Is speech recognition part of NLP?

Partly. Speech recognition is the step that turns sound into words, and natural language processing (NLP) is the wider field that studies how computers handle language. Speech recognition sits inside that wider field, so it is part of NLP in the broad sense. A dictation tool that only writes what you say uses fewer NLP methods than an assistant that acts on what you say.

Speech recognition is one of the topics inside language processing

The 2026 draft of Jurafsky and Martin's textbook opens its chapter on automatic speech recognition by saying that understanding spoken language, or transcribing it into writing, "is one of the earliest goals of computer language processing" ¹. The book treats speech and text as parts of one subject. Its title lists natural language processing, computational linguistics and speech recognition side by side ².

The two fields differ in where they start

NLP methods work on words and their meaning: grammar, what a sentence refers to and what a request means. Speech recognition begins earlier, with sound. Inside a recognizer, the language model does a job that is close to NLP. A homophone is a word that is pronounced the same as another word but differs in meaning or spelling ³. Choosing between "there" and "their" depends on context, and the language model makes that choice by scoring which word sequences are plausible.

Assistants add NLP after the words are recognized

A voice assistant passes the recognized words on for interpretation. The textbook lists digital assistants and chatbots among the places where speech is a natural interface, especially on cellphones where keyboards are less convenient ⁴. NLP can also run after the recognizer. A 2023 benchmark found that a large language model can correct words that are missing from the recognizer's N-best list, because it generates text rather than only choosing among candidates ⁵. That is a language step working on the output of a speech step.

It depends on the app: voice tools are not all one kind

Whether a voice tool uses NLP depends on its design, and the difference matters more than the label.

• Dictation into a document is mostly speech recognition with a language model. It writes what you say.
• A voice assistant that acts adds an interpretation step, which is NLP, to decide what the words ask for.
• Live captions mainly display the words the recognizer produces, so they depend most on speech recognition.

The label alone does not settle it. Ask what the tool does with the words after it recognizes them.

What to do

1. If you are studying the field, read our guide to how speech recognition works, which covers the language model step in detail.
2. If you are choosing a voice tool for typing, check whether it only transcribes or also interprets commands. That decides what it can do with a sentence.
3. Test with your own speech. Names and homophones show the language model at work, and a word error rate measured on them tells you more than a headline figure.

Related: Is speech recognition AI?, Speech vs voice recognition: what is the difference? and What is speech recognition used for?. For the model that scores word sequences, see language model (in speech recognition).