More links, resources, and information... # Natural Language Processing -- Lexers A lexer or tokenizer (e.g. a Finite State Machine or FSM) will tokenize a string of text -- splitting that text into s

I'll have more later, but here is a start (welcome to the rabbit hole). A speech synthesizer voice typically consists of two parts: 1. the text to phonemes part; 2. the phonemes to audio part. The tex

This is https://github.com/espeak-ng/espeak-ng/issues/544. There have been disagreements over various attempts to improve the Russian language (e.g. https://github.com/espeak-ng/espeak-ng/issues/474).

What is the issue/are the issues regarding espeak (I have experience converting Python 2 to 3 code)? Are you referring to espeak wanting UTF-8 (which is like/compatible with ANSI/ASCII strings), or is

You need the latest nvda development/next branch for the eSpeak NG fixes. Not as far as I am aware. I'm planning on releasing 1.49.3 in line with Unicode 11, so it has the updated Unicode characters a

The issue referenced in this discussion talked about there being too much emphasis on the "n", "as if there is an extra "g" at the end of it before the next word begins. The only change made in that r

It is not an additional "g", it is the "n" sound before a "k" or "g" in words like "sing", "king", and "sink". Within words, this process happens naturally. Across words, that process may happen (espe

It would help to read the follow up where I have modified it so it no longer does that between words. Both pronunciation styles are equally valid and I prefer the change I made, but I respect that the

Regarding your query about eSpeak NG, a lot of fixes for issues have been made in the development version of eSpeak NG. Those changes have not been merged into NVDA yet. As to why eSpeak NG? The origi

Thanks for this feedback. I'm thinking about replacing sonic, possibly using something like https://blog.demofox.org/2018/03/05/granular-audio-synthesis/ for all the audio speed up/slow down so the be

I have pushed a fix for the pronunciation of words like "possible" for the English and British English accents. Historically, unstressed "I" vowels were pronounced with that "I" sound, while over time

I have identified and fixed the issue with some Tamil characters causing eSpeak NG to freeze on Sunday. I have made other various stability fixes to the codebase as well. Hopefully, that should addres

Taking the regression in the pronunciation of "resignedly" as an example, this was caused by a fix to words like "allegedly" and "reputedly" that were misporounced in espeak. The issue is that those w

The en-GB (British English) voice for espeak-ng pronounces those words the same as it has done in version 1.49.0 (which was the first espeak-ng release, released in September 2016). I have fixed "resi