class Parrott:
def __init__(self):
'''
Births a new Parrott,
with a Naive Bayes brain
and a Solr memory.
'''
self.brain = naive_bayes.NaiveBayes()
self.twitter = membrane.twitter.api()
self.memory = Memory()
def train(self):
'''
Trains the Parrott on examples
stored in Memory (Solr)
'''
# Fetch 1000 positive and negative examples.
pos_tweets = self.memory.recall_positive(0, 1000)
neg_tweets = self.memory.recall_negative(0, 1000)
# Pull out the proper text for training data.
pos = [' '.join([tweet['tweet'], tweet['user']]) for tweet in pos_tweets]
neg = [' '.join([tweet['tweet'], tweet['user']]) for tweet in neg_tweets]
# Train the brain.
self.brain.train(pos, neg)
def test(self, pos_set, neg_set, threshold=0.8):
'''
Tests the Parrott on a set of positive
and negative examples.
Args:
pos_set (list): positive test examples
neg_set (list): negative test examples
threshold (float): testing threshold
Returns:
dict of test results.
'''
return self.brain.test(threshold, pos_set, neg_set)
def classify(self, tweet):
'''
Classifies a Tweet.
Args:
tweet (string): the Tweet content
Returns:
probability of positive classification.
'''
return self.brain.classify(tweet)
def audit(self, tweet, pos):
'''
Deletes an existing Tweet from Memory
and replaces it with the updated Tweet.
Args:
tweet (dict): the Tweet to audit
pos (bool): is the Tweet is a positive example?
'''
# Solr cannot "update", just delete & add.
self.memory.forget(tweet)
# Update the tweet.
tweet.update({'positive':pos, 'audited':True})
# Clear id and _version_ to prevent conflicts.
del tweet['id']
del tweet['_version_']
# Re-memorize.
self.memory.memorize(tweet)
