Speech recognition terms: a voice input glossary

Potok transcribes Russian speech; this site has been translated into English for convenience.

Speech recognition terms describe different parts of one technology: voice input is the general name, speech recognition (ASR) is the process of turning sound into text itself, and dictation is the same thing from the user's point of view. Potok (potok.now) has a response time of about 0.4 seconds: hold the hotkey, speak, release, and the text is ready.

Speech recognition, ASR, dictation, hotkey, WER: behind these words there are often simple things that are easier to explain in one phrase than to work out from context. Voice input as a phenomenon and the specific technical terms inside it are things of different levels, and they are easy to mix up.

Speech recognition terms in plain words

• Voice input is the general name of the technology: what is said aloud turns into text and appears in the active text field.
• Speech recognition (ASR, automatic speech recognition) is the process of turning the sound of a voice into text itself, what happens “under the hood” of voice input.
• Dictation is the same action from the user's point of view: you speak instead of typing by hand.
• Transcription is turning an already finished recording, audio or video, into text, unlike dictation, which happens in real time.
• Hotkey is the key combination that turns voice recording on and off; in Potok it is Ctrl+Space by default.
• WER (word error rate, the share of recognition errors) is the percentage of words recognized wrongly; the smaller the number, the more accurate the program.
• Language model is the part of a recognition system trained on a specific language and vocabulary; it determines how well terms, names and mixed speech are recognized.
• Personal dictionary is a list of words and names that the user adds by hand so that the program recognizes them more accurately; Potok has such a dictionary.

• Voice Access in Windows 11: what it is
• What a dialect is
• What an overlay is
• What the Bluetooth audio profile HFP/A2DP is
• What is exclusive fullscreen mode
• What is the system tray
• What WER (Word Error Rate) is

All articles in this section (181): potok.now/slovar

Which terms to start with

• What voice input of text is and how it works: in Potok it is the Ctrl+Space hotkey by default: hold, speak, release.
• What turning speech into text is called: in Potok it is cloud recognition by its own model trained for Russian speech.
• Local speech-to-text recognition: Potok works not locally but in the cloud: without the internet, recognition will not start.
• What dictation is: in Potok the free plan gives 2500 words of dictation per week on one device.
• Glossary of terms for dictating contracts and lawsuits: the Potok personal dictionary stores terms and names that are often repeated in one text.

Where Potok does not fit

These terms describe the technology as a whole, and not what Potok specifically can do. For example, offline recognition from the list above exists in some programs, but not in Potok: it is a cloud service, recognition runs on servers in Russia, and it does not work without the internet. Transcription of finished audio files and recordings is not what Potok does either: it recognizes live speech at the moment of dictation and does not transcribe what is already recorded. For finished recordings Potok has a Telegram bot, @potok_now_bot.

Frequently asked questions

Related software: terms compared

Built-in dictation and Potok

Primary sources: speakflow.ru · speakflow.ru · speakflow.ru · blog.speakflow.ru