The 2019 TALMIRI Conference took place at the Putterbury campus of the University of Bedfordshire. The acronym stands for Talent Meets the Information Retrieval Industry. The conference was conceived and organized by Ingo Frommholz and Haiming Liu with support from the University of Bedfordshire that enabled the conference to be free to attendees in a setting that was very conducive to thinking outside of the box.
The common theme of the opening papers was the management of the long tail of search results. Dyaa Albakour (Signal A.I.) highlighted highlight cases where IR research (notably from TREC ) has focused efforts on performing well on the head of the distribution and ignored ‘difficult’ long-tail cases. Dyaa then moved from academia to industry to give examples of problems and challenges he and his colleagues at Signal AI face in dealing with news stories where it is crucial to put effort into improving and evaluating performance of cases on the long tail.
I then spoke about the issues that the long tail in enterprise search presented, and the extent to which AI/ML applications were (currently!) not able to make a significant impact. My recent column in CMSWire is a summary of my presentation.
After coffee and animated discussion Steve Zimmerman (University of Essex) presented a paper on his work towards understanding user impacts and concerns related to privacy protective search engine results pages. You can read a paper on this topic by Steve and his colleagues in the proceedings of CHIIR 2019. To me the paper was memorable for alerting me to the concept of Nudging in interactive information retrieval.
Over the years I have fascinated many clients with the concept of information foraging theory. The best summary of information foraging that I have come across is by David Trepess. Over the last few years it seems to have risen in visibility. Amit Kumar Jaiswal (University of Bedfordshire) talked about applying information foraging theory to image collections, which immediately gained my attention as this is a very hot topic for digital asset management vendors.
After an excellent lunch it was time for the IRSG AGM, the minutes of which will be available on the IRSG web site in the near future. Charlie Hull then moved the discussions towards the IRI side of the conference with a very clear summary of the role of open source search applications, interspersed by his own journey from Muscat to Open Source Connections. One of the topics that Charlie touched on was the extent to which Elasticsearch was ‘open source’ We live and work in interesting times.
The final presentation came from Gabriella Kazai (Microsoft Research) on the subject of explainable search. Web search engines are examples of complex ML systems that given a high dimensional input space produce rankings of resources as a result of implicit comparisons and decisions among candidate pages. Traditionally, little explanation has been provided or indeed was required of why search results appear on a search engine result page (SERP) or at least the explanation simply equated the notion of relevance, which has been studied at great depth. Gabriella discussed what explainability means in search and how it relates to user’s mental models of relevance, SEO best practices and internal policy decisions on page quality.
To wrap up the event Ingo introduced a discussion session using humour as the facilitation technique. I have to admit I started out as a sceptic but was won over in the end as everyone in the room made some very useful contributions.