NVDA2025.1beta1+ doesn't speak ordinal numbers in Portuguese

[tiago-casal]

Until NVDA 2024.4.2, ordinal numbers were normally spoken in Brazilian Portuguese, while the beta versions of the 2025.1 series ordinal numbers in Brazilian Portuguese are no longer spoken by default, making the reading experience with NVDA bad in this type of situation. This issue also affects the Spanish language.

In English the ordinal numbers have suffixes with different letters after the digits, example, 1st, 2nd, 3rd, 4th, etc., in Portuguese it is simpler, there are only two types of suffix for ordinal numbers, feminine ordinal indicator ª (U+00AA) and masculine ordinal indicator º (U+00BA), examples, 1ª, 1º, 2ª, 2º, 3ª, 3º, 4ª, 4º, etc., the proper pronunciation of ordinal numbers with such symbols is implemented by default in speech synthesizers in Brazilian Portuguese (eSpeak NG and others).

In addition to the problem with ordinal numbers, in the Portuguese language there is an abbreviation that uses the ordinal symbol that is also no longer spoken, "nº" and "Nº", they are the abbreviations of the word "number" in Portuguese, instead of being spoken number "número" with unicode normalization started to speak "no" (in Portuguese "no" means "in the" in English and other equivalent expressions, very different from the word "number").

This problems is in the unnecessary unicode normalization of the U+00AA and U+00BA characters, which instead of passing to the synthesizers the character ª (U+00AA) and º (U+00BA) is sending in their place the letters A and O, this is not the same thing as ª and º, making the synthesizers in Portuguese start to speak "1 A" instead of first "primeira", "2 O" instead of second "segundo", etc.

Keeping unicode normalization enabled is very useful, just put an exception to not normalize the characters ª (U+00AA) and º (U+00BA) and the problem will be solved for the Portuguese and Spanish languages.

Steps to reproduce:

1. Use a voice in Brazilian Portuguese (eSpeak NG synthesizer is recommended);
2. Try to read these ordinal numbers in Portuguese (digits followed by symbols ª and/or º) with NVDA, using Say All command or control+rightArrow/leftArrow: 1ª, 1º, 2ª, 2º...

Actual behavior:

When reading the ordinal number in Portuguese "1ª" NVDA says "1 a", number "2º" it says "2 o"...

Expected behavior:

When reading the ordinal number in Portuguese "1ª" NVDA must say "primeira", number "2º" it must say "segundo" ...

NVDA logs, crash dumps and other attachments:

n/a

System configuration

NVDA installed/portable/running from source:

Installed.

NVDA version:

2025.1 beta 1 and higher, 2025.1 beta 5.

Windows version:

11, 24H2, 26100.4061.

Name and version of other software in use when reproducing the issue:

It happens globally, any text editors, any web browsers...

Other information about your system:

n/a

Other questions

Does the issue still occur after restarting your computer?

Yes.

Have you tried any other versions of NVDA? If so, please report their behaviors.

Until NVDA 2024.4.2, ordinal numbers are normally spoken in Brazilian Portuguese

If NVDA add-ons are disabled, is your problem still occurring?

Yes.

Does the issue still occur after you run the COM Registration Fixing Tool in NVDA's tools menu?

Yes.

[seanbudd]

related #17725

[CyrilleB79]

@tiago-casal, I guess this is due to "Unicode normalization" being enable by default in speech settings from 2025.1 onwards.
What if you disable this option?

And also, does the issue occur with any speech synth or only some of them? Which speech synth / voice have you tested and what is the result?

[SaschaCowley]

@seanbudd how is that issue related? This issue is about a speech synth expecting a specific character in order to speak ordinal characters. That issue is about our character descriptions being confusing.

[SaschaCowley]

@CyrilleB79 the OP wrote in the issue description that they are aware that is the problem; they are suggesting we add an exception to avoid this undesired behaviour while still allowing users to reap the benefits of unicode normalisation in other cases.

[Adriani90]

cc: @LeonarddeR

[LeonarddeR]

It might make sense to exclude the ordinal indicators from normalisation, but not always. Ideally we'd do this at the level of a symbols dictionary.

[ruifontes]

If º and ª are not normalized, when reading with arrow or control+arrow, if they are defined in symbol pronunciation, why they are normalized when reading line, phrase, sentence and say all?
And they also should not be normalized in nº or 1ª!
Because it easy to treat nº in dictionary to not be normalized, but it is not so easy to do it to all ordinal numbers!

[LeonarddeR]

When reading by line, phrase etc. normalisation happens at the level of the whole unit, so the whole line is normalised in one call.

[gerald-hartig]

While this isn't ideal, the intent of the speech is still comprehensible and there is a workaround for this, so we're not going to block the 2025.1 release on this at this very late stage.