Simple Text Classification with CNN in PyTorch
Three layers Convolution Neural Netowrk (CNN) is being trainied from scratch for a set of more than 2000 text data in question/label pairs format. The short question text is being classified in one of six possible labels and the problem is simple enough no advance model, e.g. Transformer, nor RNN is required.
The resulting model is evaluated against a hold-out test set and achieved 89% accuracy.