Received: from sleper.cwi.nl (sleper.cwi.nl [192.16.201.241]) by hera.cwi.nl with ESMTP id PAA08515 for ; Wed, 9 Feb 2000 15:18:44 +0100 (MET) Received: from localhost by sleper.cwi.nl with ESMTP id PAA54567; Wed, 9 Feb 2000 15:18:42 +0100 (MET) Message-Id: X-Mailer: exmh version 2.0.2 2/24/98 To: Lynda Hardman cc: Frank.Cornelissen@cwi.nl, Frank.Nack@cwi.nl, Jacco.van.Ossenbruggen@cwi.nl, Jim.Davis@cwi.nl, Joost.Geurts@cwi.nl, Lloyd.Rutledge@cwi.nl, Steven.Pemberton@cwi.nl, Frank.van.Harmelen@cs.vu.nl Subject: Research Description ... In-reply-to: Your message of "Wed, 09 Feb 2000 10:36:22 MET." Mime-Version: 1.0 Content-Type: text/plain; charset=us-ascii Date: Wed, 09 Feb 2000 15:18:38 +0100 From: Craig Lindley Research for the period 1 February 2000 to 1 July 2000 (within the scope of the DYNAMO project and in collaboration with Frank Nack and others): For the past two years I have been working on the representation of video semantics and the development of models and algorithms for interactively generating video sequences from video databases. This work has led to the development of an association engine for dynamically linking video components based upon patterns of similarity and dissimilarity in the semantic descriptors (or annotations) of those components. The resulting video presentations are organised by a categorical and associative video syntax. Ongoing research is concerned with developing richer models of video form (syntax and semantics), together with representation techniques and algorithms for generating videos having those forms. Candidate forms include the rhetorical (currently being investigated by the INS2 team at CWI) and narrative (that has been investigated by Frank Nack in his PhD work). Other forms are also possible, but research in discourse theory, film theory, and narrative theory have generated substantial results for rhetorical and narrative forms that might be applied to the computational task of semi-automated video sequence generation. Narrative theory itself is highly complex, and a great deal of work is required to develop effective computational models from general film aesthetics and discourse research and analysis. Ongoing work is also required to clarify the relationships between rhetorical structure and narrative structure, and to integrate these with general causal and common sense models, as well as more specific models and principles of cinematic montage. The research undertaken during the next few months at CWI will further advance understanding of these issues by concentrating upon the film theory of Sergei Eisenstein, and analysing the relationship of that theory to rhetorical structure theory (RST). Basic questions include which elements of Eisenstein's theory can be modelled using RST, or which elements of RST may be modelled using concepts developed by Eisenstein. Modelling approaches will be evaluated in terms of how easily they may support computational generation of coherent video sequences from a database of video clips. The models will also be considered in terms of the support they provide for other forms of interaction with a video database, such as summary generation, or formal exploration and/or analysis of the structure or nature of the video material. Ongoing research will extend more deeply into general narrative theory. This may begin within this period if time and progress permit.