March 29, 2017, by Lisa Chin
This post is written by Liu Haixia, PhD candidate from the School of Computer Science.
It all started by googling: natural language processing internship and the Ubiquitous Knowledge Processing (UKP) Lab retrieved as the top one option. I applied, got a chance of test and was offered a six months research intern position at UKP.
The research intern project: Automatic Image Captioning started in August 2016. This project covered computer vision and natural language processing which gave me the opportunity to explore big data and deep learning. I would like to share with you my experience from the following perspectives:
- Cultural diversity and unity. When I joined UKP, there were seven research interns from seven different countries but all of us hold similar research dreams.
- Academic atmosphere. UKP is a professional research lab and the supervision team is well organised. For example, the Head of the UKP supervises everyone. The PhD researchers usually supervise the masters’ research projects which could be their sub-PhD projects. Every Tuesday, there is a seminar at UKP where researchers or masters students report on their progress. I was able to join DeepDay seminar at UKP – an intense deep learning presentations. The academic atmosphere is everywhere, even during the PhD defense celebration and the Christmas party.
- Hardware facilities. Since the project Automatic Image Captioning dealing with big data and deep learning, it is necessary to access the high performance computer (HPC) and graphics processing unit (GPU). It is possible to access the clusters with high performance (e.g. 4x Sandy-Bridge, 32 cores [AVX]). The HPC management team and the TUD Computer Vision Station admin team are very efficient and helpful.
Overall, it was a great experience at UKP where I met professional researchers and gained deep learning skills. My suggestions for the PhD candidates who intend to do an intern outside of our university are: First, it is important to set the internship research goal very clearly before the project starts and evaluate the doability. After all, the internship is a short term project which should have a well-defined hypothesis or short term vision that leads you through. Second, it is better to do an internship project that is closely related to your PhD topic. Last but not least, do not be afraid to pick up challenging projects. You will be rewarded by the hardships you have gone through.
Acknowledgement for the support from: UKP members, my PhD supervisors, Crops For the Future (CFF), UNMC Graduate School, my BB79 fellows at UNMC, Lee Ai-Suan who gave me a German language training and not to forget, my flatmate at Darmstadt.