duee篇章级预测结果后处理

import os
import sys
import json

def read_by_lines(path):
    result = list()
    with open(path, "r", encoding="utf8") as infile:
        for line in infile:
            result.append(line.strip())
    return result

def write_by_lines(path, data):
    with open(path, "w", encoding="utf8") as outfile:
        [outfile.write(d + "\n") for d in data]

def extract_result(text, labels):
    ret, is_start, cur_type = [], False, None
    if len(text) != len(labels):
        labels = labels[:len(text)]#截断
    for i, label in enumerate(labels):
        if label != u"O":#实体
            _type = label[2:]#事件类型或角色类型
            if label.startswith(u"B-"):#以B-开头
                is_start = True
                cur_type = _type
                #这是一个实体的开始
                ret.append({"start": i, "text": [text[i]], "type": _type})
            elif _type != cur_type:#
                cur_type = _type
                is_start = True
                ret.append({"start": i, "text": [text[i]], "type": _type})#新实体
            elif is_start: #如果还在当前实体内,追加文本
                ret[-1]["text"].append(text[i])
            else:
                cur_type = None
                is_start = False
        else:#非实体
            cur_type = None
            is_start = False
    return ret

trigger_file='./ckp