Political-Tweets-Sentiment-Analysis

B.Sc Dissertation: Investigating the role of emotions and cognitive biases in political decision making through sentiment analysis of Twitter data using ML

• Architecture choices:

  • Bidirectional LSTM - to determine polarity of a tweet
  • Bayesian inference - to analyze LSTM output and make predictions

• Technologies used:

  1. Data collection:

     • Tweepy - twitter API for data scraping

  2. Data preprocessing:

     • Pandas - handling large amounts of data
     • Regex - general noise removal
     • NLTK - tokenization, stemming, lemmatization, stopword removal
     • Glob - merging multiple csvs into one

  3. Data analysis:

     • Numpy - handling high-dimensional mathematical operations
     • sklearn - feature extraction and word embedding
     • Keras - testing different architectures, picked Bidirectional LSTM after comparing accuracy of different models
     • pymc3 - Bayesian Inference with NUTS sampler