This year’s Enterprise Search Europe started early for me – I had been invited to give the opening keynote, so I made sure I arrived early enough to make sure my laptop would play nicely with the projector, always a worry! The keynote was well recieved and I’m very grateful for the opportunity to talk about Big Data Analytics and streaming search.
Next up were Hans-Josef Jeanrond of Sinequa and Steve Woodward of AstraZeneca, a return visit after an excellent presentation by Steve’s colleague Nick Brown last year. AstraZeneca have a committed approach to search led directly by their CTO office, running hackathons, pilots and larger projects to rapidly deliver a raft of applications built on their core search platform. One key feature was delivering ‘cards’ in some cases – like Google, a calculator ‘card’ when a maths query is entered, or a calendar when someone asks about booking meetings. AstraZeneca are also building mobile apps, including a ‘people search’ that allows one to call or email with a single click. It’s great to see a large company putting significant resources into enterprise search and the benefits this can bring.
Dayle Collins of PwC and Vince McNamara of Dahu were next with a talk about PwC’s Exalead-powered enterprise search across a range of business-critical content. Dayle talked about how analysis and interviews were carried out to identify recurring search patterns in the business and identify a strategic focus and Vince then explained some of the technical features developed, including custom relevance ranking. Interestingly, entity extraction is also used at query time to classify which type of query a user has entered – are they asking about a company, product or employee for example. They mentioned how a ‘gold standard’ for search relevance is being developed – it seems this is being recorded in spreadsheets currently: perhaps they should consider a more interactive tool.
The next talk came from Ian Williams of NHS Wales Informatics Service who are building a large scale patient record service using Apache Solr. Ian explained the pressures facing the NHS (austerity, difficulty with staffing, ageing populations) and how patient records are currently distributed across a number of locations and sometimes still paper-based. This exciting project (which should be an example both the the rest of the NHS and other healthcare providers) uses Solr to create a single Welsh Clinical Portal, where healthcare providers can find information on 3 million patients in 135 hospitals and 400 GP practices across Wales. We’ve been lucky enough to work with Ian’s team on this project in a small way and it was very exciting to find out more details and hear about their future plans.
After lunch, Lesley Holmes of Nottinghamshire County Council told us about how they have attempted to improve search by focussing on metadata quality – using tools from ConceptSearching to automatically apply tags. Their content is spread across many servers and often duplicated but improving search can be have huge value to their users, who often provide services for vulnerable people where accurate and up-to-date information is essential. Cedric Ulmer of FranceLabs was next, describing with Alban Ferignac of the IFCE a project to replace Exalead with Apache Solr. Interestingly this talk contained some concrete numbers – Exalead was costing €75,000 plus a €15,000 support fee for a maximum of 6 million documents (their target is 50m), with updates costing even more, and IFCE were finding it difficult to obtain reactive support. The open source Solr (supported by a worldwide community and under constant development) gave them far more flexibility, no effective limit on the number of documents indexed and the migration process cost only €15,000 – as clear an indication of the benefits of open source search as I have seen.
Next I ran a roundtable discussion on implementing open source search which was well-attended and interactive – we discussed search engine pipelines for indexing thousands of sources amongst other subjects, and the discussions continued well after we had to vacate the room! I had to rush off soon afterwards to run the evening Meetup at a local pub, where I demonstrated the Quepid search relevance tool we’ve been using for client projects recently.
Disappointed to have missed ESE, but have just read the slides for your keynote. Some really interested themes raised. I’m especially interested in the move from hadoop-based batch processing towards stream processing.
Great post, Charlie.
Thanks Chris! I did a similar presentation earlier this year which was videoed http://www.flax.co.uk/blog/2015/06/09/iko-singapore-2015-keynote-charlie-hull-the-future-of-search-fishing-the-streams-of-big-data/