The System Architecture and the Transaction Concept of the SPIDER Information Retrieval System (1996)
| Venue: | Special Issue on the Integration of Information Retrieval and Database Technology |
| Citations: | 11 - 2 self |
BibTeX
@INPROCEEDINGS{Knaus96thesystem,
author = {Daniel Knaus and Peter Schäuble},
title = {The System Architecture and the Transaction Concept of the SPIDER Information Retrieval System},
booktitle = {Special Issue on the Integration of Information Retrieval and Database Technology},
year = {1996},
pages = {pages}
}
Bookmark
 |
 |
 |
 |
 |
 |
OpenURL
Abstract
A relational database (DB) management system is extended by the SPIDER Information Retrieval (IR) system to provide IR operations on data stored in the DB system. The IR data stored in the IR system is obtained by analyzing the objects in the database. This derived data is needed for advanced IR operations such as relevance ranking and relevance feedback. It is kept consistent with the DB data by triggers and by a synchronization process. Our architecture is based on a loosely-coupled integration which is scalable to a large extent. We developed a new, more powerful transaction model for the IR system where serializability of schedules can be relaxed because the transactions have special properties. Furthermore, the ability to abort transactions is not required in the case of updates of the IR index. We present SPIDER's transaction manager which takes advantage of these properties. We also describe a query optimization strategy to achieve an optimal trade-off between retrieval effectiv...
