Semantic Analysis Technology

I attended “Semantic Analysis Technology: in Search of Categories, Concepts & Context“, the fourth ISKO UK KOnnecting KOmmunities event on 3 November 2008 at University College London.

First up were presentations from two vendors, Luca Scagliarini and Jeremy Bentley.

Scagliarini argued that information discovery suffers from information overload and information underload due to a lack of meaning-based text processing. Free text search and shallow automatic linguistic analysis did not do the job, but a ‘deep semantic analysis’ based on the analysis of relationships and ‘understanding’ the meaning that is encoded in the relationships between verbs, prepositions and nouns demonstrates potential.

Bentley reviewed key information organisation issues – unstructured information, the doubling of number of resources every 19 months, ‘findability’ problems and and the how black box solutions may not do the job. He discussed the relevance of metadata and taxonomies built specifically to reflect the way an organisation workss.

Later, practitioners presented – Rob Lee and Helen Lippell, Karen Loasby and Silver Oliver.

Lee talked about Muddy Boots, a BBC project to support the BBC’s remit to link to more external sources. Lee illustrated how structured datasets in the public domain could be used to contextualise and index BBC resources and exploit the semantic richness to link to find meaningful external links.

Lippell, Loasby and Oliver discussed three different implementations of auto-categorisation systems, demonstrating advantages and issues with each approach. The approaches were:

• using Verity Intelligent Classifier (VIC) and a taxonomy with a set of rules that could be finely tuned
• applying a rule-based automatic classification system combined with the author’s review and corrections to produce BBC content that could be described in detail. The approach
• a “statistical-based auto-categorisation” project designed to connect and cross-reference distributed BBC content and resources horizontally

McGovern makes arguments I support

In recent blogs, Gerry McGovern makes arguments close to my way of thinking.

• Choosing the right classification words
  argues for the need to ensure content creators use the words that your audience use rather than corporate/PR/government language. Of course, in many cases the “official” language has to be used, but it’s important to optimise content for the non-official terms that your audience actually uses.
• Obsessed by technology reminds us that technology isn’t the total solution to content management. However good the tools, there needs to be editorial/business input.

Semantic Analysis Technology

Former and current colleagues speaking at Semantic Analysis Technology: in search of categories, concepts & context, an ISKO UK event at UCL on 3 November 08.

Tales from the trenches of auto-categorisation: three case studies in the implementation of auto-categorisation systems

This session will look at three different implementations of auto-categorisation systems in large media organisations. We will look at common themes that have recurred time and time again. The focus will be on ‘do you really need it?’, managing expectations and tips based on the combined experience of the last 5 years. Helen Lippell, Karen Loasby and Silver Oliver are all Information Architects who have been involved in the management of controlled vocabularies and the systems that are used to automatically apply them..This session will look at three different implementations of auto-categorisation systems in large media organisations. We will look at common themes that have recurred time and time again. The focus will be on ‘do you really need it?’, managing expectations and tips based on the combined experience of the last 5 years. Helen Lippell, Karen Loasby and Silver Oliver are all Information Architects who have been involved in the management of controlled vocabularies and the systems that are used to automatically apply them..