SALOON: Predictions, Additional Analysis and Results

We experimented with the following configurations for the Hyper Parameter Tuning phase of SALOON :

Learning Rate	Parameters	Pre-Trained	No Pre-Trained
C-LR	LR	0.001	0.001
ST-LR	LRstarting LRmax Ratio Cut	0.001 0.01 32 0.1	0.001 0.01 32 0.1
ISQ-LR	LRstarting Warmup	0.01 10.000	0.01 10.000
PD-LR	LRstarting LRend Power	0.01 0.01 0.5	0.001 0.001 0.5

The results achieved by the four configuations on the evaluation set, for the classification and linking tasks, are reported in the following tables:

Classification

The following table contains the percentage of correct predictions for classification task:

Learning Rate	Pre-Trained	No Pre-Trained
C-LR	73%	53%
ST-LR	73%	54%
ISQ-LR	75%	54%
PD-LR	74%	58%

The following table contains the precision (P) and recall (R) for the linking task:

Learning Rate	Pre-Trained	No Pre-Trained
C-LR	(P, R) = (0.81, 0.91)	(P, R) = (0.34, 0.31)
ST-LR	(P, R) = (0.84, 0.84)	(P, R) = (0.39, 0.50)
ISQ-LR	(P, R) = (0.85, 0.89)	(P, R) = (0.46, 0.53)
PD-LR	(P, R) = (0.82, 0.90)	(P, R) = (0.53, 0.67)

We released all our trained models. You can find:

1. the trained SALOON model here
2. the pretrained model used as backbone for SALOON here

The predictions generated by SALOON and the three baselines can be found here.
We also release the results, in terms of precision and recall, that are available here.