The team project is a pilot project aiming to estimate all possible resources needed to train a neural machine translation (NMT) for Chinese government work reports from Chinese to English. We named our team as “Forerunners” and assumed that we were a company helping the Chinese government to identify whether they would like to use a machine translation for their work report.
First, we made a temporary proposal to clarify our objectives, the method we use, and timeline.
When we actually started to train the machine, we realized that we had to make many adjustments. Problem-solving and working with my teammates, I was able to learn how to use Microsoft Custom Translator and Systran, how to read the BLEU score, why quality assessment is necessary and how to find the optimal assessment method for different projects. Here is the updated proposal we created after one month of training the machine.
We have summarized our lesson learned in the video as follows: