Extract names of persons from filepath

Question

I'm trying to extract all the human names from a filepath. My approach is to split the filepath into individual words, then apply NTLK's Parts Of Speech tag to identify proper nouns, followed by the ne_chunk function to identify persons.

import nltk
import re


def extract_entities(y):
    #make an empty list to receive results of operation
    AggPeople = []
    #split the filepath by backslashes
    for y in y.split("\\"):
        #separate the product above into words, then attach nltk tags (ie. NNP), then attach more specific ntlk tags (ie. Person)
        for chunk in nltk.ne_chunk(nltk.pos_tag(re.findall(r"[\w]+", y))) :
            #filter out everything but the person labels
            if hasattr(chunk, 'label') and chunk.label() == "PERSON":
                #bring the results of the above into a list
                AggPeople.append(' '.join(c[0] for c in chunk.leaves()).capitalize())
                #filter out words you don't want
                AggPeople = [x for x in AggPeople if (x not in ['Schedules','Old'])]
    #get rid of duplicate words with 'set'
    return set(AggPeople)

text = "O:\Country\Province\District\city\Cricket, Jimmy (Y1617F)\Old Schedules\Cricket, Jimmy (78655) Golick doo wop 7 Sept 2016.xlsx"

print(extract_entities(text))

The problem is that the result is 'Jimmy y1617f' and I want it to say 'Jimmy'

I think the nltk.ne_chunk is grouping up words in a way that makes sense when dealing with text, but not with filepaths. To solve the problem, I tried to define my own equivalent of nltk.ne_chunk as follows:

import nltk
import re
from nltk import RegexpParser
def extract_entities(y):
    AggPeople = []
    patterns= r"<NP:{<NNP>+}"
    chunker = RegexpParser(patterns)
    print(chunker)
    for y in y.split("\\"):
        for chunk in chunker(nltk.pos_tag(re.findall(r"[\w]+", y))) :
            if hasattr(chunk, 'label') and chunk.label() == "PERSON":
                AggPeople.append(' '.join(c[0] for c in chunk.leaves()).capitalize())
                AggPeople = [x for x in AggPeople if (x not in ['Schedules','Old'])]
    return set(AggPeople)

Received an error code:

'RegexpParser' object is not callable

Full traceback:

chunk.RegexpParser with 1 stages:
RegexpChunkParser with 1 rules:
       <ChunkRule: '<NNP>'>
Traceback (most recent call last):

  File "<ipython-input-282-cb323eff63b4>", line 1, in <module>
    runfile('C:/Users//.spyder-py3/ExtractingNames.py', wdir='C:/Users//.spyder-py3')

  File "C:\spydercustomize.py", line 827, in runfile
    execfile(filename, namespace)

  File "C:\spydercustomize.py", line 110, in execfile
    exec(compile(f.read(), filename, 'exec'), namespace)

  File "C:/Users//.spyder-py3/ExtractingNames.py", line 32, in <module>
    print(extract_entities(text))

  File "C:/Users//.spyder-py3/ExtractingNames.py", line 23, in extract_entities
    for chunk in chunker(nltk.pos_tag(re.findall(r"[\w]+", y))) :

TypeError: 'RegexpParser' object is not callable

Answer 1

#looks for two proper nouns side-by-side
patterns= r"P:{<NNP>{2}}"
chunker = nltk.RegexpParser(patterns)    

def extract_entities(y):
    AggPeople = []
    for y in y.split("\\"):
        #excludes words with digits and schedules
        for chunk in chunker.parse(nltk.pos_tag(re.findall(r"\b(?!Schedules|Old)[^\d\W]+\b", y))) :
            if hasattr(chunk, 'label') and chunk.label() == "P" :
                AggPeople.append(' '.join(c[0] for c in chunk.leaves()).capitalize())
    return set(AggPeople)

text = "O:\Country\Province\District\city\Cricket, Jimmy (Y1617F)\Old Schedules\Cricket, Jimmy (78655) Golick doo wop 7 Sept 2016.xlsx"

print(extract_entities(text))

• You can make your code run faster if you place the chunker outside of the loop (otherwise, it gets regenerated with each iteration of the loop).
• If you're looking for human names, and there is usually two of them (first and last), you can specify exactly two NNPs in the pattern with the {2} symbol.
• You can exclude certain words in the regex with negative lookahead, and words with digits in it with ^\\d.