提取段落中与列表中单词相似的单词
我有以下字符串:
"The boy went to twn and bought sausage and chicken. He then picked a tddy for his sister"
要提取的单词列表:
["town","teddy","chicken","boy went"]
注意:town 和 teddy 在给定的句子中拼写错误。
我尝试了以下方法,但我得到了不属于答案的其他词:
import difflib
sent = "The boy went to twn and bought sausage and chicken. He then picked a tddy for his sister"
list1 = ["town","teddy","chicken","boy went"]
[difflib.get_close_matches(x.lower().strip(), sent.split()) for x in list1 ]
我得到以下结果:
[['twn', 'to'], ['tddy'], ['chicken.', 'picked'], ['went']]
代替:
'twn', 'tddy', 'chicken','boy went'
回答
文档中的注意事项difflib.get_closest_matches():
difflib.get_close_matches(word, possibilities, n=3, cutoff=0.6)返回最佳“足够好”匹配的列表。
word是一个需要紧密匹配的序列(通常是一个字符串),并且
possibilities是一个要与之匹配的序列列表word
(通常是一个字符串列表)。可选参数
n(默认3)是要返回的最大匹配数;n必须大于0。可选参数
cutoff(默认0.6)是范围内的浮点数[0, 1]。不得分至少与单词相似的可能性将被忽略。
目前,您正在使用默认值n和cutoff参数。
您可以指定其中一个(或两者),以缩小返回的匹配范围。
例如,您可以使用cutoff0.75的分数:
result = [difflib.get_close_matches(x.lower().strip(), sent.split(), cutoff=0.75) for x in list1]
或者,您可以指定最多只返回 1 个匹配项:
result = [difflib.get_close_matches(x.lower().strip(), sent.split(), n=1) for x in list1]
在任何一种情况下,您都可以使用列表理解来展平列表列表(因为difflib.get_close_matches()总是返回一个列表):
matches = [r[0] for r in result]
由于您还想检查二元组的接近匹配,您可以通过提取相邻“单词”的配对来实现,并将它们difflib.get_close_matches()作为possibilities参数的一部分传递给。
这是一个完整的工作示例:
import difflib
import re
sent = "The boy went to twn and bought sausage and chicken. He then picked a tddy for his sister"
list1 = ["town", "teddy", "chicken", "boy went"]
# this extracts overlapping pairings of "words"
# i.e. ['The boy', 'boy went', 'went to', 'to twn', ...
pairs = re.findall(r'(?=(b[^ ]+ [^ ]+b))', sent)
# we pass the sent.split() list as before
# and concatenate the new pairs list to the end of it also
result = [difflib.get_close_matches(x.lower().strip(), sent.split() + pairs, n=1) for x in list1]
matches = [r[0] for r in result]
print(matches)
# ['twn', 'tddy', 'chicken.', 'boy went']