Leveraging Contextual Sentences for Text Classification by Using a Neural Attention Model

Table 5

Results of models when using additional features by different ways.

Models	Accuracy	Time
Attended representation with length + TF-IDF (CNN)	72.7/83.1	379/310
Attended representation with probs (CNN)	72.8/83.0	396/321
Attended representation with probs + length + TF-IDF (CNN)	73.0/83.4	412/335
Attended representation with length + TF-IDF (BLSTM)	73.7/85.6	1084/905
Attended representation with probs (BLSTM)	73.6/85.7	1107/918
Attended representation with probs + length + TF-IDF (BLSTM)	73.9/85.9	1141/924