The National Library of Sweden and SVT, together with the FIAT/IFTA Media Management Commission invite you to the 9th edition of FIAT/IFTA’s “Changing Sceneries, Changing Roles” Seminar.
23-24 maj 2019
May 23rd-24th 2019
Garnisonen Conferens Center, Karlavägen 100, Stockholm
Number of participants
Register to the seminar here. Last day to register is May 8th 2019.
We have made reservations for 80 rooms in three various hotels, all nearby the venue.
Moderator: Elena Brodie-Kusa
Theme: Artificial Intelligence, Machine Learning and their potential for automated annotation
08.30 – 09.30 Registration and coffee
9:30–10:15 WELCOME SPEECHES
HANNA STJÄRNE, SVT, Sweden and LARS ILSHAMMAR, National Library of Sweden (NLS), Sweden. BRID DOOLEY FIAT President, VICKY PLAINE, MMC Commission.
10:15–10:45 20 YEARS OF MMC
JACQUI GUPTA, BBC, UK and EVA-LIS GREEN, NLS, Sweden
A flashback to the trends and evolutions that Archives and Media has experienced over the past 20 years. How have the Sceneries and Roles changed? What were the buzzwords 20 years ago up to the present day? What was it like in 1998 – what did we expect? What has happened through our journey of evolution, what is constant and what are the transformations?
10:45–11:15 AUDIO FORENSICS AND AUTOMATION
PATRICK AICHROTH, Fraunhofer IDMT, Germany
In this presentation, Patrick Aichroth will showcase how audio analysis Techniques developed by Fraunhofer can be used for two domains:
- Audio forensics: how signal analysis and machine learning can be used to detect suspicious locations within (potentially manipulated) speech material, hence supporting the process of validating audio material.
- Audio segment matching and visualisation: how the detection of partial matches within datasets can be used for rights tracking and to identify audio-reuse and repetitions within broadcast streams.
11.15 – 11.45 MORNING BREAK
11:45–12:15 IDENTIFYING THE BUSINESS CASES FOR AUTOMATIC METADATA IN THE FINNISH BROADCASTING COMPANY
KIM VILJANEN and ELINA SELKÄLÄ, YLE, Finland
Automatic metadata extraction technologies for audiovisual content (speech recognition, face recognition and visual feature extraction) promise new ways to address the broadcasters ever increasing demand for metadata in production, publishing and archiving. However aligning the opportunities of the new technology with the business needs of a broadcasting company turns out to be challenging due to limitations of the technology, different needs of different business cases, the need to change existing processes and systems, and the need for new skills. In this presentation we discuss our learnings from several projects in identi- fying the right business cases for automatic metadata, especially the EU funded research project MeMAD and the internal project ”Metadata machine”.
12:15–12:45 HUMANS IN THE AUTOMATED ANNOTATION LOOP (AND THEIR POTENTIAL FOR ARTIFICIAL INTELLIGENCE AND MACHINE LEARNING)
JENS EDLUND, Royal Institute of Technology (KTH), Sweden
In this age of automated annotation methods, machine learning, and the non-descript yet seemingly omnipresent AI, there is new hope that the immense media archives that can be found in many places can finally be put to better use. At the same time, the developers of AI and machine learning methods are not always interested in testing their results for usability, or even sanity. This is unsurprising – they are usually researchers of algorithms, not of archival content. Meanwhile, scholars that are used to (semi-)manually inspecting the archives and to painstaking annotation following meticulously elaborate annotation schemes, occasionally seem intimidated or at least put off by the automated methods. There are areas where these groups meet and seek to reconcile their differences – the field of digital humanities is perhaps one – but in general, there is scope for innovation when it comes to putting the human back in the automated annotation loop.
12.45 – 14.00 LUNCH
14:00–14:30 AI SOLUTIONS
LÉONARD BOUCHET and SÉBASTIEN DUCRET, RTS, Switzerland
RTS (Radio Télévision Suisse) will make a demo of the tools (AI-machine learning based) it has developed and put into production in recent months. It will also present concrete examples where these tools have proven to be very useful and powerful in solving complex or very time consuming issues. This presentation will be also the occasion to present examples of possible international collaboration and cooperation on Machine Learning related topics amongst interested broadcasters archives.
14:30–15:00 THE ARDS EXPERTS GROUP ON VIDEO MINING
GABRIELE WENGER-GLEMSER, Bayerischer Rundfunk, Germany
The cooperation of public broadcasters in Germany has established an Expert-Group on Videomining to provide information and consultation on the applicability of videomining-solutions for audiovisual archives in the cooperation in order to tackle the many open issues coming along with that task: what is the current status of audiovisual mining techno- logies, what tools and solutions are considered to be taken into account for implementation within the next years, what are the main use-cases considered to be covered by those technologies, and many, many more. Alongside representatives and experts from ARD-archives (like BR, NDR, SWR, WDR, DRA, etc.) the group included experts from the ZDF and the Austrian ORF as well. Gabrielle will present the methodologies and the approach used by the group, together with some of the first results and further goals. The presentation will give a direct view on the very practical aspects of the work in the group as well as an overview on the many obstacles and challenges faced on the run.
15:00–15:30 LEARNING FROM ARCHIVES: HOW HISTORICAL CONTENT CAN BE USED TO ENGINEER NEW CONTENT PRODUCTION TOOLS
GRAEME PHILLIPSON, BBC R&D, UK
There are now many tools to enhance video archives with extended metadata such as shot boundaries / types, face locations / identities etc. Graeme will present examples of how large quantities of historical meta- data can be used to develop new tools that would have been prohibitively difficult previously. In particular he will present work on Ed, a system which BBC R&D have developed to automate the coverage of live events with a multi-camera setup. We will show several examples of how the system has learnt how to frame, choose shot type and time changes by examining features in historical video such as face location, shot type, shot timing and audio features.
15:30–16:00 IBM AI OVERVIEW WITH SEVERAL EXAMPLES OF PROJECTS IN THE MEDIA AND LESSONS LEARNED
JAKOB ROSINSKI, IBM GBS Europe
This speech will give you an overview about client projects in the space of media archives worldwide IBM has contributed to with it’s own AI - named Watson - but also with it’s knowledge and integration capabilities. Major topics are scope definition and use case identification, further the usage of cognitive services of different kinds and vendors - with success and open problems. In such a multi-modal approach training of services is also key, and the speech should show how this can be managed both from a human and machine perspective.
16.00 – 16.30 AFTERNOON BREAK
16:30–17:00 OTMEDIA, THE TRANSMEDIA NEWS OBSERVATORY
NICOLAS HERVÉ, INA, France
From French media such as television, radio, Web, Agence France Presse (AFP) dispatches and Twitter feeds to study the flow of information, OT- Media aggregates innovative software components on indexing and data mining that analyze, quantify, visualize and understand the flow of all information, whatever their broadcast media. With nearly 12 million items analyzed, the tool is now able, in an increasingly complex and
abundant ecosystem, to track in real time the appearance, propagation and relays of information.
17:00–17:30 PANEL DISCUSSION
Moderators: Jennifer Wilson, Xavier Jacques-Jourion and Tim Manders
THEME: The challenge of content-security and rights management
9:30–10:00 NISV’S IPR FLOWCHART AND HOW TO STORE RIGHTS METADATA IN THE MAM
MARJOLEIN STEEMAN and MAARTJE HÜLSENBECK, NISV, The Netherlands
Marjolein and Maartje will talk about the research that needs to be done in order to determine the rights status of material by showing the IPR flow chart. In order to use the flowchart and draw conclusions based on copy- right, neighbouring rights and to do an ethical check based on privacy law, you need certain data regarding the material. This implies the need for specific rights information and the urgency to document this type of research. They will explain how the rights entity can be used and illustrate their concept with some use cases.
10:00–10:30 RIGHTS MANAGEMENT IN FINA’S DIGITAL ARCHIVE
JOANNA KALISZEWSKA, FINA, Poland
Comprehensive description of audiovisual resources is one of the most important tasks for the archive. Maintaining the connection between work description, its rights management metadata and linking data to the carriers is essential for the processes of building an efficient MAM system. FINA’s archive preserves wide range of materials (film, tv and radio). It can be an example of acquiring safe, standardized and efficient meta- data within file preservation system dedicated to all types of audiovisual content with connection to rights management data. During our presen- tation we would like to show how the standardization of our MAM system influenced design and functional solutions of rights management for each title of our archive. We would also like to present our answers to creating effective preservation paths by introducing chosen case studies of data description for different types of materials.
10:30–11:00 RIGHTS IN NEWS AND IPTC RIGHTSML
ANDY READ, BBC News, UK
This presentation will give an overview of how Rights Management of assets are used in BBC News workflows today and discuss the develop- ment of IPTC’s RightsML, a data model for marking up rights expressions about content of all relevant media types in a machine readable way.
11.00 – 11.30 MORNING BREAK
11:30–12:30 ’DAILY PRACTICE, DEEP DIVE’ ON RIGHTS IN MAMS
GERHARD STANZ, ORF, Austria, JOHANNE WORSAAE NIELSEN, DR, Denmark and MAJA WETTMARK, NRK, Norway
To foster exchange of information between archives by sharing best prac- tice milestones, the MMC introduces an new presentation format that we would like to call “Daily Practice, Deep Dives”. We will highlight a detailed topic from current archive workflows, and compare/discuss how different companies render them. For this seminar we have chosen “Rights Codes in MAMs”. In short presentations, ORF, NRK + DR will allow a closer look into the code systems implemented, and the extent of integration into the technical/organizational operations. By having these discussions we look to get a comprehensive view on practical experiences. Merits, flaws, and future plans shall be discussed openly , and provide genuine insights and overviews.
THEME: Megatrends and Change Management in Broadcast Archives
14:00–14:30 CLOUD BASED ARCHIVING
DAVID KLEE, A+E Television Networks, US
With the costs of storage from cloud providers like AWS, Google and Microsoft continuing to decrease (as well as the new innovations these providers continue to enable), there are use cases today where archiving media content in the “public” cloud can be practical. This talk will give an overview of cloud provider archiving capabilities, along with the pros and cons of using these providers in media supply chains. It will also explore a specific set of use cases where cloud archiving can enable agility for a global organization looking to manage content across time zones and organizational boundaries without a large physical infrastructure footprint.
14:30–15:00 UNIVERSAL IDENTIFIERS THROUGHOUT PRODUCTION CHAIN, OVERVIEW AND INTEROPERABILITY
RAYMOND DREWRY, EIDR, UK (Entertainment Identifier Registry)
Universal unique identifiers are essential throughout lifecycle of film and television works – production, distribution, redistribution, metadata management, and archiving. EIDR, a non-profit registry for the global audiovisual sector, provides such an identifier. EIDR has 100 member organizations, over 2 million records, and over 4 million cross-references to other systems. The presentation will cover the basics of what EIDR is, its features, and how it operates, and provide several examples of uses in industry, as well as heritage and archival applications.
15:00–15:30 DEVELOPMENTS AND CHANGES IN SVT ARCHIVE
HENRIK JOHANSSON, SVT, Sweden
What kind of metadata from the archive can be used for online publishing? What changes are necessary to make internal data public? Lessons learned from the project “Searchable SVT” that took place in 2018 and was a colla- boration between several departments at SVT, with the aim to develop the handling and usage of metadata.
15:30–16:00 AFTERNOON BREAK
16:00–16:30 TO ESTABLISH SOPHISTICATED 4K AND 8K WORKFLOWS THROUGH THE CASES OF NHK
ARAI HIROKAZU, NHK, Japan
The difficulty of preserving 4K and 8K is not only derived from its file size and bit rate, but also from production styles that are different from ones for SD and 2K, adoption of new broadcast standards such as wider gamut and HDR, and so on. There are still many challenges to establish sophisticated workflows, and connectivity with new distribution plat- forms and other online services also affects. In this session Arai Hirokazu will share NHK’s challenges through some actual cases.
16:30–17:00 TECHNICAL CHALLENGES AND OPPORTUNITIES FROM THE NEW EUAVSM DIRECTIVE
XAVIER JACQUES-JOURION and SEVARA IRGACHEVA, CSA, Belgium
The new EU Audiovisual Media Services Directive was finalized in the final weeks of 2018, meaning that member states will have to transpose it into national law within 21 months. We will provide a technical and strategic overview of the implications of the new directive for audiovisual media archives and related technology.
17:00-17:30 SEMINAR WRAP-UP AND CONCLUSIONS