tts-plan

We are cooperating with Helsinki University, Department of General Linguistics in making a synthetic voice for northern languages.

Status quo

Together, we have made a beta version of a North Sámi tts program.

The following files give an impression of the system:

• Our synthetic voice here reads the opening paragraphs of Johan Turi Muitalus sámiid birra ):
  • Mon lean okta sápmelaš, guhte lean bargan visot sámi bargguid ja mon dovddan visot sámi dili.
  • Ja mon lean ipmirdan, ahte Ruoŧa hállehus háliida min veahkehit nu olu go sáhttá.
  • Muhto sii eai oaččo riekta čielgasa, jur got dat lea min eallin ja dilli, deinnago sápmelaš ii sáhte jur juste čilget nu got lea.
  • Ja dasa lea dát sivva: go sápmelaš boahtá moskkus gámmirii, de son ii ipmir ii báljo maidege, go ii biegga beasa bossut njuni vuostá.
  • Su jurdagat eai golgga, go leat seainnit ja moskkus oaivvi nalde.
  • Ja ii leatge buorre sutnje orrut suhkkes vuvddiid siste, gos lea liegga ilbmi.
• As comparision: The recording of the same sentences, for the training set, by Biret-Ánne Bals Baal(thus, this is not a synthetic voice!)

The system is still far from ready, though. Here are some generated sentences without a prior recording (hence of a poorer quality):

• Vihtta bivdovuogi mun gal dušše namahin, muhto leamaš gal ovdal okta bivdovuohki vel, namalassii duhasteapmi.
• Guhtemuš dainna bivdovugiin guhkimus leaš geavahuvvon ii leat buorre diehtit, muhto buođđobivdu lea dat mii dál min áiggi eanemus ávkin lea álbmogii.

More work is needed in order to create a robust system.

Plans for future work

1. North Sámi
   1. We must improve the rule system for text processing
   2. We will record more sentences, also by more speakers
   3. We will add a component for intonation
2. Other languages
   1. We are looking into Greenlandic and the other Sámi languages

References

• A further development of the Helsinki system was presented at NODALIDA in 2007 ( abstract).
• The homepage of our Helsinki cooperation partners