Together with forty or so of my most valued and esteemed jet-lagged colleagues and friends, I attended SWIRL 2012, “the occasional talkshop on the future of information retrieval”, hosted by RMIT in Lorne, near Melbourne, earlier this month.
The broadly stated topic for this gathering was to formulate the most fruitful reasonably long range research questions for information retrieval as an academic research field. There were keynote addresses and break-out groups followed by other break-out groups followed by collective authoring sessions to compose a comprehensive consensus view of where we are and where we should be headed.
For those already convinced of the necessity of a deeper understanding of users, of the easily predictable challenges inherent in the ubiquitous presence of mobile information technology, of the slow but imminent roll-in of various aspects of ubiquitous computing, and of the value of spoken dialogue as an input modality, the final list of major issues was somewhat underwhelmingly disruptive.
But that’s what happens when a group of well synched people interact in a pleasant environment and the process is vectored towards compromise. In the end, six research questions were highlighted:
- People or finding out more about users
- Zero-word-queries or ubiquitous information access
- Mobile information retrieval
- Dialogue, including speech
- Information literacy
- Structured vs unstructured information
They will be presented as a final report from the event, redacted by the organisers. The final report will be made available to interested parties in a bit.
An observation made by one of us in session was that information retrieval tends to arrive when the infrastructure and devices are in place – in that sense information retrieval is a reactive branch of computer science. The worry I myself have is that the workshop in lifting some of these questions to top urgency pitches rocks into neighbouring fields, where no-one will be too interested in picking them up for refinement. Information retrieval as an academic field must have a readiness to address questions even before they become urgent, or we risk industrial players hacking them into a sub-standard but de-facto standard solution. Mobile informatics, social informatics, and the ubicomp field already have assumptions in place for how information retrieval should be worked.
Of the six principal issues most discussed at the workshop, the one that most interested me – because of its potentially socially disruptive nature – was that of information literacy: how to best help people to understand de-contextualised information, how to make them better searchers and critical readers, how to build technology to close the gap between the information have-nots and the informocrats, and to do this both momentarily at time of information access and in a prolonged perspective over the lifetime of a user. This issue, which involves issues far beyond technology, will stay with us as more and more of society realises the infrastructural confluence of needs and potential in which information access technology resides.
Underlying many of the discussions were some technology and methdology trends I appreciate and look forward to thinking more about, and while the top questions were less technology oriented and more application oriented, some of these technologies will be brought up in the coming report.
- Touching on the question of modelling usage and users and how to integrate knowledge of them into the system development cycle is a favourite discssion point of mine – that of use cases as a bridge between benchmarking and validation. This is, not incidentally, a main topic of the PROMISE network of excellence in which I participate, and which is in the process of publishing some reports on the matter.
- Touching on the question of ubiquitous access to information, on the push-vs-pull information provision question and in the question of structured vs unstructured information there will be the enormous and uncharted challenges brought to us by the very real approaching internet of things – when all our things will start communicating we will need to build them with communication protocols which are flexible, learning and dynamic. We will not be able to count on web services to use the same knowledge structures from version to version – yet we will expect our preferences to migrate seamlessly from workplace to home to other destinations. This calls for an introduction of search technology, but vectored towards interaction between things.
- Finally, most relevant to Gavagai and the work we do here: big data changes everything. We must build future-proof processing models and we must foresee that much of the information needs we will be providing are not about finding gold nuggets or needles in a haystack – they are about modelling the state of the world as it changes and as it is reflected in the information we process. The challenge will not be to winnow out the best sample or item but to – conversely – ingest all of it and track its meaning as a whole.
We will see where this eventually gets us, but the fresh thoughts and ideas this workshop has set in motion – not all from the workshop sessions and not all to be documented in the workshop reports – are evident in the discussions currently active in my mailbox. We need more gatherings of this type – much more productive than many large conferences are! (The availability of a beach in February didn’t hurt either).
