mounicam/sentiment_analyzer

Multilingual Sentiment Analyzer

Multilingual Sentiment Analyzer

Steps to run the code:

1. Install pytext. (Python version > 3.6)

  $ python3 -m venv pytext_venv
  $ source pytext_venv/bin/activate
  (pytext_venv) $ cd pytext 
  (pytext_venv) pip install .

2. Download Spanish glove embeddings to data/embeddings directory. You can find the embeddings here. Unzip the file.

3. Run the main file. It takes a json file as an input and assumes that 'text_m' is the field with tweet text. It adds a json object called 'extension' for each tweet with predicted sentiment and sentiment scores.

  $ source pytext_venv/bin/activate
  $ (pytext_venv) python3 main.py --input <input json file> --out <output json file>

Repo structure

1. data: Contains isol lexicon (for logistic regression classifier), tass dataset (to train Spanish sentiment analyzer) and Spanish glove embeddings.

2. logistic_regression: Code for logistic regression classifier for Spanish sentiment analysis

3. pretrained_models: Pretrained models (logisitic regression and CNN) for Spanish sentiment analysis

4. pytext: Code for CNN classfier for Spanish sentiment analysis

5. tweetment: Code for SVM classfier for English sentiment analysis

6. main.py: Code for multi-lingual sentiment analyzer (English and Spanish).

References:

1. For English sentiment analysis, we used tweetment library.

2. For Spanish sentiment analysis, we implemented this paper.