Computer Laboratory

RECOGNITION

Overview

Video/Photo rating

People in SRG

Publications

Contact


Overview

RECOGNITION is an EU 7th Framework Programme project (2010-2013). RECOGNITION concerns new approaches for embedding self-awareness in ICT systems. This will be based on the cognitive processes that the human species exhibit for self-awareness, seeking to exploit the fact that humans are ultimately the fundamental basis for high performance autonomic processes. This is due to the cognitive ability of the brain to efficiently assert relevance (or irrelevance), extract knowledge and take appropriate decisions, when faced with partial information and disparate stimuli. Using the psychological and cognitive sciences as concrete inspiration, our approach is to develop functional models of the core cognitive processes that allow humans to assert relevance and achieve knowledge from information. This involves mechanisms such as inference, belief, similarity and trust. These will be translated to the ICT domain by development of flexible RECOGNITION algorithms that can be imbedded in ICT on a flexible basis for self-awareness.

A specific topic Cambridge Systems Research Group is working on is Information Cascade. Requests for web content are known to follow a heavy-tailed distribution. For example, the top 10% of the videos in a video-on-demand system account for approximately 60% of accesses, and the remaining 90% of the videos account for 40%. This popularity pattern can make server provisioning difficult, especially for rich-media content such as video streaming, which have relatively strict delivery constraints. The problem becomes especially severe with the recent proliferation of rich-media User Generated Content (UGC) such as YouTube videos, whose popularity can vary dynamically and often dramatically.

While global replication via content delivery networks (CDNs) is efficient for the most popular contents, the majority of objects are in the tail and accessed too rarely for global replication to be practical. For instance, YouTube only uses CDNs for the most popular videos. However, objects in the tail collectively account for a sizeable fraction of accesses. Moreover, content that is not yet popular might suddenly become globally popular.

Knowledge of a UGC object can spread in two ways: broadcast highlights or viral propagation. The former happens when the UGC object is featured or highlighted on a central index page. Examples include being featured on the home page of the hosting sites (such as the featured videos list on YouTube); being promoted on an external social bookmarking site (e.g. if slashdotted, or featured on Digg, Reddit, Del.icio.us); or ranking high in a Google search. UGC objects in this class have to be popular according to the indexing algorithm used. Such high-visibility objects will likely be accessed many times and from all over the world and are best served by replicating globally via CDNs.

The second possible mean of propagation is by word-of-mouth, sharing explicitly with a group of friends. This can happen through online social networks, emails, or out-of-band (or face-to-face) conversations. This kind of viral propagation has been termed a social cascade and is considered to be an important reason for UGC information dissemination. The links between friends in an online social network explicitly captures the means of propagation for social cascades. When users access a UGC object influenced by their friends, it can be modelled as if infected by such friend's opinion. We envision that many ideas, messages, and products could be spread rapidly through our population as social epidemics.

We are interested in how the social cascades flow. In what kind of conditions does a person decide to cascade information? Can we categorise the behaviour of people who make decisions whether to pass/gossip information. We will apply cognitive psychological analytical methods to understanding decision making mechanisms.

Currently we are tackling these issues from the following angles:

  1. Understanding decision making process in micro scale with psychological aspect in people networks including extracting characteristics of user (node) and classifying the user/node behaviour. We have developed web-based applications to investigate how opinions are influenced by peers and community structures (see Video/Photo Rating).
  2. Understanding propagation of information/message through diffusion process with group based psychological aspects. In order to tackle this issue, we apply experimental, learning and discovery method, initially for understanding diffusion.

Publications

  • D. Fay, J. Kunegis, and E. Yoneki. On Joint Diagonalization for Dynamic Network Analysis. Technical Report, University of Cambridge, 2011. (UCAM-CL-TR-806).
  • N. Wang and E. Yoneki. Impact of Social Structure on Forwarding Algorithms in Opportunistic Networks. IEEE Int'l Conf. on Selected Topics in Mobile & Wireless Networking, 2011. (PDF Version).

Contact

For any queries please contact Dr Eiko Yoneki, or visit the RECOGNITION project website for more information.