Richard Boulton home

Weeknotes 5

20 April 2014

Last week was a bit different, because I spent the majority of it at a conference on Information Retrieval. This was ECIR 2014 which is a well established academic conference. It was great to catch up with the research that's been going on, and also with the researchers themselves: I've been to several ECIR conferences over the years, and first met some of the attendees when the conference was in Cambridge 14 years ago.

As ever, I've been left with a long reading list after the conference. I particularly like finding out about new techniques for improving systems which don't require lots of manual tagging work, and in that vein Colin Wilkie and Leif Azzopardi's work on a way to estimate bias in a retrieval system without performing vast amounts of queries (and without doing any manual work) was very interesting. Not only that, they'd been using as a test collection.

I'm hoping that we'll be able to distribute some of our search data, so that experiments like these can be done without needing to hit our API. As a bit of preparation work for that, I've been experimenting with building a vagrant configuration which will construct a complete working system running rummager. Apart from difficulties with vagrant itself, this seems to be working nicely.

I had one day back in the office before the long weekend, and spent it catching up with the many things that happened while I was away. The team roles have been redefined a bit, and the team members shuffled a bit, so starting from Tuesday our team will be officially responsible for "Search and Browse". I'm looking forward to this increased focus; we should be able to start delivering a lot pretty soon. We've also got a couple of great developers moving over to team, who I'm looking forward to working with more closely.