Natural Language Processing (NLP)

Definition

Natural Language Processing (NLP) is the field of artificial intelligence focused on enabling computers to understand, interpret, generate, and interact using human language, spanning tasks from text classification and translation to modern conversational AI systems.

Overview

NLP sits at the intersection of computer science, linguistics, and machine learning, and it covers a wide range of tasks: parsing grammatical structure, extracting named entities from text, classifying sentiment, translating between languages, summarizing documents, answering questions, and generating fluent text. Early NLP systems relied heavily on hand-crafted linguistic rules and statistical models built on engineered features, requiring substantial linguistic expertise to build and maintain for each new task or language. The field was transformed by deep learning, first through recurrent neural networks that processed text sequentially, and then decisively by the Transformer architecture introduced in 2017, which uses attention mechanisms to model relationships between all words in a sequence simultaneously rather than one at a time. This architecture underlies virtually all modern large-scale NLP systems, from encoder-based models like BERT, tuned for understanding tasks like classification, to decoder-based LLMs like the GPT family, tuned for generating fluent text. Central to how these models process language is tokenization, which breaks raw text into smaller units, and embeddings, which represent those units as numeric vectors capturing semantic meaning. Modern NLP applications are everywhere: search engines use it to understand queries, RAG systems use it to retrieve and synthesize relevant information, virtual assistants and chatbots like ChatGPT use it to hold conversations, and translation services use it to convert between languages in real time. NLP remains an active area of study in AI curricula, and the course on Large Language Models covers the modern Transformer-based approaches that now dominate the field; the blog post on how large language models actually work is a useful accessible companion.

Key Concepts

• Covers tasks like text classification, translation, summarization, and question answering
• Transformed by deep learning, especially the Transformer architecture introduced in 2017
• Relies on attention mechanisms to model relationships between words across a sequence
• Uses tokenization to break raw text into processable units before modeling
• Uses embeddings to represent words and phrases as numeric vectors capturing meaning
• Powers modern applications like search, chatbots, translation, and retrieval-augmented generation
• Includes both encoder-based models for understanding and decoder-based models for generation

Use Cases

Frequently Asked Questions

Related in AI Fundamentals