Back to Projects
NLP

Attention Mechanisms in Sentiment Analysis

Analysis of attention mechanism variations in recurrent neural networks for sentiment prediction.

Course: Quantitative Methods and Models
Attention Mechanisms in Sentiment Analysis

Objectives

  • 1Evaluate the effectiveness of increasing or decreasing attention levels after training.

Conclusions

  • Inverting the attention mechanism post-training worsens network performance considerably.
  • Slightly maximizing attention post-training can yield small improvements in generalization.
  • Significantly maximizing the attention mechanism worsens network performance.

Technologies

  • TensorFlow
  • Keras
  • NLTK
  • NumPy
  • Matplotlib
  • Seaborn
  • Scikit-learn
  • Pandas