Kieran McDonald

A comparison of score, rank and probability-based fusion methods for video shot retrieval

It is now accepted that the most effective video shot retrieval is based on indexing and retrieving clips using multiple, parallel modalities such as text-matching, image-matching and feature matching and then combining or fusing these parallel retrieval streams in some way. In this paper we investigate a range of fusion methods for combining based on multiple visual features (colour, edge and texture), for combining based on multiple visual examples in the query and for combining multiple modalities (text and visual). Using three TRECVid collections and the TRECVid search task, we specifically compare fusion methods based on normalised score and rank that use either the average, weighted average or maximum of retrieval results from a discrete Jelinek-Mercer smoothed language model. We also compare these results with a simple probability-based combination of the language model results that assumes all features and visual examples are fully independent.

Improving the Quality of the Personalized Electronic Program Guide

As Digital TV subscribers are offered more and more channels, it is becoming increasingly difficult for them to locate the right programme information at the right time. The personalized Electronic Programme Guide (pEPG) is one solution to this problem; it leverages artificial intelligence and user profiling techniques to learn about the viewing preferences of individual users in order to compile personalized viewing guides that fit their individual preferences. Very often the limited availability of profiling information is a key limiting factor in such personalized recommender systems. For example, it is well known that collaborative filtering approaches suffer significantly from the sparsity problem, which exists because the expected item-overlap between profiles is usually very low. In this article we address the sparsity problem in the Digital TV domain. We propose the use of data mining techniques as a way of supplementing meagre ratings-based profile knowledge with additional item-similarity knowledge that can be automatically discovered by mining user profiles. We argue that this new similarity knowledge can significantly enhance the performance of a recommender system in even the sparsest of profile spaces. Moreover, we provide an extensive evaluation of our approach using two large-scale, state-of-the-art online systems—PTVPlus, a personalized TV listings portal and Físchlár, an online digital video library system.

Design, implementation and testing of an interactive video retrieval system

In this paper we present and discuss the system we developed for the search task of the TRECVID 2002, and its evaluation in an interactive search task. To do this we will look at the strategy we used in designing the system, and we discuss and evaluate the experiments used to determine the value and effectiveness of one system incorporating both feature evidence and transcript retrieval compared to a transcript-only retrieval system. Both systems tested are built on the foundation of the Físchlár System developed and running for a number of years at the CDVP. The system is fully MPEG-7 compliant and uses XML for exchange of information within the overall architecture.

The físchlár digital video system: a digital library of broadcast TV programmes

Físchl& acute;r is a system for recording, indexing, browsing and playback of broadcast TV programmes which has been operational on our University campus for almost 18 months. In this paper we give a brief overview of how the system operates, how TV programmes are organised for browse/playback and a short report on the system usage by over 900 users in our University.

Fischlar: an on-line system for indexing and browsing broadcast television content

This paper describes a demonstration system which automatically indexes broadcast television content for subsequent non-linear browsing. User-specified television programmes are captured in MPEG-1 format and analysed using a number of video indexing tools such as shot boundary detection, keyframe extraction, shot clustering and news story segmentation. A number of different interfaces have been developed which allow a user to browse the visual index created by these analysis tools. These interfaces are designed to facilitate users locating video content of particular interest. Once such content is located, the MPEG-1 bitstream can be streamed to the user in real-time. This paper describes both the high-level functionality of the system and the low-level indexing tools employed, as well as giving an overview of the different browsing mechanisms employed.

Experiences of creating four video library collections with the Físchlár system

This paper describes how the Físchlár system, which supports indexing, browsing and searching through archives of digital video information, has been used to create four separate video libraries of information. We briefly introduce Físchlár and then describe its application in Físchlár-TV (a digital library of recorded broadcast TV content, updated regularly), Físchlár-News (a digital library of TV news, updated daily), and Físchlár-Nursing (a digital library of video teaching materials in the domain of nursing), and how Físchlár has also been used to provide searching through a collection as part of the TREC2002 video track interactive user experiments. Our experiences show that the range of user requirements for accessing video content seems to be much broader than for any other media, which makes the development of video access techniques very challenging.

Mobile access to the Físchlár-News archive

In this paper, we describe how we support mobile access to Fischlar-News, a large-scale library of digitised news content, which supports browsing and content-based retrieval of news stories. We discuss both the desktop and mobile interfaces to Fischlar-News and contrast how the mobile interface implements a different interaction paradigm from the desktop interface, which is based on constraints of designing systems for mobile interfaces. Finally we describe the technique for automatic news story segmentation developed for Fischlar-News and we chart our progress to date in developing the system.

Dublin City University at CLEF 2006: experiments for the ImageCLEF photo collection standard ad hoc task

For the CLEF 2006 Cross Language Image Retrieval (ImageCLEF) Photo Collection Standard Ad Hoc task, DCU performed monolingual and cross language retrieval using photo annotations with and without feedback, and also a combined visual and text retrieval approach. Topics are translated into English using the Babelfish online machine translation system. Text runs used the BM25 algorithm, while visual approach used simple low-level features with matching based on the Jeffrey Divergence measure. Our results consistently indicate that the fusion of text and visual features is best for this task, and that performing feedback for text consistently improves on the baseline nonfeedback BM25 text runs for all language pairs.