Google might be the biggest search resource on the Internet bar none, but it's far from perfect. In some areas it's actually worse than that, and simply fails to work at all. Now students at the University of Glasgow in Scotland are attempting to fill one of these search gaps with the development of the SMART search engine. 
Here's the problem: ask Google "How busy is it in New York?" and you will be confronted by a whole bunch of results which don't answer you query at all. In fact, I've just done exactly that and the first result unhelpfully told me what time it was in NYC, and this was followed by equally poor results including a three years old news story concerning Anne Hathaway, tripadvisor.com reviews of Washington Square and a story about the construction of the Freedom Tower (One World Trade Center) but absolutely nothing to help me gauge how bust New York City is.
The search engine for multimedia environment generated content (SMART) project aims to answer this kind of question, and provide those answers in real time. Not only would you be able to find out how busy a city center is, you'd be able to find out how busy it is right now. The computer scientists at Glasgow are attempting to build a search engine that draws the results from net-connected sensors such as cameras and microphone arrays in the physical world. By matching the search queries made with the information that these sensors provide, and then cross-referencing data from social networks, the real-time geo-location and truly personalised potential of search could be realised. The students hope that detailed answers to questions such as 'what part of the city hosts live music events which my friends have attended recently?' could become reality.
Dr Iadh Ounis from the University of Glasgow’s School of Computing Science explains that the SMART project is being built upon an open-source search engine technology called Terrier that his team has been developing for the last eight years and will exploit the existing concept of smart cities. Ounis is confident that SMART search will be up and running for testing in a real city by 2014.
Key to the success of the project will not only be the obvious presence of enough sensors combined with a method of collecting and collating the data they collect, but also the ability to process that data efficiently. Therefore students are hard at work on perceptual A/V signal processing algorithms based on sensor data and metadata (location and state for example) as well as the 'dynamic context' of the physical world as it relates to that environment and exampled by processing algorithms to deal with face detection, person tracking, crowd analysis and the classification of acoustic events. SMART will also integrate social networking data (Twitter 'tweets' primarily at this stage) to facilitate the mapping of social queries onto this physical world data.
Think of SMART as being comprised of three layers architecturally: layer one edge nodes which communicate with the physical world by collecting and combining sensor and social network feeds; layer two is the search engine itself which covers search and retrieval, and query processing of course, across multiple edge nodes; and finally layer three comprises of the applications which are expected to employ reusable mashup libraries for visualisation purposes.
Perhaps one of the more interesting use cases put forward by the SMART project team is that of live news. It sees news organisations making queries regarding the occurrence and evolution of events: "where are riots taking place?" for example. It is thought that the multimedia data acquired from city sensors could be used by the presentation layer of the system to populate news portals with live reports, personalised by correlated individual search requests from end users under a thematic banner, to create a truly dynamic and 'as it happens' news service.
Depending upon how you view the privacy and freedom debate, SMART might not be all good news though. Another use case being put forward by the project team is that of security and surveillance in an urban environment. Using the existing deployment of CCTV cameras in many urban environments for anti-terrorism and crime prevention purposes, SMART could provide a method of answering targeted queries based upon that sensor data such as the detection and surveillance of specific terrorist suspects for example.
