cdipaolo
Train on new model
Hello! I have read the closed issue about Spanish language, and I think I go idea how to retrain the model, but I am a little not sure
- I should create my dataset/train folder with pos/neg and place the data there -- that clear
- Then I create my_en.go file which walks that dataset I just created
And then I am not sure what I should do to make it Train, and save trained data and use it for analyzes?
Hi Mikhail,
Yes the easiest thing to do would be to copy the en.go procedure for training an english model in a Spanish es.go file. That function would train the model and load it into memory, but ideally you would train both the English and Spanish models in memory, then persist the models map to disk into model.json, before using go-bindata as we already do to persist the file into the library (you'd replace the current bindata.go file in the library, committing the new version). Theoretically, everything should work from there but you'd want to add tests to check.
