def update_best_hyp_words_list(orig_word, encoded_vie_units, roles, new_word, best_word_hyps_list):
#print "encoded_vie_units: " + encoded_vie_units
#print "new_word.get_encoded_unit: " + new_word.get_encoded_units()
if new_word.get_encoded_units() == encoded_vie_units:
new_word_hyp = WordHyp()
new_word_hyp.original = orig_word
new_word_hyp.labels = label_letters(orig_word)
new_word_hyp.roles = roles
new_word_hyp.reconstructed_word = new_word
new_word_hyp.mod_pen = levenshtein(new_word.to_plain_text(), orig_word)
#print new_word_hyp.get_str()
if len(best_word_hyps_list) == 0:
best_word_hyps_list.append(new_word_hyp)
return
# If the new hypothesized word has the same edit distance to the original word
# as it is of the first word in the list of current best hypothesized words list,
# add the new hypothesized word to the list
if new_word_hyp.mod_pen == best_word_hyps_list[0].mod_pen:
best_word_hyps_list.append(new_word_hyp)
# If the new hypothesized word has a LOWER edit distance to the original word
# as it is of the first word in the list of current best hypothesized words list,
# CLEAR the list and add the new hypothesized word into the list
elif new_word_hyp.mod_pen < best_word_hyps_list[0].mod_pen:
best_word_hyps_list.append(new_word_hyp)
return
solution_found = True
count = 0
targ_graph_file = open(targ_graph_file_path, "r")
for line in targ_graph_file:
start_time = time.time()
word = line.strip()
report("Decoding word: " + word)
labels = label_letters(word)
roles = [None] * len(word)
en_phones = test_phones[count]
best_hyp = WordHyp()
COORD_EPSILON = 1.0
while COORD_EPSILON >= MIN_COORD_EPSILON:
hyps_list = generate_roles_no_ref(word, labels, en_phones, search_space, COORD_EPSILON)
if len(hyps_list) > 0:
# for hyp in hyps_list:
# hyp.get_str()
best_hyp = get_most_prob_hyp(hyps_list, search_space)
reconstructed_word = str(best_hyp.reconstructed_word)
reconstructed_word = reconstructed_word[1:-1]
reconstructed_word = [part.strip().split(" ") for part in reconstructed_word.split(" ] [ ")]
reconstructed_word = " . ".join(["".join(syl) for syl in reconstructed_word])
def update_hyps_list(orig_word, roles, new_word, checked, best_word_hyps_list):
#print "encoded_vie_units: " + encoded_vie_units
#print "new_word.get_encoded_unit: " + new_word.get_encoded_units()
# print "roles: " + str(roles)
# print "new_word: " + str(new_word)
# print "are_all_subsyl_units_valid(new_word): " + str(are_all_subsyl_units_valid(new_word))
# print "are_all_letters_used(checked): " + str(are_all_letters_used(checked))
# print "checked: " + str(checked)
# print "encoded_vie_units: " + encoded_vie_units
# print "new_word.encoded_units: " + str(new_word.get_encoded_units())
# print "\n"
if not are_all_subsyl_units_valid(new_word):
return
if not are_all_letters_used(checked):
return
new_word_hyp = WordHyp()
new_word_hyp.original = orig_word
new_word_hyp.labels = label_letters(orig_word)
new_word_hyp.roles = roles
new_word_hyp.reconstructed_word = new_word
new_word_hyp.compute_mod_error()
new_word_hyp.award_hyp()
new_word_hyp.extra_letter_count = new_word.to_plain_text().count(GENERIC_VOWEL)
new_word_hyp.count_removal()
new_word_hyp.count_compound_role()
evaluate_phones_score(new_word_hyp)
best_word_hyps_list.append(new_word_hyp)
return
def update_best_hyp_words_list(orig_word, targ_syl_struct, roles, new_word,
checked, best_word_hyps_list):
if not are_all_subsyl_units_valid(new_word):
return
if not are_all_letters_used(checked):
return
if new_word.get_encoded_units() == targ_syl_struct:
new_word_hyp = WordHyp()
new_word_hyp.original = orig_word
new_word_hyp.labels = label_letters(orig_word)
new_word_hyp.roles = roles
new_word_hyp.reconstructed_word = new_word
new_word_hyp.compute_mod_error()
new_word_hyp.award_hyp()
new_word_hyp.extra_letter_count = new_word.to_plain_text().count(
GENERIC_VOWEL)
new_word_hyp.count_removal()
new_word_hyp.count_compound_role()
best_word_hyps_list.append(new_word_hyp)
# if len(best_word_hyps_list) == 0:
# best_word_hyps_list.append(new_word_hyp)
# return
# If the new hypothesized word has the same edit distance to the original word
# as it is of the first word in the list of current best hypothesized words list,
# add the new hypothesized word to the list
# if new_word_hyp.mod_pen == best_word_hyps_list[0].mod_pen:
# best_word_hyps_list.append(new_word_hyp)
# If the new hypothesized word has a LOWER edit distance to the original word
# as it is of the first word in the list of current best hypothesized words list,
# CLEAR the list and add the new hypothesized word into the list
# elif new_word_hyp.mod_pen < best_word_hyps_list[0].mod_pen:
# best_word_hyps_list.append(new_word_hyp)
return