def __init__(self, app_config):
super(InlineModel, self).__init__(app_config)
inline_models_dir = join(app_config["root"],
app_config["dirs"]["models_root"],
app_config["dirs"]["models"]["inline"])
inline_models_key = app_config["models"]["inline"]
self.dictionary = Dictionary(
join(inline_models_dir, inline_models_key["dictionary"]))
self.unigrams = Unigrams(
join(inline_models_dir, inline_models_key["unigrams"]))
self.tmp_unigrams_filename = self.unigrams.filename + app_config[
"exts"]["tmp"]
self.bigrams = Bigrams(
join(inline_models_dir, inline_models_key["bigrams"]))
self.altcase_map = AltCaseMap(
join(inline_models_dir, inline_models_key["altcase"]))
self.tmp_altcase_filename = self.altcase_map.filename + app_config[
"exts"]["tmp"]
self.ocrkey_map = OcrKeyMap(
join(inline_models_dir, inline_models_key["ocr_keys"]))
self.anagram_map = AnagramMap(
join(inline_models_dir, inline_models_key["anagrams"]))
def __init__(self, app_config):
super(InlineModel, self).__init__(app_config)
inline_models_dir = join(
app_config["root"],
app_config["dirs"]["models_root"],
app_config["dirs"]["models"]["inline"]
)
inline_models_key = app_config["models"]["inline"]
self.dictionary = Dictionary(join(inline_models_dir, inline_models_key["dictionary"]))
self.unigrams = Unigrams(join(inline_models_dir, inline_models_key["unigrams"]))
self.tmp_unigrams_filename = self.unigrams.filename + app_config["exts"]["tmp"]
self.bigrams = Bigrams(join(inline_models_dir, inline_models_key["bigrams"]))
self.altcase_map = AltCaseMap(join(inline_models_dir, inline_models_key["altcase"]))
self.tmp_altcase_filename = self.altcase_map.filename + app_config["exts"]["tmp"]
self.ocrkey_map = OcrKeyMap(join(inline_models_dir, inline_models_key["ocr_keys"]))
self.anagram_map = AnagramMap(join(inline_models_dir, inline_models_key["anagrams"]))
class InlineModel(AbstractModel):
"""Model for inline data structures
"""
def __init__(self, app_config):
super(InlineModel, self).__init__(app_config)
inline_models_dir = join(
app_config["root"],
app_config["dirs"]["models_root"],
app_config["dirs"]["models"]["inline"]
)
inline_models_key = app_config["models"]["inline"]
self.dictionary = Dictionary(join(inline_models_dir, inline_models_key["dictionary"]))
self.unigrams = Unigrams(join(inline_models_dir, inline_models_key["unigrams"]))
self.tmp_unigrams_filename = self.unigrams.filename + app_config["exts"]["tmp"]
self.bigrams = Bigrams(join(inline_models_dir, inline_models_key["bigrams"]))
self.altcase_map = AltCaseMap(join(inline_models_dir, inline_models_key["altcase"]))
self.tmp_altcase_filename = self.altcase_map.filename + app_config["exts"]["tmp"]
self.ocrkey_map = OcrKeyMap(join(inline_models_dir, inline_models_key["ocr_keys"]))
self.anagram_map = AnagramMap(join(inline_models_dir, inline_models_key["anagrams"]))
def load(self, text_data):
"""Load text data to the model
Args:
text_data (`Text`): Text data
"""
if self.is_preprocessed(text_data.filename) != 0:
self.logger.debug(text_data.filename+" already loaded: skipping it.")
return
tmp_u = Unigrams(self.tmp_unigrams_filename)
word_list = tmp_u.append_data(text_data)
self.bigrams.append_data(word_list)
tmp_ac = AltCaseMap(self.tmp_altcase_filename)
tmp_ac.append_data(tmp_u.raw_unigrams)
tmp_u.generate_low_case(tmp_ac.altcase_map)
self.ocrkey_map.append_data(tmp_u.raw_unigrams)
# Updating data
self.unigrams.raw_unigrams += tmp_u.raw_unigrams
self.unigrams.ngrams += tmp_u.ngrams
self.unigrams.prune(0.7)
self.unigrams.save()
combine_struct = {key: set() for key in tmp_ac.altcase_map.keys() + self.altcase_map.altcase_map.keys()}
for key, value in tmp_ac.altcase_map.items() + self.altcase_map.altcase_map.items():
combine_struct[key] = combine_struct[key].union(value)
self.altcase_map.altcase_map = combine_struct
self.altcase_map.prune(self.unigrams.ngrams_pruned)
self.altcase_map.save()
unlink(self.tmp_unigrams_filename)
unlink(self.tmp_altcase_filename)
self.anagram_map.append_data(self.bigrams.ngrams_pruned, self.unigrams.ngrams_pruned)
self.dictionary.append_data(self.unigrams.ngrams_pruned)
self.logger.info(text_data.filename+"'s datastructures loaded")
def correct(self, text_data):
"""Correct text data
Args:
text_data (`Text`): Text data
"""
correction_data = self.correction_data()
for paragraph in text_data.text:
for line in paragraph:
for token in line.tokens:
token[2] = init_correction_map(token[1], correction_data["dictionary"])
# Skip some correction steps if the token is too short, in the dictionary or already identified as
# garbage
if not token[2] is None and len(token[2]) == 0:
anagrams = select_anagrams(token[1], correction_data)
ocr_sims = select_ocrsims(token[1], correction_data)
token[2] = build_candidates_list(token[1], anagrams, ocr_sims, correction_data)
token[2] = correct_case(token[1], token[2], correction_data)
token[2] = rate_corrections(token[2])
if len(token[2]) == 0: # No correction has been found
token[2] = None
# Applying the bigram boost to the tokens
bigrams = extract_paragraph_bigrams(paragraph)
apply_bigram_boost(paragraph, bigrams, correction_data["occurence_map"])
# Select the appropriate correction
for line in paragraph:
for token in line.tokens:
token[2] = select_correction(token[1], token[2])
if token[2] is not None and len(token[2]) > 1:
tkn_list = [tkn for tkn, sc in token[2].items() if sc == max(token[2].values())]
if len(tkn_list) != 1:
tkn_list = select_lower_edit_distance(token[1], {tkn: token[2][tkn] for tkn in tkn_list})
if len(tkn_list) != 1:
tkn_list = [select_best_alphabetical_word(token[1], tkn_list)]
token[2] = {tkn: token[2][tkn] for tkn in tkn_list}
def correction_data(self):
"""Get the correction data
Returns:
dict: Correction data
"""
return {
"occurence_map": self.unigrams.ngrams + self.bigrams.ngrams,
"altcase": self.altcase_map.altcase_map,
"ocrkeys": self.ocrkey_map.ocrkey_map,
"anagrams": self.anagram_map.anagram_hashmap,
"alphabet": self.anagram_map.anagram_alphabet,
"dictionary": self.dictionary.dictionary
}
class InlineModel(AbstractModel):
"""Model for inline data structures
"""
def __init__(self, app_config):
super(InlineModel, self).__init__(app_config)
inline_models_dir = join(app_config["root"],
app_config["dirs"]["models_root"],
app_config["dirs"]["models"]["inline"])
inline_models_key = app_config["models"]["inline"]
self.dictionary = Dictionary(
join(inline_models_dir, inline_models_key["dictionary"]))
self.unigrams = Unigrams(
join(inline_models_dir, inline_models_key["unigrams"]))
self.tmp_unigrams_filename = self.unigrams.filename + app_config[
"exts"]["tmp"]
self.bigrams = Bigrams(
join(inline_models_dir, inline_models_key["bigrams"]))
self.altcase_map = AltCaseMap(
join(inline_models_dir, inline_models_key["altcase"]))
self.tmp_altcase_filename = self.altcase_map.filename + app_config[
"exts"]["tmp"]
self.ocrkey_map = OcrKeyMap(
join(inline_models_dir, inline_models_key["ocr_keys"]))
self.anagram_map = AnagramMap(
join(inline_models_dir, inline_models_key["anagrams"]))
def load(self, text_data):
"""Load text data to the model
Args:
text_data (`Text`): Text data
"""
if self.is_preprocessed(text_data.filename) != 0:
self.logger.debug(text_data.filename +
" already loaded: skipping it.")
return
tmp_u = Unigrams(self.tmp_unigrams_filename)
word_list = tmp_u.append_data(text_data)
self.bigrams.append_data(word_list)
tmp_ac = AltCaseMap(self.tmp_altcase_filename)
tmp_ac.append_data(tmp_u.raw_unigrams)
tmp_u.generate_low_case(tmp_ac.altcase_map)
self.ocrkey_map.append_data(tmp_u.raw_unigrams)
# Updating data
self.unigrams.raw_unigrams += tmp_u.raw_unigrams
self.unigrams.ngrams += tmp_u.ngrams
self.unigrams.prune(0.7)
self.unigrams.save()
combine_struct = {
key: set()
for key in tmp_ac.altcase_map.keys() +
self.altcase_map.altcase_map.keys()
}
for key, value in tmp_ac.altcase_map.items(
) + self.altcase_map.altcase_map.items():
combine_struct[key] = combine_struct[key].union(value)
self.altcase_map.altcase_map = combine_struct
self.altcase_map.prune(self.unigrams.ngrams_pruned)
self.altcase_map.save()
unlink(self.tmp_unigrams_filename)
unlink(self.tmp_altcase_filename)
self.anagram_map.append_data(self.bigrams.ngrams_pruned,
self.unigrams.ngrams_pruned)
self.dictionary.append_data(self.unigrams.ngrams_pruned)
self.logger.info(text_data.filename + "'s datastructures loaded")
def correct(self, text_data):
"""Correct text data
Args:
text_data (`Text`): Text data
"""
correction_data = self.correction_data()
for paragraph in text_data.text:
for line in paragraph:
for token in line.tokens:
token[2] = init_correction_map(
token[1], correction_data["dictionary"])
# Skip some correction steps if the token is too short, in the dictionary or already identified as
# garbage
if not token[2] is None and len(token[2]) == 0:
anagrams = select_anagrams(token[1], correction_data)
ocr_sims = select_ocrsims(token[1], correction_data)
token[2] = build_candidates_list(
token[1], anagrams, ocr_sims, correction_data)
token[2] = correct_case(token[1], token[2],
correction_data)
token[2] = rate_corrections(token[2])
if len(token[2]) == 0: # No correction has been found
token[2] = None
# Applying the bigram boost to the tokens
bigrams = extract_paragraph_bigrams(paragraph)
apply_bigram_boost(paragraph, bigrams,
correction_data["occurence_map"])
# Select the appropriate correction
for line in paragraph:
for token in line.tokens:
token[2] = select_correction(token[1], token[2])
if token[2] is not None and len(token[2]) > 1:
tkn_list = [
tkn for tkn, sc in token[2].items()
if sc == max(token[2].values())
]
if len(tkn_list) != 1:
tkn_list = select_lower_edit_distance(
token[1],
{tkn: token[2][tkn]
for tkn in tkn_list})
if len(tkn_list) != 1:
tkn_list = [
select_best_alphabetical_word(
token[1], tkn_list)
]
token[2] = {tkn: token[2][tkn] for tkn in tkn_list}
def correction_data(self):
"""Get the correction data
Returns:
dict: Correction data
"""
return {
"occurence_map": self.unigrams.ngrams + self.bigrams.ngrams,
"altcase": self.altcase_map.altcase_map,
"ocrkeys": self.ocrkey_map.ocrkey_map,
"anagrams": self.anagram_map.anagram_hashmap,
"alphabet": self.anagram_map.anagram_alphabet,
"dictionary": self.dictionary.dictionary
}