Concept of a Tool Wrapper Infrastructure for Supporting Services in a PLE Tobias Nelkner1 , Wolfgang Reinhardt1 , and Graham Attwell2 1 University of Paderborn, Institute of Computer Science Fuerstenallee 11, 33102 Paderborn Germany, {tobin,wolle}@upb.de 2 Pontydysgu and the Institute for Employment Research, University of Warwick graham10@mac.com Abstract. As one of the most relevant way of learning after apprentice- ship is the informal learning an implementation of a PLE should try to support the learner by mashing up services and tools of every day work, creates cross links between them and gives motivation and support for personal and individual style of learning. This paper presents implementations and ideas for the whole collection of necessary pieces of software to provide a PLE in a bottom up man- ner. A server implementation is introduced which is based on a SOA approach and which includes an extractor for metadata of file objects. This module is furthermore able to run a semantic analysis on unstruc- tured texts which results in for example in high-quality keywords and identification of persons. Taking this as technical background the social functions are explained which are identified as the functions a PLE is supposed to provide more than any knowledge management or e-learning software. Closing, these functions are converted in ideas of possible im- plementations of tools and services, back up by graphical mock-ups. Key words: service orchestration, PLE, knowledge management, learn- ing 1 Introduction The research on personal learning environments (PLEs) is a young field with few active researchers but becomes more and more relevant in context of e-learning. One of the biggest gaps of formalised learning management systems (LMS) is the point of missing support of informal learning. Moreover, it’s mainly insti- tutionalised and set up for vocational training in organisations. This results in missing incentives and motivational barriers of using these systems [5]. The more relevant part of our learning behaviour is learning in an informal and incidental way [7] which is supposed to be supported by a PLE. This kind of learning occurs mainly at work, often motivated by problems occuring at work. Therefore, one of the most important goals is to improve the individual way of learning which improves the overall performance of the organisation. This can be achieved by taking the users preferences, context and social network into account. 2 T. Nelkner, W. Reinhardt, G. Attwell The first peace of software which tried to implement basic aspects of PLEs was Colloquia in the year 2002. Later on for example, the project PLEX [1] came with slightly different solutions of service mashups. This paper tries to give a concept and ideas how an infrastructure for a PLE can look like and which services would help to support the identified social functions of a PLE. A PLE should be seen as concept of an individual customisable set of tools and services not as special application [2]. Therefore this paper doesn’t try to cover all aspects a PLE is supposed to support but provides ideas of ba- sic tools that support learning, knowledge sharing, presenting, and reflecting. The PLE concept of an orchestration of tools and services needs an underlying architecture for realisation. An existing server and a module for analysing un- structured information is basis of the provided tool ideas. Therefore the paper is structured as follows: Section 2 gives an introduction to the Knowledge Server (KNS), the implemented server architecture which manages the orchestration of different services. Section 3 illustrates MetaXsA a tool for metadata extraction and analysis of unstructured information like texts. After that, section 4 intro- duces the idea of splitting expected functionality of PLEs in media functions and social functions followed by the concept of a tool which might cover some of the social functions and services which support informal learning. Concluding this paper finishes with an outlook for future work and some points of interest where the authors see the most needs for research. 2 Architectural concept for a tool wrapper infrastructure Personal Learning Environments are characterised by a rich heterogeneity of ser- vices. A tool wrapper architecture infrastructure to support these services has to offer very flexible ways of interacting with data from different information sources. Current approaches to support workplace learning (c.f. [8, 10]) have to consider the increasing number of enterprise application, communication meth- ods and various information access methods. The productivity and efficiency of an organisation depends on how fast infor- mation can be shared with each other. In large organisational IT infrastructures employees are wasting valuable working time by searching for appropriate infor- mation in the various available information sources. One would wish to have a single access point to all the information objects within the IT infrastructure. Via this Single Point of Information (SPI) one could easily interact with the ex- isting information objects - one could read, manipulate, share or simply reflect them [9]. These typical functions of a PLE have to be supported by a technical substructure, that can easily interact with various different information sources and that provides generic services, usable by a PLE. 2.1 The KNS as exemplary SPI implementation The KNS is an entirely service oriented architecture (SOA), designed to be the technical basis for free to configure knowledge management. Its main principles Concept of a Tool Wrapper Infrastructure for Supporting Services in a PLE 3 are simplicity, extensibility, and configurability, which is the basis for personal learning. As most of the learning at work happens in an informal way [], in most cases the learning material has to be looked up by the learner itself. Further- more, a kind of understanding and of reflection happens by applying the new knowledge. Though, the way of learning is very individual which expects very individual peaces of software to support it. A PLE can be seen as a box of expert tools [4], where each of them is the best for individual tasks. And as the several tasks depend on the users work, the toolbox should be composable by the user itself, not by the software developer. These aspects implicit preconditions on the server side in a technical way and has strong influences in the user inter- face. First the server structure is described in order to summerize afterwards the implementation of the proposed flexibility. To support those principles, we decided to implement an adapter concept which allows to easily exchange information systems or re-configure the existing ones. Figure 1 shows a schematic representation of the KNS architecture. Fig. 1. Schematic representation of the KNS architecture The adapter which connects external back-end systems to the server builds the border between the zone of internal and external communication. The inter- nal communication is based on predefined messages using JMS 3 based commu- nication with the server. The communication in the external zone is back-end specific so that the adapter transform the data from the external communication into the internal one. A workflow engine controls the steps to achieve an aim. The concept is designed to provide the possibility for easily stringing together several steps to achieve the according task. Adding a new external system con- 3 Java Messaging Service 4 T. Nelkner, W. Reinhardt, G. Attwell notes implementing a new adapter and creating a new workflow which has to be deployed into the server. This is an easy way to plug new systems as the server itself doesn’t need to be changed. Due to its modular and strictly service oriented architecture, the generic concept of adapters and a self-adapting central database, the KNS is an ideal foundation for a highly configurable PLE. In fact, the adapter concept reflects the flexibility and configurability on the server side. This provides the technique to mash up knowledge intensive back-end systems and bundle them under one manager with an access for GUIs. This is the basis for providing a deeper con- nection between knowledge objects and persons, without loosing the capability for the user to work with their usual software. Moreover, it gives the possibility to develop tools that highly support the personal learning by using such connec- tions and visualizing them. The availability of a central user directory and of a tool for automatic metadata extraction and semantic analysis let arise possibil- ities of creating thematic and social networks from the data within the system. Therefore, a tool for automatic metadata extraction and semantic analysis of information objects was developed, called MetaXsA. It’s introduced in the next section. 3 Analysis of information objects To support the vision of a Single Point of Information and valuable PLEs the (semi-)automatic extraction of high-quality metadata is one of the main chal- lenges. With MetaXsA we developed a KNS-module that is responsible for the metadata extraction and semantic analysis of any information object within the system. The main MetaXsA service analyses consigned information objects and returns an extended LOM [11] file containing the extracted metadata. MetaXsA stands as abbreviation for Metadata extraction and semantic Analysis and cur- rently consists of two components: the metadata extraction and the semantic analysis. The first component uses several extractors to extract metadata in- formation included in the information object. The semantic analysis comprises semantic modules which investigate the parsed object with respect to seman- tically relevant metadata. The metadata extraction works with three different tools for metadata extraction of files. These tools analyse the files and the results are compared on consistency. In case of meanderings a majority decision is made. If three different results are the output, one of the extractors is choosen, that was evaluated to return usually the most authentic results [12]. The results of this extraction process are saved in the LOM schema and passed to the semantic analysis. The functionality of the semantic analysis is structured via modules. A set of modules forms an information extraction pipeline which can easily be extended by additional modules. Contextualisation of information needs knowledge spaces describing the context. Therefore modules can implement webservices to access knowledge spaces which allow the administration of the module databases by either user or connected databases [3]. Concept of a Tool Wrapper Infrastructure for Supporting Services in a PLE 5 MetaXsA Metadata Semantic Analysis Meeting Minute Extraction Processing Mobile Phone Processing Processing