def create_api_app():
""" Returns the backend app """
app = Flask(__name__, instance_relative_config=True)
app.register_blueprint(API_BP)
app.config.from_object(os.environ['APP_SETTINGS'])
app.config['SQLALCHEMY_TRACK_MODIFICATIONS'] = False
DB.init_app(app)
return app
class ChatState(db.Model):
__tablename__ = 'chat_state'
id = db.Column(db.Integer, primary_key=True)
client_id = db.Column(db.String)
prompt_key = db.Column(db.String)
created_at = db.Column(db.DateTime, default=datetime.datetime.now)
def __init__(self, client_id, prompt_key):
self.client_id = client_id
self.prompt_key = prompt_key
def get_vocab(vocab_size=40, return_freq=False):
db = DB()
all_messages = db.cursor.execute(
'SELECT message FROM conversation_history').fetchall()
count = dict()
for msg in all_messages:
(m, ) = msg
ml = m.lower()
for m in ml:
if m in count.keys():
count[m] += 1
else:
count[m] = 1
keys_freq = sorted(count, key=count.get, reverse=True)
vocab = []
for i in range(vocab_size):
vocab.append(keys_freq[i])
freq = []
for key in keys_freq[0:vocab_size]:
freq.append(count[key])
if return_freq:
return vocab, freq
else:
return vocab
def main():
db = DB(settings.TOP_500_MOVIE_REVIEWS)
reviews = get_reviews(db=db)
feature_finder = MostCommonWordsFinder.load('top500reviews.featureset')
def pred(review):
return review['sentiment'] == SentimentEnum.GOOD
partitioner = DistributionPartitioner(reviews, pred=pred, ratio=0.85)
partitioner.partition()
train_data = get_labeled_review_data(
partitioner.training_data,
feature_finder.find_features,
)
test_data = get_labeled_review_data(
partitioner.test_data,
feature_finder.find_features,
)
# Nu-Support Vector Classification
# nu 0.01-0.05 to work
classifier = NuSVC(nu=0.04, gamma='auto')
# 0.8547
NuSVC_classifier = SklearnClassifier(classifier)
NuSVC_classifier.train(train_data)
print("NuSVC_classifier accuracy percent:",
nltk.classify.accuracy(NuSVC_classifier, test_data))
class TherapyConfig(db.Model):
__tablename__ = 'therapy_configs'
id = db.Column(db.Integer, primary_key=True)
name = db.Column(db.String)
config = db.Column(db.JSON)
updated_at = db.Column(db.DateTime, default=datetime.datetime.now)
created_at = db.Column(db.DateTime, default=datetime.datetime.now)
def __init__(self, name, config):
self.name = name
self.config = config
def serialize(self):
return {
'id': self.id,
'name': self.name,
'config': self.config,
}
class ClientData(db.Model):
__tablename__ = 'client_data'
id = db.Column(db.Integer, primary_key=True)
client_id = db.Column(db.String)
config_id = db.Column(db.String)
session_id = db.Column(db.Integer)
prompt_key = db.Column(db.String)
prompt_text = db.Column(db.String, nullable=True)
response = db.Column(db.String)
created_at = db.Column(db.DateTime, default=datetime.datetime.now)
def __init__(self, client_id, config_id, session_id, prompt_key, response):
self.client_id = client_id
self.config_id = config_id
self.session_id = session_id
self.prompt_key = prompt_key
self.response = response
def load_data(self):
db = DB(settings.TOP_500_MOVIE_REVIEWS)
reviews = get_reviews(db=db)
feature_finder = MostCommonWordsFinder.load('top500reviews.featureset')
def pred(review):
return review['sentiment'] == SentimentEnum.GOOD
partitioner = DistributionPartitioner(reviews, pred=pred, ratio=0.85)
partitioner.partition()
self.train_data = get_labeled_review_data(
partitioner.training_data,
feature_finder=feature_finder.find_features,
)
self.test_data = get_labeled_review_data(
partitioner.test_data,
feature_finder=feature_finder.find_features,
)
ComplementNB,
MultinomialNB,
)
from sklearn.svm import SVC, LinearSVC, NuSVC, SVR, NuSVR
from sklearn.linear_model import (
Perceptron,
SGDClassifier,
RidgeClassifier,
LogisticRegression,
LogisticRegressionCV,
OrthogonalMatchingPursuit,
PassiveAggressiveClassifier,
)
logger = get_logger(__name__)
db = DB(settings.TOP_500_MOVIE_REVIEWS)
def generate_simple_markov_reviews():
gen = SimpleMarkovGenerator()
sentiments = (
SentimentEnum.GOOD,
SentimentEnum.NEUTRAL,
SentimentEnum.BAD,
)
for sentiment in sentiments:
reviews = get_reviews_by_sentiment(sentiment, db=db)
text = get_text_for_reviews(reviews)
gen.initialize_model(sentiment, text)
logger.info('Generating reviews using %s', gen)
def get_all_messeges():
db = DB()
all_messages = db.cursor.execute(
'SELECT message FROM conversation_history').fetchall()
return all_messages