Squeezing Long Sequence Data for Efficient Similarity Search

• Main Authors: SONG, Guojie, Cui, Bin, ZHENG, Baihua, Xie, Kunqing, YANG, Dongqing
• Format: text
• Language: English
• Published: Institutional Knowledge at Singapore Management University 2008
• Subjects: Computer Sciences
• Online Access: https://ink.library.smu.edu.sg/sis_research/405
• Institution: Singapore Management University

Similarity search over long sequence dataset becomes increasingly popular in many emerging applications. In this paper, a novel index structure, namely Sequence Embedding Multiset tree(SEM-tree), has been proposed to speed up the searching process over long sequences. The SEM-tree is a multi-level structure where each level represents the sequence data with different compression level of multiset, and the length of multiset increases towards the leaf level which contains original sequences. The multisets, obtained using sequence embedding algorithms, have the desirable property that they do not need to keep the character order in the sequence, i.e. shorter representation, but can reserve the majority of distance information of sequences. Each level of the tree serves to prune the search space more efficiently as the multisets utilize the predicability to finish the searching process beforehand and reduce the computational cost greatly. A set of comprehensive experiments are conducted to evaluate the performance of the SEM-tree, and the experimental results show that the proposed method is much more efficient than existing representative methods.