Media granularity - no single answer

In looking at the approaches to disecting multimedia that is being taken by Blinkx, Veotag and others it struck me that we are not facing an either/or battle for the winning approach. The two approaches are complimentary. In many ways it parallels the debate on Taxonomy versus folksonomy that emerges in discussions on tagging information.

In enterprises I believe we are seeing the emergence of a hybrid approach. Statutory compliance with Sarbanes-Oxley and other directives pushes organizations to develop taxonomies for their corporate data. This is necessary to ensure data is correctly categorized. However, the hybrid approach comes in when this taxonomy is used as the base and user generated tags are overlaid to create the folksonomy. Folksonomy have a lot of benefits because we each tend to view information from different perspectives based on the context we operate in. This is why I see multiple solutions emerging around multimedia search. Brute force indexing and speech to text conversion has a role. However, that approach may not reveal the nuances of the content. This is where the overlay of user generated, community oriented tagging comes to the fore.

To turn up the contrast on the blinkx versus veotag approach. If a video of a mime artist is indexed via blinkx there will be very little speech to index. However, if a mime fanatic time stamps the same video they may create searchable text about the type of scene, the movements, the subject that would otherwise be missed. Neither approach is "wrong". They are complimentary.

Do you see these different approaches as complimentary or competitive? share your views. Join the discussion and leave a comment below.