three_parse_structure.py 20 KB

123456789101112131415161718192021222324252627282930313233343536373839404142434445464748495051525354555657585960616263646566676869707172737475767778798081828384858687888990919293949596979899100101102103104105106107108109110111112113114115116117118119120121122123124125126127128129130131132133134135136137138139140141142143144145146147148149150151152153154155156157158159160161162163164165166167168169170171172173174175176177178179180181182183184185186187188189190191192193194195196197198199200201202203204205206207208209210211212213214215216217218219220221222223224225226227228229230231232233234235236237238239240241242243244245246247248249250251252253254255256257258259260261262263264265266267268269270271272273274275276277278279280281282283284285286287288289290291292293294295296297298299300301302303304305306307308309310311312313314315316317318319320321322323324325326327328329330331332333334335336337338339340341342343344345346347348349350351352353354355356357358359360361362363364365366367368369370371372373374375376377378379380
  1. #!/usr/bin/env/python
  2. # -*- coding:utf-8 -*-
  3. # paper3_process: 第三类word试卷模式, 题目和答案分开的情况
  4. # split2one_item:将所有行文本 按题型分大类,再在每个大类中切分每个题目
  5. # split2one_item_by_topicno:将所有行文本 按题型分大类,再在每个大类中按题号切分每个题目
  6. """
  7. 总共3种方案:1、教师用卷;2、按题号切分;3、划分试题和答案,再按题号切分
  8. """
  9. from structure.ans_structure import *
  10. from utils.insert_keywords import get_con
  11. from utils.item_resplit import resplit
  12. from utils.washutil import table_label_cleal
  13. from structure.stems_structure import stems_structure_byno
  14. from utils.item_type_line import get_item_head_info
  15. from utils.topic_no import judge_item_no_type, get_right_no
  16. from utils.stem_ans_split import stem_ans_split
  17. from collections import Counter
  18. from pprint import pprint
  19. def items_ans_reform(items_list, ans_list):
  20. """
  21. 第三种word试卷格式, 题目和答案分开的情况
  22. 答案也有几种类型:带题型?
  23. :param sent_list:
  24. :param split_point:
  25. :return:
  26. """
  27. con1 = list(filter(lambda x: x.strip() != "", items_list)) # 题目
  28. anss1 = list(filter(lambda x: x.strip() != "", ans_list)) # 答案,list中的每个元素为一行
  29. if re.match(".+?省.+?试[卷题]", con1[-1]):
  30. con1 = con1[:-1]
  31. if re.match(".+?省.+?试[卷题]|.*?答题?[卷卡页]", anss1[0]):
  32. anss1 = anss1[1:]
  33. print("-------答案页----------") # 答案页可能全是图片
  34. # pprint(anss1)
  35. print("-------答案页--end--------")
  36. #--------------答案页也包含题目的情况----------------------------
  37. ans_n = re.findall("【答案】", "\n".join(anss1))
  38. if ans_n and len(ans_n) == len(re.findall("【解析】", "\n".join(anss1)))>2: # 带相同个数的答案和解析
  39. item_res = split_by_keywords(anss1)
  40. if type(item_res) != str:
  41. return item_res
  42. # ----------------- 【解析 题目】----------------------------
  43. print('---------------解析 题目-------------------')
  44. ress = stems_structure_byno(con1)
  45. if type(ress) == str:
  46. return ress
  47. else:
  48. item_res, all_type, item_type_classify, item_no_type, item_type_num, new_item_no = ress # 全题目(不含解析)的结构化
  49. # 将空题目去掉
  50. new_res = []
  51. for k, sub_res in enumerate(item_res):
  52. if sub_res['content'].strip():
  53. sub_res['content'] = del_no(sub_res['content'])
  54. new_res.append(sub_res)
  55. item_res = new_res
  56. # pprint(item_res)
  57. # 先对题目的切分结果进行纠正!!!!!
  58. item_res = resplit(item_res)
  59. print("item_type_classify:", item_type_classify)
  60. print("item_type_num:", item_type_num)
  61. print('----------解析 答案---------------')
  62. # -------------解析 答案---------------------------
  63. # 分两种情况:1>>答案中又按题型排列, 如一、选择题 1.答案 2.答案
  64. # 2>>答案中不含题型关键字,只按序号排列
  65. # 3>>答案中不含题型关键字,且题目中也没有,all_type, item_type_classify为空
  66. # print(anss1)
  67. new_ans_no1 = []
  68. rd1_is_fail = 0
  69. have_type_line = re.search(r"[一二三四五六七八九十]\s*[、..、]\s*[^必考基础综合中等]{2,5}题", "\n".join(anss1))
  70. if have_type_line:
  71. # 这里的anss1的清洗不应该影响rd2_is_fail中的原始文本!!先不修改看看再说
  72. while re.search(r"<td><p>[A-F]</p></td>|</td><td>[A-F]</td><td>", anss1[0]) is None and (re.search(r"[\u4e00-\u9fa5]", anss1[0]) is None
  73. or re.search(r"[一二三四五六七八九十]\s*[、..、]\s*(<imgsrc.*?/>)?\s*.{2,5}题", anss1[0]) is None):
  74. del anss1[0]
  75. # 答案中的题型
  76. all_type2 = re.findall(r"\n\s*[一二三四五六七八九十]\s*[、..、]\s*([^必考基础综合中等]{2,5}题)", "\n" + "\n".join(anss1))
  77. # '本大题' 后面处理
  78. print("答案中的题型:", all_type2)
  79. ans_str = "\n" + "\n".join(anss1)
  80. # try:
  81. item_res, rd1_is_fail = anss_structure_with_type(item_res, ans_str, all_type, all_type2, item_type_num, item_type_classify)
  82. # except:
  83. # rd1_is_fail = 1
  84. # 没有题型行或第一次解析失败
  85. rd2_is_fail = 0
  86. if not have_type_line or rd1_is_fail: # 答案中没有题型行 或题型行名称不规范
  87. print('没有题型行或题目和答案的题型个数不一致或第一次解析失败')
  88. anss1 = list(
  89. map(lambda x: re.sub(r"(\n|^)\s*[一二三四五六七八九十]\s*[、..、]?\s*(<p>)?"
  90. r"(\s*.{2,5}题.+?分\s*[.。]?\s*$|.*?[((].+?[得共]\d+分.*?[))].*?$"
  91. r"|\s*.{2,5}题\s*([((].+?[))])?).*?$|(\n|^)\s*.{2,5}题(.+?分\s*[))])?\s*$", "", x), anss1))
  92. # print("anss1:", anss1)
  93. raw_item_res = item_res
  94. # try:
  95. item_res = ans_structure_step1(anss1, item_type_classify, item_res) # 答案整体结构化
  96. if str(raw_item_res) != str(item_res):
  97. rd2_is_fail = 1
  98. # except:
  99. # rd2_is_fail = 1
  100. # for i, one_item in enumerate(item_res):
  101. # item_res[i].update({'answer': "", 'parse': ""})
  102. # return item_res, item_no_type, rd2_is_fail
  103. for i, one_item in enumerate(item_res):
  104. if 'answer' not in one_item:
  105. item_res[i]['answer'] = ""
  106. if 'parse' not in one_item:
  107. item_res[i]['parse'] = ""
  108. return item_res, item_no_type, rd2_is_fail
  109. def split_by_keywords(con_list):
  110. """
  111. 第一种试卷格式:教师用卷,含答案和解析关键字
  112. 切分思路:
  113. 1.根据大题型分,再按【答案|解析】初步拆分题目,再在‘解析’和‘答案’间细分‘题干’和‘解析’
  114. :param con_list:
  115. :return: 每个切分后的题目组成的dict
  116. """
  117. # items_con = "\n" + "\n".join(con_list)
  118. # judge_item_no_type(items_con)
  119. # item_no_type = 1
  120. # all_con = table_label_cleal()
  121. # item_no = [int(no) for no in re.findall(r'\n+\s*([1-9][0-9]?)\s*[..、、]', all_con)]
  122. # if len(item_no) <= 2:
  123. # item_no_type = 2
  124. # item_no = [int(no) for no in re.findall(r'\n+\s*[((]\s*([1-9][0-9]?)\s*[))]\s*[..、、]?', all_con)]
  125. # if len(item_no) > 3:
  126. # 去掉多余空格,作用不大
  127. con2 = ["【delete】" if (k < len(con_list) - 1 and v.strip() == "" and (
  128. re.match(r"【(答案|解析)】|(答案|解析)\s*[::]|<imgsrc\d+|\s+", con_list[k + 1].strip()) or
  129. re.match(r"(([1-9]|[1-4][0-9])\s*[..、、]|[一二三四五六七八九十]\s*[、..、]\s*[^必考基础综合中等]{2,4}题)",
  130. con_list[k + 1].strip()) is None))
  131. or (k > 0 and v.strip() == "" and (
  132. re.match(r"【(答案|解析)】$|(答案|解析)\s*[::]", con_list[k - 1].strip()) or
  133. re.match(r"[a-z<>/\s]*[一二三四五六七八九十]\s*[、..、]\s*[^必考基础综合中等]{2,4}题",
  134. con_list[k - 1].strip())))
  135. else v for k, v in enumerate(con_list)]
  136. con3 = list(filter(lambda x: x != "【delete】", con2))
  137. while con3 and con3[-1].strip() == "":
  138. del con3[-1]
  139. while con3 and con3[0].strip() == "":
  140. del con3[0]
  141. con3.append("") # 不然最后一个题就漏掉了
  142. # 开头没用信息处理
  143. con3[0] = re.sub(r"([一二三四五六七八九十]\s*[、..、]\s*[^必考基础综合中等]{2,4}题)", r"\n\1", con3[0])
  144. while con3 and (re.search(r"[\u4e00-\u9fa5]", con3[0]) is None
  145. or re.search(r"[一二三四五六七八九十]\s*[、..、]\s*[^必考基础综合中等]{2,4}题", con3[0]) is None):
  146. del con3[0]
  147. # ----------------------------------开始结构化---------------------------------------------
  148. items_con = "\n" + "\n".join(con3)
  149. # 初步获取题号,题号类型
  150. items_con, item_no_info, item_no_type = judge_item_no_type(items_con)
  151. # 1、获取题型行信息、按题型行切分
  152. con4, title_info_dict, choice_class = get_item_head_info(items_con)
  153. all_type = title_info_dict["all_type"]
  154. select_type_id = title_info_dict["select_type_id"]
  155. each_item_score, each_item_score2 = title_info_dict["each_item_score"], title_info_dict["each_item_score2"]
  156. # 2、据是否有题型行分两步进行
  157. # 没有做拆图处理
  158. res = []
  159. if not all_type:
  160. print("不存在大题题型行或题型行格式有问题")
  161. if len(re.findall(r"\n\s*【答案】", items_con)) != len(re.findall(r"\n\s*【解析】", items_con)):
  162. return "不存在大题题型行或题型行格式有问题"
  163. else:
  164. item_no = []
  165. subcon = re.split(r"((?<=\n)\s*【答案】|(?<=\n)\s*【解析】)\n?", items_con.strip())
  166. pattern1 = re.compile(r"([1-9]|[1-4][0-9])\s*[..、、].+?")
  167. if re.match(pattern1, subcon[0].strip()):
  168. st_id = re.match(pattern1, subcon[0].strip()).group(1)
  169. if int(st_id) > 1:
  170. item_no.append(int(st_id))
  171. else:
  172. item_no.append(1)
  173. if len(subcon) == 5: # 只有1道题
  174. dd = dict(zip(["content", "answer", "parse"],
  175. re.split(r"(?<=\n)\s*【答案】|(?<=\n)\s*【解析】", table_label_cleal(items_con))))
  176. dd["item_topic_name"] = ""
  177. dd["content"] = re.sub(r"^\d+\s*[..、、]", "", dd["content"][:5]) + dd["content"][5:]
  178. dd["score"] = 0
  179. dd["errmsgs"] = []
  180. dd["item_id"] = item_no[0] # 要用实际id 不是索引序号
  181. res.append(dd)
  182. else:
  183. # ------在下一题【解析】在本题【答案】之间找到下一题【content】的位置--------
  184. all_item, local_item_no, errmsg_dict, count = get_con(subcon, item_no_type, index=0)
  185. item_no.extend(local_item_no)
  186. for idk, one_item in enumerate(all_item):
  187. dd = dict(zip(["content", "answer", "parse"],
  188. re.split(r"(?<=\n)\s*【答案】\n?|(?<=\n)\s*【解析】\n?",
  189. table_label_cleal(one_item))))
  190. dd["item_topic_name"] = ""
  191. dd["content"] = re.sub(r"\d+\s*[..、、]", "", dd["content"][:5]) + dd["content"][5:]
  192. dd["score"] = 0
  193. dd["errmsgs"] = [errmsg_dict[idk]] if idk in errmsg_dict else []
  194. dd["item_id"] = item_no[idk]
  195. res.append(dd)
  196. else:
  197. if len(all_type) != len(con4):
  198. print("存在题型行没有换行")
  199. return "存在题型行末尾没有换行,请在所有题型行末尾重新换行" # 放第【2】种方案中进行处理
  200. else:
  201. # if "非选择题" in all_type:
  202. # return "第" + str(all_type.index("非选择题")+1) + "大题的题型不明确"
  203. index = 0
  204. for num, one_type in enumerate(con4):
  205. count = 1
  206. if len(re.findall(r"\n\s*【答案】", one_type)) == len(re.findall(r"\n\s*【解析】", one_type)):
  207. subcon = re.split(r"((?<=\n)\s*【答案】|(?<=\n)\s*【解析】)\n?", one_type.strip())
  208. # index根据第一道题的题号进行纠正
  209. item_no = []
  210. pattern1 = re.compile(r"([1-9]|[1-4][0-9])\s*[..、、].+?")
  211. if re.match(pattern1, subcon[0].strip()):
  212. st_id = re.match(pattern1, subcon[0].strip()).group(1)
  213. if num == 0 and int(st_id) != 1:
  214. index = int(st_id) - 1
  215. item_no.append(int(st_id))
  216. else:
  217. item_no.append(index+1)
  218. if len(subcon) == 5: # 只有1道题
  219. dd = dict(zip(["content", "answer", "parse"],
  220. re.split(r"(?<=\n)\s*【答案】|(?<=\n)\s*【解析】", table_label_cleal(one_type))))
  221. dd["item_topic_name"] = all_type[num]
  222. dd["content"] = re.sub(r"^\d+\s*[..、、]", "", dd["content"][:5]) + dd["content"][5:]
  223. dd["score"] = each_item_score[num]
  224. dd["errmsgs"] = []
  225. dd["item_id"] = item_no[0] # 要用实际id 不是索引序号
  226. if not dd["score"] and each_item_score2 and str(dd["item_id"]) in each_item_score2.keys():
  227. dd["score"] = each_item_score2[str(dd["item_id"])]
  228. if select_type_id and dd["item_id"] in select_type_id:
  229. dd['is_optional'] = 'true'
  230. if dd["score"] == 0.0 and title_info_dict["total_score"][num] > 0.0:
  231. dd["score"] = title_info_dict["total_score"][num]
  232. res.append(dd)
  233. else:
  234. # ------在下一题【解析】在本题【答案】之间找到下一题【content】的位置,再按此3个关键字进行 切分--------
  235. all_item, local_item_no, errmsg_dict, count = get_con(subcon, item_no_type, all_type=all_type, num=num, index=index)
  236. item_no.extend(local_item_no)
  237. for idk, one_item in enumerate(all_item):
  238. dd = dict(zip(["content", "answer", "parse"],
  239. re.split(r"(?<=\n)\s*【答案】\n?|(?<=\n)\s*【解析】\n?",
  240. table_label_cleal(one_item))))
  241. dd["item_topic_name"] = all_type[num]
  242. dd["content"] = re.sub(r"\d+\s*[..、、]", "", dd["content"][:5]) + dd["content"][5:]
  243. dd["score"] = each_item_score[num]
  244. dd["errmsgs"] = [errmsg_dict[idk]] if idk in errmsg_dict else []
  245. dd["item_id"] = item_no[idk] # idk+1+index 为序号
  246. if choice_class:
  247. for k, v in choice_class.items():
  248. if dd["item_id"] in v:
  249. dd["item_topic_name"] = k + "选题"
  250. # elif len(choice_class) == 1:
  251. # dd["item_topic_name"] = "多选题" if k == "单" else "单选题"
  252. if not dd["score"] and each_item_score2 and str(dd["item_id"]) in each_item_score2.keys():
  253. dd["score"] = each_item_score2[str(dd["item_id"])]
  254. if select_type_id and dd["item_id"] in select_type_id:
  255. dd['is_optional'] = 'true'
  256. res.append(dd)
  257. # pprint(res)
  258. else:
  259. return "第" + str(num + 1) + "大题《" + all_type[num] + "》中【答案】或【解析】格式有误或其中某道题中出现多个相同关键字或漏关键字"
  260. index += count
  261. return res, item_no_type
  262. def split_by_topicno(con_list):
  263. """
  264. 第二种试卷格式: 不同时或都不含有{答案}和{解析}关键字
  265. 按题号切分每个题目
  266. 将所有行文本 按题型分大类,再在每个大类中切分每个题目
  267. :param con_list: 所有行文本组成的list
  268. :return: [{},{}]
  269. """
  270. con1 = list(filter(lambda x: x.strip() != "", con_list))
  271. ress = stems_structure_byno(con1) # 按题号切分后的初步结构化
  272. if type(ress) == str:
  273. return ress
  274. else:
  275. res, all_type, item_type_classify, item_no_type, item_type_num, new_item_no = ress
  276. # res, all_type, item_type_classify = stems_structure_byno(con1)
  277. print("item_type_num:", item_type_num)
  278. # pprint(res)
  279. # 可能存在有的题目有解析,有的没有
  280. for k, one_res in enumerate(res):
  281. if re.search('\n【(答案|[解分][析答]|详解|点[评睛]|考点|专题)】', one_res["content"]):
  282. case = "case1" # 默认有“答案”关键字
  283. if re.search(r'\n【答案】|答案\s*[::]', one_res["content"]) is None:
  284. # 没“答案”关键字
  285. case = "case0"
  286. dd1 = stem_ans_split(one_res, case) # 对切分后的每道题再细分
  287. one_res["content"] = dd1["content"]
  288. del dd1["content"]
  289. one_res.update(dd1)
  290. else: # 没有解析的情况
  291. one_res.update({"answer": "", "parse": ""})
  292. one_res["content"] = del_no(one_res["content"], item_no_type)
  293. if 'pic' in one_res:
  294. one_res["content"] += "\n" + "\n".join(one_res["pic"])
  295. del one_res["pic"]
  296. # 先对题目的切分结果进行纠正!!!!!
  297. res = resplit(res)
  298. # 对最后一个题后面带个别答案(无答案页)
  299. if res:
  300. pattern1 = re.search('\n\s*([1-9]|[1-4][0-9])\s*[..、、]\s*(解\s*[::]|【解析|【答案)', res[-1]["content"])
  301. if pattern1:
  302. breakp = pattern1.start()
  303. ans_str = res[-1]["content"][breakp:]
  304. ans_no_info = pre_get_item_no(ans_str, item_no_type)
  305. ans_no, ans_no_idx = get_right_no(ans_no_info)
  306. all_ans = [del_no(ans_str[i:j]) for i, j in zip(ans_no_idx, ans_no_idx[1:] + [None])]
  307. res[-1]["content"] = res[-1]["content"][:breakp]
  308. res = get_ans_match(res, all_ans, ans_no)
  309. else:
  310. ans_str = res[-1]["content"] + res[-1]["parse"]
  311. ans_no_info = pre_get_item_no(ans_str, item_no_type)
  312. ans_no, ans_no_idx = get_right_no(ans_no_info)
  313. if len(ans_no) == len(res):
  314. all_ans = [del_no(ans_str[i:j]) for i, j in zip(ans_no_idx, ans_no_idx[1:] + [None])]
  315. res[-1]["content"] = res[-1]["content"][:ans_no_idx[0]]
  316. res = get_ans_match(res, all_ans, ans_no)
  317. elif ans_no_idx:
  318. try:
  319. ans_no1, table_ans, st = get_table_ans(res[-1]["content"][:ans_no_idx[0]], flag=1)
  320. if table_ans and 0 < ans_no[0] - ans_no1[-1] < 3:
  321. all_ans = table_ans
  322. all_ans.extend([del_no(ans_str[i:j]) for i, j in zip(ans_no_idx, ans_no_idx[1:] + [None])])
  323. new_ans_no = ans_no1
  324. new_ans_no.extend(ans_no)
  325. if st >= 0:
  326. res[-1]["content"] = res[-1]["content"][:st]
  327. else:
  328. res[-1]["content"] = res[-1]["content"][:ans_no_idx[0]]
  329. res = get_ans_match(res, all_ans, new_ans_no)
  330. except:
  331. if len(ans_no)>4:
  332. all_ans = [del_no(ans_str[i:j]) for i, j in zip(ans_no_idx, ans_no_idx[1:] + [None])]
  333. res[-1]["content"] = res[-1]["content"][:ans_no_idx[0]]
  334. res = get_ans_match(res, all_ans, ans_no)
  335. # 没有识别出答案切分点的情况,很可能答案里的部分也当成题文进行拆分,所以先判断下是否有相同的id
  336. all_no = [one_res['item_id'] for one_res in res]
  337. if len(list(set(all_no))) - len(all_no) < -2:
  338. Count_no = sorted(dict(Counter(all_no)).items(), key=lambda d: d[1], reverse=True)
  339. if Count_no[0][1] > 1:
  340. split_idx = [i for i, no in enumerate(all_no) if no == Count_no[0][0]][1]
  341. for one_res in res[split_idx:]:
  342. if re.search("[((]\s+[))]|(等于|存在|[是有为])多少|求.*?[??]",
  343. one_res["content"] + "\n" + one_res["parse"]) is None:
  344. bef_no = [k for k, j in enumerate(res[:split_idx]) if j["item_id"]==one_res["item_id"]]
  345. if bef_no and not res[:split_idx][bef_no[0]]["parse"]:
  346. res[:split_idx][bef_no[0]]["parse"] = one_res["content"] + "\n" + one_res["parse"]
  347. return res[:split_idx],item_no_type
  348. return res, item_no_type