A Low-Cost Named Entity Recognition Research Based on Active Learning

Algorithm 1

The pseudocode of AL-CRF.

	Initialization: unlabelled dataset , , , , initialization data number , additional number in iteration
	// k-means clustering
	select cluster centers randomly as
	do
	for in do
	for in do
	if is then
	the cluster of is
	end
	end
	end
	update to
	while
	output the clustered dataset
	select samples from by stratified sampling
	annotate samples into
	train CRFs by
	calculate the F-value of CRFs as
	while do
	calculate entropy in
	rank the according to the entropy
	annotate top samples into
	train CRFs by
	calculate the F-value of CRFs as
	end