You are here

Using KO to interrogate stories in the news

In 2016 over 25% of Google queries were not mere information searches to retrieve matching web-pages, they were queries that asked questions with the expectation of retrieving facts including facts that might derive from multiple information sources.

Traditionally the knowledge management and information management industry has focussed on helping enterprises retrieve documents by searching, browsing, and visualizing the data stored within their content management systems. This is a very necessary endeavour, but one that is somewhat introspective. It is limited to the exclusive analysis of an enterprise’s internal content. This talk will explore innovative ways to cross-interrogate internal and external content, thereby creating new possibilities for synthetic information discovery based on natural question and answer style queries.

The talk will include practical demonstrations using a test-bench system comprising one million news articles indexed to the Associated Press taxonomies for people, places, organisations and topics, as well as crosswalk mappings to DBpedia’s Linked Open Data SPARQL endpoint.

The talk will discuss how natural language understanding (NLU) processes can help to parameterise natural language queries so that the query variables can be federated for targeted distribution to multiple content systems and knowledge-graph resources. It will demonstrate the valuable contribution provided by knowledge organization systems, including taxonomies, ontologies and Linked Data, to interrogate stories in the news.

Presentation Type: 
Presentation Audio: 
Audio Size: 
Presentation Visual: