Skip to content

Speech recognition parameter for automatic/unspoken punctuation #187

@evanbliu

Description

@evanbliu

Description
I propose adding an unspokenPunctuation (or enableUnspokenPunctuation) attribute to the SpeechRecognition interface. This Boolean attribute would control whether the speech recognition engine automatically infers and inserts punctuation marks (such as periods, commas, and question marks) based on natural pauses, grammatical structure, and prosody, without requiring the user to explicitly speak the punctuation commands.

Rationale
Customization for Different Use Cases: Different speech recognition contexts require distinct handling of text flow. Casual voice typing, automated transcription, and conversational assistants greatly benefit from automatic punctuation to produce readable, polished text out of the box. Conversely, precise dictation tools, coding via voice, or raw acoustic logging applications may require verbatim, unpunctuated streams where punctuation is strictly controlled by explicit user commands.

Enhanced User Experience: Natural, continuous speech often lacks explicit spoken punctuation commands. Allowing developers to enable automatic punctuation lowers the cognitive load for end-users, making voice input feel more intuitive and conversational while saving developers from implementing complex downstream NLP models to handle basic text formatting.

Implementation Details
Code snippet

partial interface SpeechRecognition {
  attribute boolean unspokenPunctuation;
};

Default Value: The default value of unspokenPunctuation would be false to maintain backward compatibility with existing applications and ensure deterministic, unformatted text outputs unless explicitly opted into by the developer.

Metadata

Metadata

Assignees

Labels

No labels
No labels

Type

No type

Projects

No projects

Milestone

No milestone

Relationships

None yet

Development

No branches or pull requests

Issue actions