Abschlussarbeit mit deecoob Technology GmbH

Visual data labeling using a glyph-based active learning & gamification approach

Master/Bachelor Thesis

Challenge

Labeled data are crucial for training and testing purposes in machine learning based classification processes. The creation of a significant quantity of high quality labeled data samples can be very expensive and time consuming. Depending on the subjects a machine has to learn, it might be necessary to ask experts and to model their decisions. The objective is primarily to design and implement a multi-user active learning application frontend for text mining and machine learning algorithms using gamification elements, in order to motivate people to label text documents. The application design has to consider all aspects of motivating people to carefully read repetitive, small text documents and answer questions on them. Human cognitive abilities should also be taken into account to ensure a maximum of attention during the labeling process. The application is basically designed to support and evaluate inter-rater agreements on the questions asked to the users. A visual representation of labeled and unlabeled data based on glyphs should allow an interactive evaluation of the results.

If you are curious and interested into creating visual interfaces for big data driven applications based on state of the art web technologies, please send your application to:

www.deecoob.com, thomas.thom@deecoob.com
Wilsdruffer Straße 27, 01067 Dresden, Germany
+49 (0) 351 410 5650