We sat down with Gavin Matthews, VP of Research and Max Petrenko, Senior Ontologist, to discuss their recent research paper presented at the International Conference on Artificial Intelligence to discover how NTENT’ semantic technology is leading the way in making the web a more relevant place for consumers, advertisers and online publishers.
Q: What were you presenting?
A: Our seven-page research paper explained aspects of our semantic vertical search technology, and how it is also used to power our conceptual advert matching. We went into some details of the structure of our ontology, and language-specific lexicons, and how they allow us to detect the concepts used in an article, advert or user query. Although this is always a moving target, we gave a snapshot of the sort of tools we use to development and maintain the ontology, and ensure quality in our knowledge resources. The NTENT knowledge resources is the product of 25 person-years of effort over the last decade; the size, depth and breadth of semantic coverage gives NTENT a tremendous competitive edge.
The paper was published in the proceedings of the International Conference on Artificial Intelligence, and presented in Las Vegas as part of the Intelligent Linguistic Technology track, the main workshop on Natural Language Processing at that conference.
Q: What was the significance?
A: We feel that delivering a research paper on what makes our semantic technology truly semantic and our ontology truly ontological has had significant benefits. The terms “semantics” and “ontology” have a precise technical meaning, but they are becoming used more and more loosely at analytic and business venues and are often misinterpreted. It is common to encounter systems described as “semantic” that have no direct representation of meaning, or “ontologies” that turn out either to be mere taxonomies, or to be very small and domain-specific.
One of our priorities was to demonstrate that NTENT’s knowledge resource are much more than such a toy system, and that the language-independent conceptual structure of the ontology, married with the extensively-annotated language-dependent lexicon can be used to achieve real-world tasks of understanding the concepts used in unstructured text. We also wanted to demonstrate the power of our automatic quality control and visualization toolkit for knowledge resources.
Q: What did you get out of the presentation?
A: We got some positive comments and requests for more information. Mostly, we got the impression that no-one else in the industry was doing anything like what we do. We were able to get NTENT’s name out there, and associate it with solid technology that’s solving real-world problems. We also had the opportunity to attend many presentations and tutorials from a wide range of universities and companies. These gave us some insight into how people outside NTENT are approaching the problems we face. We were also able to make some good contacts, and find a couple of interesting leads.
No comments:
Post a Comment