Which is the most efficient framework for Semantic Analysis in Machine Learning?

My product is made in Python and I need Semantic Analysis for classification of sentences into questions, complaints, etc. Which is the best framework for the same?

I think the best approach would be, try several methods, cross-validated each method using a separate validation data set (or using K-Fold cross-validation) and pick the best one.

So as a starting point you can try:

1. Simple Logistic Regression using scikit-learn
2. Random Forest or Gradient Boosting Tree
3. Recurrent Neural Networks using Keras library.