def predict_impl(doc, config, req_id=None):
text = doc["text"]
store = get_model_store()
entity_tags = None
response = {"text": text}
corpus, datasources_map, projects_map = fetch_data_mappings(doc)
truecased_text, retokenized_text, syn_processed_text, syn_processed_truecased_text, syn_processed_retokenized_text, syn_indexes = predict_preprocessing(
text, store, datasources_map)
for index, (serviceid, model_type,
pos) in enumerate(get_requested_services(doc)):
logger.info((serviceid, model_type, pos))
engine = get_engine(config['custom_entity_model'])
default_engine = get_engine(config['predefined_entity_model'])
if model_type == MODEL_TYPE_NER:
entities, entity_tags, parts_of_speech = ner_entities(
store, config, serviceid, model_type, engine, truecased_text,
retokenized_text, syn_processed_truecased_text,
syn_processed_retokenized_text, syn_indexes, default_engine,
datasources_map, projects_map, pos)
response["entities"] = entities
if (parts_of_speech
is not None) and (get(doc, 'pos', default=True) is True):
response["parts_of_speech"] = parts_of_speech
elif model_type == MODEL_TYPE_IR:
prediction, probabilities = ir_entity_tags(
store, serviceid, model_type, engine,
syn_processed_truecased_text, entity_tags, datasources_map)
logger.info("prediction......................%s" % prediction)
logger.info("probabilities......................%s" %
probabilities)
response["intent"] = {
"top_intent": prediction,
"confidence_level": probabilities
}
missed_text = missedUtterences(response, doc['serviceid'], req_id,
syn_processed_retokenized_text)
if missed_text is not syn_processed_retokenized_text:
updateDatasource(serviceid, missed_text)
if 'entities' not in response:
response['entities'] = []
if 'parts_of_speech' not in response:
response['parts_of_speech'] = []
return response
def tag(doc):
store = get_model_store()
text = get(doc, "text", default='')
text = text.replace('\xa0', ' ')
service_id = get(doc, "serviceid", default=None)
text, original_text = store.change_case(text)
query = {"serviceid": service_id}
corpus = datasource_manager.find_datasource_by_service_id(doc["serviceid"])
config = project_manager.find_model(query)
datasources_map = {
"predefined_entities": get(corpus, "predefined_entities", default=[]),
"entities": get(corpus, "entities", default=[]),
"patterns": get(corpus, "patterns", default=[]),
"phrases": get(corpus, "phrases", default=[]),
"distinct_token_list": get(corpus, "distinct_token_list", default=[]),
"intents": get(corpus, "intents", default=[])
}
projects_map = {
"custom_entity_model": get(config, "custom_entity_model",
default=None),
"ner_status": get(config, "ner.status", default=[]),
"language": get(config, 'language', 'EN')
}
default_class_name = get(config, 'predefined_entity_model', None)
custom_class_name = get(config, 'custom_entity_model', None)
engine = get_engine(custom_class_name)
last_trained = get(config, "ner.last_trained", default=None)
default_engine = get_engine(default_class_name)
model_name = get_model_name(service_id, "ner", engine)
if get_engine(custom_class_name) not in get_all_corenlp_engines():
if get(config, "ner.status", default=None) in ['trained', 'validated']:
get_model_store().store.check_trained_time_and_reload(
model_name, last_trained, service_id, "ner", engine,
custom_class_name)
response = store.tag(service_id, text, original_text, engine,
default_engine, default_class_name, datasources_map,
projects_map)
response = remove_resolved_to(response)
return response
def on_post(self, req, resp):
log_flag = False
if app_config['BOTANALYTICS_LOG_FLAG'].upper() == "ON":
log_flag, req_id, botanalytics, start_time = True, str(
uuid.uuid4()), BotAnalyticsAPI(), datetime.datetime.now()
doc = req.context['doc'] or {}
try:
store = get_model_store()
serviceid = doc["serviceid"]
datasource = manager.find_datasource_by_service_id(serviceid)
phrases = get(datasource, "phrases")
patterns = get(datasource, "patterns")
entities = get(datasource, "entities", [])
predefined_entities = get(datasource, "predefined_entities", [])
utterances = get(datasource, "utterances")
project_config = project_manager.find_config_by_id(serviceid)
default_engine = get_engine(
project_config['predefined_entity_model'])
type = doc["type"]
if type == "phrases":
utterances = tag_phrase(phrases, utterances)
if type == "patterns":
utterances = tag_pattern(patterns, utterances)
if type == "predefined":
utterances = tag_all_predefined(store, patterns, phrases,
entities, utterances,
default_engine,
predefined_entities,
project_config)
document = {"$set": {"utterances": utterances}}
manager.update_datasource_by_service_id(serviceid, document)
resp.data = json.dumps({"msg": "Successfully Updated"})
resp.set_header('X-Powered-By', 'USTGlobal ICE')
resp.status = falcon.HTTP_200
except Exception as ex:
logger.exception(ex)
resp.data = {"msg": ex}
resp.set_header('X-Powered-By', 'USTGlobal ICE')
resp.status = falcon.HTTP_500
finally:
if log_flag:
end_time = datetime.datetime.now()
total_action_time = relativedelta(end_time, start_time)
botanalytics.log(requesttype="nerrequests",
serviceid=doc['serviceid'],
req_id=req_id,
action="BULK TAG",
ner_req_timestamp=start_time.replace(
microsecond=0).isoformat(),
ner_req_end_timestamp=end_time.replace(
microsecond=0).isoformat(),
total_action_time=(
total_action_time.hours * 60 * 60 * 1000 +
total_action_time.minutes * 60 * 1000 +
total_action_time.seconds * 1000) +
(total_action_time.microseconds / 1000))
def cache_model(config,requested_services):
for service_each in requested_services:
serviceid = service_each[0]
model_type = service_each[1]
engine=""
model_class = None
if config[model_type]["status"] == 'trained' or config[model_type]["status"] == 'validated' or config[model_type]["status"] == 'validating':
if model_type=="ner":
model_class = config['custom_entity_model']
engine = get_engine(model_class)
model_name = serviceid + "-" + engine+ "-" +model_type
last_trained = config["ner"]["last_trained"]
else:
model_name = serviceid + "-" +model_type
last_trained = config["ir"]["last_trained"]
get_model_store().store.check_trained_time_and_reload(model_name, last_trained, serviceid, model_type, engine, model_class)
def retag(serviceid, utterances, new_default_model_class, predefined_entities, project_config):
store = get_model_store()
for utterances_each in utterances:
utterance = get(utterances_each, "utterance")
case_converted_utterance = get(utterances_each, "case_converted_utterance")
mapping = get(utterances_each, "mapping")
mapping = json.loads(mapping)
doc = {
"serviceid": serviceid,
"text": utterance
}
default_engine = get_engine(new_default_model_class)
def_tags = store.tag_predefined(MODEL_TYPE_NER, default_engine, case_converted_utterance, utterance)
def_tags = [tag for tag in def_tags if str(tag['tag']) in predefined_entities]
final_tags = []
cust_tags = get(mapping, "tags", [])
final_tags = remove_overlapping(def_tags, cust_tags)
mapping["tags"] = final_tags
mapping = json.dumps(mapping)
utterances_each["mapping"] = mapping
utterances_each["ir_trained"] = False
utterances_each["ner_trained"] = False
return utterances