I want to do some sequence to sequence modelling on source data that looks like this:


with target data that looks like this:

Dont be angry with the process youre going through right now

Both are of indeterminate lengths, and the lengths of target and source data aren't the same. What I'd like to do is have a prediction model where I can input similar numbers and have it generate texts based on the target training data.

I started off doing character level s2s, but the output of the model is too nonsensical even at 2-5k epochs. So I've been looking into word level s2s and NMT, but the tutorials always assume strings of text as the target and source, and I keep running into roadblocks trying to preprocess the text, when all the tutorials assume a certain syntax/set of characters. This is my first try at ML, and some of the tutorials really throw me out with the text preprocessing requirements.

Am I going down the right avenue looking at word level/NMT stuff? And is there a tutorial I've missed for something like what I'm trying to build?


Your Answer

By clicking “Post Your Answer”, you agree to our terms of service, privacy policy and cookie policy

Browse other questions tagged or ask your own question.