Luise - DREU Research Intern
This was a short week for me work-wise. My brother who studies back in California is graduating this weekend, so I am leaving for SFO early Thursday morning. Even though I made the week a few days shorter, I still had plenty of time to be productive in the lab! As I mentioned last week, I finished up processing the Egyptian data set Friday, so I could start working with the English data when I showed up Monday morning. Basically, when the scripts are written, which they are by now, most of the work is left to the computer. These scripts have to process a lot of data, so even with our powerful machines, it takes quite a long time to finish the training. So while the machine was training, aligning, decoding and all the other things it does to our data, I had time to look around online and learn more about speech recognition. After some hands on experience working with the data, the many online research articles already seem more approachable - even though that of course doesn’t mean that I understand everything… yet :) After finishing up the training, it seems like I might need to find more data to improve the performance. The audio we use is from phone conversations, and because the sound quality isn’t that great, we might have to make up for that by using more data - that would be either existing datasets or data found on blogs, forums and other places on the web. Maybe I’ll look at that next next week. But for now, I’m off to California!