Use this URL to cite or link to this record in EThOS:
Title: Interest-based segmentation of online video platforms' viewers using semantic technologies
Author: Sora, Radu
ISNI:       0000 0004 5916 0711
Awarding Body: University of Warwick
Current Institution: University of Warwick
Date of Award: 2016
Availability of Full Text:
Access from EThOS:
Access from Institution:
To better connect supply and demand for various products, marketers needed novel ways to segment and target their customers with relevant adverts. Over the last decade, companies that collected a large amount of psychographic and behavioural data about their customers emerged as the pioneers of hyper-targeting. For example, Google can infer people’s interests based on their search queries, Facebook based on their thoughts, and Amazon by analysing their shopping cart history. In this context, the traditional channel used for advertising – the media market – saw its revenues plummeting as it failed to infer viewers’ interests based on the programmes they are watching, and target them with bespoke adverts. In order to propose a methodology for inferring viewers’ interests, this study adopted an interdisciplinary approach by analysing the problem from the viewpoint of three disciplines: Customer Segmentation, Media Market, and Large Knowledge Bases. Critically assessing and integrating the disciplinary insights was required for a deep understanding of: the reasons for which psychographic variables like interests and values are a better predictor for consumer behaviour as opposed to demographic variables; the various types of data collection and analysis methods used in the media industry; as well as the state of the art in terms of detecting concepts from text and linking them to various ontologies for inferring interests. Building on these insights, a methodology was proposed that can fully automate the process of inferring viewers interests by semantically analysing the description of the programmes they watch, and correlating it with data about their viewing history. While the methodology was deemed valid from a theoretical point of view, an extensive empirical validation was also undertaken for a better understanding of its applicability. Programme metadata for 320 programmes from a large broadcaster was analysed together with the viewing history of over 50,000 people during a three-year period. The findings from the validation were eventually used to further refine the methodology and show that is it possible not only to infer individual viewers interests based on the programmes watched, but also to cluster the audience based on their content consumption habits and track the performance of various topics in terms of attracting new viewers. Having an effective way to infer viewers’ interests has various applications for the media market, most notably in the areas of better segmenting and targeting audiences, developing content that matches viewers’ interests, or improving existing recommendation engines.
Supervisor: Not available Sponsor: Norwegian State Educational Loan Fund (Statens lånekasse for utdanning)
Qualification Name: Thesis (Ph.D.) Qualification Level: Doctoral
EThOS ID:  DOI: Not available
Keywords: HF Commerce