Hi folks,

some time ago I did some research about how to classify dialogs with recurrent neural networks (GRUs, LSTM) and also 1DConvolutions. I found an architecture, which is quite new and worked pretty good on a data, which is not public. I wanted to publish it as a paper, but I think it will not happen anymore. First there is no appropiate public data set available for this task and second I just don’t have enough time at the moment.

Anyway I published details on my blog, if you are interested, take a look here.

Cheers,