353/2009 ENERGOS

Indra Software Labs S.L.U., Programa CENIT 2009, CDTI-CEN20091048

11

Jan 1, 2010 to Mar 31, 2012

This project included two main subtasks:
1) A complete study on how to use different compression techniques in order to save communication bandwith. Different scenarios and data types were considered. The development of new compression techniques, if needed, was also included. Although text compression was initially considered, it was finally discarded, as the natural language scenario was not likely to be relevant for the project. So, our research efforts were concentrated on real numbers compression, as the remaining scenarios were well supported by previously existing techniques. A new real number compressor, called FP3, was developed and implemented.

2) Another complete study on indexing techniques, considering also different scenarios and data types. Again, when no satisfactory techniques were available, new ones were expected to be developed. We started dealing with natural language text, but our efforts were concentrated, again, on more suitable datatypes with respect to the project goals. We concluded that most of the existing techniques were adequate enough, although the indexing of natural numbers (one more time) was not forcefully solved. Some different real number indexes were developed. The first one, iWCSA, was our first (failed) attempt; two more techniques, based on, respectively, inverted indexes and wavelet trees, provided us with good results, allowing us to present both ones in the world's best data compression conference: the IEEE Data Compresion Conference (DCC)