def clean(self):
"""
Clean corpus files and write the results to disk
"""
# loop through files
for corpus_file in self.corpus_files:
msg("Cleaning %s..." % corpus_file)
# get the file in a string
f = open(self.corpus_path + corpus_file, 'r')
data = f.read()
f.close()
# use an unoptimized set of arcane regular expressions to clean the data
data = re.sub(r' +(\r)?\n', '\n', data)
para_sep = r'======================================'
data = re.sub(r'([^\.])(\n+)', '\\1 ', data)
data = re.sub(para_sep, '\n'+para_sep+'\n', data)
data = re.sub(r' +\n', '\n', data)
data = re.sub(r'\n\n+', '\n', data)
data = re.sub(para_sep + r'\n' + para_sep, para_sep, data)
data = re.sub('^\n' + para_sep + '\n', '', data)
data = re.sub(r' *(\[|\]) *', ' ', data)
data = re.sub(r'\n +', '\n', data)
data = re.sub(r'^ +', '', data)
data = re.sub(para_sep + r'\n', '', data)
# write the cleaned data to a new file
new_file = corpus_file + '_cleaned'
f = open(self.corpus_path + new_file, 'w')
f.write(data)
msg("done!\n")
def clean(self):
"""
Clean corpus files and write the results to disk
"""
# loop through files
for corpus_file in self.corpus_files:
msg("Cleaning %s..." % corpus_file)
# get the file in a string
f = open(self.corpus_path + corpus_file, 'r')
data = f.read()
f.close()
# use an unoptimized set of arcane regular expressions to clean the data
data = re.sub(r' +(\r)?\n', '\n', data)
para_sep = r'======================================'
data = re.sub(r'([^\.])(\n+)', '\\1 ', data)
data = re.sub(para_sep, '\n' + para_sep + '\n', data)
data = re.sub(r' +\n', '\n', data)
data = re.sub(r'\n\n+', '\n', data)
data = re.sub(para_sep + r'\n' + para_sep, para_sep, data)
data = re.sub('^\n' + para_sep + '\n', '', data)
data = re.sub(r' *(\[|\]) *', ' ', data)
data = re.sub(r'\n +', '\n', data)
data = re.sub(r'^ +', '', data)
data = re.sub(para_sep + r'\n', '', data)
# write the cleaned data to a new file
new_file = corpus_file + '_cleaned'
f = open(self.corpus_path + new_file, 'w')
f.write(data)
msg("done!\n")
def run_test_cycles(self):
"""
Run the test cycles for training and testing the tagger.
Specifically, employ ten-fold cross-validation to train/test on different
segments of the corpus.
"""
total_time_start = time.time() # keep track of time
pct_step = int(100 / Tagger.test_cycles) # cycle steps in pct
test_pct = pct_step # percentage of the corpus to test the tagger on
train_pct = 100 - test_pct # percentage of the corpus to train the tagger on
rights = [] # array to hold number of correctly-tagged words for each test
wrongs = [] # array to hold number of incorrectly-tagged words for each test
totals = [] # array to hold number of total words for each test
all_missed = [] # array to hold incorrect tag information for each test
sep = ''.join(["-" for i in range(50)]) + "\n" # logging separator
# loop from 0-90 (step size 10)
for start_train_pct in [x*pct_step for x in range(Tagger.test_cycles)]:
msg("%sSTARTING TEST CYCLE %d\n%s" % (sep, (start_train_pct/pct_step)+1,\
sep))
# find the percent point to start collecting test sentences
# may be > 100, so circle round
start_test_pct = (start_train_pct+train_pct) % 100
# train the tagger on sentences from the corpus matching our range
training_sents = self.tb.training_sents(train_pct,start_train_pct)
self.train(training_sents)
# test the tagger on the rest of the sentences
testing_sents = self.tb.testing_sents(test_pct,start_test_pct)
(right, wrong, missed) = self.test(testing_sents)
# gather accuracy statistics for this test
total = right + wrong
rights.append(right) # store the correct count for this test cycle
wrongs.append(wrong) # store the incorrect count for this test cycle
totals.append(total) # store the total words tested for this test cycle
all_missed += missed # add incorrect tag information from this cycle
msg("Total words: %d\n" % total)
msg("Correct tags: %d (%0.2f%%)\n" % (right, right / total * 100))
msg("Incorrect tags: %d (%0.2f%%)\n" % (wrong, wrong / total * 100))
# end: test cycle
msg("%s%s" % (sep,sep))
# calculate and output statistics for the entire test
print "Total tests run: %d" % len(totals)
print "Total time taken: %0.2f seconds" % (time.time() - total_time_start)
print "Average correct tags: %0.2f%%" % (sum(rights) / sum(totals) * 100)
print "Average incorrect tags: %0.2f%%" % (sum(wrongs) / sum(totals) * 100)
print
# give the option of inspecting incorrect tags
if raw_input("Examine bad tags? ") in ['y','Y']:
self.inspect(all_missed)
def training_sents(self, train_pct, start_train_pct):
"""
Get a list of sentences for training
:param train_pct: what pct of the corpus to retrieve
:param start_train_pct: where in the corpus to begin retrieval
"""
msg("Getting training sentences...")
sents = self._sents_by_pct(train_pct, start_train_pct)
msg("done: %d%% starting at %d%%\n" % (train_pct, start_train_pct))
return sents
def training_sents(self, train_pct, start_train_pct):
"""
Get a list of sentences for training
:param train_pct: what pct of the corpus to retrieve
:param start_train_pct: where in the corpus to begin retrieval
"""
msg("Getting training sentences...")
sents = self._sents_by_pct(train_pct, start_train_pct)
msg("done: %d%% starting at %d%%\n" % (train_pct, start_train_pct))
return sents
def testing_sents(self, test_pct, start_test_pct):
"""
Get a list of untagged and tagged sentences for testing
:param test_pct: what pct of the corpus to retrieve
:param start_test_pct: where in the corpus to begin retrieval
"""
# when we retrieve testing sentences, we want to get tagged and untagged
# versions of them so we can evaluate accuracy
msg("Getting testing sentences...")
untagged_sents = self._sents_by_pct(test_pct, start_test_pct, tagged=False)
tagged_sents = self._sents_by_pct(test_pct, start_test_pct, tagged=True)
msg("done: %d%% starting at %d%%\n" % (test_pct, start_test_pct))
return (untagged_sents, tagged_sents)
def run_test_cycles(self):
"""
Run the test cycles for training and testing the tagger.
Specifically, employ ten-fold cross-validation to train/test on different
segments of the corpus.
"""
total_time_start = time.time() # keep track of time
rights = [] # array to hold number of correctly-tagged words for each test
wrongs = [] # array to hold number of incorrectly-tagged words for each test
totals = [] # array to hold number of total words for each test
all_missed = [] # array to hold incorrect tag information for each test
sep = ''.join(["-" for i in range(50)]) + "\n" # logging s
# returns tagged sentences
training_sents = self.training.tagged_sents
self.train(training_sents)
# returns untagged sentences
testing_tagged_sents = self.testing.tagged_sents
testing_untagged_sents = self.testing.sents
testing_sents = (testing_untagged_sents, testing_tagged_sents)
(right, wrong, missed) = self.test(testing_sents)
# gather accuracy statistics for this test
total = right + wrong
rights.append(right) # store the correct count for this test cycle
wrongs.append(wrong) # store the incorrect count for this test cycle
totals.append(total) # store the total words tested for this test cycle
all_missed += missed # add incorrect tag information from this cycle
msg("Total words: %d\n" % total)
msg("Correct tags: %d (%0.2f%%)\n" % (right, right / total * 100))
msg("Incorrect tags: %d (%0.2f%%)\n" % (wrong, wrong / total * 100))
msg("%s%s" % (sep,sep))
# calculate and output statistics for the entire test
print "Total tests run: %d" % len(totals)
print "Total time taken: %0.2f seconds" % (time.time() - total_time_start)
print "Average correct tags: %0.2f%%" % (sum(rights) / sum(totals) * 100)
print "Average incorrect tags: %0.2f%%" % (sum(wrongs) / sum(totals) * 100)
print
# give the option of inspecting incorrect tags
if raw_input("Examine bad tags? ") in ['y','Y']:
self.inspect(all_missed)
def __init__(self, corpus_path, corpus_files):
"""
Construct a Treebank object
:param corpus_path: path to corpus files
:param corpus_files: list of filenames for corpus text
"""
msg("Importing treebank...")
# get a corpus reader object for our corpus using NLTK
treebank = TaggedCorpusReader(corpus_path, corpus_files)
# get all sentences from corpus in a tagged format
self.tagged_sents = treebank.tagged_sents()
# get all sentences from corpus in an untagged format
self.sents = treebank.sents()
msg("done!\n")
def testing_sents(self, test_pct, start_test_pct):
"""
Get a list of untagged and tagged sentences for testing
:param test_pct: what pct of the corpus to retrieve
:param start_test_pct: where in the corpus to begin retrieval
"""
# when we retrieve testing sentences, we want to get tagged and untagged
# versions of them so we can evaluate accuracy
msg("Getting testing sentences...")
untagged_sents = self._sents_by_pct(test_pct,
start_test_pct,
tagged=False)
tagged_sents = self._sents_by_pct(test_pct,
start_test_pct,
tagged=True)
msg("done: %d%% starting at %d%%\n" % (test_pct, start_test_pct))
return (untagged_sents, tagged_sents)
def __init__(self, corpus_path, corpus_files):
"""
Construct a Treebank object
:param corpus_path: path to corpus files
:param corpus_files: list of filenames for corpus text
"""
msg("Importing treebank...")
# get a corpus reader object for our corpus using NLTK
treebank = TaggedCorpusReader(corpus_path, corpus_files)
# get all sentences from corpus in a tagged format
self.tagged_sents = treebank.tagged_sents()
# get all sentences from corpus in an untagged format
self.sents = treebank.sents()
msg("done!\n")
def pos_tags(self):
"""
Create a list of all POS tags found in the corpus
"""
msg("Getting POS tag list...")
tags = []
# loop through sentences
for sent in self.tagged_sents:
# loop through tagged words
for (word, pos) in sent:
# add tag if it's not already in list
if pos not in tags:
tags.append(pos)
msg("done\n")
return tags
def pos_tags(self):
"""
Create a list of all POS tags found in the corpus
"""
msg("Getting POS tag list...")
tags = []
# loop through sentences
for sent in self.tagged_sents:
# loop through tagged words
for (word, pos) in sent:
# add tag if it's not already in list
if pos not in tags:
tags.append(pos)
msg("done\n")
return tags
def train(self, sents):
"""
Train the tagger on a set of tagged sentences
:param sents: list of tagged sentences
"""
# collect POS tags from our corpus
self.pos_tags = self.tb.pos_tags()
# add start markers to help with bigram tagging
msg("Adjusting POS tags...")
sents = self._adjust_pos(sents)
msg("done\n")
# create 2 conditional frequency distributions (from the NLTK) that store
# observed probabilities that a given word has a certain POS, one for
# lowercase-normalized words and one for words as they appear in the text
msg("Training (Wi|Ck)...")
# create a CFD for words normalized to lowercase
self.words_given_pos = ConditionalFreqDist((wp[1], wp[0].lower()) for \
sent in sents for wp in sent)
# create a CFD for words left in their original capitalization
self.words_given_pos_upper = ConditionalFreqDist((wp[1], wp[0]) for \
sent in sents for wp in sent)
msg("done\n")
# create another CFD that stores probabilities that stores observed
# probabilities that one POS follows another POS
msg("Training (Ci+1|Ci)...")
self.pos2_given_pos1 = ConditionalFreqDist((sent[i-1][1], sent[i][1]) for \
sent in sents for i in range(1,len(sent)))
msg("done\n")
def inspect(self, missed):
"""
Inspect a testing session, and print data about tag accuracy
:param missed: list of tuples of missed tags like:
(hmm_tagged_word, gold_tagged_word, hmm_context, gold_context)
"""
# create a CFD so we can examine a matrix of incorrect vs correct tags
# ms[1][1] = tag of a gold_tagged_word
# ms[0][1] = tag of an hmm_tagged_word
cfd = ConditionalFreqDist((ms[1][1], ms[0][1]) for ms in missed)
# initialize a hash to store mistakes by frequency
mistakes = {}
# print a table showing mistake frequency
cfd.tabulate()
msg("\n")
# loop through mistake frequencies by gold standard tag, i.e., if we are
# examining gold-standard 'IN', count what we incorrectly tagged it as
conds = cfd.conditions()
for g_tag in conds:
for hmm_tag in cfd[g_tag].keys():
# how many times did we incorrectly say g_tag was hmm_tag?
count = cfd[g_tag][hmm_tag]
# add these mistakes to the count
if count not in mistakes.keys():
mistakes[count] = []
mistakes[count].append((hmm_tag, g_tag))
# get a list of all mistake types that occurred over a threshold, worst first
mistake_counts = set([count for (count, mistake_set) in \
mistakes.iteritems() if count > Tagger.mistake_threshold])
mistake_counts = reversed(sorted(mistake_counts))
# now create a list of mistake types to show the user, i.e., loop
# through all types and if they are of a high-frequency type, add to list
mistakes_to_halt = []
for count in mistake_counts:
mistake_set = mistakes[count]
for mistake_tuple in mistake_set:
mistakes_to_halt.append(mistake_tuple)
msg("%d\t%s\twas really\t%s\n" % (count, mistake_tuple[0], \
mistake_tuple[1]))
msg("\n")
# create separators used when outputting missed word contexts
sep_big = "---------------------------------------------------\n"
sep_small = "\n-----------------------------------------\n"
# loop through individual mistakes and, if they match the kind of error
# we want to halt for, show the user the mistake as well as the sentence
# context for both the gold-standard sentence and the hmm-tagged sentence
response = None
for missed_set in missed:
if response not in ['q','Q']:
(hmm_tagged_word, gold_tagged_word, hmm_tagged_sent, \
gold_tagged_sent) = missed_set
should_halt = False
# determine whether the current mistake matches a mistake type
# we want to halt for
for pair in mistakes_to_halt:
if hmm_tagged_word[1] == pair[0] and \
gold_tagged_word[1] == pair[1]:
should_halt = True
if should_halt:
msg("%sTagged '%s' with %s when it should have been %s.%s" %\
(sep_big, hmm_tagged_word[0], hmm_tagged_word[1],\
gold_tagged_word[1], sep_small))
msg("Gold: " + (' '.join([(w[0] + "/" + w[1]) for w in \
gold_tagged_sent])))
msg(sep_small)
msg("Mine: " + (' '.join([(w[0] + "/" + w[1]) for w in \
hmm_tagged_sent])))
# get user input to decide whether to keep going
response = raw_input("\n\nEnter to continue, Q to quit: ")
def train(self, sents):
"""
Train the tagger on a set of tagged sentences
:param sents: list of tagged sentences
"""
# collect POS tags from our corpus
self.pos_tags = self.training.pos_tags()
# add start markers to help with bigram tagging
msg("Adjusting POS tags...")
sents = self._adjust_pos(sents)
msg("done\n")
# create 2 conditional frequency distributions (from the NLTK) that store
# observed probabilities that a given word has a certain POS, one for
# lowercase-normalized words and one for words as they appear in the text
msg("Training (Wi|Ck)...")
# create a CFD for words normalized to lowercase
self.words_given_pos = ConditionalFreqDist((wp[1], wp[0].lower()) for \
sent in sents for wp in sent)
# create a CFD for words left in their original capitalization
self.words_given_pos_upper = ConditionalFreqDist((wp[1], wp[0]) for \
sent in sents for wp in sent)
msg("done\n")
# create another CFD that stores probabilities that stores observed
# probabilities that one POS follows another POS
msg("Training (Ci+1|Ci)...")
self.pos2_given_pos1 = ConditionalFreqDist((sent[i-1][1], sent[i][1]) for \
sent in sents for i in range(1,len(sent)))
msg("done\n")
def run_test_cycles(self):
"""
Run the test cycles for training and testing the tagger.
Specifically, employ ten-fold cross-validation to train/test on different
segments of the corpus.
"""
total_time_start = time.time() # keep track of time
pct_step = int(100 / Tagger.test_cycles) # cycle steps in pct
test_pct = pct_step # percentage of the corpus to test the tagger on
train_pct = 100 - test_pct # percentage of the corpus to train the tagger on
rights = [
] # array to hold number of correctly-tagged words for each test
wrongs = [
] # array to hold number of incorrectly-tagged words for each test
totals = [] # array to hold number of total words for each test
all_missed = [
] # array to hold incorrect tag information for each test
sep = ''.join(["-" for i in range(50)]) + "\n" # logging separator
# loop from 0-90 (step size 10)
for start_train_pct in [
x * pct_step for x in range(Tagger.test_cycles)
]:
msg("%sSTARTING TEST CYCLE %d\n%s" % (sep, (start_train_pct/pct_step)+1,\
sep))
# find the percent point to start collecting test sentences
# may be > 100, so circle round
start_test_pct = (start_train_pct + train_pct) % 100
# train the tagger on sentences from the corpus matching our range
training_sents = self.tb.training_sents(train_pct, start_train_pct)
self.train(training_sents)
# test the tagger on the rest of the sentences
testing_sents = self.tb.testing_sents(test_pct, start_test_pct)
(right, wrong, missed) = self.test(testing_sents)
# gather accuracy statistics for this test
total = right + wrong
rights.append(right) # store the correct count for this test cycle
wrongs.append(
wrong) # store the incorrect count for this test cycle
totals.append(
total) # store the total words tested for this test cycle
all_missed += missed # add incorrect tag information from this cycle
msg("Total words: %d\n" % total)
msg("Correct tags: %d (%0.2f%%)\n" % (right, right / total * 100))
msg("Incorrect tags: %d (%0.2f%%)\n" %
(wrong, wrong / total * 100))
# end: test cycle
msg("%s%s" % (sep, sep))
# calculate and output statistics for the entire test
print "Total tests run: %d" % len(totals)
print "Total time taken: %0.2f seconds" % (time.time() -
total_time_start)
print "Average correct tags: %0.2f%%" % (sum(rights) / sum(totals) *
100)
print "Average incorrect tags: %0.2f%%" % (sum(wrongs) / sum(totals) *
100)
print
# give the option of inspecting incorrect tags
if raw_input("Examine bad tags? ") in ['y', 'Y']:
self.inspect(all_missed)
def inspect(self, missed):
"""
Inspect a testing session, and print data about tag accuracy
:param missed: list of tuples of missed tags like:
(hmm_tagged_word, gold_tagged_word, hmm_context, gold_context)
"""
# create a CFD so we can examine a matrix of incorrect vs correct tags
# ms[1][1] = tag of a gold_tagged_word
# ms[0][1] = tag of an hmm_tagged_word
cfd = ConditionalFreqDist((ms[1][1], ms[0][1]) for ms in missed)
# initialize a hash to store mistakes by frequency
mistakes = {}
# print a table showing mistake frequency
cfd.tabulate()
msg("\n")
# loop through mistake frequencies by gold standard tag, i.e., if we are
# examining gold-standard 'IN', count what we incorrectly tagged it as
conds = cfd.conditions()
for g_tag in conds:
for hmm_tag in cfd[g_tag].keys():
# how many times did we incorrectly say g_tag was hmm_tag?
count = cfd[g_tag][hmm_tag]
# add these mistakes to the count
if count not in mistakes.keys():
mistakes[count] = []
mistakes[count].append((hmm_tag, g_tag))
# get a list of all mistake types that occurred over a threshold, worst first
mistake_counts = set([count for (count, mistake_set) in \
mistakes.iteritems() if count > Tagger.mistake_threshold])
mistake_counts = reversed(sorted(mistake_counts))
# now create a list of mistake types to show the user, i.e., loop
# through all types and if they are of a high-frequency type, add to list
mistakes_to_halt = []
for count in mistake_counts:
mistake_set = mistakes[count]
for mistake_tuple in mistake_set:
mistakes_to_halt.append(mistake_tuple)
msg("%d\t%s\twas really\t%s\n" % (count, mistake_tuple[0], \
mistake_tuple[1]))
msg("\n")
# create separators used when outputting missed word contexts
sep_big = "---------------------------------------------------\n"
sep_small = "\n-----------------------------------------\n"
# loop through individual mistakes and, if they match the kind of error
# we want to halt for, show the user the mistake as well as the sentence
# context for both the gold-standard sentence and the hmm-tagged sentence
response = None
for missed_set in missed:
if response not in ['q', 'Q']:
(hmm_tagged_word, gold_tagged_word, hmm_tagged_sent, \
gold_tagged_sent) = missed_set
should_halt = False
# determine whether the current mistake matches a mistake type
# we want to halt for
for pair in mistakes_to_halt:
if hmm_tagged_word[1] == pair[0] and \
gold_tagged_word[1] == pair[1]:
should_halt = True
if should_halt:
msg("%sTagged '%s' with %s when it should have been %s.%s" %\
(sep_big, hmm_tagged_word[0], hmm_tagged_word[1],\
gold_tagged_word[1], sep_small))
msg("Gold: " + (' '.join([(w[0] + "/" + w[1]) for w in \
gold_tagged_sent])))
msg(sep_small)
msg("Mine: " + (' '.join([(w[0] + "/" + w[1]) for w in \
hmm_tagged_sent])))
# get user input to decide whether to keep going
response = raw_input("\n\nEnter to continue, Q to quit: ")
