November 04, 2021 at 4 pm, the Bachelor Thesis Defence: "QualityInsight: Explanation Interface für das articlequality Modell von ORES" by Tanita Daniel will take place.

Abstract

ORES is a service that provides machine learning models for Wikipedia projects. These models are intended to help maintain a certain quality standard on Wikipedia pages and reduce the workload of Wikipedia staff. One of these models is the articlequality model, which automatically determines the quality of articles according to the grades and criteria of the respective Wikipedia.

Predictions of the articlequality model are difficult to understand without technical knowledge and machine learning experience. Therefore, they are insufficient especially for editors or other inexperienced users of Wikipedia in its original form.

The goal of this work was to create an application with a human-centered design that, with the help of visualizations and explanations, should make it possible to present the predictions of the articlequality model in a comprehensible and useful way.

To create such an application, an interface prototype was first created, which was tested in several iterations by potential users. The results from the usability tests were used to improve the prototype before the next test to end up with an intuitive design. The final prototype was then used as a template for implementing the application.

The evaluation of the created application took place via asynchronous usability tests. These revealed that while the application made the predictions interpretable and usable well, it did not yet really help in knowing what exactly needed to be improved on an article.

First assessor: Prof. Dr. Claudia Müller-Birn
Second assessor: Prof. Dr. Lutz Prechelt

Location: Takustr. 9, Raum 055
Date: November 4, 2021 at 4 pm