Table 4.

Evaluation results of AutoSentenceParser and the Snowball model.

Parser	${\mathit{\tau }}_{\mathit{s}\mathit{i}\mathit{m}}$	Total extracted	Removed	TP	Precision	Recall	F-score
AutoSentenceParser	N/A	212	53	115	72.3%	55.0%	62.5%
Snowball	95%	12	0	12	100.0%	5.7%	10.9%
Snowball	90%	25	0	25	100.0%	12.0%	21.4%
Snowball	85%	39	0	39	100.0%	18.7%	31.5%
Snowball	80%	67	5	59	95.2%	28.2%	43.5%
Snowball	75%	101	6	89	93.7%	42.6%	58.6%
Snowball	70%	123	12	102	91.9%	48.8%	63.8%
Snowball	65%	141	20	107	88.4%	51.2%	64.8%
Snowball	60%	157	27	106	81.5%	50.7%	62.5%
Snowball	55%	169	34	107	79.3%	51.2%	62.2%
Snowball	dynamic	147	21	113	89.7%	54.1%	67.5%

The fourth column “Removed” is the number of records that is deleted during post processing.