annotate g_read_sents.py @ 1:fb617586f4b2 draft

planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
author stevecassidy
date Mon, 05 Dec 2016 05:22:05 -0500
parents e991d4e60c17
children a47980ef2b96
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.corpus import PlaintextCorpusReader
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="Segments the text input into separate sentences")
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
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
12 args = the_parser.parse_args()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
13 return args
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
14
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
15 def print_out(outp, text, sentences):
1
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
16 with open(outp, 'w') as output:
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
17 curr = 0
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
18 for sent in sentences:
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
19 times = count_occurences(sent, sent[-1])
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
20 curr = text.find(sent[0], curr)
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
21 end = find_nth(text, sent[-1], times, curr) + len(sent[-1])
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
22 output.write(text[curr:end] + '\n')
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
23 curr = end
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
24
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
25 def find_nth(string, sub, n, offset):
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
26 start = string.find(sub, offset)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
27 while start >= 0 and n > 1:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
28 start = string.find(sub, start + len(sub))
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
29 n -= 1
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
30 return start
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
31
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
32 def count_occurences(lst, string):
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
33 count = 0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
34 for item in lst:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
35 if string in item:
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
36 count += 1
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
37 return count
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
38
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
39 def read_sents(inp, outp):
1
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
40 with open(inp, 'r') as fd:
fb617586f4b2 planemo upload commit a81826fe44f09a3710a35c183aa88b745aeec064-dirty
stevecassidy
parents: 0
diff changeset
41 i = fd.read()
0
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
42 corpus = PlaintextCorpusReader(os.path.dirname(inp), os.path.basename(inp))
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
43 sents = corpus.sents()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
44 print_out(outp, i, sents)
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
45
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
46 if __name__ == '__main__':
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
47 args = Parser()
e991d4e60c17 planemo upload commit 0203cb3a0b40d9348674b2b098af805e2986abca-dirty
stevecassidy
parents:
diff changeset
48 read_sents(args.input, args.output)