The Zettair Search Engine

Overview

Zettair allows you to index and search HTML (or TREC) collections. It has been designed for simplicity as well as speed and flexibility, and its primary feature is the ability to handle large amounts of text. It has a single executable, which performs both indexing and searching: when an index doesn't exist, Zettair will create one for you based on the parameters you provide, and when you do have an index, Zettair will use that index to search the indexed data. It has a simple command-line interface, and supports ranked, simple (non-nested) Boolean, and phrase queries.

Features of Zettair include:

• Speed and scalability
• Boolean, ranked and phrase querying
• Modular C API for inclusion in other projects
• Native support for TREC experiments

As of 0.6.1, the Zettair search engine has been tested on the following platforms:

• Linux
• FreeBSD
• Mac OS X (Darwin)
• Solaris
• Win32 (Windows 95, 98, NT, etc)
• Cygwin

Zettair was written in C and is licensed under a BSD-style license.

Latest News

29^th September, 2006

The Windows executables for Zettair 0.9.3 have been updated so that they identify themselves with the correct version number.

8^th September, 2006

Zettair 0.9.3 is now available, This is a bug-fix version, for the following issues:

• segfault if --stem=none is specified for indexing
• malformed output or assertion failure from zet_trec if docno-less documents are retrieved
• fix for the expanding docmap issue, where the docmap file would grow during each successive invocation of zet or zet_trec, until the index became corrupted
• summarisation bug caused lower-quality snippets to be generated
• improved handling of 'Topic:' and friends in TREC topic files, as accepted by zet_trec

10^th July, 2006

Zettair 0.9.2 is now available, which fixes a number of minor bugs found since the 0.9.1 release.

29^th June, 2006

Zettair 0.9.1 is now available, which simply fixes a number of minor bugs found since the 0.9 release. In addition to source downloads, we are now offering pre-compiled windows binaries for download.

6^th June, 2006

Zettair 0.9 is now available. There has been a huge amount of change from 0.6.1, including:

• stemming
• transparent indexing of gzipped files (though summarisation will be slow for these files)
• an overhauled metric system, including a language for composing new metrics
• the addition of the Dirichlet-smoothed language modelling metric, now as default
• optional impact-ordered operation
• greatly decreased memory consumption
• increased querying and indexing speed
• integrated effectiveness evaluation

Downloading

Documentation

Getting Started

It's also easy to get Zettair going for your TREC experiments. Here's a simple getting started with TREC guide.

FAQ

Q: How much data is Zettair capable of handling?
A: We're not really sure. We've indexed over 100GB of data without problems. As of 0.6.1, we've indexed the 426GB TREC terabyte track collection.

Q: Are indexes portable between machines?
A: Indexes are currently portable between different machines, but aren't portable between different architectures. Unfortunately, we currently don't store floating point document weights portably. You may also experience problems if your source documents are stored in different locations on different machines.

Q: I want Zettair to ... and it currently doesn't. Help?
A: You can email us at zettair@cs.rmit.edu.au. We'll try our best to help, or provide some advice.

Q: I found a bug in Zettair, where do i report it?
A: Please mail us at zettair@cs.rmit.edu.au we'll deal with it.

Credits