sentiment-analysis

Sentiment analysis task on the Amazon jewelry reviews dataset

basic-models.ipynb

• Data cleaning on the dataset
  • Convert all reviews into lowercase
  • Remove HTML/URL content from reviews
  • Remove non-alphabetic characters
  • Remove extra spaces
  • Perform contractions on the reviews
• Preprocessing steps
  • Stop word removal
  • Lemmatization
• Feature extraction using TF-IDF
• Train models such as
  • Single Layer Perceptron
  • Support Vector Machine
  • Logistic Regression
  • Multinomial Naive Bayes

advanced-models.ipynb

• Data cleaning on the dataset
  • Convert all reviews into lowercase
  • Remove HTML/URL content from reviews
  • Remove non-alphabetic characters
  • Remove extra spaces
• No preprocessing steps such as stop word removal or lemmatization
• Generate features using Word2Vec
  • Performed comparison between pretrained 'word2vec-google-news-300' model and training a Word2Vec model from scratch.
• Utilize these features and train models such as
  • Single Layer Perceptron
  • Support Vector Machine
  • Feed forward neural networks
  • Recurrent neural networks
    • Simple RNN
    • RNN with GRU