September 01, 2017 – ///,
Crowdsourcing platforms enable employers to request certain tasks to be done by unknown employees over the internet. To enhance the assignment of tasks to employees with respective to different criteria, recommendation systems are deployed on such platforms.
Deep Learning is a recent trend especially for Natural Language Processing. Tasks on platforms like Micro-Task Markets or Job-Search Engines come with title and textual descriptions to describe their requirements. With new techniques such as Word Embeddings we want to investigate how the recommendation of tasks to workers can be enhanced fundamentally.
The following tasks should be covered by the thesis:
1. Literature research: Which approaches for recommendation systems exist for crowdsourcing and related areas?
2. Use-Case analysis for crowdsourcing platforms:
a. Where is a recommendation reasonable?
b. What kind of recommendations can be applied?
c. What kind of data can be used for the calculation of the recommendation?
3. Design and Implementation of a recommendation system based on insights from (1) and (2).
4. Critical reflection of the created concept and comparison to existing approaches.
· Creativity and an independent way of working.
· Structured and scientific approach to the subject.
· Previos experience in Machine Learning
· Programming capabilities in Java, Python or R
Duration: 3-6 month (depending on the kind of the thesis)
Tutor: Steffen Schnitzer,