TR-IT-0114 :1995.5

Ali M. AlHaj, Eiichiro Sumita, Hitoshi Iida

Massively Parallel Text Retrieval

Abstract:Text retrieval systems are computer-based systems the function of which is locate user-requested documents in text databases. The requested documents which are commonly stored in electronic form may include news articles, technical abstracts, office memos, electronic mail messages, among many others. In fact, the recent widespread of such electronic documents has greatly enhanced the importance of text retrieval systems and made them vital components in modern information systems. A wide variety of operational text retrieval systems already exit, however, these systems are mostly implemented on sequential computers which lack both the storage capacity and the retrieval speed that are needed for large on-line information systems. The research reported here is based on the argument that parallel computers represent a potential tool to cope with the increasing computational and storage demands of modern text retrieval systems. In this report we describe an efficient implementation of an experimental parallel text retrieval system using an MIMD parallel computer. The system is intended to be the first step towards achieving the ultimate goal of developing an effective and efficient retrieval system to support the on-going massively parallel example-based spoken language translation project at ATR Interpreting Telecommunications Research Laboratories.