def test_custom_features_loaded(home_assistant_app_path):
nlp = NaturalLanguageProcessor(app_path=home_assistant_app_path)
dc = nlp.domain_classifier
dc.fit()
assert 'average-token-length' in dc.config.features
example = nlp.create_query('set the temperature')
feature_set = dc._model._extract_features(example)
assert 'average_token_length' in feature_set
def test_intent_classifier_logreg(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
features = {
"bag-of-words": {
"lengths": [1]
},
"freq": {
"bins": 5
},
"in-gaz": {},
"length": {},
}
ic = nlp.domains["store_info"].intent_classifier
ic.fit(model_settings={"classifier_type": "logreg"}, features=features)
features = {
"bag-of-words": {
"lengths": [1, 2]
},
"freq": {
"bins": 5
},
"in-gaz": {},
"length": {},
}
ic.fit(model_settings={"classifier_type": "logreg"}, features=features)
def test_domain_classifier_random_forest(home_assistant_app_path):
nlp = NaturalLanguageProcessor(app_path=home_assistant_app_path)
dc = nlp.domain_classifier
params = {"C": 10}
with patch("logging.Logger.warning") as mock:
dc.fit(model_settings={"classifier_type": "rforest"}, params=params)
mock.assert_any_call("Unexpected param `C`, dropping it from model config.")
def test_intent_classifier_random_forest(kwik_e_mart_app_path, caplog):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
search_grid = {
'n_estimators': [5, 10, 15, 20],
'criterion': ['gini', 'entropy'],
'warm_start': [True, False],
}
param_selection_settings = {
'grid': search_grid,
'type': 'k-fold',
'k': 10
}
ic = nlp.domains['store_info'].intent_classifier
ic.fit(model_settings={'classifier_type': 'rforest'}, param_selection=param_selection_settings)
ic.fit(model_settings={'classifier_type': 'rforest'},
param_selection={'type': 'k-fold', 'k': 10, 'grid': {'class_bias': [0.7, 0.3, 0]}})
features = {
'bag-of-words': {'lengths': [1, 2]},
'freq': {'bins': 5},
'in-gaz': {},
'length': {}
}
with patch('logging.Logger.warning') as mock:
ic.fit(model_settings={'classifier_type': 'rforest'},
features=features)
mock.assert_any_call('Unexpected param `C`, dropping it from model config.')
mock.assert_any_call('Unexpected param `fit_intercept`, dropping it from model config.')
def test_intent_classifier_svm(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
search_grid = {
"C": [0.1, 0.5, 1, 5, 10, 50, 100, 1000, 5000],
"kernel": ["linear", "rbf", "poly"],
}
param_selection_settings = {"grid": search_grid, "type": "k-fold", "k": 10}
ic = nlp.domains["store_info"].intent_classifier
ic.fit(
model_settings={"classifier_type": "svm"},
param_selection=param_selection_settings,
)
def test_intent_classifier_svm(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
search_grid = {
'C': [0.1, 0.5, 1, 5, 10, 50, 100, 1000, 5000],
'kernel': ['linear', 'rbf', 'poly'],
}
param_selection_settings = {
'grid': search_grid,
'type': 'k-fold',
'k': 10
}
ic = nlp.domains['store_info'].intent_classifier
ic.fit(model_settings={'classifier_type': 'svm'}, param_selection=param_selection_settings)
def test_domain_classifier_get_stats(home_assistant_app_path, capsys):
nlp = NaturalLanguageProcessor(app_path=home_assistant_app_path)
dc = nlp.domain_classifier
dc.fit()
eval = dc.evaluate()
eval.print_stats()
captured = capsys.readouterr()
all_elems = set([k for k in captured.out.replace("\n", "").split(" ") if k != ""])
assert "Overall" in all_elems
assert "statistics:" in all_elems
assert "accuracy" in all_elems
assert "f1_weighted" in all_elems
assert "tp" in all_elems
assert "fp" in all_elems
assert "fn" in all_elems
def test_tagger_get_stats(kwik_e_mart_app_path, capsys):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
er = nlp.domains['store_info'].intents['get_store_hours'].entity_recognizer
er.fit()
eval = er.evaluate()
eval.print_stats()
captured = capsys.readouterr()
all_elems = set(
[k for k in captured.out.replace('\n', '').split(' ') if k != ''])
assert 'Overall' in all_elems
assert 'statistics:' in all_elems
assert 'accuracy' in all_elems
assert 'f1_weighted' in all_elems
assert 'tp' in all_elems
assert 'fp' in all_elems
assert 'fn' in all_elems
def test_tagger_get_stats(kwik_e_mart_app_path, capsys):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
er = nlp.domains["store_info"].intents["get_store_hours"].entity_recognizer
er.fit()
eval = er.evaluate()
eval.print_stats()
captured = capsys.readouterr()
all_elems = set(
[k for k in captured.out.replace("\n", "").split(" ") if k != ""])
assert "Overall" in all_elems
assert "statistics:" in all_elems
assert "accuracy" in all_elems
assert "f1_weighted" in all_elems
assert "tp" in all_elems
assert "fp" in all_elems
assert "fn" in all_elems
def test_domain_classifier_get_stats(home_assistant_app_path, capsys):
nlp = NaturalLanguageProcessor(app_path=home_assistant_app_path)
dc = nlp.domain_classifier
dc.fit()
eval = dc.evaluate()
eval.print_stats()
captured = capsys.readouterr()
all_elems = set(
[k for k in captured.out.replace('\n', '').split(' ') if k != ''])
assert 'Overall' in all_elems
assert 'statistics:' in all_elems
assert 'accuracy' in all_elems
assert 'f1_weighted' in all_elems
assert 'tp' in all_elems
assert 'fp' in all_elems
assert 'fn' in all_elems
def test_memm_model(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
er = nlp.domains['store_info'].intents['get_store_hours'].entity_recognizer
params = {
'type': 'k-fold',
'k': 5,
'scoring': 'accuracy',
'grid': {
'penalty': ['l1', 'l2'],
'C': [0.01, 1, 100, 10000, 1000000, 100000000]
},
}
with patch('logging.Logger.warning') as mock:
er.fit(param_selection=params)
assert 'C' in er.config.param_selection['grid']
assert 'penalty' in er.config.param_selection['grid']
mock.assert_not_called()
def test_memm_model(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
er = nlp.domains["store_info"].intents["get_store_hours"].entity_recognizer
params = {
"type": "k-fold",
"k": 5,
"scoring": "accuracy",
"grid": {
"penalty": ["l1", "l2"],
"C": [0.01, 1, 100, 10000, 1000000, 100000000],
},
}
with patch("logging.Logger.warning") as mock:
er.fit(param_selection=params)
assert "C" in er.config.param_selection["grid"]
assert "penalty" in er.config.param_selection["grid"]
mock.assert_not_called()
def test_intent_classifier_random_forest(kwik_e_mart_app_path, caplog):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
search_grid = {
"n_estimators": [5, 10, 15, 20],
"criterion": ["gini", "entropy"],
"warm_start": [True, False],
}
param_selection_settings = {"grid": search_grid, "type": "k-fold", "k": 10}
ic = nlp.domains["store_info"].intent_classifier
ic.fit(
model_settings={"classifier_type": "rforest"},
param_selection=param_selection_settings,
)
ic.fit(
model_settings={"classifier_type": "rforest"},
param_selection={
"type": "k-fold",
"k": 10,
"grid": {
"class_bias": [0.7, 0.3, 0]
},
},
)
features = {
"bag-of-words": {
"lengths": [1, 2]
},
"freq": {
"bins": 5
},
"in-gaz": {},
"length": {},
}
with patch("logging.Logger.warning") as mock:
ic.fit(model_settings={"classifier_type": "rforest"},
features=features)
mock.assert_any_call(
"Unexpected param `C`, dropping it from model config.")
mock.assert_any_call(
"Unexpected param `fit_intercept`, dropping it from model config.")
def test_intent_classifier_logreg(kwik_e_mart_app_path):
nlp = NaturalLanguageProcessor(app_path=kwik_e_mart_app_path)
features = {
'bag-of-words': {'lengths': [1]},
'freq': {'bins': 5},
'in-gaz': {},
'length': {}
}
ic = nlp.domains['store_info'].intent_classifier
ic.fit(model_settings={'classifier_type': 'logreg'},
features=features)
features = {
'bag-of-words': {'lengths': [1, 2]},
'freq': {'bins': 5},
'in-gaz': {},
'length': {}
}
ic.fit(model_settings={'classifier_type': 'logreg'},
features=features)
from mindmeld.components.nlp import NaturalLanguageProcessor
nlp = NaturalLanguageProcessor('.')
nlp.build()
from mindmeld.components.nlp import NaturalLanguageProcessor
import spacy
import os.path
spacyNlp = spacy.load("en_core_web_sm")
jobTitleNlp = spacy.load("JOB_TITLE")
dateTimeNlp = spacy.load("DATE_TIME")
input_text = input("Enter text to process:\n")
path = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
nlp = NaturalLanguageProcessor(app_path=path)
nlp.build(incremental=True)
sentences = input_text.split(".")
reply = "\nThe contract has been summarized into the key details, as seen below \n\n"
personCounter = 0
for s in sentences:
if s is not "":
processedWord = nlp.process(s)
intent = processedWord.get('intent')
domain = processedWord.get('domain')
print("Intent is : " + intent)
print("Domain is : " + domain)
# handle salary case
result_salary = ''
if intent == 'get_salary':
for c in s:
if c.isdigit():
result_salary += c
reply += "The SALARY of the employee is " + result_salary + "/month\n"
def handle_message(message):
input_text = message.text
path = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
nlp = NaturalLanguageProcessor(app_path=path)
nlp.build(incremental=True)
sentences = input_text.split(".")
reply = "\nThe contract has been summarized into the key details, as seen below \n\n"
for s in sentences:
if s is not "":
processedWord = nlp.process(s)
intent = processedWord.get('intent')
domain = processedWord.get('domain')
print("Intent is : " + intent)
print("Domain is : " + domain)
# handle salary case
result_salary = ''
if intent == 'get_salary':
for c in s:
if c.isdigit():
result_salary += c
reply += "The SALARY of the employee is " + result_salary + "/month\n"
doc = spacyNlp(s)
jobDoc = jobTitleNlp(s)
dateDoc = dateTimeNlp(s)
for ent in doc.ents:
if ent.label_ == 'PERSON' and intent == 'employment_details':
reply += "The EMPLOYEE in the contract is : " + ent.text + "\n"
elif ent.label_ == 'PERSON' and intent == 'employed_by':
reply += "The EMPLOYER in the contract is : " + ent.text + "\n"
elif ent.label_ == 'ORG':
reply += "The COMPANY that the employee will report to is : " + ent.text + "\n"
elif ent.label_ == 'GPE':
reply += "APPLICABLE LAW : " + ent.text + "\n"
for ent in jobDoc.ents:
if intent == 'job_details':
reply += "The JOB TITLE of the employee is : " + ent.text + "\n"
#print(ent.text, ent.start_char, ent.end_char, ent.label_)
for ent in dateDoc.ents:
if intent == 'contract_date':
reply += "The contract was SIGNED on : " + ent.text + "\n"
elif intent == 'effective_date':
reply += "The STARTING date of employment is : " + ent.text + "\n"
elif intent == 'last_date':
reply += "The TERMINATION date of employment is : " + ent.text + "\n"
# path = os.path.abspath(os.path.join(os.path.dirname(__file__), ".."))
# nlp = NaturalLanguageProcessor(app_path=path)
# nlp.build(incremental=True)
# processWord = message.text
# processedWord = nlp.process(processWord)
# intent = processedWord.get('intent')
# domain = processedWord.get('domain')
# logging.info('intent: ' + intent)
# logging.info('domain: ' + domain)
# if domain == 'greeting':
# reply = handleGreeting(intent)
# else: # domain = nanocore_faq
# reply = handleFaq(intent)
bot.reply_to(message, reply) # responds with success message