I hadn’t realized that it has been nearly a year since my last post. It is wild how fast time can fly by.
I am starting to expand a project I began the summer before last. The goal of this project was to use a convolutional neural network to explore differences and similarities between various languages.
While I found some really cool patterns in the hidden layer activations across the different languages, a major pitfall of that the initial project was the limited nature of the training and test sets. I only had a single speaker for each language. So really, the network was performing talker recognition (with excellent performance I might add).
The goal of the current project is the same as before, but with an expanded dataset, one that includes multiple speakers and dialects of the tested languages. In addition, I have also included more languages to further expand the diversity of the dataset.
I will be blogging my progress on this project on this site, while maintaining an active development repository on my GitHub page, most likely under LangNet-2.0, though I haven’t really settled on a name.