This repo will be like series of parts from classical nlp to modern nlp and using the machine learning and deep learning models to apply the data after we get the conversation from text to numbers.
This studied from different courses related to classical nlp stanford, nlp in actions book, most valuable first 2 videos from DR:Ahmed El Sallab and others.
https://www.youtube.com/watch?v=oWsMIW-5xUc&list=PLLssT5z_DsK8HbD2sPcUIDfQ7zmBarMYv
First two videos have a lot and valuable information in Arabic related to Part 1.
https://www.youtube.com/watch?v=VWwvPgP2Fb0&list=PLX2D7RnWrLv7qGcUl_MOpC74tH5f30mCC
To get in touch of what's here follow steps below
First part of the repo associated with bag of words along with text preprocessing we should care about when we handle text.
What you will get from first part:
Most of classical nlp pipelines from preprocessing to see difference effect of each stage beside all the requirements for each stage.
Some Neural network frameworks are used but we have mentioned explanation about.
-
Text Preprocessing stages
- Cleaning
- Normalization
- Tokenization
- Padding
- Others
-
Text Preparation stages (BOW)
- One-Hot Matrix represent document
- Binary Vectorization
- Count Vecotrization
- Frequency
- TF-IDF
-
Chapter 2 direction from nlp in actions Book
- read first ch2 posts in Arabic.
- then see the notebook
-
Chapter 3 direction from nlp in actions Book
- read first ch3 posts in Arabic.
- then see the notebook
-
Apps direction
- First see introduction about nlp with deep learning (Heavily notebook)
- Apply on Kaggle dataset (with out text prepreocessing)
- Apply on Kaggle dataset (with text prepreocessing)