Sentence-level Sentiment Classification

My implementation of Tsinghua University's Deep Learning Summer School Individual Lab

Given a natural language text, sentiment was classified with five-class sentiment labels:

• 0 (very negative)
• 1 (negative)
• 2 (neutral)
• 3 (positive)
• 4 (very positive)

GRU, LSTM and self-attention mechanism were used and compared.