No PunctuationBoundary reported when a sentence ends with the word sun.

[fresswolf]

I couldn't find any exact definition of the word boundary types reported by the API. So I'm not sure if this is a bug.

I found that usually when a sentence end with a dot "." a word boundary event with the type SpeechSynthesisBoundaryType.Punctuation and the text "." is fired.

However, if the sentence ends with "sun", only an event with type SpeechSynthesisBoundaryType.Word is fired and the punctuation gets included in that word "sun." The issue can't be reproduced with similar short words like "son", where it seems to work as expected.

The following function reproduces the unexpected behaviour.

const wordBoundaryTest = async() => {
    const text = "Here comes the sun. Enjoy it!";
    const speechConfig = SpeechConfig.fromSubscription(process.env.AZURE_SPEECH_KEY, process.env.AZURE_SPEECH_REGION);
    speechConfig.speechSynthesisOutputFormat = SpeechSynthesisOutputFormat.Audio24Khz96KBitRateMonoMp3;
    speechConfig.speechSynthesisVoiceName = "en-US-EricNeural";
    const synthesizer = new SpeechSynthesizer(speechConfig);

    synthesizer.wordBoundary = function (s, e) {
        console.log(`Word "${e.text}" with type ${e.boundaryType}"`);
    }

    synthesizer.speakTextAsync(text);
}

Actual behaviour:
Word "Here" with type WordBoundary"
Word "comes" with type WordBoundary"
Word "the" with type WordBoundary"
Word "sun." with type WordBoundary"
Word "Enjoy" with type WordBoundary"
Word "it" with type WordBoundary"
Word "!" with type PunctuationBoundary"

Expected behaviour:
Word "Here" with type WordBoundary"
Word "comes" with type WordBoundary"
Word "the" with type WordBoundary"
Word "sun" with type WordBoundary"
Word "." with type PunctuationBoundary"
Word "Enjoy" with type WordBoundary"
Word "it" with type WordBoundary"
Word "!" with type PunctuationBoundary"

Also, the SpeechSynthesisBoundaryType.Sentence seems to never fire, but as there is no documentation about these events it's unclear what the event actually means.

[glharper]

@yulin-li Could you take a look at this? Thanks.

[yulin-li]

@fresswolf thanks for reporting this bug. I can confirm this is a BUG, I guess the reason is sun. is recognized by the system as a single word. We are investigating.

[fresswolf]

@yulin-li Meanwhile I came across situations where it doesn't only fail to report a punctuation, but it's also ignored by the speech synthesis.
Example:
<speak version='1.0' xml:lang='en-US'><voice xml:lang='en-US' name='en-US-TonyNeural'>The brothers stand. Mario doesn't.</voice></speak>

In the resulting audio the word "stand" and "Mario" are weirdly melted together in a new word "Standmario" and the punctuation isn't reported.