Acoustic Model with Multiple Lexicon Types for Indonesian Speech Recognition

Table 4

The dataset of Indonesian audio after first validation (approx. size 10,000 utterances).

Category	Utterances before validation		Utterances after validation
	Number of utterances	Total duration (hours)	Number of utterances	Total duration (hours)
Autos and vehicles	2,234	1.3289	29	0.0214
Comedy	6,017	4.0768	71	0.0498
Education	47,477	38.4886	5,973	4.5254
Entertainment	39,155	25.2972	470	0.2994
Film and animation	6,475	4.5591	75	0.0567
Gaming	124	0.0987	1	0.0000
Howto and style	28,726	41.8189	450	0.5933
Music	1,409	2.5783	2	0.0011
News and politics	19,295	16.8386	303	0.2502
People and blogs	32,205	22.7193	2,432	1.7630
Pets and animals	57	0.0433	1	0.0000
Science and technology	31,492	22.9787	515	0.3849
Sports	137	0.1131	1	0.0000
Travel and events	78	0.0633	1	0.0008
Uncategorized	410	0.2515	9	0.0069
Total	215,291	181.2543	10,333	7.9529