class Guesser():
def __init__(self, feed, user, config):
import os.path
self.user = user
self.filename = config['bayes_dir']
self.filename += "/users/%s" % user.id
if not os.path.exists(self.filename):
os.makedirs(self.filename)
self.filename += '/feed_%s.bayes' % str(feed.id)
log.debug("filename:%s" % self.filename)
stopwords = meta.Session\
.query(model.Stopword)\
.filter_by(feed_id=feed.id).all()
self.stopwords = map(lambda x: x.word, stopwords)
self.trainer = Bayes()
self.trainer.getTokens = lambda x: my_tokenize(x, self.stopwords)
if os.path.exists(self.filename):
self.trainer.load(self.filename)
else:
self.trainer.newPool('ham')
self.trainer.newPool('spam')
def save(self):
self.trainer.save(self.filename)
def clear(self):
self.trainer = Bayes()
# self.trainer.getTokens = my_tokenize
self.trainer.getTokens = lambda x: my_tokenize(x, self.stopwords)
self.trainer.newPool('ham')
self.trainer.newPool('spam')
def is_spam(self, entry, use_classified=True):
if use_classified:
classy = meta.Session\
.query(model.Classification)\
.filter_by(user_id = self.user.id, entry_id=entry.id).first()
if classy:
if classy.pool == 'spam':
return True
elif classy.pool == 'ham':
return False
else:
raise "bad pool"
g = self.guess(entry)
if g['spam'] and not g['ham']:
return True
if not g['spam'] and g['ham']:
return False
return (g['spam'] > g['ham'])
def guess(self, entry):
from rssmonster.controllers.bayes import __relevant__
log.debug("__relevant__(entry) %s" % __relevant__(entry))
log.debug("__relevant__(entry) %s" % self.trainer.guess(__relevant__(entry)))
log.debug('self.filename: %s' % self.filename)
# ret = dict(self.trainer.guess(__relevant__(entry)))
ret = dict(self.trainer.guess(__relevant__(entry)))
log.debug("ret: %s" % ret)
if not 'spam' in ret:
ret['spam'] = None
if not 'ham' in ret:
ret['ham'] = None
return ret
