The effect of the Stream line Inter-relevant Successive Trees (SIRST) is great in the field of text search. In this paper, the binary SIRST will be extended to a ternary one. Whats more, we propose the method of calculating the K of K-ary SIRST. Finally, a sliding window retrieval algorithm is proposed. Experimental results show that the K-ary SIRST is more efficient than the binary one in the mass of information retrieval, and sliding window search algorithm is more efficient than the normal one.