Generated annotations for a 2M song lyrics corpus using the NLP methods I developed in the course of my PhD thesis. Annotations are on different levels such as structure, summary, explicit lyrics, topics, emotions.

[Research paper and code on request]

Comparison of different text classification methods to detect explicit lyrics, i.e. texts containing profane or offensive language.

[Research paper and code on request]

Developed text summarization method for song lyrics. The method takes into account the audio properties of the song lyrics to generate better summaries.

[Research paper and code on request]

Developed text features to improve performance in playlist recommendation in the RecSys 2018 challenge.

[Research paper and code on request]

Implemented convolutional neural network architecture for bimodal text segmentation, using both text and audio of song lyrics. Implementation was done in Tensorflow and resulted in a state-of-the-art performance in the task.

[Research paper and code on request] 

Located in the automotive sector in an R&D team, working on a next-gen natural language dialogue system for in-car use (clients e.g. Audi and Porsche).

• facilitation of compositional semantic parsing based on data-driven dependency parsing
• compositional natural language generation, and the NLP pipeline in general
• Implementations were mostly performed in Java.