jep4 / bert-political-tweet-analysis

CMPT-353-Prj

Overview

This is a python-based project that classifies sentences' political tendency with BERT language model. Around 45% of accuracy is achieved.

Required libraries

tensorflow, torch, transformers, sklearn, pandas, numpy__

Order of execution

Recommend to use Google Collab notebook__

to test BERT model: run BERT.ipynb to test BERT model with a new sentence: run BERT_withTest.ipynb to see a plot: run data_analysis.ipynb & word_ratio.py (Not required) data.csv / data2.csv / data3.csv :: Sample data files

Expected results

BERT.ipynb would print the training process and the accuracy score BERT_withTest.ipynb would print the R / D / N according to the test sentences