Natural language processing (NLP)

About this course

This course treats text as data: turning raw text into features, classifications, extracted entities, and generated outputs you can actually measure and trust. The goal isn't to "call an NLP library" but to understand what happens to text at every step — and where meaning gets lost along the way.

The course moves from tokenization and normalization across languages, through bag-of-words and embeddings, to classification with labeled data, named entity and field extraction, summarization, and evaluating generated text without a reference answer. Each block ends with a report built on your own data — a text representation report with a count-based baseline and a list of strong features, or an extraction pipeline for one type of your documents.

Built for developers and analysts who already work with text data and want to move from calling a ready-made API to understanding what's inside it. You'll need basic Python and a general sense of how machine learning models are built — this isn't a course on programming from scratch.

What you'll learn

• Tokenize and normalize text across multiple languages
• Build statistical and embedding-based text representations
• Tune and evaluate a text classifier on imbalanced data
• Extract named entities, fields, and relations from documents
• Evaluate summarization and generated text without a reference answer
• Diagnose where a text pipeline breaks on real-world data
• Account for language drift, privacy, and cost when choosing a model
• Compare a purpose-built classifier against a general-purpose model

What you'll end up with

• A set of reports for each block, from text representation to a pipeline operations review
• A working text classifier with measured metrics on imbalanced data
• An extraction pipeline for one type of your own documents
• A certificate on paid plans

Not for you if you want a ready-made chatbot prompt without understanding what happens to the text underneath it.