I saw an interesting presentation today on web personalisation and profiling. Delivered by Dr Nikolaos Nanas from LiSys in Greece. He talked about the way that the Internet was moving away from being a digital library of content and moving towards more user-generated (and user ‘broadcast’) content. He described the increasing number of tweets, status updates and sharing as ‘the real-time web’ leading to a broadcast web that might challenge traditional mass media.
With the development of ‘the real-time web’ he sees a key feature being the need to personalise the information stream through a profile. He suggested that profiles need to be:
Dr Nanas described an information filtering system that has been inspired by how biological systems, in particular the immune system, function. This autopoietic view describes how organisms distinguish between self and non-self as a way of determining essentially what belongs and what doesn’t. He went on to draw an analogy with Adaptive Information Filtering. The work led to the development of the Nootropia system.
One of the demonstration systems from the work is available at http://noo.lisys.gr/demoNoo3/noo.jsp This allows you to pull data from RSS feeds while the profile ranks the results according to the profile you build up by clicking on news items. You can create an account on the system and try it for yourself.
Demo systems have also been developed for search, using widgets (http://observatories.cereteth.gr) and as a news reader for iphones.
The profiling system uses an analysis of the text in selected articles to build your profile of your interests. It appears to use some statistical analysis of the frequency of words, the distance between words or phrases being repeated and a hierachical weighting system to calculate the relevance of articles. That relevance algorithm is used to rank articles. Dr Nanas did suggest that there was some bias in-built within the system towards the sources that you choose, but as this is learnt by the profile as part of the process then that probably reflects the user preferences for particular sources.
A fascinating session that illustrated how technologists are building search systems that can learn from user behaviour to improve the relevance of search results. Comparing these sorts of developments with the simple search and catalogue systems that libraries currently reply upon it is evident quite how much scope there is to improve library search. In many cases library search hasn’t really scratched the surface of personalisation. Even the new harvested data discovery services such as Summon http://www.serialssolutions.com/summon don’t yet have these sort of in-built profiling systems. Even where organisations do have data about their users (such as which course they are on) this personal data isn’t being used to inform their interactions with the Library search systems.