def train(self, texts, solutions):
try:
Y = encode_one_hot(get_solution_vec(solutions, self.tags))
X = pipelines.get_embedding_indices(texts,
maxlen=self.sequence_length)
self.clf.train(X, Y)
except RuntimeError:
print("Training model failed")
raise
def train(self, texts, solutions):
X = pipelines.get_wv_vec(texts, self.sequence_length)
solution_vec = get_solution_vec(solutions, tags=self.tags)
stance_indices = np.where(solution_vec > 0) # NEUTRAL = 0
X_stance = X[stance_indices]
stance_solution = solution_vec[stance_indices]
stance_solution -= 1
Y_stance = encode_one_hot(stance_solution)
np.place(solution_vec, solution_vec == 2, 1)
Y_arg = encode_one_hot(solution_vec)
self.clf_argumentative.train(X, Y_arg)
self.clf_stance.train(X_stance, Y_stance)
def test(self, texts, solutions, avg_metrics=False):
X = pipelines.get_wv_vec(texts, self.sequence_length)
solution_vec = get_solution_vec(solutions, tags=self.tags)
stance_indices = np.where(solution_vec > 0) # NEUTRAL = 0
X_stance = X[stance_indices]
stance_solution = solution_vec[stance_indices]
stance_solution -= 1
Y_stance = encode_one_hot(stance_solution)
np.place(solution_vec, solution_vec == 2, 1)
Y_arg = encode_one_hot(solution_vec)
self.clf_argumentative.train(X, Y_arg)
self.clf_stance.train(X_stance, Y_stance)
X = pipelines.get_wv_vec(texts, self.sequence_length)
arg_prediction_vec = self.clf_argumentative.predict(X)
stance_prediction_vec = self.clf_stance.predict(X)
return None
def test(self, texts, solutions, avg_metrics=False):
Y = encode_one_hot(get_solution_vec(solutions, self.tags))
X = pipelines.get_wv_vec_sequence(texts)
return super()._evaluate_metrics(X, Y, avg_metrics)
def train(self, texts, solutions):
Y = encode_one_hot(get_solution_vec(solutions, tags=self.tags))
X = pipelines.get_wv_vec_sequence(texts)
self.clf.train(X, Y)
def test(self, texts, solutions, avg_metrics=False):
Y = encode_one_hot(get_solution_vec(solutions, self.tags))
X = pipelines.get_embedding_indices(texts, maxlen=self.sequence_length)
return super()._evaluate_metrics(X, Y, avg_metrics)
def train(self, texts, solutions):
Y = encode_one_hot(get_solution_vec(solutions, tags=self.tags))
X = pipelines.get_embedding_indices(texts, maxlen=self.sequence_length)
self.clf.train(X, Y)
def test(self, texts, solutions, avg_metrics=False):
Y = get_solution_vec(solutions, self.tags)
X = pipelines.get_feature_vec(texts)
return super()._evaluate_metrics(X, Y, avg_metrics)
def train(self, texts, solutions):
Y = get_solution_vec(solutions, self.tags)
X = pipelines.get_feature_vec(texts)
self.clf.train(X, Y)
def train(self, texts, solutions):
Y = encode_one_hot(get_solution_vec(solutions, self.tags))
X = pipelines.get_feature_vec(texts)
self.clf.train(X, Y)