What is punctuation restoration? It is the step that adds punctuation and capital letters to raw recognised words. Speech has no commas or full stops, so a recogniser first outputs a flat run of lowercase words. A second pass guesses sentence breaks from pauses and wording.
Transcribe a recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram
Files up to 2 GB · export to TXT, MD, DOCX, PDF, SRT and VTT · speaker labels and timestamps · about 30 languages
Key facts
- Pauses help but do not decide: people stop mid-sentence to think and run sentences together when they talk fast.
- Question marks are hard because spoken questions are often marked only by rising pitch, which the plain words do not show.
- Names, places and the start of each sentence need capital letters, and unusual names are the most likely to stay lowercase.
- The step does not change the words, so a recognition error stays an error with a full stop after it.
- Raw text without punctuation is much harder to read and also harder for later tools to split into sentences or subtitles.
What it looks like
// raw recogniser output
so we start at nine and then dana will talk about the budget any questions
// after punctuation restoration
So we start at nine, and then Dana will talk about the budget. Any questions?
More
See the full list of languages, the tools and the glossary.
Frequently asked questions
Why does raw speech recognition output have no punctuation?
The audio contains sounds and pauses, not symbols. A recogniser maps sounds to words, and punctuation has to be added afterwards as a separate step.
Is the punctuation always correct?
No. It is a best guess. Read the result for misplaced full stops, missing question marks and sentences that run on.
Do dictated commands such as full stop and comma count?
That is a different method, where the speaker says the marks aloud. Restoration works without any such words and adds the marks itself.
Does the Potok bot add punctuation?
Transcripts from the bot arrive as readable text, and long ones come as a file in TXT, MD, DOCX, PDF, SRT or VTT. Still read them over before you publish.
How can I improve the result?
Speak in clear sentences with short pauses between them, and record in a quiet room. Clean audio gives the guesser more to work with.
Sources
Each source was opened by hand on the date shown. Pages change, so open the source if a detail matters to you.
- Speech recognition (Wikipedia), Wikipedia · checked
- Transcription (linguistics) (Wikipedia), Wikipedia · checked
Try it on your own recording
Send your voice message, audio or video to @potok_now_bot on Telegram and the text comes back in the chat.
Open @potok_now_bot in Telegram