8. Assignment VIII: Sentiment Analysis Using Deep Learning#

8.1. Question 1#

Build a movie review classifier using the dataset in demo_data/movie_reviews.csv. The objective of the classifier is to automatically classify a movie review into either positive or negative category.

The dataset is the famous IMBd moview reviews dataset. You can take a look at the SOTA classification performance on this dataset here.

In your experiments, please include the following strategies in your considerations:

  • Please use sequence models for this task (you may experiment with networks of different topologies).

  • For embedding layers, please try both self-trained embedding layer along with the sentiment classifier, as well as pre-trained embeddings (either provided in spacy, or available on GloVe website).

  • Please include dropout and regularization layers to avoid overfitting.