Luise - DREU Research Intern
This week I got a lot of different things done. As expected, we needed more data to improve our model, so I spent a good deal of the week getting a better idea of how this new data is structured, and how I can get it in the right format. Since we are going to use both the new and the old data, and need to get them structured in the right way and merge it all. This is a bit more complicated than the previous data formatting, because I no longer have example scripts where I can find inspiration. It’s a cool challenge though. I need to have a really good overview of the data and a great understanding of what Kaldi (the toolkit we use for the training) requires.
Writing the scripts is a fun process. I start out trying to get an understanding of the bigger picture, then I think I figure it out and know what to do. I dive into the coding, write a bunch, and then perhaps I get to run it right away, or maybe I realize that the files are not as I thought they were, that I need different data, or maybe that the two datasets are structured differently (e.g. use different phonetic alphabets). And then I need to rethink my approach. And this goes on for a while. It might sound tedious, but I actually think that it is a lot of fun. It is almost like a puzzle, and I get super caught up in challenges like this.
Most of the work I have done previous weeks has been about eventually training models, so that the computer will be able to transcribe audio files - speech to text, that is. This week I also got to play a bit with Festival, a piece of software that can work the other way around - namely, it will synthesize speech based on a given piece of text. I am going to work more with this later, but for now, I had fun doing one of their tutorials where I created a small “clock” where for any given time, my voice would say something like “the time is now almost half past five in the afternoon” (if I typed 17:28 for instance).
Besides from work, this week was characterized by soccer and good food. I went out for delicious Mexican food in East Village, and today, Sunday, two of my roommates and I went to Queens to have what is supposed to be some of the best Colombian food in NYC. I have to admit that it really was delicious - and my Colombian roommate agreed which is probably a better signifier. The meal was the perfect way to get in the right mood for the Copa America game between Colombia and Peru which we subsequently went to watch at a busy Mexican bar nearby. The women’s soccer World Cup is also happening, and Tuesday, another friend and I went to see Sweden playing at a bar near Union Square. When my fellow Danes didn’t make the World Cup, I guess I have to root for our neighboring countrymen :)