from yatk import ir
from yatk.ml.svm import SVM as Classifier
# from yatk.ml.nb import NaiveBayes as Classifier
con = sqlite3.connect('test.db')
con.row_factory = sqlite3.Row
cur = con.cursor()
docs = []
cur.execute('select class, text from docs')
for row in cur.fetchall():
docs.append((row['class'], row['text']))
index = ir.SentimentIndex('delta', 'bogram')
index.get_class = lambda x: x[0]
index.get_text = lambda x: x[1]
index.build(docs)
x = []
y = []
for doc in docs:
x.append(index.weight(index.features(doc)))
y.append(doc[0])
cl = Classifier()
cl.train(x, y)
cl.save('test.svm')
index.save('test.index')
con.close()
class Guesser:
def __init__(self):
self._cl = Classifier()
def is_candidate(self, word):
return True
def make_class(self, word):
pass
def traverse(self, sentences):
x = []
y = []
for sentence in sentences:
for w in range(0, len(sentence)):
word = sentence[w]
if not self.is_candidate(word):
continue
x.append(self.gen_features(sentence, w))
y.append(self.make_class(word))
return (x, y)
def train(self, sentences):
(train_x, train_y) = self.traverse(sentences)
self._cl.train(train_x, train_y)
def predict(self, sentences):
(test_x, test_y) = self.traverse(sentences)
return (self._cl.predict(test_x), test_y)
def test(self, sentences):
(estim_y, test_y) = self.predict(sentences)
return self._cl.evaluate(test_y, estim_y)
def guess(self, word):
return self._cl.predict([self.gen_features([(word, )], 0)])[0]
def gen_features(self, sentence, w):
word = sentence[w][0]
x = {}
x['p3:' + word[0:3]] = 1
x['p4:' + word[0:4]] = 1
x['p5:' + word[0:5]] = 1
x['p6:' + word[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x['s2:' + word[-2:]] = 1
x['s3:' + word[-3:]] = 1
x['s4:' + word[-4:]] = 1
x['s5:' + word[-5:]] = 1
x['w:' + word] = 1
for i in range(1, 4):
if w > i - 1:
word = sentence[w - i][0]
# x[str(i) + 'p3:' + prev[0:3]] = 1
# x[str(i) + 'p4:' + prev[0:4]] = 1
# x[str(i) + 'p5:' + prev[0:5]] = 1
# x[str(i) + 'p6:' + prev[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x[str(i) + 's2:' + word[-2:]] = 1
x[str(i) + 's3:' + word[-3:]] = 1
x[str(i) + 's4:' + word[-4:]] = 1
# x[str(i) + 's5:' + prev[-5:]] = 1
x[str(i) + 'w:' + word] = 1
for i in range(1, 2):
if w + i < len(sentence) - 1:
word = sentence[w + i][0]
# x[str(i) + 'p3:' + prev[0:3]] = 1
# x[str(i) + 'p4:' + prev[0:4]] = 1
# x[str(i) + 'p5:' + prev[0:5]] = 1
# x[str(i) + 'p6:' + prev[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x[str(i) + '+s2:' + word[-2:]] = 1
x[str(i) + '+s3:' + word[-3:]] = 1
x[str(i) + '+s4:' + word[-4:]] = 1
# x[str(i) + 's5:' + prev[-5:]] = 1
x[str(i) + '+w:' + word] = 1
return x
def save(self, path):
self._cl.save(path)
@staticmethod
def load(path):
obj = Guesser()
obj._cl = Classifier.load(path)
return obj
from yatk import ir
from yatk.ml.svm import SVM as Classifier
# from yatk.ml.nb import NaiveBayes as Classifier
con = sqlite3.connect('test.db')
con.row_factory = sqlite3.Row
cur = con.cursor()
docs = []
cur.execute('select class, text from docs')
for row in cur.fetchall():
docs.append((row['class'], row['text']))
index = ir.SentimentIndex('delta', 'bogram')
index.get_class = lambda x: x[0]
index.get_text = lambda x: x[1]
index.build(docs)
x = []
y = []
for doc in docs:
x.append(index.weight(index.features(doc)))
y.append(doc[0])
cl = Classifier()
cl.train(x, y)
cl.save('test.svm')
index.save('test.index')
con.close()
class Guesser:
def __init__(self):
self._cl = Classifier()
def is_candidate(self, word):
return True
def make_class(self, word):
pass
def traverse(self, sentences):
x = []
y = []
for sentence in sentences:
for w in range(0, len(sentence)):
word = sentence[w]
if not self.is_candidate(word):
continue
x.append(self.gen_features(sentence, w))
y.append(self.make_class(word))
return (x, y)
def train(self, sentences):
(train_x, train_y) = self.traverse(sentences)
self._cl.train(train_x, train_y)
def predict(self, sentences):
(test_x, test_y) = self.traverse(sentences)
return (self._cl.predict(test_x), test_y)
def test(self, sentences):
(estim_y, test_y) = self.predict(sentences)
return self._cl.evaluate(test_y, estim_y)
def guess(self, word):
return self._cl.predict([self.gen_features([(word,)], 0)])[0]
def gen_features(self, sentence, w):
word = sentence[w][0]
x = {}
x['p3:' + word[0:3]] = 1
x['p4:' + word[0:4]] = 1
x['p5:' + word[0:5]] = 1
x['p6:' + word[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x['s2:' + word[-2:]] = 1
x['s3:' + word[-3:]] = 1
x['s4:' + word[-4:]] = 1
x['s5:' + word[-5:]] = 1
x['w:' + word] = 1
for i in range(1, 4):
if w > i - 1:
word = sentence[w - i][0]
# x[str(i) + 'p3:' + prev[0:3]] = 1
# x[str(i) + 'p4:' + prev[0:4]] = 1
# x[str(i) + 'p5:' + prev[0:5]] = 1
# x[str(i) + 'p6:' + prev[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x[str(i) + 's2:' + word[-2:]] = 1
x[str(i) + 's3:' + word[-3:]] = 1
x[str(i) + 's4:' + word[-4:]] = 1
# x[str(i) + 's5:' + prev[-5:]] = 1
x[str(i) + 'w:' + word] = 1
for i in range(1, 2):
if w + i < len(sentence) - 1:
word = sentence[w + i][0]
# x[str(i) + 'p3:' + prev[0:3]] = 1
# x[str(i) + 'p4:' + prev[0:4]] = 1
# x[str(i) + 'p5:' + prev[0:5]] = 1
# x[str(i) + 'p6:' + prev[0:6]] = 1
# x['s1:' + word[-1:]] = 1
x[str(i) + '+s2:' + word[-2:]] = 1
x[str(i) + '+s3:' + word[-3:]] = 1
x[str(i) + '+s4:' + word[-4:]] = 1
# x[str(i) + 's5:' + prev[-5:]] = 1
x[str(i) + '+w:' + word] = 1
return x
def save(self, path):
self._cl.save(path)
@staticmethod
def load(path):
obj = Guesser()
obj._cl = Classifier.load(path)
return obj