Table 11.
P-values, recall |
Medical |
Conjoint |
Clinical |
Medical |
Conjoint |
Disjoint |
---|---|---|---|---|---|---|
(synonym task) | space | corpus | ensemble | ensemble | corp. ens. | corp. ens. |
Clinical space |
0.011 |
1.000 |
0.057 |
0.885 |
0.003 |
0.000 |
Medical space |
- |
0.004 |
0.000 |
0.000 |
0.000 |
0.000 |
Conjoint corpus |
- |
- |
0.210 |
1.000 |
0.001 |
0.000 |
Clinical ensemble |
- |
- |
- |
0.480 |
0.189 |
0.001 |
Medical ensemble |
- |
- |
- |
- |
0.047 |
0.000 |
Conjoint corp. ens. | - | - | - | - | - | 0.041 |
P-values for the differences between the recall results on the synonym task for the semantic spaces/ensembles presented in Table 10. P-values showing a statistically significant difference (p-value < 0.05) are presented in bold-face.
P-values for the post-processing and for the abbr →exp and exp →abbr are not shown in the table. However, for the significance level p-value < 0.05, there were no statistically significant recall difference between the standard Disjoint Corpora Ensemble and the post-processing version for any of the three tasks (p-value = 0.25 for abbr →exp and p-value = 0.062 for exp →abbr). When testing the recall difference between the pairs of semantic spaces/ensembles shown in Table 10 for the abbr →exp task, there was only a significant difference for the pairs Medical Space vs. Clinical Ensemble (p-value = 0.039), Medical Space vs. Disjoint Corpora Ensemble (p-value = 0.004) and Medical Ensemble vs. Disjoint Corpora Ensemble (p-value = 0.039). For the exp →abbr task, there were no statistically significant differences.