Kundin Bayanai na Murya na Hausa na Gaskiya, na Gida don Horar da AI
Ƙarfafa LLM, ASR, da aikace-aikacen murya naka da sama da sa’o’i 50,000 na sahihin bayanan magana na Hausa da aka shirya don AI.
Ƙarfafa LLM, ASR, da aikace-aikacen murya naka da sama da sa’o’i 50,000 na sahihin bayanan magana na Hausa da aka shirya don AI.
A matsayin wani ɓangare na faɗaɗɗen rufin harsuna, GeoPoll tana samar da kundin bayanan sauti na harshen Hausa da aka riga aka yi musu lakabi, masu inganci, waɗanda aka tanadar musamman don horar da samfuran basira ta na’ura (AI). Ba kamar kundin bayanan da aka kirkira ta na’ura ko waɗanda aka rubuta aka karanta ba, namu bayanai suna fitowa ne daga tattaunawar waya ta gaskiya tare da ‘yan asalin masu magana da Hausa a ƙasashe daban-daban. Ana tsara waɗannan tattaunawa bisa takardun rubutaccen shiri na fannoni na musamman don samun daidaiton jigo, tare da barin damar amsoshin halitta da ba a tsara ba.
Kowane rikodi ana rubuta shi a rubuce (transcribe) kuma a ware masu magana daban-daban (diarize) ta hannun masu nazarin harshe da ke iya bambance bambancin Hausa na yankuna. Daga nan kuma ana yi musu alamar bayanai masu faɗi kamar shekaru, jinsi, lahja, da wuri. Sakamakon haka shi ne babban ɗakin adana tattaunawar Hausa na gaskiya, wanda aka tsara don a yi amfani da shi wajen daidaita manyan samfuran harshe (LLM fine-tuning), horar da tsarin gane magana (ASR training), ƙirƙirar murya (TTS synthesis), da kuma aikace-aikacen AI na harsuna masu yawa.
Muna da sama da sa’o’i 50,000 na Hausa na gida daga sama da masu magana 30,000 na daban-daban a fadin nahiyar Afirka. Ga ƙasashen da aka rufe*
Tambayi game da ƙwarewar mu a sauran ƙasashen da ake magana da Hausa.
Cika wannan fom don tuntuɓar mu game da bayanan samfurin, tsarin fayil, cikakken rufi, ko buƙatu na musamman.
