Pronunciation Examples

• How is my last name "Ran" pronounced?
• Same word in different phoneme: The farm was used to produce produce;
• An amusing story from my English teacher. What is the "Two to two to two two"?

The Problem

Text-to-speech pronunciation is often inaccurate and inconsistent because of technology limitations.

The Need

Accurate, consistent pronunciation of content spoken by text-to-speech (TTS) synthesis

• Screen readers used by people who are blind
• Read Aloud tools that assist language learners and people with cognitive and learning disabilities
• Voice-based assistants

Goals

• Define a standard way for content authors to provide pronunciation information in HTML
• Leverage Speech Synthesis Markup Language (SSML), if possible
• Solution that will be used by screen readers

Features

SSML features critical for implementation:

• Language
• Voice Family / Gender
• Phonetic Pronunciation of String Values
• String Substitution
• Rate / Pitch / Volume
• Emphasis
• Say As
• Pausing

Technical Approaches Considered

Two front runners:

1. In-line SSML within web content
2. Attribute-based model of SSML

Examples on last page of Slides

Working Draft Publications

• Explainer: Improving Spoken Presentation on the Web overview of the work
• Pronunciation User Scenarios provides examples of:
  • End-users, including screen reader users
  • Content providers, including educators
  • Software developers, including content managements systems
• Pronunciation Gap Analysis and Use Cases more detailed context, analysis of potential technical approaches, and more.

Start Here

Pronunciation Overview page:

https://www.w3.org/WAI/pronunciation/

Join the Task Forces

Interested? Join the Task Forces

Email Roy Ran: ran@w3.org

Your Input

• Questions?
• Comments?

Technical Examples

In-line SSML in an HTML fragment is shown below:

Attribute based model of SSML: