Glossary
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.
Updated October 17, 20263 cited facts, each traced to its source
Also called ML.
Machine learning (ML) is a way of building software whose behavior comes from data. Instead of writing a rule for every case, an engineer supplies many examples, and a training procedure adjusts a model until its outputs match them. Google's Machine Learning Glossary defines it as a program or system that trains a model from input data 1. The trained model then makes predictions about new input it has not seen before.
Training replaces hand-written rules
Imagine a program that must tell a spoken "yes" from a spoken "no". A rule-based program needs someone to list the sound patterns of each word and every exception to them. A machine learning system is shown thousands of recordings labeled with the right word. It adjusts its internal numbers, called parameters, until it gets most of them right, and then it applies what it learned to recordings it has never heard.
Three steps: data, training, prediction
- Collect examples with the correct answers, such as recordings paired with verified transcripts.
- Train the model by repeatedly comparing its output with the answer and adjusting its parameters to reduce the error.
- Predict on new input, such as the sentence you dictate today.
The result depends mostly on the examples. A model trained on clean studio recordings may struggle in a noisy kitchen, and a model trained mostly on one accent may miss others. That is why the same dictation software can work well for one speaker and poorly for another.
Why it matters for dictation
Modern speech recognition is machine learning. In a 2012 review, a deep neural network reduced the word error rate of an older statistical baseline from 27.4 to 18.5 percent on a Switchboard telephone task 2. That was a telephone test, not everyday dictation, but it shows why the field moved to learned models.
Machine learning also affects where your audio is processed. Apple says on-device machine learning benefits privacy because the data for inference workloads stays on the device rather than on a server 3. A cloud dictation service runs a similar kind of model on its own servers, so your audio leaves your device. When you compare dictation apps, the model's training data is hidden from you, so judge the result instead: test the same paragraph in each app and compare the errors, using word error rate.
How it relates to the other terms
Machine learning is the broad field. Generative AI refers to the models that create new content, and most of them are built with machine learning. A large language model is a very large text model trained with the same methods. Speech recognizers use machine learning to map sound to words, and the hidden Markov model was the statistical method that older recognizers used before neural networks took over.
Related terms
- Generative AI: models that create new content, often built with machine learning.
- Large language model (LLM): a very large text model trained on language data.
- Automatic speech recognition (ASR): the speech task that dictation software solves with machine learning.
- Hidden Markov model: the statistical method used by many older speech recognizers.
- Word error rate (WER): the measure used to compare recognizers.
Guides that cover it
Evidence
Every number on this page, with where it comes from. Open a fact to see the exact words in its source.
- 1.
- 2.
- 3.
Related terms
- Speech to Text AI: Dictation, Assistants and Voice AgentsThe topic this term belongs to
- AI text detectorAn 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.
- Generative AIGenerative 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.
- Intent recognitionIntent 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.
- 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.
- 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.
- Speech language modelA 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-to-speech modelA 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.
- 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.
- Turn detectionTurn 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.
- Voice agentA 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 assistantA 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.