| Home News Links Groups Projects Registration | Educational Technologies WS 06/071. Visualisation of the pedagogical knowledge domainActiveMath's course generator uses the HTN (Hierarchical Task Network) Planner JShop2. This planner uses a desciption of the domain he is planning on. In the case of ActiveMath, this domain description is like "A guided tour usually starts with an introduction. An introduction looks like ... After that, an example and the main definition follows. At some point, the student is provided with a concept map exercise in order to get an overview of the topic. And so on ..." As the notion 'hierarchical' and 'network' promise, the domain of jshop is somehow structured. The student's task is to implement a tool which reads in the jshop-domain-description, analyses the structure (hierarchy, network, etc.) of tasks, and shows a graph representing the knowledge of the planner about the pedagogical domain. A (well documented) prototype is provided by Philipp which can serve as a basis. Information about jshop2 can be found on the SHOP-website or in the detailed JSHOP2-documentation.
Supervisor: Carsten Ullrich
2. Generate problem space in SLOPERT and transform to an exerciseLet SLOPERT explore the Problem Space including misconceptions. Take the result and transform it into ActiveMath's exercise encoding format.
Supervisor: Claus Zinn
3. Learner Model for iCMapAlthough iCMap issues events about user actions there is no user model that evaluates those events. This project suggests to implement a basic user model for iCMap:
Supervisor: Arndt Faulhaber
4. Interactive e-portfolio viewerBased on the IMS e-Portfolio specification IMS e-Portfolio specification develop a viewer that is able to read and export an e-portfolio description. An interactive e-portfolio allows to alter the portfolio by adding, deleting, or modifying single items. To get the IMS eP classes, apply the IMS eP specification XSD description to JaxMe. As a start, a tree view is sufficient. An optimal candidate is one with Eclipse RCP experience. A bachelor/master thesis is highly wished as a follow up activity.
Supervisor: Martin Homik
5. Generating Parametrized ExercisesActiveMath Exercise Susbsystem has defined an architecture for enhancing manually authored exercises on fly, using so-called generators. A generator is a java programm, that is transforming the graph of interactions of an exercise in some way in the process of delivery. One of the examples of such generators is a "Randomizer". This generator allows for authoring parametrized exercises. Authors create an abstract exercise, using variables for the problem statement and solution and provide the set of possible values for these variables. The Randomizer in each instance of an exercise is substituting the variables with random values from the given value set. Until now, Randomizer was only working with given finite sets of possible values. The current task is aiming to enhance the Randomizer to randomize over any (possibly infinite) intevals, discrete or continous. Additionally to that, it should be enhanced to support randomizing over sets of elementary functions and their compositions. Additional step is to make this process adaptive, i.e. only the functions that the learner already knows about, will be considered. Information about the prior knowledge of the learner has to be exctracted from the learner model.
Supervisor: George Goguadze
6. Domain ViewerBeim Autorieren wird Euch auffallen, wie mühselig es ist, die Beziehungen zwischen den einzelnen Lernobjekten herauszufinden. Es wäre doch so schön, wenn man das mit einem Blick sehen könnte. Solch ein Tool fehlt uns und Ihr sollt es schreiben. Im Prinzip sind das zwei Schritte:
Supervisor: Martin Homik
7. Mathematical Rendering TesterThe world wide web is based on a relatively mild description of the content which, when clean, may leave much freedom to the browser for rendering. This freedom is visible in HTML and CSS but also in MathML, an encoding of formulae where the ``reading intesity`` is much bigger than normal text. We wish to provide a web-based tool that takes a set of MathML expression, lets them be displayed (big) by the browser, and takes screenshots of them, then uploads these screenshots (properly cut) into a storage on the server which authors or implementors can use for quality assessment of rendering implementations and of content. If things go well, this should be coupled with an automatic recognition tool from the University of Birmingham. Technologies and could be hosted at the W3C Math web-pages:
7.bis. MathML test suite runnerThe world wide web is based on a relatively mild description of the content which, when clean, may leave much freedom to the browser for rendering. This freedom is visible in HTML and CSS but also in MathML, an encoding of formulae where the ``reading intesity`` is much bigger than normal text. To test players of MathML, design-science and Wolfram Research have developed a test-suite that can be found at the W3C Math site. The test-suite is presented on the web with ideal renderinges. We wish to provide a runnable tool, hopefully web-startable, that recreates the pictures by rendering each of the elements of the test-suite on any given client and stores it, integrating them into another copy of the test-suite which can be used to compare, visually, the renderings of the given client, a browser, to the reference rendering. If everything goes well, this is integrated into a web-start-and-publish framework.
8. Analyzing Online, Collaborative DataAn important area of e-Learning is "educational data mining" or the process of using Machine Learning techniques to learn student behavior patterns and then to use that data subsequently to help students learn in an e-Learning system. One of the course lectures will focus on some of these Machine Learning techniques. A great advantage of distance and e-Learning is the abundance of data that is generated for empirical analysis, but it is only in recent years that researchers and practitioners have begun to harness the power of this data to help in the process of e-Learning. On this project, you will generate Machine Learning classifiers from log data for an actual online, collaborative system called Argunaut using a software environment called YALE. You will also write the code that integrates the resultant machine learning classifiers with a demonstration version of the Argunaut system. Finally, you'll write some code to help in analyzing the inter-rater reliability of data that has been annotated by experts. These software components will give you an overview of the important pieces of an educational data mining system. The system and integration you develop will help moderators intervene in and support collaborative discussions between children in subjects ranging from social issues to philosophy.
Supervisor: Bruce McLaren
9. Generating Exercises Interacting with Domain ReasonerActiveMath Exercise Susbsystem has defined an architecture for automatically generating exercises, using so-called generators. A generator is a java programm, that is able to generate each step of an exercise. In this task, the exercise system will communicate with an external Domain Reasoner for providing diagnosis on learner's action and generate the consequent exercise step, based upon this diagnosis information.
Supervisor: George Goguadze
|
|
Webmaster, |