
            for start in [role_start, type_start]:
                _text = _text.replace(start, left_bracket)
            for end in [role_end, type_end]:
                _text = _text.replace(end, right_bracket)
            return _text

        if gold_list is None or len(gold_list) == 0: # 不存在标注信息的情况下根据pred_list 全部置为空
            gold_list = ["%s%s" % (type_start, type_end)] * len(pred_list)

        if text_list is None:
            text_list = [None] * len(pred_list)

        if raw_list is None:
            raw_list = [None] * len(pred_list)

        for gold, pred, text, raw_data in zip(gold_list, pred_list, text_list, raw_list):
            # print(gold)
            # print("*************************************")
            gold = convert_bracket(gold)
            pred = convert_bracket(pred)

            gold = clean_text(gold)
            pred = clean_text(pred)
            # print(gold)

            et = None # 事件类型
            rt = None # 角色类型
            if text and specical_str in text:
                et = text.split(specical_str)[0].strip()
                rt = text.split(specical_str)[1].strip()
                text = text.split(specical_str)[-1].strip()

            instance = {'gold': gold,
                        'pred': pred,
                        'gold_tree': None,
                        'text': text,
                        'raw_data': raw_data
                        }