Looking back at the ten papers and two panel discussions at the BCS IRSG Search Solutions 2021 conference last week there was a very strong theme that emerged with most of the papers providing a set of variations. The theme at its most basic was that ‘one size does not fit all’, especially when it comes to the user interface. The underlying issues were presented by Katriina Byström (University of Oslo) based on work by Katriina and her colleagues on vertical and horizontal tasks that was published last year. In 2017 Katriina, together with Marianne Lykke, Ann Bygholm and Louise Søndergaard, undertook a very detailed study that categorized the search queries in a medium-sized firm by theme and by functional department. This study will be published in due course in the Journal of Documentation with the title “The role of historical and contextual knowledge in enterprise search” . (An open access version will be available on the Oslo University server once the paper has been published.) One of the important outcomes was that searching for people was by far the dominant use of the SharePoint search application. This should not be taken as a use case for ‘expertise search’ but highlights the need to pay a significant amount of attention to resolving personal names.
As the day progressed the importance of understanding user requirements emerged as the dominant theme. Steve Sale at AstraZeneca talked about the need to optimize the UI for specific user groups and query types, and (using Sinequa) enabling users to define their optimum UI. Another facet of this theme pervaded the presentation by Tim Gollins on providing access to national and local archives at National Records of Scotland. The challenge here is that the language used in documents and other artifacts has changed over time. Knowing when there was a change in language is important in maximizing recall. Similar problems arise in organisations as departments change their names and products have different branding, the latter point being raised by Theresa Regli in her presentation on digital asset management and search
UI design sits at the junction of cognitive psychology and information retrieval. The links between these two disciplines were very well presented by Olivia Foulds, covering not just the arrangement of functional elements but also the range of perceptive capabilities of search users. Olivia’s paper was voted Best Paper by delegates. At the heart of the issue is ‘clutter’ which can significantly decrease the notional usability of the interface, a situation that Theresa Regli had also come across. I ended up having to present a paper at the last moment because a speaker was not available and focused on my view that a specification for a search application needs to start with not just a nominal identification of user needs but also of how these need to be accommodated in the user interface.
Over the last few years there has been a significantly increased attention paid to the information requirements of professional users, such as lawyers, journalists and clinicians. ‘Good enough’ search is certainly not going to be good enough. Andy MacFarlane presented the outcomes of the DMINR project (sponsored by Google News Initiative) which aims to develop applications for news research and applications. At FullFact the staff monitor the outcomes of statements made journalists, politicians and others to ensure that they do not inadvertently convey false and misleading information. David Corney talked about the extensive amount of systems research being undertaken to support staff in (for example) recognizing repeat occurrences of the same claim but presented in a different context or language structure.
Another important group of professional users are information scientists and clinicians working on systematic reviews which aim to valid claims made for medical treatment, especially using drugs. Rene Spijker gave a comprehensive account of how the impact of Covid 19 has resulted in the need to undertake these reviews more quickly, and yet there are challenges from information being behind a firewall and of being able to deconstruct and index scientific papers in PDF format.
Technology did get its place in the panel discussions, where there was general agreement that dense vectors represented a major advance in retrieval performance, but as a result of this and other technical advances universities were having to radically restructure undergraduate and masters courses.
The conference concluded with the presentations of the BCS Search Industry Awards
Martin White