ADS Research Assistant Vacancy: Multimodal Evidence of Concept Drift

We are looking for a highly motivated research assistant who is working towards a Master in Computer Science, AI or Information Science. The ideal candidate has a strong software development background, knowledge of NLP and Linked Data technologies and a keen interest in deep learning.

You will work under the supervision of Desmond Elliott (ILLC, UvA) and Laura Hollink (CWI) to develop methods that combine evidence from text, structured data and images for concept drift detection. For example, images of Hillary Clinton as the First Lady may depict her engaging in humanitarian issues, compared to her current visual representation: giving speeches behind podiums, wearing strong suits, and meeting hundreds of people at rallies.

You will use deep neural networks to learn features over multiple modalities, based on the hypothesis that a joint model of textual, structural and visual concepts will predict better concepts than a unimodal or a bimodal model.

Desirable experience:

• NLP or Computer Vision background (and willingness to learn about the other topic.)
• Experience with Theano / Torch programming for Deep Learning
• Web development experience with Django / Flask
• Excellent writing skills

Send applications to d.elliott@uva.nl with the subject “ADS 2016 Application”. Include a statement of motivation, your CV, and your academic transcript.