Averbis Internship Program
What’s in it for you?
Practical
Experience the Averbis culture and build up practical knowledge and experience in real projects and their challenges
Lead
Discover your strengths as you face challenges that prepare you to make a difference
Learn
Gain real knowledge through practical experience in one of the exciting technology areas of our time
Open Positions
Internship Text Mining / NLP (m/F/D)
Abstract & Purpose
Studies show that doctors spend as much as 6 hours out of an 11-hour work day on documentation1. A substantial part of this time is spent on documenting diagnoses and medications of the patients, among others. In this thesis, we explore the idea to automatically extract these entities of interest from the initial patient-doctor dialogue. To accomplish this goal, current state-of-the-art speech-to-text algorithms should be combined with our text analysis engine that can extract many relevant entities from medical documents. A special focus will be put on dealing with this special question & answering structure that is typically present in a patient-doctor dialogue. The master student is encouraged to try different nature language processing approaches and to combine them with our existing framework.
What you bring along
- You are a student or graduate of a study program (Uni/FH) of computer science, business informatics, mathematics or comparable
- You are enthusiastic about machine learning methods and/or text mining
- You have good knowledge in Python or Java and are familiar with frameworks like TensorFlow, scikit-learn, PyTorch (or similar frameworks)
- German language skills are an advantage
Internship Text Mining / NLP (m/F/D)
Abstract & Purpose
Machine learning models for document classification often yield impressive results, yet their decision are hard to comprehend. This black-box behavior has some major downsides: For practitioners, it may lead to a lack of trust in machine learning models. For text mining engineers, it makes it very hard to improve incorrect classification decisions. In the thesis, the student should experiment with explainability and visualization methods for support vector machines, convolutional neural network and other recent deep learning approaches in the context of a document-classification task. The goal is to identify tokens or groups of tokens that are responsible for the model’s decision.
What you bring along
- You are a student or graduate of a study program (Uni/FH) of computer science, business informatics, mathematics or comparable
- You are enthusiastic about machine learning methods and/or text mining
- You have good knowledge in Python or Java and are familiar with frameworks like TensorFlow, scikit-learn, PyTorch (or similar frameworks)
- German language skills are an advantage