Synonym extraction and abbreviation expansion with ensembles of semantic spaces

Table 10 Results on clinical + medical evaluation set

Evaluation configuration	Abbr →Exp		Exp →Abbr		Syn
	Clinical	Medical	Clinical	Medical	Clinical	Medical
	RI_4	RI_4	RI_4	RI_8	RI_20	RI_20
	RP_4_sw	RP_4_sw	RP_4_sw	RP_8_sw	RP_4_sw	RP_2_sw
	SUM, False		SUM, False		SUM, False
	P	R	P	R	P	R
Clinical space	0.03	0.17	0.03	0.19	0.05	0.29
Medical space	0.01	0.06	0.01	0.08	0.03	0.18
Conjoint corpus space	0.03	0.19	0.01	0.08	0.05	0.30
Clinical ensemble	0.04	0.24	0.03	0.19	0.06	0.34
Medical ensemble	0.02	0.11	0.01	0.11	0.05	0.33
Conjoint corpus ensemble	0.03	0.19	0.02	0.14	0.07	0.40
Disjoint corpora ensemble	0.05	0.30	0.03	0.19	0.08	0.47
+Post-processing (top 10)	0.07	0.39	0.06	0.33	0.08	0.47
+Dynamic cut-off (top ≤ 10)	0.28	0.39	0.31	0.33	0.08	0.45

Results (P = weighted precision, R = recall, top ten) of the best semantic spaces and ensembles on the three tasks. The results are based on the clinical + medical evaluation set and are grouped according to the number of semantic spaces employed: one, two or four. The disjoint corpus ensemble is performed with and without post-processing. A dynamic cut-off allows less than ten terms to be suggested in an attempt to improve precision. Results for tests of statistical significance are shown in Table 11.

ISSN: 2041-1480