Friday, March 30, 2007

Group Meeting March 30, 2007

presenter: Sergey Sosnovsky
topic: Ontological Technologies for User Modeling

once upon the time....while ontologies meet user modelling

firstly, Sergey briefly described the definition/design/history of user modelling and introduce the most frequent used technologies in user modelling.
he mentioned that ontoligies may specifically help in representation, elicitation, application technologies.
secondly, Sergey explained ontologies in semantic web, light-weight/heavy-weight ontologies and multi domain ontologies.

so, how ontologies can help user modelling???
a. ontology overlay UM
b. perosonal ontology views, including subset of concepts from the domain ontology, part of the domain ontology and individual network of concepts
he also brought up the discussion regarding to share ontologies and gave examples
1. OWL-OLM system visualizes the personal ontology view
2. Foxtrot system visualize the sets of concepts
c. ontologies for UM mining

Experiment Design:
a student has some c knowledge and then shift to learn java. there are some share concepts between these two. the suggestion is to use ontologies mapping. use GLUE algorism to calculate the relevancy between C and java. (due to time limit, Sergey didn't drill into this section, please refer to ppt file from paws website)
the point is to find out is it better in the transition from C to java manually than automatically?
1. why not try pairwised comparison??
2. this is the case of C and java, they have similar case, but how to make it more generalizable?
3. if the domain is the same, how to make the ontology mapping much more accurately?

Monday, March 26, 2007

Group Meeting March 23, 2007

First Presenter: Michael Yudelson
Topic: Digital Object Repositories centered on FEDORA

Michael presented digital object repositories centered on FEDORA. The digital object can have data stream and data stream has its own disseminator that operates on contents type. There was a debate about defining digital object. In our systems, specific quiz can be a digital object and the function of disseminator is to create and display contents. He also introduced FEDORA and it makes possible to store and discover by indexing digital objects.

In terms of collaborating and managing learning contents, FEDORA can be helpful.
Beyond Fedora how we can all objects with navigation support and handle digital object user-adaptively is our problem to solve.

Second Presenter: Jaewook Ahn
Topic: Personalized search and visualization for YourNews

Jaewook presented some newly added functions in his YourNews system. The system has short-term and long-term user profile and generates user profile by observing user’s actual view. It adds top five keywords in entire user profile to every query and Rosta pointed that there is some possibility to lose user’s intention by disrupting their point of view. Dr. Brusilovsky added that query expansion is good technique but we should be careful when apply it to system. In long-term model, the range of focus can be sparse than short-term model, hence it can be meaningful to cluster data in long-term model into meaningful group.

Sunday, March 18, 2007

Group Meeting March 16, 2007

Presentation 1: Danielle

Title: Learning SQL with a computerized Tutor

Danielle presented projects related to SQL tutorial which is a new project for several members of the group. The main goal of the project is developing a personalized SQL tutorial using different techniques from adaptive hypermedia.

Danielle presented an existing SQL tutor which is developed by University of Canterbury which is designed to address students’ problems regarding to learning SQL such as problem to memorize database schemas, or misconceptions in students' understanding of the elements of SQL. The tutor tries to provide useful hints and helps when students make any error regarding a specific SQL command instead of general messages that are hard for students to understand.

Danielle provided a demo of the SQL tutorial from University of Canterbury in which students get a SQL question and have to fill out empty boxes. While working on the problem, students can ask for different kind of hints and helps.

We had a discussion about the difference of “bug model” and model of students’ evaluation. Basically, bug model is the model of students’ misconceptions while evaluation model deals with students’ evaluations of their responses to a problem.

Peter also described briefly about the new SQL project which will be similar to our C project but for Databases. It means the system will have different components such as quizzes, examples, and script based evaluation. It also includes components developed in different places and the main challenge of the project is a meaningful integration of the components and a design of a centralized user model that communicates with all the components. So in this SQL system, the teacher will be able to structure the course and assigns activities to each lecture. Every link in the system will provide guidance and the guidance can be social or knowledge based.

Presentation 2: Sharon

Sharon presented her example authoring system. The system is a community based example authoring in a way that the teacher puts the example into the system but then every line of the example will get annotated by the students. Students are doing the annotations as an educational activity. They are also able to criticize annotations of other students. In the current design, once a student start annotating an example, it will be locked for that student and other student can only rate the annotations. This might be changed in later design.

Sharon is planning to finish the implementation as soon as possible to run the first study which will be done in a classroom with 9 students. The study will allow half of the examples to go through two rounds of annotation and ratings and the other half through only one round of annotations and ratings. At the end an expert will evaluate all the annotations and ratings. This part is only for the purpose of the study and evaluation.

In longer term the project will let use to provide guidance on open core examples since the content of the examples are not enough for providing the guidance but added annotation can help to provide guidance or even provide better search for the examples.

Friday, March 02, 2007

PAWS Meeting 2007-03-02

1. English for Dummies: Searching for Rules versus Exploiting Examples

Presented by Angela Brunstein

In our recent study on the Chemnitz Internet Grammar, we replicated and extended our earlier results on that system by presenting one and the same version of the grammar system to all students and by unifying tests and instructions. Altogether 200 students processed two chapters of the grammar on Present Continuous and Present Perfect for 30 minutes. They either searched for specific contents in the grammar or explored the grammar on their own. Before and after processing the grammar, they answered factual questions and mastered application tests. Corresponding with literature and our earlier results, students with the specific search task answered questions related to the contents they had searched for more in detail. In contrast, students with the unspecific exploration task answered more questions but less in detail. More interestingly, both groups differed remarkably when applying their knowledge. When learning by examples, students who explored the grammar on their own improved their skills more than students who had searched for specific contents. This effect can also be found in tendency when learning by rules and examples, but not for learning by rules only. In that case, a guiding instruction highlighting most relevant facts in the grammar seems to enhance learning.

2. PEBL: Web Page Classification without Negative Examples

Presented by Chirayu Wongchokprasitti

Web page classification is one of the essential techniques for Web mining because classifying Web pages of an interesting class is often the first step of mining the Web. However, constructing a classifier for an interesting class requires laborious pre- processing such as collecting positive and negative training examples. For instance, in order to construct a “homepage” classifier, one needs to collect a sample of homepages (positive examples) and a sample of non homepages (negative examples). In particular, collecting negative training examples requires arduous work and caution to avoid bias. This paper presents a framework, called Positive Example Based Learning (PEBL), for Web page classification which eliminates the need for manually collecting negative training examples in preprocessing. The PEBL framework applies an algorithm, called Mapping-Convergence (M-C), to achieve high classification accuracy (with positive and unlabeled data) as high as that of a traditional SVM (with positive and negative data). M-C runs in two stages: the mapping stage and convergence stage. In the mapping stage, the algorithm uses a weak classifier that draws an initial approximation of “strong” negative data. Based on the initial approximation, the convergence stage iteratively runs an internal classifier (e.g., SVM) which maximizes margins to progressively improve the approximation of negative data. Thus, the class boundary eventually converges to the true boundary of the positive class in the feature space. We present the M-C algorithm with supporting theoretical and experimental justifications. Our experiments show that, given the same set of positive examples, the M-C algorithm outperforms one-class SVMs, and it is almost as accurate as the traditional SVMs.

3. Discussion of job-recommendation system Proactve

presented by Daniela Hyunsook Lee