Computational Cognitive Science

Table 4 Number of Battig categories found: results for strict (S) and lax (L) condition for different model sizes for ICA, LDA 1 (ppmi-ceil), LDA 2 (ppmi-ceil long), and LDA 3 (no weighting)

From: Exploratory analysis of semantic categories: comparing data-driven and human similarity judgments

		Model size
	Model type		10	20	30	40	50	60	80	100
	ICA		1.0	8.3	10.7	16.2	17.0	16.6	17.1	14.9
	LDA 1		0	1.2	3.9	10.3	14.8	16.3	18.4	16.2
S	LDA 2		0	1.2	3.6	10.8	14.8	17.8	19.9	16.8
	LDA 3		0	1.4	3.7	9.7	14.0	18.7	18.8	16.5
	ICA-uniq		1.0	8.3	10.7	16.2	17.0	16.6	17.1	14.9
	LDA 1-uniq		0	1.2	3.9	10.3	14.8	16.2	18.3	16.0
	LDA 2-uniq		0	1.2	3.6	10.8	14.8	17.7	19.6	16.5
	LDA 3-uniq		0	1.4	3.7	9.7	14.0	18.7	18.7	16.3
	ICA		7.7	14.3	23.0	34.0	39.2	36.8	37.9	40.0
	LDA 1		4.8	12.0	22.7	28.9	36.2	38.1	41.3	43.1
L	LDA 2		4.1	11.9	21.5	29.6	36.4	39.7	42.9	43.0
	LDA 3		4.1	12.3	21.8	30.9	36	38.9	41.8	43.1
	ICA-uniq		7.7	14.3	23.0	33.0	37.3	36.4	37.4	38.4
	LDA 1-uniq		4.8	12.0	22.7	28.8	35.0	36.3	37.0	36.7
	LDA 2-uniq		4.1	11.9	21.5	29.5	35.8	38.2	39.0	35.9
	LDA 3-uniq		4.1	12.3	21.8	30.9	35.2	37.6	37.8	36.1

The highest value for each model size and condition is marked in boldface. The number of different category types is 53.

Back to article page