Multitask Learning for Aspect-Based Sentiment Classification

Table 2

Statistics of the experiment datasets.

Dataset	Positive	Negative	Neutral	Conflict	Total	Opinion
14Lap-Train	987	866	460	45	2358	2504
14Lap-Test	341	128	169	16	654	674
14Res-Train	2164	805	633	91	3693	3484
14Res-Test	728	196	196	14	1134	1008
15Res-Train	902	252	34	—	1315	1210
15Res-Test	319	319	27	—	685	510