An Improved BERT and Syntactic Dependency Representation Model for Sentiment Analysis

Algorithm 1

ARC_model ().

	{//initial configuration
	STACK stack = [root]
	BUFFER buffer_words = [,…,]
	POS pos = [.pos, …, .pos]//POS = {NN, JJ, VBZ, …}
	ARCS arc_set = {empty}//ARCS = {dobj, amod, nsubj, …}
	while (the buffer is not empty and the stack contains more than one node)
	{//The following operations are performed according to the words and their POS of the top stack and the buffer
	if (the shift will be used in the transition)
	then { run op_shift operation }
	if (the left arc will be used in the transition)
	then { run op_left_arc operation;
	add a left arc about the two words to the arc_set.}
	if (the right arc will be used in the transition)
	then { run op_right_arc operation;
	add a right arc about the two words to the arc_set.}
	}
	}//the end.