Natural language processing maps text or speech to labels, extracted facts, retrieval results, or generated language. Meaning depends on context, labeling policy, and the errors tolerated by the application.
Choose a starting point
Begin with text boundaries and representation. Follow the path through classification, sequence tasks, retrieval, language models, evaluation, and deployment with a realistic collection of inputs.
- Natural Language Processing Tutorial
- Text corpus contracts: identity, label timing and annotation rules
- Text validation: split conversations, duplicates and time together
- BIO labels and subword alignment for entity extraction
- Script profiles and code-switching boundaries in text intake
- Text embeddings: pair labels, hard negatives and versioned vectors
- Document summaries with sentence-level evidence contracts
- Coreference chains: track mentions without guessing identity
- PII detection and redaction on original text offsets
Common Mistakes
A score averaged across clean examples can hide failures on names, rare terms, new domains, or long documents. The linked sections treat data and evaluation as part of the model contract.