I'm searching a method to classify a text (5 classes) with a little corpus and which is domain specific (in pharmaceutical area). I tried some models on python like random forest or multinomial naive bayes (i had some good results like an accuracy of 0.75) but i saw the models in deep learning were performant in the domain of classification. I tried a multi-layer perceptron with 2 hiddens layers but i'm not familiar in this domain. The results were equivalent with those i tried (accuracy of 0.72).
Do you have any idea which model i can try in deep learning considering my corpus (not a lot of data and domain specific)?
I have like 450 obervations in english with 35 tokens in average.
