Table 5:

Performance of the best model (CRF-based model trained with Similarity Sampling) at the span level. Characteristics with * next to their name are fine-grained items, while others have subtypes. For characteristics with subtypes, the results are aggregated for brevity. For example, Trial Design:Type results include predictions for Parallel Group, Factorial, etc. Similarly, Sample Size:Type aggregates the results for different sample size calculations: Required, Targeted, Actual at Enrollment, and Actual at Outcome Analysis.

Domain	Characteristics	Strict			Partial
		P	R	F1	P	R	F1
Trial Design	Type	0.43	0.46	0.44	0.41	0.87	0.55
	Phase	1.00	1.00	1.00	1.00	1.00	1.00
	Comparative Intent	0.33	0.60	0.63	0.33	0.60	0.63
	*Crossover Period/Treatment	0.67	1.00	0.80	0.67	1.00	0.80
	*Factorial Factor/Treatment	0.43	0.60	0.50	0.43	0.60	0.50
Blinding	Type	0.72	0.75	0.73	0.76	0.80	0.78
	Objects	0.32	0.50	0.39	0.35	0.54	0.43
Randomization	Type	0.64	0.38	0.48	0.79	0.49	0.62
	*Block Size	0.20	0.33	0.25	0.40	0.67	0.50
	*Minimization Criteria	0.67	0.50	0.57	0.83	0.63	0.71
	*Stratification Criteria	0.71	0.83	0.77	0.75	0.88	0.81
	*Personnel	0.00	0.00	0.00	0.13	0.40	0.20
	*Ratio	0.93	0.88	0.90	1.00	0.94	0.97
	*Sequence Generation	0.32	0.43	0.36	0.68	0.89	0.77
Sample Size	Type	0.42	0.65	0.51	0.52	0.81	0.63
	*Alpha	0.42	0.50	0.45	0.79	0.95	0.86
	*Dropout Rate	0.27	0.43	0.33	0.45	0.71	0.56
	*Power	0.91	1.00	0.95	0.91	1.00	0.95
Settings	Type	0.61	0.69	0.65	0.69	0.78	0.74
	*Location	0.37	0.69	0.48	0.45	0.85	0.59
Allocation Concealment	*Allocation Concealment Methods	0.00	0.00	0.00	0.00	0.00	0.00
OVERALL		0.48	0.64	0.55	0.57	0.78	0.66