Mercurial > repos > stevecassidy > nltktools
annotate g_tokenize.py @ 0:e991d4e60c17 draft
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
author | stevecassidy |
---|---|
date | Wed, 12 Oct 2016 22:17:53 -0400 |
parents | |
children | fb617586f4b2 |
rev | line source |
---|---|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
1 import nltk |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
2 import string |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
3 import argparse |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
4 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
5 def arguments(): |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
6 parser = argparse.ArgumentParser(description="tokenize a text") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
7 parser.add_argument('--input', required=True, action="store", type=str, help="input text file") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
8 parser.add_argument('--output', required=True, action="store", type=str, help="output file path") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
9 parser.add_argument('--lower', required=False, action="store_true", help="lowercase all words") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
10 parser.add_argument('--nopunct', required=False, action="store_true", help="remove all punctuation characters") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
11 args = parser.parse_args() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
12 return args |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
13 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
14 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
15 def strip_punct(text): |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
16 table = string.maketrans("","") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
17 return text.translate(table, string.punctuation) |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
18 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
19 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
20 def tokenize(in_file, out_file, lower=False, nopunct=False): |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
21 text = open(in_file, 'r').read() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
22 if lower: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
23 text = text.lower() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
24 if nopunct: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
25 text = strip_punct(text) |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
26 result = [] |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
27 text = unicode(text, errors='ignore') |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
28 sentences = nltk.sent_tokenize(text) |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
29 for sentence in sentences: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
30 tokens = nltk.word_tokenize(sentence) |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
31 result.append(tokens) |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
32 output = open(out_file, 'w') |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
33 # write one token per line |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
34 for sentence in result: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
35 for token in sentence: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
36 output.write(token + "\n") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
37 output.close() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
38 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
39 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
40 if __name__ == '__main__': |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
41 args = arguments() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
42 tokenize(args.input, args.output, lower=args.lower, nopunct=args.nopunct) |