Mercurial > repos > stevecassidy > nltktools

annotate g_collocation.py @ 0:e991d4e60c17 draft

Find changesets by keywords (author, files, the commit message), revision number or hash, or revset expression.
planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
author stevecassidy
date Wed, 12 Oct 2016 22:17:53 -0400
parents
children fb617586f4b2
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 sys
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
2 import os
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
3 import nltk
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
4 from nltk.collocations import *
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
5 import argparse
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
6
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
7 def Parser():
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
8 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
9 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
10 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
11 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
12 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
13 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
14 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
15
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
16 args = the_parser.parse_args()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
17 return args
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
18
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
19 def collocation(inp, outp, freq_filter, results, coll_type, pos):
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
20 pos = bool(pos == 'true')
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
21 i = str(unicode(open(inp, 'r').read(), errors='ignore'))
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
22 o = open(outp, 'w')
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
23 all_words = []
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
24 if pos:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
25 text = i.split(' ')[:-1]
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
26 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
27 all_words = [x.strip(' ').strip('\n') for x in all_words]
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
28 else:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
29 sents = nltk.sent_tokenize(i)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
30 for sent in sents:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
31 all_words += nltk.word_tokenize(sent)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
32 if coll_type == 'bigram':
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
33 measures = nltk.collocations.BigramAssocMeasures()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
34 finder = BigramCollocationFinder.from_words(all_words)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
35 else:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
36 measures = nltk.collocations.TrigramAssocMeasures()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
37 finder = TrigramCollocationFinder.from_words(all_words)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
38 finder.apply_freq_filter(int(freq_filter))
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
39 colls = finder.nbest(measures.pmi, int(results))
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
40 for coll in colls:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
41 o.write("%s\t%s" % coll)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
42 o.write('\n')
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
43 o.close()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
44
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
45 if __name__ == '__main__':
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
46 args = Parser()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
47
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
48 collocation(args.input, args.output, args.freq_filter, args.results, args.coll_type, args.pos)