Towards cross media document annotation

The research work presented in this thesis focuses on manual and semi automatic methods for achieving semantic (single and cross media) document annotation. By adding "semantics" we mean explicitly specifying a meaning for the content being annotated and introducing a structure to the document based upon a shared common set of concepts and their relationships which conceptualize the domain in which is of interest. The motivation for this research work has been derived after observing the pitfalls in current practices for document lifecycle management and then drafting the requirements for the need to annotate (both in single and across media) documents. Firstly we identify the need for semantic document enrichment within the use cases where we applied our research. This is after observing the problems encountered by knowledge workers while following current work practises for document creation, storage and retrieval i.e. the document lifecycle. The research work carried out during the course of the thesis tries to address these pitfalls by introducing semantic annotation as a means to enrich the document during its lifecycle. Semantic enrichment of documents enables computer programs to comprehend the meaning of the content being annotated. This is possible due to the emergence of the next generation of the web called the Semantic Web. The Semantic Web the web acts as a universal medium for exchange of data. The standards and technologies developed as part of the Semantic Web research enable us to create linked data from structure sources with an underlying semantic meaning associated to them. In this thesis we introduce Semantic Annotation from a Semantic Web perspective i.e. using the core set of standards and technologies for enabling the Semantic Web. Then we identify the need for semantically annotating documents across media. Web documents consist of resources collated from various media (images, text etc) which complement each other in the given context. Annotating documents in single media does not take the context of these collated resources into account. For example in a web page the textual material could be used to complement the image which is being shown. At this stage of the research we will define what cross media annotation means, what its requirements are and how it will address the pitfalls in the practices followed in the use cases for managing a documents lifecycle. Further we will detail the enhancements cross media annotation provides over annotation in single media. After the definition stage we dwell into the technical requirements for creating a cross media annotation framework, how the complexity of the annotation process could be eased through the use of semi automated and socio collaborative approaches. Next we show how we have addressed the requirements identified during the definition stage and present the results achieved after a detailed user and statistical evaluation of the cross media framework (AKTive Media) which we have developed. The primary research questions we address in thesis deal with defining what cross media annotation is, what the requirements for cross media annotation (CM) are, identifying the technical requirements for creating a CM framework , specifying how the process of C;M annotation can be supported and the annotation complexity reduced. The dissemination and exploitation efforts to end users and the impact our research has made also presented in detail during the course of this thesis. Finally we describe how our research in knowledge acquisition using cross media annotation strategies has evolved into a new prototype for acquiring knowledge with the help of automated web forms which help knowledge workers in normal data entry operations while implicitly populating semantic knowledge bases without explicit effort from the user. This forms based system has been deployed into the use cases and also formed the basis for the formation of a spin out start up company.EThOS - Electronic Theses Online ServiceGBUnited Kingdo