Table 2 Best results for top n. For window sizes 1+1 and 8+8
From: Expansion of medical vocabularies using distributional semantics on Japanese patient blogs
% Window size 1+1 | Best strategy | Average | 2.5 % percentile | 97.5 % percentile | Variance |
Medical Finding | summed similarity | 16.3 % | 13.9 % | 19.2 % | 0.0002 |
Pharmaceutical Drug | summed similarity | 24.8 % | 20.1 % | 29.4 % | 0.0006 |
Body Part | cluster level 83 | 16.2 % | 7.5 % | 21.9 % | 0.0011 |
Window size 8+8 | Best strategy | Average | 2.5 % percentile | 97.5 % percentile | Variance |
Medical Finding | cluster level 12 | 12.1 % | 9.1 % | 14.9 % | 0.0002 |
Pharmaceutical Drug | cluster level 1 | 11.1 % | 7.3 % | 14.7 % | 0.0004 |
Body Part | luster level 2 | 23.1 % | 18.4 % | 28.1 % | 0.0006 |