Milestones


Number Name Date
MS1 Requirements December 2011
MS2 Initial system architecture March 2012
MS3 Initial prototypes September 2012
MS4 Updated prototypes May 2013
MS5 Integrated system December 2013
MS6 System validation March 2014

WP1: Project Management

The overall objective of the project management is to lead TOSCA-MP so that it reaches all scientific, organisational and financial goals and fulfils all its contractual obligations. Therefore, the project management will set up efficient structures and procedures with a clear distribution of responsibilities. A clear definition of tasks and a work distribution which, on the one hand, avoids excessive interactions between the participants but, on the other hand, ensures that all the necessary communication between the participants will be established. The project management of TOSCA-MP is responsible for ensuring the overall project quality. A Quality Assurance (QA) Plan will be applied to all TOSCA-MP activities and deliverables. QA will ensure mediation of errors as early in the life cycle as possible. The QA plan consists of planned and systematic activities to determine and ensure achievement of TOSCA-MP quality objectives.

 

WP2: Automatic Metadata Extraction and Enrichment

The overall objectives of this work package are extraction, generation and aggregation of metadata information from various audio-visual content using advanced media analysis techniques. The technologies to be developed in WP2 should process digital content and generate required metadata automatically. Furthermore, metadata extracted from audio and visual content will be linked and interpreted by semantic enrichment. The main challenge addressed by this work package is the adaptive usage of the above mentioned WP2 objectives in multilingual and genre dependent processing. Multilingual speech processing will be addressed through spoken language identification, automatic transcription in several languages and automatic translation. For genre adaptive processing, a workflow will be provided that allows genre dependent scene interpretation. This will be reached by metadata extraction and semantic enrichment adapted for the respective content. The expected result of this work package is a workflow for automatic audio visual metadata extraction, enrichment and linking which is fully adjustable to the special needs of the processed content and distributed search applications.

 

WP3: Networked Search and Result Presentation

The overall objective of this work package is to provide tools and mechanisms for a networked search over distributed repositories and an aggregated set of visual paradigms for representing results for the different users. In order to provide a unique experience in content retrieval based on different genres, topics or contexts, we have to go further and also provide an unique way to show results and let the user retrieve and browse distributed repositories based on different retrieval paths. Finally, this work package aims at the ability of the system of learn of successful and unsuccessful retrievals based on relevant and reliable user feedback. Among the challenges that will be tackled by this work package, we can find the integration of semantic search engines beyond text (able to retrieve AV content based not only in the audio transcription or other textual metadata) with the ability to seamlessly search over distributed repositories and be able to aggregate and perform filtering based on characteristics or topics among the results. Another challenge addressed by this work package is the system ability to learn how to perform better searches after explicit and implicit feedback from the user. The expected results of this work package are a service-based networked semantic search engine, able to deal with several ontologies/contexts, visualisation paradigms for result presentation based not only on relevance ranking. Further results are a hybrid mechanism for implicit and explicit user feedback for system ability to improve searches, adapting to task description and context changes.

 

WP4: Task Modelling and Benchmarking Methods

The overall objective of this work package is to make benchmarking of indexing and search tools an integral part of the production workflow by developing novel benchmarking methods based on real-world tasks of user in production rather than laboratory tasks. The aim of these benchmarking tools is not only to assess the fitness of indexing and search tools for certain tasks in the production workflow but also to provide detailed feedback that can be used to train and adapt indexing and search tools to new domains and data sets. The challenges addressed by this work package are the definitions of models that formalise tasks of users in the production workflow and to research how the result quality of a task can be cast into quantifiable metrics, and to calibrate the proposed benchmarking methods to match user expectations of quality. The expected results of this work package are models of relevant user tasks in the production workflow and novel benchmarking methods, as well as large data sets, publicly available for benchmarking purposes beyond the end of the project.

 

WP5: Distributed Repository Frameworks and Data Models

The requirements for a repository in media production systems are different for metadata generation, semantic annotation and linking, task adaptive content enrichment, as well as search and browsing. Hence, the overall objective of this work package is to develop a distributed repository solution in which several specific repositories are linked together and are cooperating to solve sophisticated tasks in the area of distributed audiovisual content processing and archiving. The main goal is to develop an open and scalable repository framework which statically and dynamically combines separately stored audiovisual content files, metadata describing the content, semantic metadata annotations, relations between metadata (ontologies) and content enrichments. Figure 4 depicts an overview about such a distributed repository. Challenges in this work package are the definition of data models for different types of repositories and for the statically and dynamically generated links between individual elements in the various repositories, the definition of appropriate interfaces and protocols for communication and data transport and the performance aspect against the background of real time content processing. The expected result of the work package is a prototype implementation based on a service-oriented architecture.

 

WP6: Integration

This objective of this work package is to form a technical umbrella for TOSCA-MP. In the starting phase of the project usage scenarios will be defined and the requirements will be systematically elicited. In the next step the system architecture will be defined. During the runtime of the project, WP6 is responsible for monitoring technical decisions and outside developments. Based on Service Oriented Architecture (SOA) principles, WP6 will make use of a central Enterprise Service Bus (ESB) within the Metadata Production Management Framework (MPMF) for the actual integration of the components developed in work packages 2 to 5 into a proof-of-concept prototype. System integration is often particularly challenging in large R&D projects; the partners are well-aware of this and through their extensive combined experience on previous projects of this nature they are confident that a fully-integrated system will be produced. The expected results of this work package are usage scenarios, a set of relevant requirements, a high-level description of the system, an integrated prototype system and validation of the overall system with reference to the original scenarios and requirements.

 

WP7: Standardisation, Dissemination, Exploitation

The main objective of this work package is to support activities which help the project to have an impact, i.e. to facilitate the dissemination of projects results, prepare their exploitation and coordinate contributions to standardisation. The challenge for this work package is to ensure that the relevant external stakeholders, fora and research communities are informed about research and results of TOSCA-MP. It aims to ensure that the logistics for internal dissemination are set up from the start of the project, to have high paper acceptance rates in relevant conferences and journals and to offer the consortium sufficient room to perform standardisation efforts. The expected results are multiple research and whitepapers on the (intermediate) outcome of the project, at conferences and in scientific journals. Furthermore, contributions to relevant standardisation bodies are expected. Finally, best practice and knowledge will be shared with the media production and archive communities for the largest possible exploitation of the project results. The end result will be an increased awareness of all project activities both internally, among the project partners, as well as externally, e.g. amongst the industry.

© 2017 TOSCA-MP - Task-Oriented Search and Content Annotation for Media Production
The research leading to the presented results has received funding from the European Union's
Seventh Framework Programme (FP7/2007-2013) under grant agreement n° 287532. - Imprint