Text-Classification 📝
Text classification is one of the most crucial classifications as it involves various parameters in the consideration. I have tried to use the most accurate model BERT, which is a deep learning model which helps us to classify the words in their context, in this way the classwise distinguishment will be easier.
before directly feeding data to the model we have to perform a series of steps to clean and process the data.
-
Data – Cleaning and Pre-processing: Text Cleaning – conversion to lowercase , removing punctuations and characters ( using ‘re’ library ). Removing Stop Words – used NLTK English ‘stopword’ corpus to detect stop words in a statement. Applying Lemmatization – used NLTK wordnet and WordNetLemmatizer.
-
Data – Augmentation: Applied Word Level Augmentation based on contextual word embeddings – Used NLP-AUG library and BERT to achieve the same. Made 4 embeddings for each statement.
Steps to run the project:
- The code can be run on any jupyter Notebook or Google Colab ( preferable use Google colab with GPU as hardware accelerator ).
- Run each cell individually
This project is created with:
Python 3.6/3.8 libraries: BERT, TensorFlow, Hugging Face transformers, scikit-learn, pandas, numpy, seaborn, pyLDAvis, NLTK, MatPlotLib, NLPAug, JSON .