NLPyramid ( A simplest way to Learn NLP)
3 min readJul 5, 2022
A path to learn Natural language processing
Natural language processing is the process to build machines that understand and respond to text or voice data and respond with text or speech of their like the humans do. NLP can be used at various problem statements like , Text Classification, Summerization, Question Answering, Image Segmentation, Text generation etc.
So the question is where to start. Here, i am trying to make a path from where the learning can be started.
So, what this pyramid illustrates. Here, we can see there are multiple steps.
- Text Preprocessing — When we received the text from any of the source it is not easily understood by a Machine so we need to do preprocessing on that. For that we can implement various steps like tokenize, stemming, lemma. Next, we will prepare Bag Of Words to work on the text. Then we will work on embedding which means we have to convert text in Machine understandable format. For that we can use TFIDF, Word2vec, GloVe, FastText, ELMO etc. After doing these steps we can use the converted text for the Machine to learn context of that text.
- Neural Network — Once the text is converted we have to use Neural network to train the model and learn the context behind that. For this we can use ANN but it can not learn the context of the text. So, we have to move on RNN, LSTM, Bi-directional LSTM and Encoder Decoder and we can use these architecture to train the model.
- Transformer — As RNN, LSTM and Encoder Decoder have the limit of attention we need to move on attention mechanism which is used in transformers. So, we will learn various architectures like BERT, GPT, BART etc. These architecture are having various use cases as they are trained according to that. For example, Generative Pre-trained Transformer ( GPT ) is trained on unlabeled data so it is used to generate the text.
Let’s take a look in each steps one by one. I will append link in below section in future.
