The festival galleries – GaFes

Festivals are gaining an increasing success and some of them, such as Cannes or Avignon, produce a significant activity on the Web. This project proposes to analyze this activity to better understand the festival practices and to develop access and visualization methods for the contents generated during and around the festivals. These two objectives are complementary and the underlying scientific problems concern both information technologies and humanities, which are teamed in this project. This project focuses on two axes: the study of usages through the data collected on the Web and the repurposing of content captured or produced by internet users. These two axes rely on the collection and structuration of data linked to a festival through several sources (Twitter, blogs, forums, etc...). The structuration of these collections will allow us to process investigations, to extract knowledge about the public and about the cultural practices, to show specific view points for defined categories of users:
Structuration: the task will be to extract descriptors, to identify similarities, to produce condensed representation of the content. We will propose paradigms for robust representation of these documents, in particular using partial least square techniques that allow to easily combine a priori knowledge and unsupervised analysis of large quantities of data.
Numerical landscape: the objective of this axis is to revisit the investigation tools on numerical landscape and to follow users in their choices and interactions (choosing a show, advising, commenting,...). The invention of a numerical landscape raises the question of the observation protocol for users, of the quantity of information, of the strategies for data collection and selection. The automatic structuration of collected data will offer a technological and methodological framework for their exploration and analysis.
Repurposing: The objective is to exploit collected multimedia data to compose synthetic views of these collections. In practise, the task is to automatically produce video clips intended for a precise need of users (for example, a shot presentation of the next day shows). We will evaluate approaches based on extraction and assembly of video segments, through the fundamental questions of the definition of interest functions adapted to the cultural context and to the composition logic, which are open problems which will be addressed jointly by computer scientists and sociologists. The breakthrough potential of the project lies in the answers that it will propose for the issue of numerical landscapes and on the manner we envision to include the knowledge of the public and of the impact of the shows in the structuration and repurposing of video content.
Experiments will be performed on 4 famous festivals: the Avignon theater festival, the Festival Lumière, the Transmusicales and the Cannes festival. The project has already been presented to the organizers of the first three and to the company UniversCiné, which covers extensively the Cannes Festival. All have formally approved their support to the project. The valorization of the project results will notably include the development of a “Festival Observatory”, an open application intended both for the general public, the organizers and the sociologists. This application will offer a synchronous view of the indicators automatically extracted, navigation inside the collection of data and the generated structure, to display video clips automatically generated from a user request.