alaakh42 / Sentiment_Model_Template

The ultimate template to build a Sentiment Model starting from Linear SVM to Word2Vec and GLove

Snetiment_Model_Template

The ultimate template to build a Sentiment Model starting from Linear SVM to Word2Vec The sentiment classification model was trained on IMDB Reviews data this a Kaggle Competition

I built two consecutive classifiers, the first is Subjective/ Objective and the second is Positive/ Negative tweets classifier The text vectorization using BoW, Tf-Idf, Word2vec, and AraVec CBoW and SkipGram word vectors and sometimes stacked with some statistical numerical features engineeerd from each review I used LR, Linear SVM, Linear SGD, Multinomial NB, Linear SGD and XGBoost Results varied between 72-82% Precision and Recall

To install the required packages, run the following command

pip install -r requirements.txt

This repo is written in Python 2

This repo contains the following files

1. Sentiment_Models_Template.ipynb : Which contains the sentiment model meat
2. utils.py : A python script that conatins some function to plot the learning curve, confussion matrix, and validation curve to assest the model
3. logistic_reg_inspection.py : A python script that contains code to check what the model learnt using logistic regression coeffecients To use it, you need to do the following:

importance = get_most_important_features(vectorizer, log_clf, 10) # log_clf is the LogisitcRegressin Model
top_scores = [a[0] for a in importance[0]['tops']]
top_words = [a[1] for a in importance[0]['tops']]
bottom_scores = [a[0] for a in importance[0]['bottom']]
bottom_words = [a[1] for a in importance[0]['bottom']]

plot_important_words(top_scores, top_words, bottom_scores, bottom_words, "Most important words for Sentiment")

4. data : Is the folder that will contain your data

5. cleaned_data : Contains the data after cleaning

NOTE:: To RUN this code you will need to download both GLove and Goggle News Word2Vec Models