NLPlay with Transformers

Progress report of 'NLPlay with Transformers'

---

Week-1

• introduction to google colab
• learned basic python programming
• learned libraries for ML : numpy, pandas, matplotlib

---

Week-2

• learned to use nltk library for NLP and DL
• learned theory behind neural networking
• learned to use pyTorch library for making classifiers

---

Week-3

• made a numerical classifier using MNIST dataset for practice
• saw numerous advancements in deep NLP in detail
• further explored the applications of deep NLP
• learned to use various types of word embeddings
• made a sentiment classifier using Feed Forwarded NN
  for this IMDB movie review dataset was used
  for this task simple BOW(bag-of-words) representation was used reached accuracy of 80.02%

  ---

Post mid-sem week (Week-4)

• learned how a recurrent neural network works
• made a sentiment classifier using RNN
  for this the same IMDB movie review dataset was used
  reached accuracy of 87.28%
• learned the LSTM and GRU models
• implemented these in the sentiment classifier
  reached accuracy of 86.51% and 89.86%

  ---

Week-5

• learned what a transformer is and how it works

---

Week-6

• made sentiment classifiers using BERT
• reached accuracy of 90.36%

---

Week-7

• made sentiment classifiers using RoBERTa
• reached accuracy of 91.02%

---

Week-8

• saw what GPT-2 and T5 are
• learned how to generate text using transformers
• created a custom dataset
• generated text using GPT-2
• analysed output by calculating the BLEU Score

---