I want to build an autocomplete model using RNN where input is article names (documents title).
X: ['Billing', 'Loan status', 'Filling loan application', 'Contact Info', ...]
The article name can be 1-5 word long and there is an even distribution of article names among word count i.e. no. of 1 word articles names is almost same to no. of 2 word article name.
I want my model to start predicting article names based on first 3 typed characters (as I do not want to miss one word article names) and then make predict call based on each next typed characters.
I am not able to decide what should be the target output (y) here while training the model? Should I right shift my input sequence by 3 characters and train the model like below?
PS: Please mind my drawing.