1. Building by Open
Collaboration
Craig Macdonald
University of Glasgow
craigm@dcs.gla.ac.uk
2. Motivations
• Why should we have IR platforms?
– They facilitate research, by bringing researchers close
to the state‐of‐the‐art
• Why should we have !quot;#$%&!'()# IR platforms?
– They facilitate greater research potential
• Keep in contact with the commercial companies
– They provide visibility
– By contributing towards open source platforms, we
can all reap the benefits
3. History of Terrier
• 2001‐2004: Terrier started as EPSRC (British
research council) project: Iadh Ounis, Gianni
Amati, Ben He & Vassilis Plachouras form
platform. I join later.
• 2005: v1.0.0 core released by the Univ. of
Glasgow as open source under the MPL
license, followed later by v.1.0.1 and v1.0.2
• 2007: v1.1.0, v1.1.1 released
• 2008: v2.0, v2.1, v2.2 released
4. Current Terrier Core
• A framework for doing IR experimentation and
building IR apps
• Supports indexing commonly used IR research
collections (e.g. TREC). Indexing options are
direct indexing, single‐pass indexing & Map
Reduce‐based indexing
• Standard retrieval facilities, with many
weighting models, including Query Expansion
• Sample desktop search application
12. Acceptable Patches
• Fits with Terrier’s style
– E.g. comments, javadoc, documentation
– Reuses existing code
• Can be cleanly applied (up to date with
current code)
• Does not break existing functionality
– All test cases pass
• Is agreeable to the committer
14. Our limitations
• Our aim is to do research
• The platform facilitates that. By having the core
platform open source we help &'$!()%do research
also
• Open source works best when &'$!()%can
contribute back as well
• We need your help to build a community – we
can’t spend all year developing code to release,
and helping users
– We can all take a share of the development and of
assisting users on the forum