Mercurial > repos > stevecassidy > nltktools
annotate g_stemmer.py @ 2:a47980ef2b96 draft
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
author | stevecassidy |
---|---|
date | Wed, 01 Nov 2017 01:19:55 -0400 |
parents | fb617586f4b2 |
children |
rev | line source |
---|---|
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
1 import nltk |
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
2 from nltk.stem import PorterStemmer, LancasterStemmer, snowball |
0
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 |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
6 def arguments(): |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
7 parser = argparse.ArgumentParser(description="Segments the text input into separate sentences") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
8 parser.add_argument('--input', required=True, action="store", type=str, help="input text file") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
9 parser.add_argument('--output', required=True, action="store", type=str, help="output file path") |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
10 parser.add_argument('--stemmer', required=False, action="store", type=str, help="output file path") |
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 |
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
14 |
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
15 def stem_file(in_file, out_file, stemmer_type): |
1
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
16 with open(in_file, 'r') as fd: |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
17 unsegmented = fd.read() |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
18 |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
19 with open(out_file, 'w') as output: |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
20 sentences = nltk.sent_tokenize(unsegmented) |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
21 stemmer = get_stemmer(stemmer_type) |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
22 for sentence in sentences: |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
23 words = nltk.word_tokenize(sentence) |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
24 for word in words: |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
25 stemmed_word = stemmer.stem(word) |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
26 output.write(stemmed_word) |
fb617586f4b2
planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents:
0
diff
changeset
|
27 output.write('\n') |
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
28 |
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
29 |
0
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
30 def get_stemmer(stemmer_type): |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
31 if stemmer_type == 'lancaster': |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
32 stemmer = LancasterStemmer() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
33 elif stemmer_type == 'porter': |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
34 stemmer = PorterStemmer() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
35 else: |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
36 stemmer = snowball.EnglishStemmer() |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
37 return stemmer |
e991d4e60c17
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff
changeset
|
38 |
2
a47980ef2b96
planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents:
1
diff
changeset
|
39 |
0
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 stem_file(args.input, args.output, args.stemmer) |