Similarity search is a fundamental operation for applications that deal with unstructured data sources. In this paper we propose a new pivot-based method for similarity search, called sparse spatial selection (SSS). This method guarantees a good pivot selection more efficiently than other methods previously proposed. In addition, SSS adapts itself to the dimensionality of the metric space we are working with, and it is not necessary to specify in advance the number of pivots to extract. Furthermore, SSS is dynamic, it supports object insertions in the database efficiently, it can work with both continuous and discrete distance functions, and it is suitable for secondary memory storage. In this work we provide experimental results that confirm the advantages of the method with several vector and metric spaces

The development of applications that manage large text collections needs indexing methods which allow efficient retrieval over text. Several indexes have been proposed which try to reach a good trade-off between the space needed to store both the text and the index, and its search efficiency. Self-indexes are becoming more and more popular. Not only they index the text, but they keep enough information to recover any portion of it without the need of keeping it explicitly. Therefore, they actually replace the text. In this paper, we focus in a self-index known as wavelet tree. Being originally organized as a binary tree, it was designed to index the characters from a text. We present three variants of this method that aim at reducing its size, while keeping a good trade-off between space and performance, as well as making it well-suited for indexing natural language texts. The first approach we describe joins Huffman compression and wavelet trees. The other two new variants index words instead of characters and use two different word-based compressors.