Library

Glossary of dictation and speech recognition

191 terms, each defined in a sentence and explained on its own page.

A

Accent: An accent is the way a speaker pronounces a language, shaped by where they grew up, their first language or their community. Accents change pronunciation, not vocabulary, and speech recognition error rates can differ sharply from one accent to another.
Access to Work: Access to Work is a UK government grant for people with a disability or health condition who need support to do their job. GOV.UK says it can help pay for specialist equipment and assistive software, among other support, and that the grant does not affect other benefits and does not have to be paid back.
Accessibility permission (macOS): Accessibility permission is a macOS setting that lets a specific app read and control other apps through the system's accessibility features. macOS keeps it under Privacy & Security, and an app gets no access until you allow it there.
Acoustic model: An acoustic model is the part of a speech recognizer that scores how well a short window of audio matches the sounds of speech. Since about 2012 most acoustic models are deep neural networks, and their training data strongly affects accuracy.
AI dictation: AI dictation is dictation in which a language model edits the recognized text after speech recognition, removing filler words, fixing punctuation and sometimes rewording a sentence. It can change your meaning, so check the result before you send it.
AI text detector: An AI text detector is software that estimates whether a passage was written by a language model, usually from the words alone. Detectors make mistakes in both directions, so a score is weak evidence about how a text was made.
All-party consent: All-party consent is a recording rule under which every participant in a conversation must agree before it can be recorded or intercepted. California, Florida and Maryland use it in their statutes, and it is stricter than one-party consent.
Allow Full Access: Allow Full Access is an iPhone setting that you switch on for a third-party keyboard before it can have open access, such as network access. Apple warns that a keyboard with this setting on is permitted to collect and transmit the text you type.
Apple Dictation: Apple Dictation is the free, built-in feature that types what you say into any text field on a Mac. Apple also calls it Mac Dictation, macOS Dictation or keyboard dictation.
Apple Intelligence: Apple Intelligence is the name Apple gives to its AI features across its devices. Its on-device models can improve Dictation on newer Macs, and Apple says apps can use the same models.
Apple silicon: Apple silicon is the family of chips Apple designs for its own computers, starting with certain Macs in late 2020. Each chip puts several computer components on one piece of silicon, which Apple calls a system on a chip.
Articulation rate: Articulation rate is the number of syllables a speaker produces per second, leaving out pauses. It measures how fast the speech itself comes out, so it is usually faster than the overall speaking rate that counts every pause.
Assistive technology (AT): Assistive technology (AT) is any item, piece of equipment or product system that increases, maintains or improves what a person with a disability can do. Speech recognition, voice control and screen readers are common examples.
Attorney-client privilege: Attorney-client privilege protects confidential communications between a lawyer and their client about seeking legal advice or services. A third party who is present, or a tool that handles the conversation, can put that protection at risk.
Augmentative and alternative communication (AAC): Augmentative and alternative communication (AAC) is a clinical practice area that supplements or compensates for impairments in speaking or understanding language. It uses tools such as picture boards, letter boards and speech-generating devices, and it is augmentative when used alongside speech and alternative when used instead of speech.
Automatic punctuation: Automatic punctuation is when speech software infers commas, periods and question marks from your speech and inserts them without a spoken command. It can be switched on or off, and it works only in supported languages.
Automatic speech recognition (ASR): Automatic speech recognition (ASR) is software that converts spoken audio into written words. Modern systems are neural networks trained on large amounts of transcribed speech, and they run in the cloud or on the device.
Azure AI Speech: Azure AI Speech is Microsoft's cloud speech service. Microsoft now calls it Azure Speech in Foundry Tools, and its speech to text powers the online recognition behind Windows voice typing, which needs an internet connection.

B

Back-end speech recognition: Back-end speech recognition is a workflow in which the dictated audio is recognized and then edited by someone else, usually a transcriptionist, before it is signed. Its main trade-off is the cost and time of the second pass against the writer's own editing effort.
Biometric data: Biometric data is personal data produced by technical processing of a person's physical, physiological or behavioral characteristics that allows or confirms who that person is. A voice recording becomes biometric data when it is processed to identify the speaker.
Bring your own key (BYOK): Bring your own key (BYOK) is a setup in which you supply your own API key for an AI or speech provider. The app then sends requests through your account with that provider, so your agreement with the provider governs the data.
Business associate agreement (BAA): A business associate agreement (BAA) is the written contract that HIPAA requires before a covered entity lets a vendor create, receive, maintain or transmit protected health information on its behalf. Dictation tools used in clinics often need one, and a BAA alone does not make a product compliant.

C

CART captioning: CART captioning is a live text service in which a trained captioner types what is said, usually on a stenotype machine, and the words appear on a screen as they are spoken. CART stands for Communication Access Realtime Translation, and it is also called real-time captioning.
Character error rate (CER): Character error rate (CER) is the number of character insertions, substitutions and deletions needed to turn a recognizer's output into the correct text, divided by the number of characters in the correct text. It is the usual measure where words are not separated by spaces.
Characters per minute (CPM): Characters per minute (CPM) is the number of characters typed in one minute, counting spaces and punctuation. Words per minute is the same count divided by five, so a typist at 50 WPM enters 250 characters per minute.
Clean verbatim: Clean verbatim is a transcript that removes filler words, stutters, repetitions and false starts while keeping the speaker's wording and meaning. It is the usual choice for reading, captions and coding, where hesitations add noise rather than information.
Closed captions (CC): Closed captions (CC) are a text version of a program's dialogue and relevant sounds that the viewer can turn on or off. The word closed means the viewer chooses whether to see them, which separates them from open captions.
Cloud dictation: Cloud dictation sends your recorded speech to a remote server, which recognizes it and returns the text. It usually needs an internet connection, and your audio leaves your device, so check the provider's policy before you dictate sensitive material.
Code-switching: Code-switching is the habit of moving between two or more languages within a conversation or a single sentence. Speech recognizers built around one language often misread the switched words, so mixed speech is harder to dictate than speech in one language.
Colemak: Colemak is an alternative English keyboard layout, introduced in 2006 and designed for efficient, comfortable touch typing. Its supporters describe it as the third most popular layout for English touch typing, after QWERTY and Dvorak, but that ranking has no stated method.
Conformer: A Conformer is a neural encoder for speech recognition that combines self-attention, which sees the whole input, with convolution, which captures local sound patterns. FastConformer is NVIDIA's faster variant of the same design, used in streaming models.
Connected experiences (Microsoft 365): Connected experiences are Microsoft 365 features backed by cloud services, and Dictate is one of the experiences that analyze your content. You can turn them off in Manage Settings, but that setting also turns off other services.
Connectionist temporal classification (CTC): Connectionist temporal classification (CTC) is a way to train a neural network to turn audio into text without knowing in advance which moment of sound matches which letter. The network labels every short slice of audio with a character or a blank, and the blanks and repeated labels are then removed to leave the transcript.
Copilot+ PC: A Copilot+ PC is a class of Windows 11 hardware with a neural processing unit (NPU) that Microsoft says can run at 40 or more trillion operations per second. Some newer Windows AI features, including fluid dictation, require this hardware.
Copytaker: A copytaker was a typist in a newspaper office who took down stories that reporters dictated over the telephone. The Collins English Dictionary defines the role this way, and it predates the computer-based filing that replaced it.
Cursorless: Cursorless is a spoken language for structural code editing that lets developers code by voice. Every command is an action performed on a target in the code, and it runs on Talon with a VS Code extension.
Custom vocabulary: A custom vocabulary is a list of words, names and phrases that you add to dictation software so it recognizes and spells them as you intend. It matters most for names, jargon and product terms the recognizer has not learned.

D

De Quervain's tenosynovitis: De Quervain's tenosynovitis is a painful condition in which tendons that move the thumb become irritated where they pass through a narrow tunnel on the thumb side of the wrist. Repeated thumb and wrist movements, such as texting, can worsen it.
Diacritics: Diacritics are marks added to a letter, such as the acute accent in é, the tilde in ñ or the vowel marks in Arabic, that change how a letter sounds or which word it spells. Speech tools and everyday writers often leave them out, so a dictated word can come out spelled differently from what you meant.
Dialect: A dialect is a regional or social variety of a language, with its own vocabulary, grammar and pronunciation. Speech recognition is usually less accurate on dialects than on the standard form, because labeled training data for dialects is often scarcer.
Dictaphone: A Dictaphone is a machine that records a speaker's voice so that another person can play it back and type it later. The name began as a brand, and it became the everyday word for dictation machines.
Dictate (Microsoft 365): Dictate is the speech-to-text feature in Microsoft 365 apps such as Word, Outlook and PowerPoint. It needs a Microsoft 365 subscription, a microphone and a reliable internet connection.
Dictation: Dictation is speaking your words so that software writes them as text, usually into the app that has the cursor. Phones and computers now include it, and people also call it voice dictation or voice typing.
Dictation (ChromeOS): Dictation on ChromeOS is a built-in accessibility feature that types recognized speech into an editable field. You turn it on with Search + D or the microphone icon in the status tray.
Dictation foot pedal: A dictation foot pedal is a footswitch that controls playback of a recording, so a typist can start, pause and rewind audio without using their hands. It is the traditional control for transcribing dictation.
Dictation software: Dictation software is a program that recognizes your speech and types the words where your cursor is, as you say them. It comes built into many phones and computers and is also sold as separate apps.
Dictation template: A dictation template is a block of text saved under a short spoken phrase, so one phrase inserts the whole block. Dragon calls these auto-texts, and Microsoft's Dragon Medical One guide documents the command that creates one from selected text.
Digital dictation: Digital dictation is recording speech as digital audio, then sending the file to a typist or to speech recognition software for transcription. It replaced tape with files that can be encrypted, uploaded or routed for processing.
Disabled Students' Allowance (DSA): Disabled Students' Allowance (DSA) is UK government support for students with a disability, mental health problem or long-term illness, to cover study costs that come from the condition. It depends on individual needs rather than household income, and it does not have to be paid back.
Dragon Medical One: Dragon Medical One, abbreviated DMO, is Nuance's speech recognition product for clinicians who dictate notes into health records. Microsoft now combines it with DAX Copilot in a product called Dragon Copilot.
Dragon NaturallySpeaking: Dragon NaturallySpeaking, usually called Dragon, is a speech recognition program for dictating text on a computer. Dragon Systems released it in 1997, and Nuance has sold it in several editions since.
Dragon Professional Anywhere: Dragon Professional Anywhere is a subscription edition of Dragon that Nuance sells for professional and enterprise users. It is often shortened to Dragon Anywhere, a name that also belongs to Nuance's mobile app, so check which product a listing means.
Dvorak Simplified Keyboard: The Dvorak Simplified Keyboard is an alternative English keyboard layout that places the most common letters on the home row. Dvorak and Dealey applied for a patent on it in 1932, but the studies that claimed it was faster than QWERTY have been criticized, and the evidence for a speed gain is weak.
Dysarthria: Dysarthria is a speech disorder that happens when brain or nerve damage changes the way the muscles used for speaking work, so speech can sound slow, slurred or hard to understand. Speech recognition trained on typical voices can make more errors with this speech, so accuracy depends on the person and the software.
Dysgraphia: Dysgraphia is a difficulty with writing that does not match a person's cognitive level and age, even after adequate instruction. It can affect letter formation, spacing, spelling, writing speed and composition, and speech to text lets a person get their words onto the page without writing each letter by hand.

E

Electronic health record (EHR): An electronic health record (EHR) is a digital record of a patient's health information, built up over time from visits across care settings. The federal eCQI glossary also calls it an electronic medical record (EMR), and it is where dictated and typed notes end up.
Enhanced Dictation: Enhanced Dictation was a Mac option, available from OS X Mavericks, that lets you dictate without an internet connection after a download. Apple's help for macOS Mojave describes it, but Apple's later Dictation pages do not mention it.

F

False accept (accidental activation): A false accept, also called a false wake or accidental activation, is when a voice assistant or dictation feature starts listening or recording although the wake word was not said. Smart speakers have been measured to wake by accident, and that can capture speech the user did not mean to share.
Federated learning: Federated learning is a way to train one machine learning model on data spread across many devices. Each device computes an update from its own data, and only those updates are sent to a server, so the raw data does not have to be collected in one place.
Filler words: Filler words are sounds and words such as um, uh, like and you know that speakers produce while they plan what to say next, without adding content. Many dictation tools can remove them automatically, and research on transcripts found they can be deleted without compromising meaning.
Fluid dictation: Fluid dictation is a Windows feature on Copilot+ PCs that corrects grammar, punctuation and filler words as you speak. Microsoft says it is powered by small language models that run on the device, and that you can undo a correction by saying Revert or Undo that.
Forced alignment: Forced alignment is the process of matching a known transcript to its audio so that each word or phone gets a start and end time. Unlike open speech recognition, it does not decide what was said; it only finds when each word was said.
Free tier: A free tier is the no-cost version of a dictation app, with limits such as a weekly word cap, fewer modes or no AI processing. Read those limits before you rely on it, because a free plan can run out in the middle of your work.
Front-end speech recognition: Front-end speech recognition is a workflow in which the person who dictates also reviews, corrects and signs the text, instead of passing the audio to someone else. The term is most common in radiology, where it is set against back-end transcription.

G

Gboard: Gboard is Google's keyboard app for Android and iPhone, and its microphone key turns speech into text in any app where you can type. On Android it can run speech recognition on the phone, while on iPhone the voice input is handled inside the Gboard app and sent to Google.
Generative AI: Generative AI is the class of AI models that create new content, such as text, audio or images, from patterns in their training data. The NIST profile defines it as models that emulate the structure and characteristics of input data to generate derived synthetic content.
Generative error correction (GER): Generative error correction (GER) passes a speech recognizer's output, often its N-best list of alternative transcripts, to a large language model, which rewrites the transcript to fix errors. It can recover words the recognizer never offered, but it can also change words that were already right.
Globe key (Fn key): The Globe key, also called the Fn key or function key, is a modifier on Apple keyboards. In Keyboard settings you choose what a press does, such as switching input source, starting Dictation or doing nothing.
Google voice typing: Google voice typing is the speech-to-text input that Google provides on Android, so you can speak words into a text field. Its name and location vary by phone: Gboard shows a microphone key, while Samsung Galaxy phones list Google Voice Typing next to their own voice input.

H

Hidden Markov model (HMM): A hidden Markov model (HMM) is a statistical model in which a sequence of hidden states produces the observations you can measure, and each state changes to the next with set probabilities. In speech recognition, the hidden states are units of sound and the observations are short frames of audio.
Home row: The home row is the middle row of keys on a keyboard, where the fingers rest when you touch type. On a QWERTY keyboard it holds A, S, D, F for the left hand and J, K, L and semicolon for the right, and every other key is reached by moving a finger away from it and back.
Homophone: A homophone is a word pronounced the same as another word but different in meaning, spelling or both, such as to, too and two. Dictation software has to choose the right spelling from context, so homophones are one of the most common errors in dictated text.
Hunt and peck: Hunt and peck is a typing method in which the typist finds each key by sight and presses it with one or two fingers, usually the index fingers. It is also called two-finger typing, and it differs from touch typing, which relies on memorized key positions and the home row.

I

Improve Siri & Dictation: Improve Siri & Dictation is the Mac setting that controls whether Apple may store and review audio from your Siri and Dictation requests. When you first turn on Dictation, the same choice is labeled Share Audio Recordings.
Individualized Education Program (IEP): An Individualized Education Program (IEP) is a written statement for a child with a disability, developed, reviewed and revised under the federal IDEA regulations. For US students, it is the school's plan for the special education and services the child needs, and the team must consider whether the child needs assistive technology.
Input method editor (IME): An input method editor (IME) is software that supports text input for characters that cannot be typed easily on a standard keyboard, such as many East Asian characters. Speech input can also be an input method, and Android allows only one input method to be enabled at a time.
Intent recognition: Intent recognition is the step that decides what a spoken request is asking for, such as setting a timer or sending a message, and which details the action needs. Voice assistants and voice agents rely on it before they act.
Inverse text normalization (ITN): Inverse text normalization (ITN) is the step that converts a speech recognizer's spoken-form output into written form, so that 'two hundred' becomes 200. It changes how the text reads, not which words the recognizer heard.

K

Keystroke dynamics: Keystroke dynamics is the identification of a person from the rhythm of their typing, using the timing between key presses as a pattern much like a handwritten signature. It is a behavioral biometric, so it supports identity checks through statistical matching rather than proving who typed.
Keystrokes per character (KSPC): Keystrokes per character (KSPC) is the average number of keystrokes a text entry method needs to produce one character of text. A basic QWERTY keyboard scores 1, methods that need more presses score above 1, and word prediction can score below 1.
Keystrokes per hour (KPH): Keystrokes per hour (KPH) is the number of key presses made in one hour. It is the standard unit for numeric data entry, where many jobs require 8,000 or 10,000 keystrokes per hour, and it counts keys rather than words.

L

Language model (in speech recognition): In speech recognition, a language model (LM) scores how probable a word sequence is in a language. The recognizer combines that score with the acoustic evidence to choose between words that sound alike, such as homophones.
Large language model (LLM): A large language model (LLM) is a language model with a very large number of parameters, trained on huge amounts of text to predict and generate language. Google's glossary defines the minimum as a language model having a very high number of parameters.
LibriSpeech: LibriSpeech is a free corpus of about 1,000 hours of read English speech, taken from LibriVox audiobooks, and it is one of the most widely used benchmarks for English speech recognition. Its scores describe clean, read speech, which differs from everyday dictation.
Live Captions: Live Captions is a feature that shows text for speech and audio as it happens, using the phone's own processing. Google calls it Live Caption on Pixel phones, and Apple calls it Live Captions on iPhone. It reads what others say, so it is the reverse of dictation.
Live Transcribe: Live Transcribe is a Google app for Android phones that turns the speech of people around you into captions as they talk. It helps you follow a conversation. It does not type your own words into a text field, which is what dictation does.
Locale: A locale is a language combined with a region, and sometimes a script, such as en-GB for English as used in the United Kingdom. Dictation and speech settings use locales because the same language can differ in spelling, punctuation and speech patterns from one region to another.
Low-resource language: A low-resource language is one with little digital speech and text data for training speech recognition and other language tools. There is no fixed cutoff, and a language can be well resourced for one task and poorly resourced for another.

M

Machine learning: Machine learning is a method in which a program learns its behavior from example data rather than from rules written by hand. Google's glossary defines it as a program or system that trains a model from input data.
Median nerve: The median nerve is one of the main nerves of the arm. It runs from the forearm through the wrist into the hand, giving feeling and movement to parts of the hand.
Medical scribe: A medical scribe is a person, or more recently an AI tool, that records what happens in a clinical visit and writes it into the chart so the clinician can focus on the patient. The clinician remains responsible for the note and must check and sign it.
Mel spectrogram: A Mel spectrogram is a grid of sound energy over time, with frequency bands spaced to match human hearing. Many speech recognizers compute its logarithm, the log-Mel spectrogram, and feed it to a neural network instead of the raw waveform.
Microphone key (dictation key): The Microphone key is a key on some Mac keyboards that starts Apple Dictation when you press and release it. Pressing and holding it activates Siri instead, if Siri is turned on.
Modern Standard Arabic (MSA): Modern Standard Arabic (MSA), also called Fusha, is the standardized form of Arabic used in writing and formal speech, such as news broadcasts and official documents. Speakers use regional dialects in everyday conversation, and speech tools recognize formal Arabic more accurately than dialect speech.

N

Neural Engine: The Apple Neural Engine (ANE) is a block of hardware inside Apple's chips that runs machine learning models efficiently. Speech models can run on it, and Apple uses it for on-device machine learning inference.
Neural processing unit (NPU): A neural processing unit (NPU) is a specialized chip for AI-intensive processes, such as real-time translation and image generation. Microsoft says many new Windows AI features require an NPU that can run at 40 or more trillion operations per second, and a speech model can run faster on one.
Noise-canceling microphone: A noise-canceling microphone is designed to filter ambient noise, so it picks up mainly the speaker's voice. It can help dictation in noisy rooms, but it does not remove all noise, and it is different from the active noise cancellation in headphones.
Non-native speech: Non-native speech is speech in a language that is not the speaker's first language, often called L2 speech or foreign-accented speech. Recognizers can transcribe read non-native English well, but they handle hesitations, repetitions and revisions differently, and results depend on the system.
Nuance Communications: Nuance Communications was a speech technology company best known for the Dragon family of speech recognition products. Microsoft announced that it had completed its acquisition of Nuance on March 4, 2022.
Number and grid overlays (Voice Access): Number and grid overlays are Voice Access views for clicking things you cannot name. Show numbers puts a number on each clickable item so you can say it, and Show grid divides the screen into numbered tiles that narrow the mouse pointer step by step.

O

On-device speech recognition: On-device speech recognition converts speech to text with a model that runs on your own computer or phone, so the audio does not need to be sent to a server. It is also called local, offline or on-device dictation.
One-party consent: One-party consent is a recording rule under which a conversation can be recorded lawfully when one of its participants agrees, which is usually the person doing the recording. The US federal wiretap law uses this rule, and individual states may set stricter ones.
Online speech recognition: Online speech recognition sends your voice to a cloud service that turns it into text, instead of processing it on your PC. Microsoft says voice typing on Windows uses online speech recognition, which Microsoft describes as giving more accurate recognition.
Open captions: Open captions are text drawn onto the frames of a video, so every viewer sees them and nobody can switch them off. They are also called burned-in captions or hardcoded subtitles.
Oral composition: Oral composition is working out and shaping a text by speaking it, whether to a dictation tool, to a recorder or silently in thought. Research on letters finds that planning takes most of the time, however the text is produced, so speaking mainly changes how the words get onto the page.
Oral reading fluency (ORF): Oral reading fluency (ORF) measures how many words a reader gets right in one minute of reading a passage aloud. Schools record the result as words correct per minute (WCPM), which subtracts the words read wrongly from the total.

P

Pajama time: Pajama time is the record-keeping that clinicians finish after their clinic day ends, often at home in the evening. A 2017 study of primary care clinicians measured 86 minutes of EHR work per weekday after clinic hours.
Perpetual license: A perpetual license is a software license you pay for once. It lets you keep using that version without renewing. Updates, support and new major versions are often separate, time-limited or paid extra.
Phi Silica: Phi Silica is Microsoft's on-device language model, which runs on the neural processing unit of Copilot+ PCs. Microsoft says it is being replaced by Aion Instruct, and its voice typing help names it as the download to watch while fluid dictation gets ready.
Pinyin: Pinyin, also called Hanyu Pinyin, is the system that spells Mandarin Chinese syllables with the Latin alphabet and marks the tones with diacritics. People type Chinese by spelling the sound in Pinyin and choosing the character they want, and speech tools must choose among characters that share one Pinyin spelling.
PowerScribe: PowerScribe is a family of radiology reporting products in which a radiologist dictates an imaging report by voice. It began as a Nuance dictation application, and Microsoft completed its purchase of Nuance in 2022, so Microsoft now sells the line.
Prompt (AI): In AI, a prompt is the input you give a model to get a response, such as a question, an instruction or pasted text. Google's glossary defines it as any text entered as input to a large language model to condition the model to behave in a certain way.
Protected health information (PHI): Protected health information (PHI) is individually identifiable health information that HIPAA protects when it is held or sent in electronic or other form by a covered entity or its business associate. Patient names with diagnoses, in dictated notes or voice recordings, are PHI. It is also called ePHI when held electronically.
Push-to-talk: Push-to-talk is a dictation mode in which you hold a key while you speak and release it to stop, so the software writes only the speech made while the key is down. Some tools also offer tap-to-start and tap-to-stop.

Q

QWERTY: QWERTY is the standard letter arrangement on English keyboards, named for the first six letters of the top row. Christopher Latham Sholes devised it in the early 1870s for typewriters, and it is still the layout most people type on, even though faster layouts have been proposed.

R

Real-time factor (RTF): The real-time factor (RTF) is the time a system takes to process audio divided by the length of that audio. An RTF of 1 or lower means the system keeps up with real-time speech. Some leaderboards report the inverse, RTFx, where higher is faster.
Real-time text (RTT): Real-time text (RTT) is a way of sending text as it is typed, so the other person can read each word before the sender finishes. It is used in conversation by many deaf and hard of hearing people, and it has no send step like an ordinary text message. It is a typing channel, not a speech technology.
Reasonable accommodation: A reasonable accommodation is a change to the job, the workplace or the equipment that lets a qualified person with a disability do the work. Under U.S. law, employers with 15 or more employees must provide one unless it would impose an undue hardship.
Registered Healthcare Documentation Specialist (RHDS): The Registered Healthcare Documentation Specialist (RHDS) is a voluntary credential from AHDI for people who create healthcare documentation. It is awarded after an exam of 120 questions, and it is a prerequisite for the CHDS.
Repetitive strain injury (RSI): Repetitive strain injury (RSI) is an umbrella term for pain and damage to muscles, tendons and nerves that builds up from repeated movements, held postures or sustained effort, felt in the hands, wrists, arms, shoulders and neck.
Right-to-left (RTL) script: A right-to-left (RTL) script is a writing system, such as Arabic, Hebrew, Persian or Urdu, whose letters run from right to left. Text that mixes these letters with digits or Latin words is bidirectional, so software needs a specific algorithm to order and display it.
RNN transducer (RNN-T): An RNN transducer (RNN-T) is a neural speech model that reads audio in order and emits words as it goes, so text can appear while the speaker is still talking. It has an encoder for the sound, a prediction network for the words already written, and a joint network that chooses the next output.
Rollover typing: Rollover, or key rollover, is the number of key presses a keyboard can detect when they overlap. Two-key rollover reliably detects only two keys at once, and a matrix keyboard without diodes can register a ghost key when three keys are pressed together.

S

Samsung voice input: Samsung voice input is the voice-to-text feature built into the Samsung Keyboard on Galaxy phones and tablets. It works only with that keyboard, so it is a separate choice from Google Voice Typing.
Scopist: A scopist edits the transcripts of official proceedings that court reporters create. Scopists are typically paid by the page, starting from $1.00 to $1.50 per page, and the role is also called transcript editor.
Screen reader: A screen reader is software that reads what is on a computer or phone screen aloud, or sends it to a braille display, so that people who are blind or have low vision can use the device. The most used screen readers on Windows are JAWS and NVDA, and VoiceOver is the one built into Apple devices.
Section 504 plan: A Section 504 plan is the written record of the accommodations a US public school provides to a student with a disability, under Section 504 of the Rehabilitation Act of 1973. It is a civil rights arrangement and differs from an Individualized Education Program (IEP), which comes from special education law.
Shorthand: Shorthand is an abbreviated symbolic writing method that increases the speed of writing compared with longhand. Trained writers used it to take dictation before recording machines, and it still underpins court reporting.
Siri: Siri is Apple's built-in voice assistant. You speak a request and it answers or carries out an action, while Dictation types what you say into a text field instead.
Site microphone permission: Site microphone permission is the browser's setting for each website that decides whether that site may use your microphone, usually chosen when the browser prompts you. It works alongside the operating system's own microphone setting, so both must allow the microphone before a site can listen.
SOAP note: A SOAP note is a structured clinical note with four parts: Subjective, Objective, Assessment and Plan. The format was developed by Larry Weed, and a StatPearls reference article describes it as a widely used documentation method for clinicians.
Speaker diarization: Speaker diarization is the task of working out who spoke when in a recording, by splitting the audio into turns and grouping turns that come from the same voice. It is separate from transcribing the words, and a transcript gains speaker labels only when diarization is run alongside recognition.
Speaker verification: Speaker verification is a voice biometric check in which a system accepts or rejects a claimed identity by comparing a voice with stored models of the person claimed. It is also called voice authentication, voice ID or voice biometrics.
Special category data (GDPR): Special category data is the group of personal data that Article 9 of the GDPR bans organizations from processing unless a listed exception applies. It includes health data and biometric data used to identify a person, and a voice recording can fall into it when it is processed that way.
Speech framework (Apple): The Speech framework is Apple's developer toolkit for turning audio into text inside an app. It includes the older SFSpeechRecognizer class and the newer SpeechAnalyzer, which Apple says transcribes entirely on the device once its models are downloaded.
Speech language model: A speech language model (speech LLM) is a large language model that accepts audio as input, so it can process speech directly instead of working only on a text transcript. The 2023 SALMONN model is one example that lets an LLM process general audio inputs.
Speech pack (Windows): A speech pack is the set of on-device language files that Windows needs before it can recognize speech in a language offline. Microsoft says you must download a speech pack for a language in Voice access before you can use it, and the first download needs an internet connection.
Speech Services by Google: Speech Services by Google is the Android speech engine from Google that turns speech into text for other apps and reads on-screen text aloud. Google's Play Store listing now uses the title Speech Recognition & Synthesis, and reviews on that listing still call the app by the older name.
Speech translation: Speech translation converts speech in one language into the same meaning in another language, as text or as synthesized speech. Most systems first recognize the speech and then translate the text, so a recognition error carries into the translation.
Speech-to-speech model: A speech-to-speech model takes audio in and produces audio out with one model, without writing the conversation out as text in between. Most voice systems instead chain speech recognition, a language model and speech synthesis.
Spoken language identification: Spoken language identification is the task of working out which language is being spoken in a recording. Products call it automatic language detection. It usually chooses from a short list of candidate languages and decides for a segment of audio rather than word by word.
Spoken punctuation: Spoken punctuation is saying the name of a punctuation mark, such as comma or question mark, so that dictation software writes that mark into the text. It is the manual alternative to automatic punctuation.
Stenotype: A stenotype, also called a steno machine, is a chorded keyboard that stenographers use for shorthand, pressing several keys together to produce a syllable or word. The NCRA's court reporter skills test sets a speed of 225 words per minute, and voice writing is a newer method that uses speech recognition instead.
SubRip subtitle file (SRT): A SubRip subtitle file, or SRT, is a plain-text file of numbered caption blocks. Each block has a start and end time in hours, minutes, seconds and milliseconds, followed by the caption text. It is one of the most common caption formats on the internet.
Subtitles for the deaf and hard of hearing (SDH): SDH stands for subtitles for the deaf and hard of hearing. They show the dialogue plus the sound effects, music and speaker identification that a hearing viewer would get from the soundtrack.

T

Talon Voice: Talon Voice, usually called Talon, is hands-free input software for writing, coding, browsing and gaming. It runs alongside a speech-recognition engine, and it is used for voice coding, often together with the Cursorless editing language.
Tendinitis: Tendinitis is inflammation of a tendon, the tissue that connects muscle to bone. It can occur as a result of injury or overuse, and repeated movements can trigger the overuse kind.
Text insertion: Text insertion is the step where dictation software places recognized words into the text field that has the cursor. It is done either by simulating keystrokes or by pasting the text from the clipboard.
Text to speech (TTS): Text to speech (TTS), also called speech synthesis, is software that converts written text into spoken audio. Modern systems use neural networks, and they can run on the device or in the cloud.
Third-party keyboard: A third-party keyboard is a keyboard app from a developer other than the phone's maker, which you install and switch to in place of the built-in keyboard. Apple calls the iPhone version a custom keyboard, and Apple's rules limit what it can do, such as network access and microphone use.
Timestamp (transcription): A timestamp in a transcript is a time marker, such as 00:12:15, that shows when the nearby text was spoken in the recording. Timestamps let readers jump to a passage, and they can be attached to whole segments or to single words, with different accuracy.
Transcribe (Microsoft 365): Transcribe is a Word for Microsoft 365 feature that turns an uploaded audio file into text in the Transcribe pane, rather than typing your speech into the document. Its monthly limits depend on your license.
Transcript Assist (Galaxy AI): Transcript Assist is a Samsung Galaxy AI feature that converts a saved recording in the Voice Recorder app into text. It works on select Galaxy models and needs a working internet connection. It transcribes recordings you already made, so it is not live dictation into a text field.
Transcription foot pedal: A transcription foot pedal is a switch that controls audio playback with the foot, so a transcriber can start, pause and rewind a recording while typing. It is also called a foot control or foot pedal.
Transcription line rate: A transcription line is a unit of 65 characters, and the line count is the number of such units in a document. The transcription line rate is the price paid for each line, so the unit controls how transcribed work is measured.
Transcriptionist: A transcriptionist is a person who listens to recorded speech and types it as text, such as a clinician's dictated report. Speech recognition now writes first drafts, so much of the work has shifted to checking and correcting machine text.
Transliteration: Transliteration represents the text of one language in the writing system of another, for example by writing a word in Latin letters. It keeps the spelling or sound and does not translate the meaning, which is what separates it from translation.
Turn detection: Turn detection is the step that decides when a speaker has finished a turn, so a voice system knows when to respond. A pause can be a hesitation rather than an ending, so the decision trades speed against cutting the speaker off.
Turnaround time (TAT): Turnaround time (TAT) is how long a transcriber or service takes to deliver a finished transcript after an order is received. Court schedules define it in calendar days or hours, counted from the order.
Typing accuracy: Typing accuracy is the share of typed characters or words that are correct, usually given as a percentage. A typing score should always report accuracy beside speed, because a fast typist who leaves many uncorrected errors has not typed well.

V

Verbatim transcription: Verbatim transcription writes down exactly what is said, including filler words, false starts, repetitions and non-speech sounds, with no tidying. Researchers and legal teams use it when the way someone speaks is part of the record.
Vibe coding: Vibe coding is building software by describing what you want to an AI model and accepting the code it writes, without reviewing that code. Andrej Karpathy introduced the term, and the developer Simon Willison argues for a narrower meaning than the broad one now in use.
Voice Access (Android): Voice Access (Android) is a Google feature that lets you operate an Android phone or tablet by speaking commands, such as opening an app or tapping a labelled item. Google built it for people who have difficulty using a touch screen. It is a control tool that can also enter text, and it is separate from Windows Voice Access on PCs.
Voice Access (Windows): Voice Access is the Windows 11 feature that lets you control the whole PC and write text by voice, not only dictate. Microsoft says it works without an internet connection after setup, and it replaces Windows Speech Recognition.
Voice activity detection (VAD): Voice activity detection (VAD) is a component that decides whether a stretch of audio contains speech. Speech recognizers use it to split audio into turns, skip silence and decide when a speaker has finished.
Voice agent: A voice agent is a voice system that listens, reasons with a language model, uses tools, and speaks back across several turns to finish a task, such as checking an order and rescheduling it. Benchmarks show that agents still fail many tasks on realistic audio.
Voice assistant: A voice assistant is software that listens for a spoken request, works out what you want, and then acts or answers, such as setting a timer or turning off a light. Dictation is different: it turns your speech into text for you to edit.
Voice cloning: Voice cloning is the use of a model trained on recordings of one person to generate new speech in that person's voice. Some products need only minutes of clear audio, and cloned voices have been used in fraud, so recordings of your voice carry a real risk.
Voice coding: Voice coding, also called programming by voice or coding by voice, is writing and editing source code by speaking instead of typing. It needs a speech-recognition engine and a separate platform that turns spoken words into code editing actions.
Voice command: A voice command is a spoken instruction that the software carries out, such as delete or stop listening, rather than writing the words as text. Dictation tools must separate commands from dictated words, and the rules differ by product.
Voice Control (Apple): Voice Control is Apple's feature for controlling a Mac with spoken commands, including opening apps, clicking items and editing text. When it is on, it also handles dictation, so standard macOS Dictation is not available.
Voice message transcripts: A voice message transcript is written text of the words in an audio message, produced by the messaging app so you can read instead of listen. Apple, Google and WhatsApp each offer one, with different defaults and different privacy arrangements. It turns audio into text, the reverse of dictation.
Voice profile: A voice profile is the set of files that dictation software keeps for one user, with settings such as the microphone they use and their custom words. It is not the same as speaker recognition, which identifies who is speaking.
Voice typing: Voice typing is dictation built into an operating system or app, so that speech is typed into a text field as you talk. Talk to text and speech typing are other names for the same act, and each platform uses its own name.
Voice typing (Google Docs): Voice typing in Google Docs is a built-in speech-to-text tool that types as you speak. You start it from the Tools menu in a supported browser, and the browser controls the speech-to-text service.
Voice typing launcher: The voice typing launcher is the Windows voice typing setting that lets you start dictation quickly when the cursor is in a text box. Microsoft lists it among the settings you open from the gear icon in the voice typing flyout.
Voice writing: Voice writing is a transcription method in which a trained writer repeats speech into speech recognition software through a stenomask, a mask that keeps the writer's voice from being heard in the room. It is used in court reporting, medical transcription and captioning.
VoiceOver: VoiceOver is the screen reader built into macOS, which speaks what is on the screen so you can use the Mac without seeing it. It does the reverse of dictation: it reads text aloud rather than writing your speech.
Voiceprint: A voiceprint is a stored model of a person's voice, built from measurable features of their speech and used to recognize or verify that person. Illinois law lists voiceprints among biometric identifiers, and other vendors call the same idea a voice template or speaker model.

W

Wake word: A wake word is a fixed phrase, such as Hey Google or Hey Siri, that a voice assistant listens for on the device before it starts handling a request. Until the phrase is heard, the assistant is only checking for that phrase.
wav2vec 2.0: wav2vec 2.0 is a speech model that first learns the structure of speech from large amounts of unlabeled audio, then is fine-tuned on a much smaller set of transcribed recordings. Because transcribed speech is expensive to make, this approach helps when labeled data is scarce.
Wayland: Wayland is the display system that Linux desktops are moving to in place of the X Window System, called X11. Its compositor delivers input to each app, so a typing tool that worked on X11 may not type on a Wayland desktop.
Web Speech API: The Web Speech API is a set of browser interfaces that let a web page start speech recognition and receive the recognized words as text, through an object called SpeechRecognition. Where the recognition runs, on your device or on a server, depends on the browser.
WebVTT (VTT): WebVTT (Web Video Text Tracks, usually shortened to VTT) is a plain-text file format for timed captions and subtitles on the web. A .vtt file is a list of cues, and each cue gives a start time, an end time and the text to show.
Windows key + H (Win+H): Windows key + H, written Win+H, is the keyboard shortcut that starts voice typing in Windows, so you can dictate into the text box where your cursor is. Microsoft says it does not start voice typing while Voice Access is running.
Windows on Arm: Windows on Arm is Windows running on processors that use the Arm instruction set instead of x86 or x64 chips. It runs native Arm apps and many unmodified x86 and x64 apps, and Microsoft says Arm-native apps give the best performance and battery life.
Windows Speech API (SAPI): The Windows Speech API (SAPI) is Microsoft's interface that lets Windows apps use speech recognition and text-to-speech without building the speech engine themselves. Windows.Media.SpeechRecognition is a separate API for command and control in Windows Runtime apps.
Windows Speech Recognition (WSR): Windows Speech Recognition (WSR) is the older speech feature built into Windows for voice control and dictation. Microsoft says it is deprecated and no longer being developed, and that Voice Access replaces it.
Word error rate (WER): Word error rate (WER) is the number of word substitutions, deletions and insertions in a transcript, divided by the number of words in the correct transcript. Lower is better, and a score above 100 percent is possible.
Words per minute (WPM): Words per minute (WPM) is the number of words typed in one minute. Typing tests count a word as five characters, including spaces, so WPM is a convention rather than a count of real words. The average in the largest online typing test was about 52 WPM.

Z

Zero data retention (ZDR): Zero data retention (ZDR) is a setting or contract term under which a service keeps none of your data once it has handled your request. Each vendor sets its own scope, so check what the label excludes before you rely on it.

Library

Glossary of dictation and speech recognition

191 terms, each defined in a sentence and explained on its own page.

A

Accent: An accent is the way a speaker pronounces a language, shaped by where they grew up, their first language or their community. Accents change pronunciation, not vocabulary, and speech recognition error rates can differ sharply from one accent to another.
Access to Work: Access to Work is a UK government grant for people with a disability or health condition who need support to do their job. GOV.UK says it can help pay for specialist equipment and assistive software, among other support, and that the grant does not affect other benefits and does not have to be paid back.
Accessibility permission (macOS): Accessibility permission is a macOS setting that lets a specific app read and control other apps through the system's accessibility features. macOS keeps it under Privacy & Security, and an app gets no access until you allow it there.
Acoustic model: An acoustic model is the part of a speech recognizer that scores how well a short window of audio matches the sounds of speech. Since about 2012 most acoustic models are deep neural networks, and their training data strongly affects accuracy.
AI dictation: AI dictation is dictation in which a language model edits the recognized text after speech recognition, removing filler words, fixing punctuation and sometimes rewording a sentence. It can change your meaning, so check the result before you send it.
AI text detector: An AI text detector is software that estimates whether a passage was written by a language model, usually from the words alone. Detectors make mistakes in both directions, so a score is weak evidence about how a text was made.
All-party consent: All-party consent is a recording rule under which every participant in a conversation must agree before it can be recorded or intercepted. California, Florida and Maryland use it in their statutes, and it is stricter than one-party consent.
Allow Full Access: Allow Full Access is an iPhone setting that you switch on for a third-party keyboard before it can have open access, such as network access. Apple warns that a keyboard with this setting on is permitted to collect and transmit the text you type.
Apple Dictation: Apple Dictation is the free, built-in feature that types what you say into any text field on a Mac. Apple also calls it Mac Dictation, macOS Dictation or keyboard dictation.
Apple Intelligence: Apple Intelligence is the name Apple gives to its AI features across its devices. Its on-device models can improve Dictation on newer Macs, and Apple says apps can use the same models.
Apple silicon: Apple silicon is the family of chips Apple designs for its own computers, starting with certain Macs in late 2020. Each chip puts several computer components on one piece of silicon, which Apple calls a system on a chip.
Articulation rate: Articulation rate is the number of syllables a speaker produces per second, leaving out pauses. It measures how fast the speech itself comes out, so it is usually faster than the overall speaking rate that counts every pause.
Assistive technology (AT): Assistive technology (AT) is any item, piece of equipment or product system that increases, maintains or improves what a person with a disability can do. Speech recognition, voice control and screen readers are common examples.
Attorney-client privilege: Attorney-client privilege protects confidential communications between a lawyer and their client about seeking legal advice or services. A third party who is present, or a tool that handles the conversation, can put that protection at risk.
Augmentative and alternative communication (AAC): Augmentative and alternative communication (AAC) is a clinical practice area that supplements or compensates for impairments in speaking or understanding language. It uses tools such as picture boards, letter boards and speech-generating devices, and it is augmentative when used alongside speech and alternative when used instead of speech.
Automatic punctuation: Automatic punctuation is when speech software infers commas, periods and question marks from your speech and inserts them without a spoken command. It can be switched on or off, and it works only in supported languages.
Automatic speech recognition (ASR): Automatic speech recognition (ASR) is software that converts spoken audio into written words. Modern systems are neural networks trained on large amounts of transcribed speech, and they run in the cloud or on the device.
Azure AI Speech: Azure AI Speech is Microsoft's cloud speech service. Microsoft now calls it Azure Speech in Foundry Tools, and its speech to text powers the online recognition behind Windows voice typing, which needs an internet connection.

B

Back-end speech recognition: Back-end speech recognition is a workflow in which the dictated audio is recognized and then edited by someone else, usually a transcriptionist, before it is signed. Its main trade-off is the cost and time of the second pass against the writer's own editing effort.
Biometric data: Biometric data is personal data produced by technical processing of a person's physical, physiological or behavioral characteristics that allows or confirms who that person is. A voice recording becomes biometric data when it is processed to identify the speaker.
Bring your own key (BYOK): Bring your own key (BYOK) is a setup in which you supply your own API key for an AI or speech provider. The app then sends requests through your account with that provider, so your agreement with the provider governs the data.
Business associate agreement (BAA): A business associate agreement (BAA) is the written contract that HIPAA requires before a covered entity lets a vendor create, receive, maintain or transmit protected health information on its behalf. Dictation tools used in clinics often need one, and a BAA alone does not make a product compliant.

C

CART captioning: CART captioning is a live text service in which a trained captioner types what is said, usually on a stenotype machine, and the words appear on a screen as they are spoken. CART stands for Communication Access Realtime Translation, and it is also called real-time captioning.
Character error rate (CER): Character error rate (CER) is the number of character insertions, substitutions and deletions needed to turn a recognizer's output into the correct text, divided by the number of characters in the correct text. It is the usual measure where words are not separated by spaces.
Characters per minute (CPM): Characters per minute (CPM) is the number of characters typed in one minute, counting spaces and punctuation. Words per minute is the same count divided by five, so a typist at 50 WPM enters 250 characters per minute.
Clean verbatim: Clean verbatim is a transcript that removes filler words, stutters, repetitions and false starts while keeping the speaker's wording and meaning. It is the usual choice for reading, captions and coding, where hesitations add noise rather than information.
Closed captions (CC): Closed captions (CC) are a text version of a program's dialogue and relevant sounds that the viewer can turn on or off. The word closed means the viewer chooses whether to see them, which separates them from open captions.
Cloud dictation: Cloud dictation sends your recorded speech to a remote server, which recognizes it and returns the text. It usually needs an internet connection, and your audio leaves your device, so check the provider's policy before you dictate sensitive material.
Code-switching: Code-switching is the habit of moving between two or more languages within a conversation or a single sentence. Speech recognizers built around one language often misread the switched words, so mixed speech is harder to dictate than speech in one language.
Colemak: Colemak is an alternative English keyboard layout, introduced in 2006 and designed for efficient, comfortable touch typing. Its supporters describe it as the third most popular layout for English touch typing, after QWERTY and Dvorak, but that ranking has no stated method.
Conformer: A Conformer is a neural encoder for speech recognition that combines self-attention, which sees the whole input, with convolution, which captures local sound patterns. FastConformer is NVIDIA's faster variant of the same design, used in streaming models.
Connected experiences (Microsoft 365): Connected experiences are Microsoft 365 features backed by cloud services, and Dictate is one of the experiences that analyze your content. You can turn them off in Manage Settings, but that setting also turns off other services.
Connectionist temporal classification (CTC): Connectionist temporal classification (CTC) is a way to train a neural network to turn audio into text without knowing in advance which moment of sound matches which letter. The network labels every short slice of audio with a character or a blank, and the blanks and repeated labels are then removed to leave the transcript.
Copilot+ PC: A Copilot+ PC is a class of Windows 11 hardware with a neural processing unit (NPU) that Microsoft says can run at 40 or more trillion operations per second. Some newer Windows AI features, including fluid dictation, require this hardware.
Copytaker: A copytaker was a typist in a newspaper office who took down stories that reporters dictated over the telephone. The Collins English Dictionary defines the role this way, and it predates the computer-based filing that replaced it.
Cursorless: Cursorless is a spoken language for structural code editing that lets developers code by voice. Every command is an action performed on a target in the code, and it runs on Talon with a VS Code extension.
Custom vocabulary: A custom vocabulary is a list of words, names and phrases that you add to dictation software so it recognizes and spells them as you intend. It matters most for names, jargon and product terms the recognizer has not learned.

D

De Quervain's tenosynovitis: De Quervain's tenosynovitis is a painful condition in which tendons that move the thumb become irritated where they pass through a narrow tunnel on the thumb side of the wrist. Repeated thumb and wrist movements, such as texting, can worsen it.
Diacritics: Diacritics are marks added to a letter, such as the acute accent in é, the tilde in ñ or the vowel marks in Arabic, that change how a letter sounds or which word it spells. Speech tools and everyday writers often leave them out, so a dictated word can come out spelled differently from what you meant.
Dialect: A dialect is a regional or social variety of a language, with its own vocabulary, grammar and pronunciation. Speech recognition is usually less accurate on dialects than on the standard form, because labeled training data for dialects is often scarcer.
Dictaphone: A Dictaphone is a machine that records a speaker's voice so that another person can play it back and type it later. The name began as a brand, and it became the everyday word for dictation machines.
Dictate (Microsoft 365): Dictate is the speech-to-text feature in Microsoft 365 apps such as Word, Outlook and PowerPoint. It needs a Microsoft 365 subscription, a microphone and a reliable internet connection.
Dictation: Dictation is speaking your words so that software writes them as text, usually into the app that has the cursor. Phones and computers now include it, and people also call it voice dictation or voice typing.
Dictation (ChromeOS): Dictation on ChromeOS is a built-in accessibility feature that types recognized speech into an editable field. You turn it on with Search + D or the microphone icon in the status tray.
Dictation foot pedal: A dictation foot pedal is a footswitch that controls playback of a recording, so a typist can start, pause and rewind audio without using their hands. It is the traditional control for transcribing dictation.
Dictation software: Dictation software is a program that recognizes your speech and types the words where your cursor is, as you say them. It comes built into many phones and computers and is also sold as separate apps.
Dictation template: A dictation template is a block of text saved under a short spoken phrase, so one phrase inserts the whole block. Dragon calls these auto-texts, and Microsoft's Dragon Medical One guide documents the command that creates one from selected text.
Digital dictation: Digital dictation is recording speech as digital audio, then sending the file to a typist or to speech recognition software for transcription. It replaced tape with files that can be encrypted, uploaded or routed for processing.
Disabled Students' Allowance (DSA): Disabled Students' Allowance (DSA) is UK government support for students with a disability, mental health problem or long-term illness, to cover study costs that come from the condition. It depends on individual needs rather than household income, and it does not have to be paid back.
Dragon Medical One: Dragon Medical One, abbreviated DMO, is Nuance's speech recognition product for clinicians who dictate notes into health records. Microsoft now combines it with DAX Copilot in a product called Dragon Copilot.
Dragon NaturallySpeaking: Dragon NaturallySpeaking, usually called Dragon, is a speech recognition program for dictating text on a computer. Dragon Systems released it in 1997, and Nuance has sold it in several editions since.
Dragon Professional Anywhere: Dragon Professional Anywhere is a subscription edition of Dragon that Nuance sells for professional and enterprise users. It is often shortened to Dragon Anywhere, a name that also belongs to Nuance's mobile app, so check which product a listing means.
Dvorak Simplified Keyboard: The Dvorak Simplified Keyboard is an alternative English keyboard layout that places the most common letters on the home row. Dvorak and Dealey applied for a patent on it in 1932, but the studies that claimed it was faster than QWERTY have been criticized, and the evidence for a speed gain is weak.
Dysarthria: Dysarthria is a speech disorder that happens when brain or nerve damage changes the way the muscles used for speaking work, so speech can sound slow, slurred or hard to understand. Speech recognition trained on typical voices can make more errors with this speech, so accuracy depends on the person and the software.
Dysgraphia: Dysgraphia is a difficulty with writing that does not match a person's cognitive level and age, even after adequate instruction. It can affect letter formation, spacing, spelling, writing speed and composition, and speech to text lets a person get their words onto the page without writing each letter by hand.

E

Electronic health record (EHR): An electronic health record (EHR) is a digital record of a patient's health information, built up over time from visits across care settings. The federal eCQI glossary also calls it an electronic medical record (EMR), and it is where dictated and typed notes end up.
Enhanced Dictation: Enhanced Dictation was a Mac option, available from OS X Mavericks, that lets you dictate without an internet connection after a download. Apple's help for macOS Mojave describes it, but Apple's later Dictation pages do not mention it.

F

False accept (accidental activation): A false accept, also called a false wake or accidental activation, is when a voice assistant or dictation feature starts listening or recording although the wake word was not said. Smart speakers have been measured to wake by accident, and that can capture speech the user did not mean to share.
Federated learning: Federated learning is a way to train one machine learning model on data spread across many devices. Each device computes an update from its own data, and only those updates are sent to a server, so the raw data does not have to be collected in one place.
Filler words: Filler words are sounds and words such as um, uh, like and you know that speakers produce while they plan what to say next, without adding content. Many dictation tools can remove them automatically, and research on transcripts found they can be deleted without compromising meaning.
Fluid dictation: Fluid dictation is a Windows feature on Copilot+ PCs that corrects grammar, punctuation and filler words as you speak. Microsoft says it is powered by small language models that run on the device, and that you can undo a correction by saying Revert or Undo that.
Forced alignment: Forced alignment is the process of matching a known transcript to its audio so that each word or phone gets a start and end time. Unlike open speech recognition, it does not decide what was said; it only finds when each word was said.
Free tier: A free tier is the no-cost version of a dictation app, with limits such as a weekly word cap, fewer modes or no AI processing. Read those limits before you rely on it, because a free plan can run out in the middle of your work.
Front-end speech recognition: Front-end speech recognition is a workflow in which the person who dictates also reviews, corrects and signs the text, instead of passing the audio to someone else. The term is most common in radiology, where it is set against back-end transcription.

G

Gboard: Gboard is Google's keyboard app for Android and iPhone, and its microphone key turns speech into text in any app where you can type. On Android it can run speech recognition on the phone, while on iPhone the voice input is handled inside the Gboard app and sent to Google.
Generative AI: Generative AI is the class of AI models that create new content, such as text, audio or images, from patterns in their training data. The NIST profile defines it as models that emulate the structure and characteristics of input data to generate derived synthetic content.
Generative error correction (GER): Generative error correction (GER) passes a speech recognizer's output, often its N-best list of alternative transcripts, to a large language model, which rewrites the transcript to fix errors. It can recover words the recognizer never offered, but it can also change words that were already right.
Globe key (Fn key): The Globe key, also called the Fn key or function key, is a modifier on Apple keyboards. In Keyboard settings you choose what a press does, such as switching input source, starting Dictation or doing nothing.
Google voice typing: Google voice typing is the speech-to-text input that Google provides on Android, so you can speak words into a text field. Its name and location vary by phone: Gboard shows a microphone key, while Samsung Galaxy phones list Google Voice Typing next to their own voice input.

H

Hidden Markov model (HMM): A hidden Markov model (HMM) is a statistical model in which a sequence of hidden states produces the observations you can measure, and each state changes to the next with set probabilities. In speech recognition, the hidden states are units of sound and the observations are short frames of audio.
Home row: The home row is the middle row of keys on a keyboard, where the fingers rest when you touch type. On a QWERTY keyboard it holds A, S, D, F for the left hand and J, K, L and semicolon for the right, and every other key is reached by moving a finger away from it and back.
Homophone: A homophone is a word pronounced the same as another word but different in meaning, spelling or both, such as to, too and two. Dictation software has to choose the right spelling from context, so homophones are one of the most common errors in dictated text.
Hunt and peck: Hunt and peck is a typing method in which the typist finds each key by sight and presses it with one or two fingers, usually the index fingers. It is also called two-finger typing, and it differs from touch typing, which relies on memorized key positions and the home row.

I

Improve Siri & Dictation: Improve Siri & Dictation is the Mac setting that controls whether Apple may store and review audio from your Siri and Dictation requests. When you first turn on Dictation, the same choice is labeled Share Audio Recordings.
Individualized Education Program (IEP): An Individualized Education Program (IEP) is a written statement for a child with a disability, developed, reviewed and revised under the federal IDEA regulations. For US students, it is the school's plan for the special education and services the child needs, and the team must consider whether the child needs assistive technology.
Input method editor (IME): An input method editor (IME) is software that supports text input for characters that cannot be typed easily on a standard keyboard, such as many East Asian characters. Speech input can also be an input method, and Android allows only one input method to be enabled at a time.
Intent recognition: Intent recognition is the step that decides what a spoken request is asking for, such as setting a timer or sending a message, and which details the action needs. Voice assistants and voice agents rely on it before they act.
Inverse text normalization (ITN): Inverse text normalization (ITN) is the step that converts a speech recognizer's spoken-form output into written form, so that 'two hundred' becomes 200. It changes how the text reads, not which words the recognizer heard.

K

Keystroke dynamics: Keystroke dynamics is the identification of a person from the rhythm of their typing, using the timing between key presses as a pattern much like a handwritten signature. It is a behavioral biometric, so it supports identity checks through statistical matching rather than proving who typed.
Keystrokes per character (KSPC): Keystrokes per character (KSPC) is the average number of keystrokes a text entry method needs to produce one character of text. A basic QWERTY keyboard scores 1, methods that need more presses score above 1, and word prediction can score below 1.
Keystrokes per hour (KPH): Keystrokes per hour (KPH) is the number of key presses made in one hour. It is the standard unit for numeric data entry, where many jobs require 8,000 or 10,000 keystrokes per hour, and it counts keys rather than words.

L

Language model (in speech recognition): In speech recognition, a language model (LM) scores how probable a word sequence is in a language. The recognizer combines that score with the acoustic evidence to choose between words that sound alike, such as homophones.
Large language model (LLM): A large language model (LLM) is a language model with a very large number of parameters, trained on huge amounts of text to predict and generate language. Google's glossary defines the minimum as a language model having a very high number of parameters.
LibriSpeech: LibriSpeech is a free corpus of about 1,000 hours of read English speech, taken from LibriVox audiobooks, and it is one of the most widely used benchmarks for English speech recognition. Its scores describe clean, read speech, which differs from everyday dictation.
Live Captions: Live Captions is a feature that shows text for speech and audio as it happens, using the phone's own processing. Google calls it Live Caption on Pixel phones, and Apple calls it Live Captions on iPhone. It reads what others say, so it is the reverse of dictation.
Live Transcribe: Live Transcribe is a Google app for Android phones that turns the speech of people around you into captions as they talk. It helps you follow a conversation. It does not type your own words into a text field, which is what dictation does.
Locale: A locale is a language combined with a region, and sometimes a script, such as en-GB for English as used in the United Kingdom. Dictation and speech settings use locales because the same language can differ in spelling, punctuation and speech patterns from one region to another.
Low-resource language: A low-resource language is one with little digital speech and text data for training speech recognition and other language tools. There is no fixed cutoff, and a language can be well resourced for one task and poorly resourced for another.

M

Machine learning: Machine learning is a method in which a program learns its behavior from example data rather than from rules written by hand. Google's glossary defines it as a program or system that trains a model from input data.
Median nerve: The median nerve is one of the main nerves of the arm. It runs from the forearm through the wrist into the hand, giving feeling and movement to parts of the hand.
Medical scribe: A medical scribe is a person, or more recently an AI tool, that records what happens in a clinical visit and writes it into the chart so the clinician can focus on the patient. The clinician remains responsible for the note and must check and sign it.
Mel spectrogram: A Mel spectrogram is a grid of sound energy over time, with frequency bands spaced to match human hearing. Many speech recognizers compute its logarithm, the log-Mel spectrogram, and feed it to a neural network instead of the raw waveform.
Microphone key (dictation key): The Microphone key is a key on some Mac keyboards that starts Apple Dictation when you press and release it. Pressing and holding it activates Siri instead, if Siri is turned on.
Modern Standard Arabic (MSA): Modern Standard Arabic (MSA), also called Fusha, is the standardized form of Arabic used in writing and formal speech, such as news broadcasts and official documents. Speakers use regional dialects in everyday conversation, and speech tools recognize formal Arabic more accurately than dialect speech.

N

Neural Engine: The Apple Neural Engine (ANE) is a block of hardware inside Apple's chips that runs machine learning models efficiently. Speech models can run on it, and Apple uses it for on-device machine learning inference.
Neural processing unit (NPU): A neural processing unit (NPU) is a specialized chip for AI-intensive processes, such as real-time translation and image generation. Microsoft says many new Windows AI features require an NPU that can run at 40 or more trillion operations per second, and a speech model can run faster on one.
Noise-canceling microphone: A noise-canceling microphone is designed to filter ambient noise, so it picks up mainly the speaker's voice. It can help dictation in noisy rooms, but it does not remove all noise, and it is different from the active noise cancellation in headphones.
Non-native speech: Non-native speech is speech in a language that is not the speaker's first language, often called L2 speech or foreign-accented speech. Recognizers can transcribe read non-native English well, but they handle hesitations, repetitions and revisions differently, and results depend on the system.
Nuance Communications: Nuance Communications was a speech technology company best known for the Dragon family of speech recognition products. Microsoft announced that it had completed its acquisition of Nuance on March 4, 2022.
Number and grid overlays (Voice Access): Number and grid overlays are Voice Access views for clicking things you cannot name. Show numbers puts a number on each clickable item so you can say it, and Show grid divides the screen into numbered tiles that narrow the mouse pointer step by step.

O

On-device speech recognition: On-device speech recognition converts speech to text with a model that runs on your own computer or phone, so the audio does not need to be sent to a server. It is also called local, offline or on-device dictation.
One-party consent: One-party consent is a recording rule under which a conversation can be recorded lawfully when one of its participants agrees, which is usually the person doing the recording. The US federal wiretap law uses this rule, and individual states may set stricter ones.
Online speech recognition: Online speech recognition sends your voice to a cloud service that turns it into text, instead of processing it on your PC. Microsoft says voice typing on Windows uses online speech recognition, which Microsoft describes as giving more accurate recognition.
Open captions: Open captions are text drawn onto the frames of a video, so every viewer sees them and nobody can switch them off. They are also called burned-in captions or hardcoded subtitles.
Oral composition: Oral composition is working out and shaping a text by speaking it, whether to a dictation tool, to a recorder or silently in thought. Research on letters finds that planning takes most of the time, however the text is produced, so speaking mainly changes how the words get onto the page.
Oral reading fluency (ORF): Oral reading fluency (ORF) measures how many words a reader gets right in one minute of reading a passage aloud. Schools record the result as words correct per minute (WCPM), which subtracts the words read wrongly from the total.

P

Pajama time: Pajama time is the record-keeping that clinicians finish after their clinic day ends, often at home in the evening. A 2017 study of primary care clinicians measured 86 minutes of EHR work per weekday after clinic hours.
Perpetual license: A perpetual license is a software license you pay for once. It lets you keep using that version without renewing. Updates, support and new major versions are often separate, time-limited or paid extra.
Phi Silica: Phi Silica is Microsoft's on-device language model, which runs on the neural processing unit of Copilot+ PCs. Microsoft says it is being replaced by Aion Instruct, and its voice typing help names it as the download to watch while fluid dictation gets ready.
Pinyin: Pinyin, also called Hanyu Pinyin, is the system that spells Mandarin Chinese syllables with the Latin alphabet and marks the tones with diacritics. People type Chinese by spelling the sound in Pinyin and choosing the character they want, and speech tools must choose among characters that share one Pinyin spelling.
PowerScribe: PowerScribe is a family of radiology reporting products in which a radiologist dictates an imaging report by voice. It began as a Nuance dictation application, and Microsoft completed its purchase of Nuance in 2022, so Microsoft now sells the line.
Prompt (AI): In AI, a prompt is the input you give a model to get a response, such as a question, an instruction or pasted text. Google's glossary defines it as any text entered as input to a large language model to condition the model to behave in a certain way.
Protected health information (PHI): Protected health information (PHI) is individually identifiable health information that HIPAA protects when it is held or sent in electronic or other form by a covered entity or its business associate. Patient names with diagnoses, in dictated notes or voice recordings, are PHI. It is also called ePHI when held electronically.
Push-to-talk: Push-to-talk is a dictation mode in which you hold a key while you speak and release it to stop, so the software writes only the speech made while the key is down. Some tools also offer tap-to-start and tap-to-stop.

Q

QWERTY: QWERTY is the standard letter arrangement on English keyboards, named for the first six letters of the top row. Christopher Latham Sholes devised it in the early 1870s for typewriters, and it is still the layout most people type on, even though faster layouts have been proposed.

R

Real-time factor (RTF): The real-time factor (RTF) is the time a system takes to process audio divided by the length of that audio. An RTF of 1 or lower means the system keeps up with real-time speech. Some leaderboards report the inverse, RTFx, where higher is faster.
Real-time text (RTT): Real-time text (RTT) is a way of sending text as it is typed, so the other person can read each word before the sender finishes. It is used in conversation by many deaf and hard of hearing people, and it has no send step like an ordinary text message. It is a typing channel, not a speech technology.
Reasonable accommodation: A reasonable accommodation is a change to the job, the workplace or the equipment that lets a qualified person with a disability do the work. Under U.S. law, employers with 15 or more employees must provide one unless it would impose an undue hardship.
Registered Healthcare Documentation Specialist (RHDS): The Registered Healthcare Documentation Specialist (RHDS) is a voluntary credential from AHDI for people who create healthcare documentation. It is awarded after an exam of 120 questions, and it is a prerequisite for the CHDS.
Repetitive strain injury (RSI): Repetitive strain injury (RSI) is an umbrella term for pain and damage to muscles, tendons and nerves that builds up from repeated movements, held postures or sustained effort, felt in the hands, wrists, arms, shoulders and neck.
Right-to-left (RTL) script: A right-to-left (RTL) script is a writing system, such as Arabic, Hebrew, Persian or Urdu, whose letters run from right to left. Text that mixes these letters with digits or Latin words is bidirectional, so software needs a specific algorithm to order and display it.
RNN transducer (RNN-T): An RNN transducer (RNN-T) is a neural speech model that reads audio in order and emits words as it goes, so text can appear while the speaker is still talking. It has an encoder for the sound, a prediction network for the words already written, and a joint network that chooses the next output.
Rollover typing: Rollover, or key rollover, is the number of key presses a keyboard can detect when they overlap. Two-key rollover reliably detects only two keys at once, and a matrix keyboard without diodes can register a ghost key when three keys are pressed together.

S

Samsung voice input: Samsung voice input is the voice-to-text feature built into the Samsung Keyboard on Galaxy phones and tablets. It works only with that keyboard, so it is a separate choice from Google Voice Typing.
Scopist: A scopist edits the transcripts of official proceedings that court reporters create. Scopists are typically paid by the page, starting from $1.00 to $1.50 per page, and the role is also called transcript editor.
Screen reader: A screen reader is software that reads what is on a computer or phone screen aloud, or sends it to a braille display, so that people who are blind or have low vision can use the device. The most used screen readers on Windows are JAWS and NVDA, and VoiceOver is the one built into Apple devices.
Section 504 plan: A Section 504 plan is the written record of the accommodations a US public school provides to a student with a disability, under Section 504 of the Rehabilitation Act of 1973. It is a civil rights arrangement and differs from an Individualized Education Program (IEP), which comes from special education law.
Shorthand: Shorthand is an abbreviated symbolic writing method that increases the speed of writing compared with longhand. Trained writers used it to take dictation before recording machines, and it still underpins court reporting.
Siri: Siri is Apple's built-in voice assistant. You speak a request and it answers or carries out an action, while Dictation types what you say into a text field instead.
Site microphone permission: Site microphone permission is the browser's setting for each website that decides whether that site may use your microphone, usually chosen when the browser prompts you. It works alongside the operating system's own microphone setting, so both must allow the microphone before a site can listen.
SOAP note: A SOAP note is a structured clinical note with four parts: Subjective, Objective, Assessment and Plan. The format was developed by Larry Weed, and a StatPearls reference article describes it as a widely used documentation method for clinicians.
Speaker diarization: Speaker diarization is the task of working out who spoke when in a recording, by splitting the audio into turns and grouping turns that come from the same voice. It is separate from transcribing the words, and a transcript gains speaker labels only when diarization is run alongside recognition.
Speaker verification: Speaker verification is a voice biometric check in which a system accepts or rejects a claimed identity by comparing a voice with stored models of the person claimed. It is also called voice authentication, voice ID or voice biometrics.
Special category data (GDPR): Special category data is the group of personal data that Article 9 of the GDPR bans organizations from processing unless a listed exception applies. It includes health data and biometric data used to identify a person, and a voice recording can fall into it when it is processed that way.
Speech framework (Apple): The Speech framework is Apple's developer toolkit for turning audio into text inside an app. It includes the older SFSpeechRecognizer class and the newer SpeechAnalyzer, which Apple says transcribes entirely on the device once its models are downloaded.
Speech language model: A speech language model (speech LLM) is a large language model that accepts audio as input, so it can process speech directly instead of working only on a text transcript. The 2023 SALMONN model is one example that lets an LLM process general audio inputs.
Speech pack (Windows): A speech pack is the set of on-device language files that Windows needs before it can recognize speech in a language offline. Microsoft says you must download a speech pack for a language in Voice access before you can use it, and the first download needs an internet connection.
Speech Services by Google: Speech Services by Google is the Android speech engine from Google that turns speech into text for other apps and reads on-screen text aloud. Google's Play Store listing now uses the title Speech Recognition & Synthesis, and reviews on that listing still call the app by the older name.
Speech translation: Speech translation converts speech in one language into the same meaning in another language, as text or as synthesized speech. Most systems first recognize the speech and then translate the text, so a recognition error carries into the translation.
Speech-to-speech model: A speech-to-speech model takes audio in and produces audio out with one model, without writing the conversation out as text in between. Most voice systems instead chain speech recognition, a language model and speech synthesis.
Spoken language identification: Spoken language identification is the task of working out which language is being spoken in a recording. Products call it automatic language detection. It usually chooses from a short list of candidate languages and decides for a segment of audio rather than word by word.
Spoken punctuation: Spoken punctuation is saying the name of a punctuation mark, such as comma or question mark, so that dictation software writes that mark into the text. It is the manual alternative to automatic punctuation.
Stenotype: A stenotype, also called a steno machine, is a chorded keyboard that stenographers use for shorthand, pressing several keys together to produce a syllable or word. The NCRA's court reporter skills test sets a speed of 225 words per minute, and voice writing is a newer method that uses speech recognition instead.
SubRip subtitle file (SRT): A SubRip subtitle file, or SRT, is a plain-text file of numbered caption blocks. Each block has a start and end time in hours, minutes, seconds and milliseconds, followed by the caption text. It is one of the most common caption formats on the internet.
Subtitles for the deaf and hard of hearing (SDH): SDH stands for subtitles for the deaf and hard of hearing. They show the dialogue plus the sound effects, music and speaker identification that a hearing viewer would get from the soundtrack.

T

Talon Voice: Talon Voice, usually called Talon, is hands-free input software for writing, coding, browsing and gaming. It runs alongside a speech-recognition engine, and it is used for voice coding, often together with the Cursorless editing language.
Tendinitis: Tendinitis is inflammation of a tendon, the tissue that connects muscle to bone. It can occur as a result of injury or overuse, and repeated movements can trigger the overuse kind.
Text insertion: Text insertion is the step where dictation software places recognized words into the text field that has the cursor. It is done either by simulating keystrokes or by pasting the text from the clipboard.
Text to speech (TTS): Text to speech (TTS), also called speech synthesis, is software that converts written text into spoken audio. Modern systems use neural networks, and they can run on the device or in the cloud.
Third-party keyboard: A third-party keyboard is a keyboard app from a developer other than the phone's maker, which you install and switch to in place of the built-in keyboard. Apple calls the iPhone version a custom keyboard, and Apple's rules limit what it can do, such as network access and microphone use.
Timestamp (transcription): A timestamp in a transcript is a time marker, such as 00:12:15, that shows when the nearby text was spoken in the recording. Timestamps let readers jump to a passage, and they can be attached to whole segments or to single words, with different accuracy.
Transcribe (Microsoft 365): Transcribe is a Word for Microsoft 365 feature that turns an uploaded audio file into text in the Transcribe pane, rather than typing your speech into the document. Its monthly limits depend on your license.
Transcript Assist (Galaxy AI): Transcript Assist is a Samsung Galaxy AI feature that converts a saved recording in the Voice Recorder app into text. It works on select Galaxy models and needs a working internet connection. It transcribes recordings you already made, so it is not live dictation into a text field.
Transcription foot pedal: A transcription foot pedal is a switch that controls audio playback with the foot, so a transcriber can start, pause and rewind a recording while typing. It is also called a foot control or foot pedal.
Transcription line rate: A transcription line is a unit of 65 characters, and the line count is the number of such units in a document. The transcription line rate is the price paid for each line, so the unit controls how transcribed work is measured.
Transcriptionist: A transcriptionist is a person who listens to recorded speech and types it as text, such as a clinician's dictated report. Speech recognition now writes first drafts, so much of the work has shifted to checking and correcting machine text.
Transliteration: Transliteration represents the text of one language in the writing system of another, for example by writing a word in Latin letters. It keeps the spelling or sound and does not translate the meaning, which is what separates it from translation.
Turn detection: Turn detection is the step that decides when a speaker has finished a turn, so a voice system knows when to respond. A pause can be a hesitation rather than an ending, so the decision trades speed against cutting the speaker off.
Turnaround time (TAT): Turnaround time (TAT) is how long a transcriber or service takes to deliver a finished transcript after an order is received. Court schedules define it in calendar days or hours, counted from the order.
Typing accuracy: Typing accuracy is the share of typed characters or words that are correct, usually given as a percentage. A typing score should always report accuracy beside speed, because a fast typist who leaves many uncorrected errors has not typed well.

V

Verbatim transcription: Verbatim transcription writes down exactly what is said, including filler words, false starts, repetitions and non-speech sounds, with no tidying. Researchers and legal teams use it when the way someone speaks is part of the record.
Vibe coding: Vibe coding is building software by describing what you want to an AI model and accepting the code it writes, without reviewing that code. Andrej Karpathy introduced the term, and the developer Simon Willison argues for a narrower meaning than the broad one now in use.
Voice Access (Android): Voice Access (Android) is a Google feature that lets you operate an Android phone or tablet by speaking commands, such as opening an app or tapping a labelled item. Google built it for people who have difficulty using a touch screen. It is a control tool that can also enter text, and it is separate from Windows Voice Access on PCs.
Voice Access (Windows): Voice Access is the Windows 11 feature that lets you control the whole PC and write text by voice, not only dictate. Microsoft says it works without an internet connection after setup, and it replaces Windows Speech Recognition.
Voice activity detection (VAD): Voice activity detection (VAD) is a component that decides whether a stretch of audio contains speech. Speech recognizers use it to split audio into turns, skip silence and decide when a speaker has finished.
Voice agent: A voice agent is a voice system that listens, reasons with a language model, uses tools, and speaks back across several turns to finish a task, such as checking an order and rescheduling it. Benchmarks show that agents still fail many tasks on realistic audio.
Voice assistant: A voice assistant is software that listens for a spoken request, works out what you want, and then acts or answers, such as setting a timer or turning off a light. Dictation is different: it turns your speech into text for you to edit.
Voice cloning: Voice cloning is the use of a model trained on recordings of one person to generate new speech in that person's voice. Some products need only minutes of clear audio, and cloned voices have been used in fraud, so recordings of your voice carry a real risk.
Voice coding: Voice coding, also called programming by voice or coding by voice, is writing and editing source code by speaking instead of typing. It needs a speech-recognition engine and a separate platform that turns spoken words into code editing actions.
Voice command: A voice command is a spoken instruction that the software carries out, such as delete or stop listening, rather than writing the words as text. Dictation tools must separate commands from dictated words, and the rules differ by product.
Voice Control (Apple): Voice Control is Apple's feature for controlling a Mac with spoken commands, including opening apps, clicking items and editing text. When it is on, it also handles dictation, so standard macOS Dictation is not available.
Voice message transcripts: A voice message transcript is written text of the words in an audio message, produced by the messaging app so you can read instead of listen. Apple, Google and WhatsApp each offer one, with different defaults and different privacy arrangements. It turns audio into text, the reverse of dictation.
Voice profile: A voice profile is the set of files that dictation software keeps for one user, with settings such as the microphone they use and their custom words. It is not the same as speaker recognition, which identifies who is speaking.
Voice typing: Voice typing is dictation built into an operating system or app, so that speech is typed into a text field as you talk. Talk to text and speech typing are other names for the same act, and each platform uses its own name.
Voice typing (Google Docs): Voice typing in Google Docs is a built-in speech-to-text tool that types as you speak. You start it from the Tools menu in a supported browser, and the browser controls the speech-to-text service.
Voice typing launcher: The voice typing launcher is the Windows voice typing setting that lets you start dictation quickly when the cursor is in a text box. Microsoft lists it among the settings you open from the gear icon in the voice typing flyout.
Voice writing: Voice writing is a transcription method in which a trained writer repeats speech into speech recognition software through a stenomask, a mask that keeps the writer's voice from being heard in the room. It is used in court reporting, medical transcription and captioning.
VoiceOver: VoiceOver is the screen reader built into macOS, which speaks what is on the screen so you can use the Mac without seeing it. It does the reverse of dictation: it reads text aloud rather than writing your speech.
Voiceprint: A voiceprint is a stored model of a person's voice, built from measurable features of their speech and used to recognize or verify that person. Illinois law lists voiceprints among biometric identifiers, and other vendors call the same idea a voice template or speaker model.

W

Wake word: A wake word is a fixed phrase, such as Hey Google or Hey Siri, that a voice assistant listens for on the device before it starts handling a request. Until the phrase is heard, the assistant is only checking for that phrase.
wav2vec 2.0: wav2vec 2.0 is a speech model that first learns the structure of speech from large amounts of unlabeled audio, then is fine-tuned on a much smaller set of transcribed recordings. Because transcribed speech is expensive to make, this approach helps when labeled data is scarce.
Wayland: Wayland is the display system that Linux desktops are moving to in place of the X Window System, called X11. Its compositor delivers input to each app, so a typing tool that worked on X11 may not type on a Wayland desktop.
Web Speech API: The Web Speech API is a set of browser interfaces that let a web page start speech recognition and receive the recognized words as text, through an object called SpeechRecognition. Where the recognition runs, on your device or on a server, depends on the browser.
WebVTT (VTT): WebVTT (Web Video Text Tracks, usually shortened to VTT) is a plain-text file format for timed captions and subtitles on the web. A .vtt file is a list of cues, and each cue gives a start time, an end time and the text to show.
Windows key + H (Win+H): Windows key + H, written Win+H, is the keyboard shortcut that starts voice typing in Windows, so you can dictate into the text box where your cursor is. Microsoft says it does not start voice typing while Voice Access is running.
Windows on Arm: Windows on Arm is Windows running on processors that use the Arm instruction set instead of x86 or x64 chips. It runs native Arm apps and many unmodified x86 and x64 apps, and Microsoft says Arm-native apps give the best performance and battery life.
Windows Speech API (SAPI): The Windows Speech API (SAPI) is Microsoft's interface that lets Windows apps use speech recognition and text-to-speech without building the speech engine themselves. Windows.Media.SpeechRecognition is a separate API for command and control in Windows Runtime apps.
Windows Speech Recognition (WSR): Windows Speech Recognition (WSR) is the older speech feature built into Windows for voice control and dictation. Microsoft says it is deprecated and no longer being developed, and that Voice Access replaces it.
Word error rate (WER): Word error rate (WER) is the number of word substitutions, deletions and insertions in a transcript, divided by the number of words in the correct transcript. Lower is better, and a score above 100 percent is possible.
Words per minute (WPM): Words per minute (WPM) is the number of words typed in one minute. Typing tests count a word as five characters, including spaces, so WPM is a convention rather than a count of real words. The average in the largest online typing test was about 52 WPM.

Z

Zero data retention (ZDR): Zero data retention (ZDR) is a setting or contract term under which a service keeps none of your data once it has handled your request. Each vendor sets its own scope, so check what the label excludes before you rely on it.