figshare
Browse

The audio files of the synthetic speech spontaneity test

media
posted on 2023-04-05, 10:34 authored by Heete Sahkai, Liisi Piits, Liis Ermus, Indrek Hein, Indrek Kiissel, Meelis Mihkla, Kristjan Suluste, Egert Männisalu, Rene Altrov, Hille PajupuuHille Pajupuu, Jaan Pajupuu

Project:  Turning podcasts into a training corpus for conversational text-to-speech synthesis


Format of the audio file: S1/S2/S3_1/2/3/4/5/6/7/8_SPON/READ.wav

Synthesis techniques:

S1

S2

S3

Training corpus:

SPON - spontaneous speech corpus

READ - read aloud speech corpus

1--8 Passages in the listening test:

  •   Clean input texts (without fillers and disfluencies)

1. Ja ka oskus kuulata sageli ju aitab. Vähemalt ma arvan, et mures olijat aitab väga palju see, kui keegi teda lihtsalt kuulab ja on olemas. Aga inimesed ei oska seda teha sageli. [And the ability to listen often also helps. At least I think that someone who is anxious is significantly aided by someone just listening to them and being there. But      often people don’t know how to do that.]

2. Mul on tunne, et tegelikult me kõik vajame mingi hetk seda kohta, kus me saamegi vastutuse ära anda. See ei ole tegelikult üldse halb. See on osa elust ja me igaüks vajame seda. [I have a feeling that actually at some moment, we all need a place where we can surrender our responsibilities. And that is not actually all that bad. It is a part of life and we all need it.]

3. Mulle meeldib ka otsustada, mulle meeldib ka vastutada. Mul ei ole midagi selle vastu, aga vahepeal on tunne, et ma tahan, et keegi lihtsalt minu eest mingid asjad ära teeks. [I also like to make decisions, I also like to take responsibility. I have nothing against it, but sometimes I feel that I want that someone would just take care of some things for me.]

4. Sest tegelikult, kui ma vaatan praegu tagasi, meil on nii ägedad kaks kuud olnud selles maamajakeses. Teeme kaminat, ja me ei vaata enam nii palju telekat. [Because actually, now when I look back, we had two awesome months in this cottage. We build a fire in the fireplace, and we don’t watch as much TV anymore.]

  • Natural input texts (with fillers and disfluences)

5. *e noh kindlasti sina oled üks väheseid *e pigem väheseid inimesi, kes oskab kuulata ja oskab seda nii-öelda mingisugust nõu anda või lohutust anda. Aga paljud ei oskagi, nad jäävad kohmetuks. [*eh, well, surely you are one of the few, *uh, of rather few people, who knows how to listen and knows how to so-to-say give some kind of advice or comfort. But many don’t even know and remain awkward.]

6. Aga kui nüüd *e mõelda vananemise peale, siis *e noh, me muutume aasta-aastalt vanemaks, targemaks ka muidugi. Ilusamaks ka, ma loodan. *e mina küll tunnen, et ma nagu, mõnes mõttes lähen aastatega paremaks. [But now if we, *uh, think of aging, then, *uh well we become older year-by-year, and wiser too of course. More beautiful too, I hope. *Uh I feel that I kind of, in some sense, improve with the years.]

7. *e ja siis mul oli küll muide see mõte et *e. Ma olen alati mõelnud, et ma tahaks *e seda inimest nagu rohkem *e tunda, või nagu saanud tunda tol hetkel. Et selles suhtes on, see kirjaidee on küll noh päris *e tore. [*uh and then I did by the way have the idea that *uh. I have always thought that I would like *uh to know *uh this person more, or have known at that moment. That in that way, this letter idea is, well, quite *uh fun.]

8. Ta rääkis täpselt seda, mida nagu mina olin nagu mõelnud, aga tema lihtsalt pani selle palju paremini nagu sõnadesse ja oskas seda nii ilusasti nagu serveerida. Ma siiamaani mõtlen selle peale. [He said exactly that, what, like, I had like thought, but he just put it into like words much better, and could present it so beautifully. I still think of it.]


History

Usage metrics

    Licence

    Exports

    RefWorks
    BibTeX
    Ref. manager
    Endnote
    DataCite
    NLM
    DC