Mercurial > repos > stevecassidy > nltktools

annotate g_collocation.py @ 2:a47980ef2b96 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
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
Ignore whitespace changes - Everywhere: Within whitespace: At end of lines:
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.collocations import BigramCollocationFinder, BigramAssocMeasures
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
3 from nltk.collocations import TrigramCollocationFinder, TrigramAssocMeasures
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
4 import argparse
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
5
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
6 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
7
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
8
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
9 def Parser():
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
10 the_parser = argparse.ArgumentParser(description="Parse the sentence using Chart Parser and a supplied grammar")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
11 the_parser.add_argument('--input', required=True, action="store", type=str, help="input text file")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
12 the_parser.add_argument('--output', required=True, action="store", type=str, help="output file path")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
13 the_parser.add_argument('--freq_filter', required=True, action="store", type=str, help="The minimum number of required occurrences in the corpus")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
14 the_parser.add_argument('--results', required=True, action="store", type=str, help="The maximum number of collocations to show in the results")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
15 the_parser.add_argument('--coll_type', required=True, action="store", type=str, help="Type of collocations to find")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
16 the_parser.add_argument('--pos', required=True, action="store", type=str, help="Data input is a set of POS tags")
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
17
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
18 return the_parser.parse_args()
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
19
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
20
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
21 def collocation(inp, outp, freq_filter, results, coll_type, pos):
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
22 pos = bool(pos == 'true')
1
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
23 with open(inp, 'r') as fd:
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
24 i = 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 all_words = []
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
27 if pos:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
28 text = i.split(' ')[:-1]
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
29 all_words = [x[0:x.index('/')] if x != '\n' else x for x in text]
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
30 all_words = [x.strip(' ').strip('\n') for x in all_words]
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
31 else:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
32 sents = nltk.sent_tokenize(i)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
33 for sent in sents:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
34 all_words += nltk.word_tokenize(sent)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
35 if coll_type == 'bigram':
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
36 measures = BigramAssocMeasures()
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
37 finder = BigramCollocationFinder.from_words(all_words)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
38 else:
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
39 measures = TrigramAssocMeasures()
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
40 finder = TrigramCollocationFinder.from_words(all_words)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
41 finder.apply_freq_filter(int(freq_filter))
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
42 # score the ngrams and get the first N
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
43 colls = finder.score_ngrams(measures.pmi)[:int(results)]
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
44 with open(outp, 'w') as output:
1
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
45 for coll in colls:
2
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
46 (a, b), score = coll
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
47 output.write("%s\t%s\n" % (a, b))
a47980ef2b96 planemo upload for repository https://github.com/Alveo/alveo-galaxy-tools commit b5b26e9118f2ad8af109d606746b39a5588f0511-dirty
stevecassidy
parents: 1
diff changeset
48
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
49
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
50 if __name__ == '__main__':
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
51 args = Parser()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
52
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
53 collocation(args.input, args.output, args.freq_filter, args.results, args.coll_type, args.pos)