MiSPOT: dynamic product placement for digital TV through MPEG-4 processing and semantic reasoning

Exploring synergies between digital tv recommender systems and electronic health records

“…This is indeed an evolution of the approach to receiver-side semantic reasoning we presented in [30] to deliver recommendations of TV programs and/or commercial products that the viewers may be interested in, though not involving any kind of health-related information. With regard to the server-based architecture of Fig.…”

“…6 This is indeed a filtering process, though not driven by the profile of an individual viewer, but rather by a set of stereotypes that average the preferences and needs of different groups of viewers (using the same data structures as for individual profiles). At this point, we use a classical (cumulative) semantic similarity metrics that looks at relationships between the attributes of the TV program being broadcast and those of the items available to recommendation (indeed, it is the same metrics of [30]) -thus, we hitherto treat the health-related features of the items just like any other attributes, because TV audiences may involve people with very diverse health conditions; the enabling/disabling nature 5 http://diseaseontology.sourceforge.net/ 6 Of course, knowing the TV schedule, these computations can be performed in advance for the programs that will be broadcast in the future. of those data will be considered later, when filtering for individual viewers.…”

“…For reasons of scale, however, the ontologies still have to be kept in dedicated servers. It is worth noting that, in order to introduce health-related concerns in the recommendation logic, we have enhanced the ontologies formerly used in [30] to match TV programs with commercial products and services, which involved metadata from the TV-Anytime, MPEG-7 and eCl@ss standards. Specifically, following the methodology of [17], we have incorporated the openEHR internal vocabularies plus a subset of the Disease Ontology 5 to reason about health conditions, diseases, symptoms and signs using well-adopted terminologies.…”

“…These systems exploit data gathered in web navigation or TV watching records [5,46], learning histories [9,39], consumption profiles [26,28], electronic health records [6,25,29] and so on. Various authors have even combined several such sources to exploit synergies between different areas, for example, to enhance TV programs with educational contents related by topic [43], to situate advertising material within TV shows and thereupon launch interactive commercials [30], or to augment electronic programming guides with tutoring or commercial advising functionalities [2]. Only telemedicine remains an isolated area, inasmuch as the systems one can find in literature work exclusively with information from the healthcare domain (e.g.…”

“…We have explored some of these functionalities by incorporating our proposal in MiSPOT, a recommender system devised to embed selected advertisements in TV programs [30].…”