InfoChartQA / eval /compare_sequence.py
Jietson's picture
Upload 8 files
f020aef verified
import json
from typing import List, Hashable, Union
from collections import Counter
def is_sequence_valid(sequence: List[Union[Hashable, str]],
case_sensitive: bool = False,
strip_spaces: bool = True,
fuzzy_duplicates: bool = False,
fuzzy_threshold: float = 0.6) -> bool:
"""
检查序列是否合法(无重复元素)
参数:
sequence: 待检查的序列
case_sensitive: 是否区分大小写(仅适用于字符串)
strip_spaces: 是否去除字符串两端空格
fuzzy_duplicates: 是否启用模糊查重(仅适用于字符串)
fuzzy_threshold: 模糊匹配阈值(0-1)
返回:
bool: True表示无重复(合法),False表示有重复(非法)
示例:
>>> is_sequence_valid(["A", "B", "C"]) # True
>>> is_sequence_valid(["A", "a"], case_sensitive=False) # False
>>> is_sequence_valid([" apple ", "apple"]) # False
"""
if not sequence:
return True
processed = []
# print(sequence)
for item in sequence:
# print(item)
if isinstance(item, str):
# 字符串预处理
processed_item = item
if not case_sensitive:
processed_item = processed_item.lower()
if strip_spaces:
processed_item = processed_item.strip()
processed.append(processed_item)
else:
processed.append(item)
# 常规检查(精确匹配)
# print(processed)
if not fuzzy_duplicates:
return len(processed) == len(set(processed))
# 模糊查重模式
for i in range(len(processed)):
for j in range(i + 1, len(processed)):
if isinstance(processed[i], str) and isinstance(processed[j], str):
# 使用difflib进行模糊匹配
from difflib import SequenceMatcher
similarity = SequenceMatcher(None, processed[i], processed[j]).ratio()
if similarity >= fuzzy_threshold:
return False
else:
# 非字符串类型退化为精确匹配
if processed[i] == processed[j]:
return False
return True
def extract_answers_from_file(file_path):
"""
从JSON文件中读取数据并提取answer序列
参数:
file_path: str - JSON文件路径
返回:
dict - 包含提取序列和元数据的字典
"""
try:
# 读取JSON文件
with open(file_path, 'r', encoding='utf-8') as f:
input_data = json.load(f)
# 初始化结果字典
result = {
"sequences": [],
"details": []
}
# 遍历每个条目
for key, item in input_data.items():
# 检查answer字段是否存在
if 'answer' not in item:
continue
# 提取answer并按逗号分割成序列,去除前后空格
answer_sequence = [x.strip() for x in str(item['answer']).split(',')]
# 存储序列和相关信息
result["sequences"].append(answer_sequence)
result["details"].append({
"question_id": item.get("question_id", ""),
"figure_path": item.get("figure_path", ""),
"qtype": item.get("qtype", -1),
"question": item.get("question", ""),
"sequence_length": len(answer_sequence)
})
return result
except FileNotFoundError:
print(f"错误:文件 {file_path} 未找到")
return None
except json.JSONDecodeError:
print("错误:文件内容不是有效的JSON格式")
return None
except Exception as e:
print(f"处理文件时发生错误:{str(e)}")
return None
from difflib import SequenceMatcher
from typing import List, Union, Optional
def fuzzy_match(s1: str, s2: str, threshold: float = 0.6) -> bool:
"""
模糊字符串匹配(基于相似度阈值)
:param s1: 字符串1
:param s2: 字符串2
:param threshold: 相似度阈值(0-1)
:return: 是否匹配
"""
flag = False
flag |= SequenceMatcher(None, s1.lower().strip(), s2.lower().strip()).ratio() >= threshold
flag |= s1 in s2
flag |= s2 in s1
# print(s1 , s2 , SequenceMatcher(None, s1.lower().strip(), s2.lower().strip()).ratio(),flag)
return flag
def is_sequence_match_ordered(
seq1: List[str],
seq2: List[str],
fuzzy: bool = False,
threshold: float = 0.6
) -> bool:
"""
检查两个序列是否顺序完全一致
:param seq1: 序列1
:param seq2: 序列2
:param fuzzy: 是否启用模糊匹配
:param threshold: 模糊匹配阈值
:return: 是否匹配
"""
if len(seq1) != len(seq1):
return False
if not is_sequence_valid(seq1, case_sensitive=True):
return False
if not is_sequence_valid(seq2, case_sensitive=True):
return False
# print(seq1 , seq2)
if fuzzy:
return all(fuzzy_match(x, y, threshold) for x, y in zip(seq1, seq2))
else:
return all(x.strip().lower() == y.strip().lower() for x, y in zip(seq1, seq2))
def is_sequence_match_unordered(
seq1: List[str],
seq2: List[str],
fuzzy: bool = False,
threshold: float = 0.8
) -> bool:
"""
检查两个序列是否元素一致(不考虑顺序)
:param seq1: 序列1
:param seq2: 序列2
:param fuzzy: 是否启用模糊匹配
:param threshold: 模糊匹配阈值
:return: 是否匹配
"""
if len(seq1) != len(seq2):
return False
seq1_processed = [s.lower().strip() for s in seq1]
seq2_processed = [s.lower().strip() for s in seq2]
if fuzzy:
# 构建双向最佳匹配
matched_indices = set()
for i, s1 in enumerate(seq1):
for j, s2 in enumerate(seq2):
if j not in matched_indices and fuzzy_match(s1, s2, threshold):
matched_indices.add(j)
break
return len(matched_indices) == len(seq1)
else:
return sorted(seq1_processed) == sorted(seq2_processed)
# 测试用例
if __name__ == "__main__":
A = "Russia, DR Congo, Ethiopia, Bangladesh, Iraq, Yemen, Pakistan, India"
B = "Russia: 2 \nD.R. Congo: 3 \nEthiopia: 5 \nBangladesh: 5 \nIraq: 7 \nYemen: 7 \nPakistan: 12 \nIndia: 134"
B = B.replace("\n", ",")
B = B.replace(" ", "")
A = A.replace(" ", "")
print(is_sequence_match_ordered(A.split(","), B.split(","), fuzzy=True))
# 测试数据
exact_ordered = ["Apple", "Banana", "Orange"]
exact_unordered = ["Banana", "Orange", "Apple"]
fuzzy_ordered = [" Apple ", "banana", "Orang"]
fuzzy_unordered = ["banan", "orang", " apple"]
# 精确顺序匹配测试
print("精确顺序匹配:")
print(exact_ordered, exact_ordered, is_sequence_match_ordered(exact_ordered, exact_ordered)) # True
print(exact_ordered, exact_unordered, is_sequence_match_ordered(exact_ordered, exact_unordered)) # False
# 精确无序匹配测试
print("\n精确无序匹配:")
print(exact_ordered, exact_unordered, is_sequence_match_unordered(exact_ordered, exact_unordered)) # True
print(exact_ordered, ["Apple", "Banana"], is_sequence_match_unordered(exact_ordered, ["Apple", "Banana"])) # False
# 模糊顺序匹配测试
print("\n模糊顺序匹配:")
print(exact_ordered, fuzzy_ordered, is_sequence_match_ordered(exact_ordered, fuzzy_ordered, fuzzy=True)) # True
print(exact_ordered, fuzzy_unordered,
is_sequence_match_ordered(exact_ordered, fuzzy_unordered, fuzzy=True)) # False
# 模糊无序匹配测试
print("\n模糊无序匹配:")
print(exact_ordered, fuzzy_unordered,
is_sequence_match_unordered(exact_ordered, fuzzy_unordered, fuzzy=True)) # True
print(exact_ordered, ["App", "Banan"],
is_sequence_match_unordered(exact_ordered, ["App", "Banan"], fuzzy=True)) # False
answer = "Trondheim,Munich,TheHague,Muscat,RasAlKhaimah,Dubai,Taipei,Doha,Ajman,AbuDhabi"
response = "Trondheim,Munich,TheHague,Muscat,RasAlKhaimah,Dubai,Taipei,Doha,Ajman,AbuDhabi"
print(is_sequence_match_ordered(answer.split(","), response.split(","), fuzzy=True))
assert is_sequence_valid(["A", "B", "C"]) == True
assert is_sequence_valid(["A", "A"]) == False
# 大小写测试
assert is_sequence_valid(["A", "a"], case_sensitive=False) == False
assert is_sequence_valid(["A", "a"], case_sensitive=True) == True
# 空格处理测试
assert is_sequence_valid(["apple", " apple "]) == False
assert is_sequence_valid(["apple", " apple "], strip_spaces=False) == True
# 模糊匹配测试
assert is_sequence_valid(["apple", "applee"], fuzzy_duplicates=True) == False
assert is_sequence_valid(["apple", "aple"], fuzzy_duplicates=True, fuzzy_threshold=0.8) == False
assert is_sequence_valid(["apple", "orange"], fuzzy_duplicates=True) == True
# 混合类型测试
assert is_sequence_valid([1, "1"]) == True
assert is_sequence_valid([1, 1]) == False
# 边界情况
assert is_sequence_valid([]) == True
assert is_sequence_valid([None, None]) == False