The concept, design, and implementation of a Hardware Assisted Pruned Index (HAPI) component is proposed, discussed, and evaluated. Previously, hardware support for information retrieval was limited to only special purpose filtering components for full text search applications. HAPI, in contrast, assists in the inverted index processing as thus can be applied to larger scale collections. Furthermore, reconfigurable computing is a relatively new trend in hardware design and was not previously explored in any form to support information retrieval applications. Using feature-oriented and precision-oriented design techniques, the HAPI component developed accommodates terms of different lengths and a variable number of posting entries per term. Via emulation, experimental results of the HAPI component verify that indeed HAPI achieves high reusability and performance for information retrieval applications. |