THESEUS Basic Technologies
The THESEUS research programme represents an important contribution to the creation of a new internet-based knowledge infrastructure that means faster and more effective online knowledge processing in future. One of the main subjects being researched is the development of semantic technologies that are capable of recognising and classifying the content and meaning of information (words, pictures or sounds). These technologies allow ‘smart’ computer programmes to recognise and replicate the context in which data has been stored. In addition, computers can also draw logical conclusions from content by applying rules and classification principles, and subsequently recognise and construct links between various items of information from diverse sources.
THESEUS is an umbrella organisation that unites researchers from the public and industry sectors, who are now collaborating to develop and design innovative basic technologies and technical standards. Consortium members from the field of industry will develop prototypes of the new technologies and test them in six application scenarios. The purpose of the tests is to find short-term ways of converting the new technologies into innovative tools, commercially-viable services and potentially profitable business models for the World Wide Web and other internet-based networks.
Some of the basic technologies being developed by the research partners are functions for the automatic creation of metadata for audio, video, 2D and 3D picture files and their application-specific combination, as well as mechanisms for the semantic processing of multimedia documents and their associated services. Another main focus of the research is the development of tools for the management of ontology-supported knowledge representation. The consortium is also working on new machine learning algorithms, situation-sensitive smart dialogue processing systems, and innovative user interfaces. New DRM (Digital Rights Management) processes to provide the holders of intellectual property and marketing rights to multimedia content with better protection in future will also be developed.
a) Automatic generation of metadata for multimedia files
The research partners are looking for ways of registering content and creating interoperable metadata for audio-visual data, with the aim of improving the indexing, archiving and accessing of multimedia sources. The main focus here is on semantic indexing and annotation processes, automatic metadata generation and context-based video recognition, and the target is to integrate semantic information in various forms (text, speech, photographs, audio, video etc.) into an object group with higher information quality. Metadata combination is application-specific i.e. the system provides an optimal and individual combination that is able to make a specific search of the source databank statistics. Adaptive, content-based processes for the efficient compression of video signals, offering users who want to access image and video information a higher level of efficiency, are also in the pipeline.
b) Semantic processing of multimedia documents
The demand for reasonable response times when searching through highly complex multimedia databanks has prompted the development of particularly efficient indexing and retrieval algorithms for metadata; the new algorithms will make content-based image and video searches in databanks with hundreds of thousands of objects (pictures, video shots, video clips) possible.
c) Ontologies
The THESEUS research partners are engaged in the development of integrated methods and innovative tools for the management of ontologies in software applications with extremely high quantities of stored resources. The object is to cover all the phases normally associated with ontology creation – from design right down to mapping with other interoperable ontologies or automated logical deduction (reasoning).
d) Machine learning
The research partners are working on scalable technology for shaping and structuring ontologies. In the course of this work, they will utilise innovative processes based on machine learning principles; these will help them to discover logical-deterministic and probabilistic relationships in text, audio, image and video data. This, in turn, will allow statistical Machine Learning methods to be developed that will make the extraction of knowledge from relational data possible; this data can then be used to create or extend existing ontologies. Machine Learning methods are also being used in the development of a system for the automatic annotation of text documents. The semantic information is stored in efficient Rich Structures, enabling optimal searching and clustering at a later stage. The discovery and use of Web Services are supported by making use of text annotation technology in their descriptions.
e) Situation-sensitive dialogue processing
Research into flexible components for dialogue processing – components that can be specially configured to suit different application scenarios - is in progress. Basically, these components provide functional modules for the creation of systems that enable humans and machines to interact in dialogue form. The research partners are developing analysis and generation methods and algorithms that react specifically to texts, graphics, speech and gestures; these will be used for what are known as multimodal interfaces, which are controlled by gestures and human speech. Users can phrase their queries intuitively and refine them in the course of a free-format dialogue with the system, whose linguistic performance is modelled using a discourse ontology that describes standard dialogue rules and behaviour. A special component called a Semantic Mediator provides the interface between the representations of the multimodal user entry sources on the one hand, and the various metadata sources on the other. This is a highly efficient means of determining the correct semantic data source for an enquiry. One of the requirements when incorporating these semantic data sources is a set of methods for the semantic annotation of any Web Services that must be consulted or accessed interactively.
This multimodal interaction allows the user to navigate semantically and provides support in the form of frequent suggestions or associative remarks. The creation of user enquiries and questions is made even more efficient with the additional support offered by tutorials and other help functions.
f) Innovative user interfaces
Another aim of the THESEUS programme is to create GUIs that allow easy and immediate recognition of the relationship between different data, metadata and documents, which enables them to be handled intuitively. In THESEUS, the results of search requests will be displayed in what are known as ‘knowledge networks’, which show not only the relationship between results and the keyword used to search for them, but also between the results themselves. These knowledge networks enable users to search more intuitively, provide a broader view of the entire subject range and help to locate the required information more quickly.
The user-friendliness of THESEUS methods will be enhanced even further by the creation of a harmonised framework that gives all THESEUS developments a similar ‘look’. Users will not have to waste time and effort familiarising themselves with the system again every time they use it, but will find that all results have an identical structure and similar functions. Another advantage of this framework is that the user interface can be adapted to suit a variety of display options. Thanks to ‘Triple Play’ technology, THESEUS developments can be viewed on mobile phones, monitors and TV screens; display quality is always of the same high standard, and optimal use is made of the respective devices and their technical possibilities.
The researchers are not only concerned with the end consumer needs, but also want to create conditions that will make technology, developed in THESEUS, available to corporate users who want to offer fast and easy access to their applications. With this in mind, a framework is being constructed that will enable corporate users to adapt this technology to their own requirements with a minimum of effort, instead of having to create new products. This will mean that, in future, companies will be able to save time and money, and also improve the quality of products by partially or completely ‘recycling’ them.
The main emphasis in all the research is on the processing of knowledge. In view of this, work is in progress to design models for data or metadata entry that can be adapted to a large variety of application scenarios with a minimum of fuss. Front-end technology provides visual tools for the creation and editing of ontology models and instances. ‘Roles’ can be defined to regulate access rights to the tools e.g. for the creation of metadata templates. The interface design will be particularly intuitive, ensuring that very little information will need to be entered in order to edit data.
g) Evaluation
The THESEUS research programme also involves the design of methods and tools, and the compilation and administration of data, required for testing and evaluating the technological innovations it will produce. The aim is to develop new test methods for image and speech analysis technology, and to measure sound and picture quality (after data has been encrypted and compressed, for instance). Special attention will also be paid to the evaluation of usability and information presentation, user-oriented design and accessibility.
Downloads