How to use regular expressions in python to split articles based on punctuation-CodePudding

I need to divide the article into sentences by punctuation. I use the following regular expression:

re.split(r'[,|.|?|!]', strContent)

It does work, but there is a problem. It will separate the following Latin names that should not be split (such as G. lucidum):

Many studies to date have described the anticancer properties of G. lucidum,

The abbreviation of this Latin name is a capital letter followed by a dot and a space. So I try to modify the above regular expression as follows:

re.split(r'[,|(?:[^A-Z].)|?|!]', strContent)

However, the following error prompt was received:

re.error: unbalanced parenthesis

How can I modify this regular expression?

CodePudding user response：

You should use a negative lookbehind, and put it before the character set that matches the sentence ending.

The negative lookbehind should match a word that's just a single capital letter. This can be done by matching a word boundary before the letter with \b.

You also don't need | inside the character set. That's used for alternative patterns to match.

re.split(r'(?<!\b[A-Z])[,.?!]', strContent)

CodePudding user response：

Using pure regex to find complete sentences is difficult, because of edge cases such as abbreviations, which you have been seeing. You should use an NLP library like NLTK instead.

from nltk.tokenize import sent_tokenize
text = "Many studies to date have described the anticancer properties of G. lucidum.  The studies are vast."
print(sent_tokenize(text))

# ['Many studies to date have described the anticancer properties of G. lucidum.', 'The studies are vast.']