I need to improve my Siamese LSTM model with high F1 score.

This Model is used for detecting similarity between two sentences. It is Multi-class problem There are 5400 sentence pair in training data-set. I got 76 accuracy but F1 score is really bad. I need a freelancer which can improve my model in a sense that my F1 score, precision and recall have good value(F1Score Should be more than 85).