My Skill Lab
Machine LearningIntermediate

Natural Language Processing Fundamentals

Teach machines to work with human language.

Taught by Bibek Sharma · Data Scientist

What you'll learn

  • Clean, tokenise, and normalise raw text using both Python and scikit-learn pipelines
  • Convert text into numerical features with bag-of-words, n-grams, and TF-IDF weighting
  • Understand and use word embeddings to capture semantic meaning between words
  • Build, evaluate, and tune text classifiers and sentiment models with scikit-learn
  • Read a confusion matrix and choose the right metric for imbalanced text data
  • Explain how transformers, attention, and pretrained models power modern NLP

About this course

Go from raw text to working language models in this hands-on course. You will learn how to clean and tokenise text, turn words into numbers with TF-IDF and embeddings, build sentiment and text classifiers in scikit-learn, and finish with a clear, modern picture of how transformers like BERT and GPT actually work.

Course content

8 lessons
  1. 1What NLP Is and Why Text Is HardThe NLP landscape, why language is messy, and the standard text pipeline.Preview
  2. 2Tokenisation: Splitting Text Into PiecesWord, sentence, and subword tokenisation, plus the edge cases that break naive splitting.Preview
  3. 3Cleaning Text: Stopwords, Stemming, and LemmatisationNormalising text by removing noise words and reducing words to their base forms.
  4. 4Bag-of-Words and TF-IDF: Turning Text Into NumbersCounting words into vectors and weighting them by importance with TF-IDF.
  5. 5Word Embeddings: Meaning as GeometryDense vectors that place similar words near each other, from Word2Vec to modern embeddings.
  6. 6Building a Text Classifier With scikit-learnAn end-to-end TF-IDF plus classifier pipeline for categorising documents.
  7. 7Sentiment Analysis and Evaluating NLP ModelsDetecting positive vs negative sentiment and measuring models with precision, recall, and F1.
  8. 8Transformers and How Modern NLP WorksAttention, pretraining, and using pretrained transformer models with a few lines of code.