This technology aims to commercialise academic research in semantic picture index and search undertaken by two departments – Computing & Mathematics and Art & Design - at the University of Ulster. The work is quite long standing so the two departments have built up relevant academic expertise over the years. This technology represents the combination of those two discplines.
· ‘Recommender’ algorithms are already used commercially. Trying to get computers to make decisions using semantics is very much at the leading edge. So the development in this project of a semantic recommender system is consistent with commercial and academic directions in computing. Images and videos are generally more difficult to process analytically than text. So working with photos is a worthwhile and challenging area of commercial applications.
· The specific niche application is a ‘one off’ for which it is not yet possible to identify a market or market size. That said, computers and even robots are starting to be used or envisaged in the context of the care of vulnerable people – normally children and the elderly – where they are seen as a route to cost saving. The value of reminiscing is established by academic research.
The niche application would be seen to be highly valuable in terms of numbers of people in care homes.
The drivers for semantic systems, image-based decision making and ‘recommender’ systems are all strong, based on the explosion in content generally, the so called semantic web, and e-commerce. Recommenders are used in e-commerce (e.g. by Amazon in recommending books or music) and in helping people navigate cultural products such as films where preferences are highly nuanced, e.g. TV programmes.
The main driver for recommendation engines is e-commerce – used for books, music, DVD, events, travel, wine, perfume, cosmetics. The Netflix (an online DVD rental company) Prize was an open competition for the best collaborative filtering algorithm to predict user ratings for films, based on previous ratings without any other information about the users or films. The $1m prize was indicative of the potential importance of recommendation engines in relevant e-commerce sectors.
That said, this particular product does not target the mainstream ‘recommender’ space - instead it targets a niche application which is the increase in the numbers of elderly people, new ideas about how to enrich the later years of life etc..
This project has two related but quite different objectives.
· The first is a ‘recommender’ engine which uses semantic processing to link pictures through an analysis of their tags. The recommender is called SmartiPix. This general application is an approach to dealing with the problem of huge photo collections. For example, taking 2000 photos a year for 20 years yields 40,000 photos - too many to deal with. It is asserted that the recommender engine would provide a way of helping organise these photos
· The second is a niche application of this recommender engine to be used with elderly and vulnerable people.
Where an elderly person is in a care home, relatives would contribute some dozens of photos from the elderly person’s collection (or the family’s collection). These would be used as a source of material to go through with a carer, adding tags as appropriate. The tags would then allow relevant photos taken by other people to be accessed. For example, if there is a photo of inside a particular church, maybe someone else has a photo of the outside.
University of Ulster