emission.py

#!/usr/bin/env python

__author__ = "vaishaks <[email protected]>"
__date__ = "Mar 25, 2013"

import sys
from collections import defaultdict

emission_counts = defaultdict(float)
unigram_counts = defaultdict(float)

def create_counts(filename):
	"""The emission counts from gene.count is read from the file and 
	is stored in a defaultdict in the following format.
	{'word+tag':number-of-times-word-was-tagged-with-tag}
	The unigram counts from gene.count is read from the file and 
	is stored in a defaultdict in the following format.
	{'tag':total-number-of-times-tag-occured}
	"""
	try:
		count_file = open(filename, 'r')
	except IOError:
		sys.stderr.write("ERROR: Cannot read inputfile %s.\n" % filename)
		sys.exit(1)
	for line in count_file:
		line_list = line.split()
		if line_list[1] == 'WORDTAG':
			if not emission_counts.has_key(line_list[3]+"O"):
				emission_counts[line_list[3]+"O"] = 0.0
			if not emission_counts.has_key(line_list[3]+"I-GENE"):
				emission_counts[line_list[3]+"I-GENE"] = 0.0
			emission_counts[line_list[3]+line_list[2]] = float(line_list[0])
		elif line_list[1] == '1-GRAM':
			unigram_counts[line_list[2]] = float(line_list[0])

def map_infrequent_words():
	"""Words occuring less than 5 times in the training data 
	are marked as _RARE_ to account for unknown words.
	"""
	training_file = open('gene.train', 'r')
	new_training_file = open('gene_rare.train', 'w')
	for line in training_file:
		line_list = line.split()
		if line == '\n':
			new_training_file.write(line)
		elif emission_counts[line_list[0]+line_list[1]] < 5:
			new_training_file.write("_RARE_ "+line_list[1]+"\n")
		else:
			new_training_file.write(line)

def emission(word, tag):
	"""Returns e(x|y), where 'e' is the emission parameter, x is the 
	word and y is the tag.
	e(x|y) = Count(y->x)/Count(y)g
	Count(y->x) is the emission count where x is tagged as y.
	Count(y) is the unigram count or the total number of times words 
	are tagged as y.
	"""
	if(emission_counts.has_key(word+tag)):
		return (emission_counts[word+tag]/unigram_counts[tag])
	else:
		return(emission_counts['_RARE_'+tag]/unigram_counts[tag])

def tagger(word):
	"""It takes a word as input and returns the tag with 
	the maximum probability of being paired with that word.
	"""
	tag_counts = defaultdict(float)
	tags = unigram_counts.keys()
	for tag in tags:
		tag_counts[tag] = emission(word, tag)
	max_tag_emission = max(tag_counts.values())
	for tag in tags:
		if tag_counts[tag] == max_tag_emission:
			return tag

def file_tagger():
	"""Reads a file which is not tagged and tags each word 
	with its corresponding tag and writes it into another file.
	"""
	inputfile = open('gene.dev', 'r')
	outputfile = open('gene_dev.p1.out', 'w')
	for word in inputfile:
		if word == '\n':
			outputfile.write('\n')
		else:
			tag = tagger(word[:len(word)-1])
			outputfile.write(word[:len(word)-1]+" "+tag+"\n")

create_counts("gene.count")
map_infrequent_words()
create_counts("gene_rare.count")
#print emission_counts['_RARE_O']
#print emission_counts['_RARE_I-GENE']
#print emission('Pol', 'O')
#print tagger('vai')
file_tagger()