def encode_itranslate_decode(i, num_map, tp_tokenizer, num_hypotheses=3):
try:
log_info("Inside encode_itranslate_decode function", MODULE_CONTEXT)
model_path, sp_encoder, sp_decoder = get_model_path(i['id'])
translator = load_models.loaded_models[i['id']]
i['src'] = str(sp.encode_line(sp_encoder, i['src']))
i_final = format_converter(i['src'])
if 'target_prefix' in i and len(
i['target_prefix']) > 0 and i['target_prefix'].isspace(
) == False:
log_info("target prefix: {}".format(i['target_prefix']),
MODULE_CONTEXT)
i['target_prefix'] = misc.convert_digits_preprocess(
i['tgt_lang'], i['target_prefix'])
i['target_prefix'] = replace_num_target_prefix(i, num_map)
if tp_tokenizer is not None:
i['target_prefix'] = tp_tokenizer(i['target_prefix'])
i['target_prefix'] = str(
sp.encode_line(sp_decoder, i['target_prefix']))
tp_final = format_converter(i['target_prefix'])
tp_final[-1] = tp_final[-1].replace(']', ",")
m_out = translator.translate_batch([i_final],
beam_size=5,
target_prefix=[tp_final],
num_hypotheses=num_hypotheses,
replace_unknowns=True)
else:
m_out = translator.translate_batch([i_final],
beam_size=5,
num_hypotheses=num_hypotheses,
replace_unknowns=True)
translation = multiple_hypothesis_decoding(m_out[0], sp_decoder)
return translation
except Exception as e:
log_exception(
"Unexpexcted error in encode_itranslate_decode: {} and {}".format(
e,
sys.exc_info()[0]), MODULE_CONTEXT, e)
raise
def interactive_translation(inputs):
out = {}
i_src, tgt = list(), list()
tagged_tgt = list()
tagged_src = list()
sentence_id = list()
tp_tokenizer = None
try:
for i in inputs:
sentence_id.append(i.get("s_id") or "NA")
if any(v not in i for v in ['src', 'id']):
log_info("either id or src missing in some input",
MODULE_CONTEXT)
out = CustomResponse(Status.ID_OR_SRC_MISSING.value,
inputs)
return out
log_info("input sentence:{}".format(i['src']), MODULE_CONTEXT)
i_src.append(i['src'])
i['src'] = i['src'].strip()
i['src_lang'], i['tgt_lang'] = misc.get_src_tgt_langauge(
i['id'])
i['src'] = misc.convert_digits_preprocess(
i['src_lang'], i['src'])
if special_case_handler.special_case_fits(i['src']):
log_info(
"sentence fits in special case, returning accordingly and not going to model",
MODULE_CONTEXT)
translation = special_case_handler.handle_special_cases(
i['src'], i['id'])
translation = [translation]
tag_tgt, tag_src = translation, i['src']
else:
log_info(
"Performing interactive translation on:{}".format(
i['id']), MODULE_CONTEXT)
i['src'], date_original, url_original, num_array, num_map = tagger_util.tag_number_date_url(
i['src'])
tag_src = i['src']
if i['id'] == 56:
"english-hindi"
if i['src'].isupper():
log_info(
"src all Upper case hence Tital casing it",
MODULE_CONTEXT)
i['src'] = i['src'].title()
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
elif i['id'] == 7:
"english-tamil"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 10:
"english-gujarati"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 15:
"english-kannada"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 16:
"english-telugu"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 17:
"english-malayalam"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 18:
"english-punjabi"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 42:
"english-marathi"
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
elif i['id'] == 50:
"telugu-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 6:
"hindi-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 62:
"marathi-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 8:
"tamil-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 55:
"punjabi-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 48:
"kannada-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 60:
"malayalam-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 52:
"gujarati-english"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 65:
"english-bengali 4th"
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
elif i['id'] == 66:
"bengali-english 3rd"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 67:
"ta-en 3rd"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 68:
"en-ta 5th"
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
elif i['id'] == 69:
"hi-en 3rd"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 70:
"en-hi 15th"
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
elif i['id'] == 71:
"te-en 2nd"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 72:
"en-te 3rd"
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
elif i['id'] == 73:
"ml-en 2nd"
tp_tokenizer = sentence_processor.moses_tokenizer
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.moses_detokenizer(i)
for i in translation
]
elif i['id'] == 74:
"en-ml 3rd"
tp_tokenizer = sentence_processor.indic_tokenizer
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation = encode_itranslate_decode_v2(
i, num_map, tp_tokenizer)
translation = [
sentence_processor.indic_detokenizer(i)
for i in translation
]
else:
log_info(
"unsupported model id: {} for given input".format(
i['id']), MODULE_CONTEXT)
raise Exception(
"Unsupported Model ID - id: {} for given input".
format(i['id']))
translation = [i.replace("▁", " ") for i in translation]
translation = [
misc.regex_pass(i, [
patterns['p8'], patterns['p9'], patterns['p4'],
patterns['p5'], patterns['p6'], patterns['p7']
]) for i in translation
]
tag_tgt = translation
translation = [
tagger_util.replace_tags_with_original(
i, date_original, url_original, num_array, num_map)
for i in translation
]
translation = [
misc.convert_digits_postprocess(i['tgt_lang'], item)
for item in translation
]
log_info(
"interactive translation-experiment-{} output: {}".format(
i['id'], translation), MODULE_CONTEXT)
tgt.append(translation)
tagged_tgt.append(tag_tgt)
tagged_src.append(tag_src)
out['response_body'] = [{
"tgt": tgt[i],
"tagged_tgt": tagged_tgt[i],
"tagged_src": tagged_src[i],
"s_id": sentence_id[i],
"src": i_src[i]
} for i in range(len(tgt))]
out = CustomResponse(Status.SUCCESS.value, out['response_body'])
except Exception as e:
status = Status.SYSTEM_ERR.value
status['why'] = str(e)
log_exception(
"Unexpected error:%s and %s" % (e, sys.exc_info()[0]),
MODULE_CONTEXT, e)
out = CustomResponse(status, inputs)
return out
def translate_func(inputs):
inputs = inputs
out = {}
pred_score = list()
sentence_id, node_id = list(), list()
input_subwords, output_subwords = list(), list()
i_src, tgt = list(), list()
tagged_tgt, tagged_src = list(), list()
s_id, n_id = [0000], [0000]
i_s0_src, i_s0_tgt, i_save = list(), list(), list()
i_tmx_phrases = list()
try:
for i in inputs:
s0_src, s0_tgt, save = "NA", "NA", False
if all(v in i for v in ['s_id', 'n_id']):
s_id = [i['s_id']]
n_id = [i['n_id']]
if any(v not in i for v in ['src', 'id']):
log_info("either id or src missing in some input",
MODULE_CONTEXT)
out = CustomResponse(Status.ID_OR_SRC_MISSING.value,
inputs)
return out
if any(v in i for v in ['s0_src', 's0_tgt', 'save']):
s0_src, s0_tgt, save = handle_custome_input(
i, s0_src, s0_tgt, save)
i_s0_src.append(s0_src), i_s0_tgt.append(
s0_tgt), i_save.append(save)
log_info("input sentences:{}".format(i['src']), MODULE_CONTEXT)
i_src.append(i['src'])
i['src'] = i['src'].strip()
src_language, tgt_language = misc.get_src_tgt_langauge(i['id'])
if src_language == 'English' and i['src'].isupper():
i['src'] = i['src'].title()
i['src'] = misc.convert_digits_preprocess(
src_language, i['src'])
if special_case_handler.special_case_fits(i['src']):
log_info(
"sentence fits in special case, returning accordingly and not going to model",
MODULE_CONTEXT)
translation = special_case_handler.handle_special_cases(
i['src'], i['id'])
scores = [1]
input_sw, output_sw, tag_tgt, tag_src = "", "", translation, i[
'src']
else:
log_info("translating using NMT-model:{}".format(i['id']),
MODULE_CONTEXT)
prefix, i['src'] = special_case_handler.prefix_handler(
i['src'])
i['src'], date_original, url_original, num_array, num_map = tagger_util.tag_number_date_url(
i['src'])
tag_src = (prefix + " " + i['src']).lstrip()
i['src'], is_missing_stop_punc = special_case_handler.handle_a_sentence_wo_stop(
src_language, i['src'])
if i['id'] == 6:
"hi-en_exp-2 05-05-20"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 7:
"english-tamil"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 10:
"english-gujrati"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 15:
"english-kannada"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 16:
"english-telgu"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 17:
"english-malayalam"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 18:
"english-punjabi"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 42:
"english-marathi exp-2"
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
elif i['id'] == 56:
"09/12/19-Exp-5.6:"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 8:
"ta-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 44:
"eng-mr-3rd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 47:
"en-kn 2nd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 48:
"kn-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 49:
"en-tel 2nd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 50:
"tel-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 51:
"en-guj 2nd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 52:
"guj-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 53:
"en-punjabi 2nd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 55:
"punjabi-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 57:
"en-bengali 3rd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 58:
"bengali-en 2nd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 59:
"en-malay 2nd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 60:
"malay-en 1st"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 62:
"mr-to-en 2nd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 65:
"en-bengali 4th"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 66:
"bengali-en 3rd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 67:
"ta-en 3rd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 68:
"en-ta 5th"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 69:
"hi-en 3rd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 70:
"en-hi 15th"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 71:
"te-en 2nd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 72:
"en-te 3rd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.indic_detokenizer(
translation)
elif i['id'] == 73:
"ml-en 2nd"
i['src'] = sentence_processor.indic_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.moses_detokenizer(
translation)
elif i['id'] == 74:
"en-ml 3rd"
i['src'] = sentence_processor.moses_tokenizer(i['src'])
translation, scores, input_sw, output_sw = encode_translate_decode_v2(
i)
translation = sentence_processor.indic_detokenizer(
translation)
else:
log_info(
"Unsupported model id: {} for given input".format(
i['id']), MODULE_CONTEXT)
raise Exception(
"Unsupported Model ID - id: {} for given input".
format(i['id']))
translation = oc.postprocess_a_sentence_wo_stop(
tgt_language, translation, is_missing_stop_punc)
translation = (prefix + " " + translation).lstrip()
translation = translation.replace("▁", " ")
translation = misc.regex_pass(translation, [
patterns['p8'], patterns['p9'], patterns['p4'],
patterns['p5'], patterns['p6'], patterns['p7']
])
tag_tgt = translation
translation = tagger_util.replace_tags_with_original(
translation, date_original, url_original, num_array,
num_map)
translation = oc.cleaner(tag_src, translation, i['id'])
translation = misc.convert_digits_postprocess(
tgt_language, translation)
log_info(
"translate_function-experiment-{} output: {}".format(
i['id'], translation), MODULE_CONTEXT)
tgt.append(translation)
pred_score.append(scores)
sentence_id.append(s_id[0]), node_id.append(n_id[0])
input_subwords.append(input_sw), output_subwords.append(
output_sw)
tagged_tgt.append(tag_tgt), tagged_src.append(tag_src)
i_tmx_phrases.append(i.get("tmx_phrases", []))
out['response_body'] = [{
"tgt": tgt[i],
"pred_score": pred_score[i],
"s_id": sentence_id[i],
"input_subwords": input_subwords[i],
"output_subwords": output_subwords[i],
"n_id": node_id[i],
"src": i_src[i],
"tagged_tgt": tagged_tgt[i],
"tagged_src": tagged_src[i],
"save": i_save[i],
"s0_src": i_s0_src[i],
"s0_tgt": i_s0_tgt[i],
"tmx_phrases": i_tmx_phrases[i]
} for i in range(len(tgt))]
out = CustomResponse(Status.SUCCESS.value, out['response_body'])
except ServerModelError as e:
status = Status.SEVER_MODEL_ERR.value
status['why'] = str(e)
log_exception(
"ServerModelError error in TRANSLATE_UTIL-translate_func: {} and {}"
.format(e,
sys.exc_info()[0]), MODULE_CONTEXT, e)
out = CustomResponse(status, inputs)
except Exception as e:
status = Status.SYSTEM_ERR.value
status['why'] = str(e)
log_exception(
"Unexpected error:%s and %s" % (e, sys.exc_info()[0]),
MODULE_CONTEXT, e)
out = CustomResponse(status, inputs)
return out
def batch_translator(input_dict):
model_id = input_dict['id']
src_list = input_dict['src_list']
num_sentence = len(src_list)
input_subwords_list = [None] * num_sentence
output_subwords_list = [None] * num_sentence
tagged_src_list = [None] * num_sentence
tagged_tgt_list = [None] * num_sentence
tgt_list = [None] * num_sentence
score_list = [None] * num_sentence
out = {}
date_original_array = [None] * num_sentence
url_original_array = [None] * num_sentence
num_array_array = [None] * num_sentence
num_map_array = [None] * num_sentence
prefix_array = [None] * num_sentence
sp_encoder, translator, sp_decoder = get_models(model_id)
input_sentence_array_prepd = [None] * num_sentence
special_case_sentence_indices = []
src_language, tgt_language = misc.get_src_tgt_langauge(model_id)
try:
for i, sent in enumerate(src_list):
input_sentence = sent.strip()
input_sentence = misc.convert_digits_preprocess(
src_language, input_sentence)
if special_case_handler.special_case_fits(input_sentence):
special_case_sentence_indices.append(i)
log_info(
"sentence fits in special case, capturing index to process at last",
MODULE_CONTEXT)
else:
prefix_array[
i], input_sentence = special_case_handler.prefix_handler(
input_sentence)
input_sentence,date_original_array[i],url_original_array[i],num_array_array[i],num_map_array[i] = \
tagger_util.tag_number_date_url(input_sentence)
tagged_src_list[i] = (prefix_array[i] + " " +
input_sentence).lstrip()
input_sentence_array_prepd[i] = input_sentence
input_sentence_array_prepd, sent_indices_wo_stop = \
special_case_handler.handle_sentences_wo_stop(src_language,input_sentence_array_prepd)
log_info("translating using NMT-model:{}".format(model_id),
MODULE_CONTEXT)
if model_id == 5:
"hi-en exp-1"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 6:
"hi-en_exp-2 05-05-20"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 7:
"english-tamil"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 10:
"english-gujrati"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
# translation = translation.replace("ન્યાય માટે Accessક્સેસને","ન્યાયની પહોંચને")
elif model_id == 11:
"english-bengali"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 15:
"english-kannada"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
# translation = translation.replace("uc","")
elif model_id == 16:
"english-telgu"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 17:
"english-malayalam"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 18:
"english-punjabi"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 32:
"29/10/2019 Exp-12: old_data_original+lc_cleaned+ ik names translated from google(100k)+shabdkosh(appended 29k new),BPE-24K,50knmt,shuff,pretok"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 42:
"english-marathi exp-2"
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
elif model_id == 56:
"09/12/19-Exp-5.6:"
input_sentence_array_prepd = [
sentence.title() if sentence.isupper() else sentence
for sentence in input_sentence_array_prepd
]
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 8:
"ta-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 44:
"eng-mr-3rd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 45:
"en-ta 4th"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 47:
"en-kn 2nd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 48:
"kn-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 49:
"en-tel 2nd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 50:
"tel-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 51:
"en-guj 2nd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 52:
"guj-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 53:
"en-punjabi 2nd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 55:
"punjabi-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 57:
"en-bengali 3rd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 58:
"bengali-en 2nd"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 59:
"en-malay 2nd"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 60:
"malay-en 1st"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 61:
"ta-to-en 3rd"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 62:
"mr-to-en 2nd"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
elif model_id == 65:
"en-bengali 4th"
input_sentence_array_prepd = [
sentence_processor.moses_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.indic_detokenizer(translation)
for translation in translation_array
]
elif model_id == 66:
"bengali-en 3rd"
input_sentence_array_prepd = [
sentence_processor.indic_tokenizer(sentence)
for sentence in input_sentence_array_prepd
]
translation_array, input_subwords_list, output_subwords_list, score_list = \
encode_translate_decode(input_sentence_array_prepd,sp_encoder,translator,sp_decoder,input_subwords_list,output_subwords_list,score_list)
translation_array = [
sentence_processor.moses_detokenizer(translation)
for translation in translation_array
]
else:
log_info(
"Unsupported model id: {} for given input".format(
model_id), MODULE_CONTEXT)
raise Exception(
"Unsupported Model ID - id: {} for given input".format(
model_id))
translation_array = oc.postprocess_sentences_wo_stop(
tgt_language, translation_array, sent_indices_wo_stop)
for i in range(num_sentence):
if i in special_case_sentence_indices:
log_info(
"sentence fits in special case, returning output accordingly and not from model",
MODULE_CONTEXT)
tgt_list[i] = special_case_handler.handle_special_cases(
src_list[i].strip(), model_id)
score_list[i] = 1
input_subwords_list[i],output_subwords_list[i],tagged_tgt_list[i],tagged_src_list[i] = \
"","",tgt_list[i],src_list[i].strip()
else:
translation_array[i] = (prefix_array[i] + " " +
translation_array[i]).lstrip()
translation_array[i] = translation_array[i].replace(
"▁", " ")
translation_array[i] = misc.regex_pass(
translation_array[i], [
patterns['p8'], patterns['p9'], patterns['p4'],
patterns['p5'], patterns['p6'], patterns['p7']
])
tagged_tgt_list[i] = translation_array[i]
translation_array[i] = tagger_util.replace_tags_with_original(translation_array[i],\
date_original_array[i],url_original_array[i],num_array_array[i],num_map_array[i])
translation_array[i] = oc.cleaner(tagged_src_list[i],
translation_array[i],
model_id)
tgt_list[i] = translation_array[i]
log_info(
"translate_function-experiment-{} output: {}".format(
model_id, translation_array[i]), MODULE_CONTEXT)
tgt_list[i] = misc.convert_digits_postprocess(
tgt_language, tgt_list[i])
if (not tgt_list[i]) or (tgt_list[i].isspace()):
tgt_list[i] = src_list[i]
out = {
"tagged_src_list": tagged_src_list,
"tagged_tgt_list": tagged_tgt_list,
"tgt_list": tgt_list
}
except ServerModelError as e:
log_exception(
"ServerModelError error in TRANSLATE_UTIL-translate_func: {} and {}"
.format(e,
sys.exc_info()[0]), MODULE_CONTEXT, e)
raise e
except Exception as e:
log_exception(
"Exception caught in NMTTranslateService:batch_translator:%s and %s"
% (e, sys.exc_info()[0]), MODULE_CONTEXT, e)
raise e
return out