Following transformer architecture to develop an intent detection and slot filling model. The shared encoder includes a convolution layer to better extract embedding within window context and the co-interactive module as joint learning for intent detection and slot filling tasks
