Mercurial > repos > stevecassidy > nltktools
annotate g_tokenize.py @ 3:0df72a8ab095 draft default tip
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit f2432aaedd36ae7662873623d8861d0982dffdd2
| author | stevecassidy |
|---|---|
| date | Mon, 20 Nov 2017 22:52:11 -0500 |
| parents | a47980ef2b96 |
| children |
| 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 |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
5 nltk.download('punkt', quiet=True) |
|
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
6 |
|
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
7 |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
8 def arguments(): |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
9 parser = argparse.ArgumentParser(description="tokenize a text") |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
10 parser.add_argument('--input', required=True, action="store", type=str, help="input text file") |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
11 parser.add_argument('--output', required=True, action="store", type=str, help="output file path") |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
12 parser.add_argument('--lower', required=False, action="store_true", help="lowercase all words") |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
13 parser.add_argument('--nopunct', required=False, action="store_true", help="remove all punctuation characters") |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
14 return parser.parse_args() |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
15 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
16 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
17 def strip_punct(text): |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
18 table = text.maketrans("", "") |
|
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
19 return text.translate(table, string.punctuation) |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
20 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
21 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
22 def tokenize(in_file, out_file, lower=False, nopunct=False): |
|
1
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
23 with open(in_file, 'r') as fd: |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
24 text = fd.read() |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
25 |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
26 if lower: |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
27 text = text.lower() |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
28 result = [] |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
29 # text = unicode(text, errors='ignore') |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
30 sentences = nltk.sent_tokenize(text) |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
31 for sentence in sentences: |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
32 tokens = nltk.word_tokenize(sentence) |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
33 if nopunct: |
|
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
34 tokens = filter(lambda w: w not in string.punctuation, tokens) |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
35 result.append(tokens) |
|
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
36 |
|
1
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
37 with open(out_file, 'w') as output: |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
38 # write one token per line |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
39 for sentence in result: |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
40 for token in sentence: |
|
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
41 output.write(token + "\n") |
|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
42 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
43 |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
44 if __name__ == '__main__': |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
45 args = arguments() |
|
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
46 tokenize(args.input, args.output, lower=args.lower, nopunct=args.nopunct) |
