21,500 hours CC-BY 4.0 Training Ngā reo katoa

Ngā kaupapa

voice audionlp

He huinga raraunga kōrero reo-maha nui a VAANI i whakawhanakehia e te Indian Institute of Science (IISc), Bangalore, e mau ana i te āhua 21,500 hāora oro i kohia mai i te 110,000 pea o ngā kaikōrero puta noa i ngā takiwā 120 i ngā kāwanatanga 22 o Inia. E kapi ana i ngā reo me ngā mita e 86 e kōrerohia ana puta noa i Inia, tae atu ki ngā reo matua kua whakarārangitia ā-ture me te tokomaha hoki o ngā momo reo ā-rohe, ā-iwi taketake hoki, ā, e 835 hāora o te kōrero kua tuhia ā-tuhi.

I hangaia te huinga raraunga nei hei tautoko i te whānuitanga o ngā mahi kōrero me ngā mahi reo, tae atu ki te whakamōhiotanga kōrero aunoa, te waihanga kupu-ki-te-kōrero, te tautohu kaikōrero, te tautohu reo, te whakapai ake i te kōrero, me te whakawhanake tauira reo aratau-maha. I tukuna i raro i te raihana CC BY 4.0, ā, ko tōna tino kaupapa he whakangungu, he whakataurite paerewa hoki i ngā pūnaha AI, otirā he tino hāngai ki ngā kairangahau me ngā kaiwhakawhanake e mahi ana ki ngā reo Inia iti rawa ngā rauemi me ngā reo kāore e tino whai māngai ana.

Horopaki me te Whanaketanga

I hangaia a VAANI (arā, ko te tikanga he "reo" he "kōrero" rānei i roto i ētahi reo Inia) e te Indian Institute of Science (IISc) i Bangalore hei wāhanga o tētahi whakapau kaha ki te whakatika i te āputa nui o ngā rauemi raraunga kōrero mō ngā reo o Inia. Ko te āhua reo o Inia tētahi o ngā mea tino kanorau i te ao, e kapi ana i ngā rau reo me ngā mano mita reo, heoi anō, ko te nuinga o ngā huinga raraunga kōrero e aro ana ki tētahi huinga whāiti noa o ngā reo kua pai kē te whai rauemi. I whakawhanakehia a VAANI hei whakawhānui i te wātea o ngā raraunga whakangungu mō tēnei wāhi reo kāore i tino tautokona, mā te kohikohi raraunga ā-mara i te tauine nui puta noa i ngā rohe matawhenua, reo hoki, e tino kanorau ana o te motu.

I kohia ngā raraunga i te tata ki te 110,000 ngā kaikōrero kua marara ki ngā takiwā 120 i roto i ngā kāwanatanga 22 o Inia, e whakaatu ana i tētahi whakapau kaha āta whakarite kia mau ngā rerekētanga tūturu ā-rohe, ā-mita reo hoki, kaua e whakawhirinaki noa ki ngā hopunga oro taiwhanga kua āta whakahaerehia. E toro ana te huinga raraunga ki ngā reo me ngā mita reo 86, mai i ngā reo matua kua whakarārangitia pēnei i te Hindi, Tamil, Telugu, Bengali, Kannada, me Malayalam tae atu ki ngā momo ā-rohe maha me ngā reo iwi taketake tae atu ki te Gondi, Santali, Kurukh, Wancho, me Tenyidie, me te tini noa atu.

Te Hanga o te Huinga Raraunga me ngā Āhuatanga Matua

E 21,500 hāora pea te katoa o ngā oro kei roto i VAANI, ā, nā konei ka noho ia hei tētahi o ngā kohinga kōrero reo-maha nui rawa e aro ana ki ngā reo o Inia. I roto i tēnei, kua tuhia ā-tuhi te 835 hāora, e whakarato ana i ngā tuhinga pono-taketake mō tētahi wāhanga o te kohinga. Kei roto hoki i te huinga raraunga ētahi āhuatanga aratau-maha, e āhei ai te whakamahi i tua atu i ngā mahi kōrero tikanga noa.

Ko ngā āhuatanga matua o te huinga raraunga ko ēnei:

  • Te kapi o ngā reo me ngā mita reo 86, tae atu ki te maha o ngā momo iti-rauemi me ngā reo iwi taketake
  • Ngā takoha mai i te tata ki te 110,000 ngā kaikōrero nō ngā horopaki matawhenua me ngā taupori tino kanorau
  • 21,500 hāora katoa o te oro, ā, 835 hāora o te kōrero kua tuhia ā-tuhi
  • Ngā hopunga ā-mara e toro ana ki ngā takiwā 120 i roto i ngā kāwanatanga 22 o Inia
  • I tukuna i raro i te raihana CC BY 4.0, e whakaae ana kia whānui te whakamahi anō me te whakahua i te pūtake
  • Te tautoko mō ngā whakamahinga rangahau aratau-kotahi me ngā whakamahinga aratau-maha

Ngā Mahi e Tautokona Ana me ngā Whakamahinga

I hoahoatia a VAANI hei rauemi mō te whānuitanga o ngā mahi tukatuka kōrero me te reo. Ka taea e ngā kairangahau me ngā kaiwhakawhanake te whakamahi mō te whakangungu me te aromātai i ngā pūnaha automatic speech recognition (ASR), te waihanga text-to-speech (TTS), te tautohu kaikōrero, me ngā tauira tautohu reo. Ka tautoko hoki te huinga raraunga i ngā mahi whakapai kōrero me te whakawhanake i ngā multimodal large language models (LLMs). Nā te whānui o tana kapi reo, ka tino hāngai ia mō te whakataurite paerewa i ngā pūnaha e whakaarotia ana kia mahi puta noa i te taiao reo tino kanorau o Inia.

I te mea kua whakaurua mai te maha o ngā reo me ngā mita reo iti-rauemi kāore pea he raraunga kōrero o mua, he iti rawa rānei, he tino whai take a VAANI mō ngā kairangahau e aro ana ki te hanga hangarau reo whakauru. Ka whakarato ia i tētahi tūāpapa mō te whakawhanake i ngā taputapu ASR me ngā taputapu tukatuka reo māori e taea ai te tautoko i ngā hapori kua roa nei ō rātou reo e ngaro ana i te whanaketanga AI auraki.

Te Hiranga mō te Hangarau Reo o Inia

Ko te rahi me te kanorau reo o VAANI e noho ai ia hei takoha nui ki te ao rangahau kōrero reo-maha, otirā i roto i te horopaki o ngā reo o Āhia ki te Tonga. Mā te tuhi i ngā momo kōrero ā-waha mai i ngā hapori iwi taketake, taiwhenua, ā-rohe hoki, i te taha o ngā reo e whānui ake ana te kōrerotia, ka mau i te huinga raraunga tētahi āhuatanga o te taonga tuku iho reo o Inia kāore e tino kitea i roto i ngā rauemi rorohiko. Nā tana raihana tuwhera ka māmā ake te whakamahi mā ngā whare wānanga, ngā rōpū kāwanatanga, me ngā kairangahau ahumahi e whai ana kia hangaia he hangarau kōrero e nui ake te whakauru me te māngai tika.

Pūrongo