Table 7.

Effect of different embedding representation methods on model performance.

	Digital music		Baby		Office products		Beauty		Yelp
	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE	MSE	MAE
DeepCoNN-Glove14	1.202	0.722	1.440	0.873	0.909	0.707	1.453	0.922	1.687	1.361
DeepCoNN-BERT-base	1.185	0.706	1.417	0.869	0.873	0.684	1.424	0.907	1.650	1.322
DeepCoNN-RoBERTa-base	1.172	0.698	1.403	0.856	0.856	0.675	1.406	0.897	1.633	1.309
MPRS-Glove43	0.947	0.678	1.282	0.845	0.772	0.653	1.361	0.800	1.548	0.981
MPRS-BERT-base	0.925	0.676	1.262	0.833	0.760	0.648	1.358	0.804	1.537	0.971
MPRS-RoBERTa-base	0.919	0.664	1.258	0.827	0.756	0.644	1.350	0.798	1.525	0.964
AFMRUI-Glove	0.934	0.675	1.280	0.845	0.767	0.658	1.365	0.809	1.540	0.979
AFMRUI-BERT-base	0.918	0.662	1.266	0.827	0.751	0.644	1.347	0.799	1.517	0.965
AFMRUI	0.910	0.657	1.256	0.821	0.740	0.638	1.341	0.786	1.502	0.954

Significant values are in [bold].