[docs]
def get_question_type(q):
"""
Determines if a question is multiple or single choice and strips the choice indicator.
Args:
q (str): The question text containing choice type indicators
Returns:
tuple: (stripped_question, question_type)
- stripped_question (str): Question text with choice indicators removed
- question_type (str): Either 'multiple choice' or 'single choice'
"""
if '多选' in q: # multiple choice
question_type = "multiple choice"
stripped_question = q.rstrip('（多选）').lstrip('（多选）').rstrip('(多选)').lstrip('(多选)').rstrip('[多选]').lstrip('(多选)')
else:
question_type = 'single choice'
stripped_question = q.rstrip('（单选）').lstrip('（单选）').rstrip('(单选)').lstrip('(单选)').rstrip('[单选]').lstrip('[单选]')
return stripped_question, question_type
[docs]
def split_ans(ans_text):
"""
Converts answer choices (A, B, C, D, E) to their corresponding indices.
Args:
ans_text (str): String containing answer choices (e.g., 'A', 'BC', 'A,B,C')
Returns:
list: List of integer indices corresponding to the answer choices (e.g., [0] for A, [1,2] for BC)
"""
selects_map = {'A': 0, 'B': 1, 'C': 2, 'D': 3, 'E': 4}
# ans_text = ans_text.split('：')[-1].strip()
# if ', ' in ans_text:
# ans_lis = ans_text.split(', ')
# elif '、' in ans_text:
# ans_lis = ans_text.split('、')
# elif ',' in ans_text:
# ans_lis = ans_text.split(',')
# elif ' ' in ans_text:
# ans_lis = ans_text.split(' ')
# else: # single or ABC
# ans_lis = []
# for c in ans_text:
# ans_lis.append(c)
ans_index_lis = []
# for ans in ans_lis:
# ans_index_lis.append(selects_map[ans])
for i in ans_text:
if i in selects_map:
ans_index_lis.append(selects_map[i])
return ans_index_lis
[docs]
def strip_select(select_text):
"""
Removes answer choice prefixes (A., B., C., etc.) from selection text.
Args:
select_text (str): Text containing answer choice prefix
Returns:
str: Clean text with prefix removed
"""
return select_text.lstrip('A. ').lstrip('B. ').lstrip('C. ').lstrip('D. ').lstrip('E. ').lstrip('A.').lstrip('B.').lstrip('C.').lstrip('D.').lstrip('E.').lstrip('A ').lstrip('B ').lstrip('C ').lstrip('D ').lstrip('E ').lstrip('A').lstrip('B').lstrip('C').lstrip('D').lstrip('E')
import copy
[docs]
def parse_qa(questions):
"""
Parses a formatted string of questions into structured question-answer objects.
Args:
questions (str): Multi-line string containing questions, choices, answers, and references
Returns:
list: List of dictionaries containing parsed QA data with keys:
- question (str): The question text
- question_type (str): Type of question ('multiple choice' or 'single choice')
- selects (list): List of answer choices
- answer (list): List of correct answer indices
- reference (str): Reference text for the question
"""
theme_qas = []
cnt = 0
questions_list = questions.split("\n\n")
q = "q"
selects = []
for ques in questions_list: # '问题': quesiton
if '问题' in ques and 'A' in ques and 'B' in ques and 'C' in ques and 'D' in ques:
ques = ques.split('问题', 1)
ques = ques[1]
ques = ques.lstrip('问题').lstrip("：")
ques_list = ques.split('\n')
if "答案：" in ques: # '答案': answer
if 'A' not in ques_list[1] and len(ques_list[0]) < 10:
q = ques_list[1]
q_list = q.split("：", 1)
if len(q_list) > 1:
if len(q_list[0]) < 2:
q = q_list[1]
else:
if '单选' in q_list[1] or '多选' in q_list[1]:
q = q_list[0] + q_list[1]
else:
q = q_list[0]
else:
q = q_list[0]
ques_index = 2
if '单选' or '多选' in ques_list[0]:
_, question_type = get_question_type(q)
else: # '（单选）' or '（多选）' in ques
q, question_type = get_question_type(q)
else:
q = ques_list[0]
q_list = q.split("：", 1)
if len(q_list) > 1:
if len(q_list[0]) < 2:
q = q_list[1]
else:
if '单选' in q_list[1] or '多选' in q_list[1]:
q = q_list[0] + q_list[1]
else:
q = q_list[0]
else:
q = q_list[0]
ques_index = 1
q, question_type = get_question_type(q)
selects = []
ans_index = 0
for i, que in enumerate(ques_list[ques_index:]):
if "答案：" in que:
ans = que.lstrip('答案：')
ans = split_ans(ans)
ans_index = i + ques_index
break
selects.append(strip_select(que))
if "引用文本：" in ques: # reference
for que in ques_list[ans_index + 1:]:
if "引用文本：" in que:
ref = que.lstrip('引用文本：')
break
theme_qas.append({'question': q, 'question_type': question_type, 'selects': copy.deepcopy(selects), 'answer': ans, 'reference': ref})
cnt += 1
else:
if 'A' not in ques_list[1] and len(ques_list[0]) < 10:
q = ques_list[1]
q_list = q.split("：", 1)
if len(q_list) > 1:
if len(q_list[0]) < 2:
q = q_list[1]
else:
if '单选' in q_list[1] or '多选' in q_list[1]:
q = q_list[0] + q_list[1]
else:
q = q_list[0]
else:
q = q_list[0]
ques_index = 2
if '单选' or '多选' in ques_list[0]:
_, question_type = get_question_type(q)
else: # '（单选）' or '（多选）' in ques
q, question_type = get_question_type(q)
else:
q = ques_list[0]
q_list = q.split("：", 1)
if len(q_list) > 1:
if len(q_list[0]) < 2:
q = q_list[1]
else:
if '单选' in q_list[1] or '多选' in q_list[1]:
q = q_list[0] + q_list[1]
else:
q = q_list[0]
else:
q = q_list[0]
ques_index = 1
q, question_type = get_question_type(q)
selects = [strip_select(select) for select in ques_list[ques_index: ]]
elif "答案：" in ques:
if q == "q":
continue
if "引用文本" in ques:
que_list = ques.split('\n')
ans = que_list[0].lstrip('答案：')
ans = split_ans(ans)
ref = que_list[1].lstrip('引用文本：')
theme_qas.append({'question': q, 'question_type': question_type, 'selects': selects, 'answer': ans, 'reference': ref})
cnt += 1
else:
ans = ques.lstrip("答案：")
ans = split_ans(ans)
elif "引用文本：" in ques:
if q == "q":
continue
ref = ques.lstrip('引用文本：')
theme_qas.append({'question': q, 'question_type': question_type, 'selects': selects, 'answer': ans, 'reference': ref})
cnt += 1
else:
continue
return theme_qas