This paper presents current work on a video retrieval project at Cambridge University and Olivetti Research Limited (ORL). We show that statistical methods developed for text retrieval are also effective for retrieving and browsing multimedia documents. These methods allow rapid retrieval of news broadcasts by information content determined from teletext subtitles. Information retrieval results for experiments performed on a large archive of news broadcasts are presented. This is made possible by the ORL Medusa system, which allows practical recording, storage, and playback of tens of gigabytes of multimedia data. This work is a step towards practical retrieval of multimedia documents, where the information content is determined from speech recognition performed on the audio soundtrack. We describe the project background, the ORL Medusa multimedia system, and retrieval application, as well as the news broadcast corpus and methods of browsing the retrieved news stories.

Introduction The Information Age is fully upon us. A recent article noted that there are perhaps 50 million people using the Internet on a regular basis, and that "the current growth rate is about 15% per month (!) and this could well continue until almost all of those in the `developed world' are connected" [Fenn94, p. 30]. In addition, the digital domain consists not only of text but increasingly of other media representations, from graphics images to audio to motion video. As the amount of information and number of users exponentially escalate, more attention focuses on the basic problems of information management: How do you digitize information? How can you then visualize it and find what you need? How do you use and manipulate it effectively? How is it stored and managed? The proliferation of technical articles and special issues addressing these questions underscore their importance; see for example the special issue on content-based retrieval [Narasimhalu95] or digital

Abstract not found

This paper describes an approach for detecting dialogue scenes in movies. The approach uses automatically extracted low- and mid-level visual features that characterise the visual content of individual shots, and which are then combined using a state transition machine that models the shotlevel temporal characteristics of the scene under investigation. The choice of visual features used is motivated by a consideration of formal film syntax. The system is designed so that the analysis may be applied in order to detect different types of scenes, although in this paper we focus on dialogue sequences as these are the most prevalent scenes in the movies considered to date. 1

In this paper, we study the automatic construction of personalized TV News programs, where we want to build a program with predefined duration and maximum content value for a specific user. We combine video indexing techniques to parse TV News recordings into stories, and information filtering techniques to select stories which are most adequate given the user profile. We formalize the selection process as an optimization problem, and we study how to take into account duration in the selection of stories. Experiments show that a simple heuristic can provide high quality selection with little computation. We also describe two prototypes, which implement two different mechanisms for the construction of user profiles: . explicit specification, using a category-based model, . implicit specification, using a keyword-based model.

The development of Digital Television opens new perspectives in the distribution of audio-visual material to the general public. The immediate advantages of digital broadcast are the improvement in transmission quality and the increase in transmission capacity. But the major change for users will come from the future capacity of the video delivery chain to process this digital information to build new interaction paradigms, such as Interactive Television. In particular, it is expected that one important paradigm will be the construction of customized programs, programs that are specifically designed to fit the needs of each user. This presentation will describe some of the technologies that can be used for this type of processing, for example automatic audio-video analysis and parsing, information filtering algorithms, user profile creation and update, and recommendation systems. Experiments and prototypes that have been developed by the Eurecom Multimedia Communications Department will be presented.

INTRODUCTION La disponibilit croissante de documents multimdia sous forme digitale pose le dlicat problme de l'accs ce gigantesque volume d'information. Si la recherche documentaire a apport depuis longtemps des mcanismes pour retrouver de l'information textuelle, les documents audio-visuels posent des difficults particulires. En effet, pour dfinir automatiquement le contenu d'un document audio-visuel, il faut faire appel des technologies de reconnaissance de formes, soit pour le son, soit pour l'image, dont la complexit est souvent trs grande, et dont les performances souffrent de nombreuses limitations. Dans cet article, nous nous intressons l'analyse et l'indexation automatique de journaux tlviss. Le but est de reconnatre la structure de journal de faon identifier les diffrents lments qui le constituent (prsentateur, reportages, interviews, publicits...). A partir de ces lments, il est ensuite facile de construire, par exemple, un interface utilisateur permettant un accs hypermdia

The automatic analysis of the contents of multimedia documents requires to combine informations coming from various data types (audio, video, text...). In this paper, we propose an architecture that describes agents for processing flows of information. These agents can be applied to elementary data types (audio, video...), but also on the results produced by other agents. The architecture includes a Multimedia Flow Browser, which is able to display simultaneously visual representations of the various information flows produced by agents, and an Agent Editor, which provides a graphical interface to create new agents by combining existing agents. The architecture is open, so that it is possible to add new data types (and the procedures to visualize them) and new agents. A simulated example is presented to show the possible usage of this tool in an application based on TV News recordings.