Table 3. Mini-expert system performance evaluation results.

	Combined + weighted (mini-expert system)	Combined + unweighted	Cosine + weighted	Cosine + unweighted
MRR	0.72	0.70	0.72	0.68
% success at 1	62	59	63	57
% success at 5	86	85	85	83
% success at 10	90	91	90	89
% success at 20	93	92	92	91

Mean reciprocal rank (MRR) measures how close the correct match is to the top rank on average. It ranges from 0 to 1, and values close to 1 indicate that correct matches appear closer to the top on average. % success at N = % of cases with correct diagnoses within top N ranks. Combined = combined cosine and semantic similarity. Cosine = cosine similarity only.