def evaluate(self, line, aspect):
line = str(line)
aspect = str(aspect)
text_left, _, text_right = [
s.lower().strip() for s in line.partition("$T$")
]
aspect = aspect.lower().strip()
text_raw_indices = self.tokenizer.text_to_sequence(text_left + " " +
aspect + " " +
text_right)
aspect_indices = self.tokenizer.text_to_sequence(aspect)
aspect_len = np.sum(aspect_indices != 0)
text_bert_indices = self.tokenizer.text_to_sequence('[CLS] ' +
text_left + " " +
aspect + " " +
text_right +
' [SEP] ' +
aspect + " [SEP]")
bert_segments_ids = np.asarray([0] *
(np.sum(text_raw_indices != 0) + 2) +
[1] * (aspect_len + 1))
bert_segments_ids = pad_and_truncate(bert_segments_ids,
self.tokenizer.max_seq_len)
inp = np.array([[text_bert_indices], [bert_segments_ids]])
inp = torch.from_numpy(inp)
inp = inp.to(self.opt.device)
t_outputs = self.model(inp)
t_probs = F.softmax(t_outputs, dim=-1).cpu().detach().numpy()
return t_probs
def evaluate(self, text, aspect):
aspect = aspect.lower().strip()
text_left, _, text_right = [
s.strip() for s in text.lower().partition(aspect)
]
text_indices = self.tokenizer.text_to_sequence(text_left + " " +
aspect + " " +
text_right)
context_indices = self.tokenizer.text_to_sequence(text_left + " " +
text_right)
left_indices = self.tokenizer.text_to_sequence(text_left)
left_with_aspect_indices = self.tokenizer.text_to_sequence(text_left +
" " +
aspect)
right_indices = self.tokenizer.text_to_sequence(text_right,
reverse=True)
right_with_aspect_indices = self.tokenizer.text_to_sequence(
aspect + " " + text_right, reverse=True)
aspect_indices = self.tokenizer.text_to_sequence(aspect)
left_len = np.sum(left_indices != 0)
aspect_len = np.sum(aspect_indices != 0)
aspect_boundary = np.asarray([left_len, left_len + aspect_len - 1],
dtype=np.int64)
text_len = np.sum(text_indices != 0)
concat_bert_indices = self.tokenizer.text_to_sequence('[CLS] ' +
text_left + " " +
aspect + " " +
text_right +
' [SEP] ' +
aspect +
" [SEP]")
concat_segments_indices = [0] * (text_len + 2) + [1] * (aspect_len + 1)
concat_segments_indices = pad_and_truncate(concat_segments_indices,
self.tokenizer.max_seq_len)
text_bert_indices = self.tokenizer.text_to_sequence("[CLS] " +
text_left + " " +
aspect + " " +
text_right +
" [SEP]")
aspect_bert_indices = self.tokenizer.text_to_sequence("[CLS] " +
aspect +
" [SEP]")
data = {
'concat_bert_indices': concat_bert_indices,
'concat_segments_indices': concat_segments_indices,
'text_bert_indices': text_bert_indices,
'aspect_bert_indices': aspect_bert_indices,
'text_indices': text_indices,
'context_indices': context_indices,
'left_indices': left_indices,
'left_with_aspect_indices': left_with_aspect_indices,
'right_indices': right_indices,
'right_with_aspect_indices': right_with_aspect_indices,
'aspect_indices': aspect_indices,
'aspect_boundary': aspect_boundary
}
t_inputs = [
torch.tensor([data[col]], device=self.opt.device)
for col in self.opt.inputs_cols
]
t_outputs = self.model(t_inputs)
t_probs = F.softmax(t_outputs, dim=-1).cpu().numpy()
return t_probs
key_entity = []
for e in entities:
content = title
index = content.find(e)
if index + (tokenizer.max_seq_len // 2) < len(content):
content = content[:index + (tokenizer.max_seq_len // 2)]
if len(content) > tokenizer.max_seq_len:
content = content[-(tokenizer.max_seq_len - len(e) - 3):]
text_raw_indices = tokenizer.text_to_sequence(content)
aspect_indices = tokenizer.text_to_sequence(e)
aspect_len = np.sum(aspect_indices != 0)
text_bert_indices = tokenizer.text_to_sequence('[CLS] ' + content + ' [SEP] ' + e + ' [SEP]')
bert_segments_ids = np.asarray([0] * (np.sum(text_raw_indices != 0) + 2) + [1] * (aspect_len + 1))
bert_segments_ids = pad_and_truncate(bert_segments_ids, tokenizer.max_seq_len)
text_bert_indices = torch.tensor([text_bert_indices], dtype=torch.int64).to(opt.device)
bert_segments_ids = torch.tensor([bert_segments_ids], dtype=torch.int64).to(opt.device)
inputs = [text_bert_indices, bert_segments_ids]
outputs = model(inputs)
t_probs = F.softmax(outputs, dim=-1).cpu().numpy()
sentiment = t_probs.argmax(axis=-1)
if sentiment == 1:
key_entity.append(e)
# remove infered key entities that can be substring of other entities.
final_res = []
for e1 in key_entity:
flag = 0