Research InterestsPrototyping of Interactive Systems and DevicesMulti-modal InteractionFabricationInternet of ThingsHuman Computer InteractionRoboticsChatbots
Open Thesis Topics
Voice assistants are ubiquitous. Whether it's Alexa, Siri or the Google Assistant, they are all widespread and are now used by many daily. These assistants are being used in an increasing number of fields and play a major role in research. However, most of them depend on a cloud-based service to perform the NLU (Natural Language Understanding), and you have little control over them, also in terms of data protection. There are many (open source) systems for speech-to-text (e.g. DeepSpeech(2), Vosk) and intent recognition on text (e.g. RASA). The goal of this thesis is to implement one of these local systems as an alternative path to an existing cloud-based NLU, and to perform a qualitative comparison.
In our group, we already use Cerence Studio, an intent recogniser for spoken language that runs in the cloud. Your system should be compatible with the training format of Cerence Studio (or build your training format that can be exported to Cerence Studio). For a good result, the text-to-speech and intention recognisers component must work together. You would then integrate your offline system into our existing dialogue platform stepDP.
The focus will be on the concept and implementation - so you should have good programming skills. For a Master’s Thesis, there will be an additional focus on the qualitative comparison of the two NLU services. The result is to be evaluated later in a suitable way (in a Master's thesis in the form of a user study).
- Good programming skills in JAVA or C#
- Understanding of basic speech processing concepts
- Ideal: Previous experience in existing (open source) solutions in the field of speech recognition and NLU (intention detection)
- Basic understanding of machine learning
How to apply
Please send an application to Julian Wolter containing the following information. (if you do not answer every point, your application will not be considered):
- When you plan to start the thesis
- When you plan to finish the thesis
- A short motivational statement (max 0.5 pages) why this topic is interesting for you
- A brief summary of your prior knowledge
- Your transcript of records and your CV