Stance-Detection-In-Text-Data
Predicting whether the tweet is Favored or Against using TF-IDF and Global Vector Models.
Running Instructions
Prerequisites:
-
Python 2.7
-
Pandas
-
Numpy
-
Sci-kitlearn(sklearn)
-
Natural Language Processing Toolkit (nltk)
-
Natural Language Processing Toolkit Data(words,punkt)
To download the above package the following commands need to be executed from python shell Import nltk nltk.download(‘punkt’) nltk.download(‘words’) -
Download Global Vector prebuilt model from http://nlp.stanford.edu/data/glove.twitter.27B.zip
Running Instructions
-
Unzip the folder downloaded from above url and place all the files in code folder provided with the submission.
-
To run the program type the below command after navigating to code foler
python biasDetectionMain.py --trainingfile training.txt --testfile test-gold.txt --choice g --glovefile glove.twitter.27B.25d.txt --evaluation k --classifier svc parameters need to be supplied : a) –trainingfile : path to training.txt file in code folder b) –testfile :path to test-gold.txt file code folder c) –glovefile: path to global vector file. We can give file with 25 dimensions or 50 dimensions or 100 dimensions or 200 dimensions eg for 50 dimensions --glovefile glove.twitter.27B.25d.txt d) –choice : to run the model with glove file enter ‘g’ or to run the model with tf idf enter t e) –evaluation : enter ‘k’ for k fold cross validation or ‘a’ for accuracy f) –classifier : enter svc for support vector machines ,gbc for gradient boosting classifier, rfc for random forest classifier