Description
Date depot: 1 janvier 1900
Titre: Exploiting Network Content-awareness to provide novel added value services
Directrice de thèse:
Renata TEIXEIRA (Inria-Paris (ED-130))
Domaine scientifique: Sciences et technologies de l'information et de la communication
Thématique CNRS : Non defini
Resumé:
Modern telecommunications involve a carrier infrastructure through which end devices interconnect to communicate with each other. This ecosystem has known major changes evolving from (1) islands of connectivity that provide only the voice service to (2) a more and more converging global “network of networks” that provides various types of services (mobile voice, web, data exchange, computing etc). With the advent of the web, some end devices connected to the Internet became themselves service providers that could reach, thanks the Internet’s global reachability, millions of users around the world. This change marked the beginning of a new era in the telecommunications industry: the birth of over-the-top (OTT) content providers (e.g. Yahoo, Google, Facebook etc). In this new ecosystem, composed of end users, carriers (telcos, ISPs, CDNs) and content providers, carriers are slowly loosing services in favour of content providers. To give an example, voice, video on demand, and videoconferencing can be provided on the application layer in a way that is transparent to the network. This is the case of Skype, Viber or Netflix to name few. In order to counter this effect, ISPs are desperately seeking to provide new services in every way (e.g. invest in their own content), but this quest remains elusive. It is even almost fair to assert the following today: Any service that can be consumed on an end user device (e.g. screen), and that can be provided by the carrier, can be provided as well at the application layer by using the network as a “dumb” commodity that “moves the bits around. The ISPs (carriers) seem therefore stuck with the “data carriage” and connectivity hat.
The thesis aims at exploring possible exceptions to the above mentioned assertion. If such an exception exists, it must most likely relate to services that use the knowledge that only the network (of the carrier) has. Compared to content providers, ISPs have an invaluable asset: a cross content provider view on the content traffic of a wide variety of Internet users. In this thesis, we would like to explore new services that can make profit from this knowledge.
One of such novel services, with which the thesis will start, is content and media curation. Content and media curation is the act of using both automatic and human resources, for the sake of aggregating, sorting, organizing and presenting interesting content to users. It has emerged during the last years as a promising trend in the web, and has remained exclusively an over-the-top service. The rationale behind the advent of such a new concept is that the Internet today contains much more content than what we can individually consume and sort by ourselves. We need therefore tools that “cure” the content for us so that we have only what “deserves” to be seen. Examples of such tools include Reddit Error: Reference source not found, Digg Error: Reference source not found, and PinterestError: Reference source not found; even social networks like twitter can be seen as media curation tools[4].
There are two main families of content curation tools. The first relies on individuals' expertise or passion for a given topic. Such individuals will act as ``content curators'' for their followers and will sort and organize the content on their behalf. For the web, Pinterest falls mostly under the first category: Pinterest users organize by topic their favorite web contents by putting them into different boards made available to their followers. Thus, by following a board that fits her interests, a user is likely to receive future interesting feeds. The second family of curation tools uses mainly the so-called ``wisdom of the crowd'', the postulate that the collective answer of a large group of individuals to a given question is often better than individuals' responses. Reddit and Digg fall under this second category. In Reddit, for instance, users submit their favorite contents (e.g., a link for video, or a news) to the portal, and let ``the crowd'' (other subscribers) rate them. Highly rated URLs will be published on the frontpage, which results in an entertainment platform where it is easy to find the hottest content of the moment. Such tools can be therefore perceived as recommendation systems that recommend to the Internet users the content that is likely to attract their attention. Finally, another example is “Google Trends” which provides the hottest searched words in the Internet at the moment (according to Google Search), thus providing an automatic curation tool.
As a starting point, the thesis will focus on the latter family of curation tools. We think that ISPs are better positioned to provide such a service thanks to the Internet-wide view that they have on the traffic. In fact, existing curation platforms rely on relatively small communities of users (e.g., Reddit) or perform curation only for the services they are told about (e.g., Google Search). By ``simply'' mon
Doctorant.e: Scavo Giuseppe