Automatic enrichment

Enabling personalised recommendation services requires knowledge about the content, and knowledge about the users of this content. To automate the recommendation services this knowledge has to be made understandable for computers. Metadata enrichment is a core process in doing this.

Semantic annotation is the process of identifying knowledge elements in text, and mapping them to instances and entities in a given knowledge base.

Why was this of interest to NoTube?

The Open Linked Data cloud makes available vast amounts of readily available human knowledge that has been accumulated on the Web: for example Wikipedia contains information on a large number of proper nouns, domain-specific technical terms, and many topics that are usually not covered in other encyclopaedias.

In using semantics to make connections between TV and Web content, NoTube was interested in enriching TV content metadata (from text and from audio/video sequences) with related concepts and background knowledge from Open Linked Data resources such as DBpedia. By enriching the original TV content in this way, new connections can be made between programmes, and between programmes and related entities on the Web: such as related activities, places, people, subjects and organisations.

What NoTube has done in this area

NoTube’s programme data enrichment services perform named entity recognition on EPG (electronic programme guide) data to annotate plain text with semantic links, i.e. links to concepts in existing Open Linked Data repositories such as DBpedia, SKOS and IMDB (for people’s names and roles). By adding semantic links to the EPG data, the programmes can be placed in a broader context allowing for interesting links to be made between items based on related data such as genre, location, actor or director.

NoTube’s Beancounter services are concerned with aggregating existing social activity data from various sources (such as Twitter and LastFM) and enriching it with related data available from the Linked Data cloud to generate a machine-readable weighted interests User Profile.

Automatically matching the enriched user interests profile to the enriched programme data potentially allows for more personal, accurate and interesting TV recommendations.

Find out more: See the Things to use section of this site

Who was involved? Vrije Universiteit in AmsterdamInstitut fuer Rundfunktechnik GmbH and Ontotext, in collaboration with other project partners.


Click here for further information on automatic enrichment

More research topics