Amazon yakaburitsa dataset yekunzwisisa kutaura mumitauro makumi mashanu neimwe

Amazon yakaburitsa pasi peCC BY 4.0 rezinesi iyo "MASSIVE" (Mullingual Amazon SLURP yeSlot Kuzadza, Intent Classification, uye Virtual-assistant Evaluation) dhatabheti, mhando dzemakina ekudzidza masisitimu, uye maturusi ekudzidzisa ako mamodheru anogona kushandiswa kunzwisisa ruzivo rwemutauro wechisikigo (NLU, Natural Language Kunzwisisa). Seti iyi inosanganisira anopfuura miriyoni emashoko akatsanangurwa uye akakamurwa akagadzirirwa mitauro 51.

Iyo SLURP muunganidzwa, yaimbovepo kuChirungu, yakashandiswa sereferensi yekuvaka iyo MASSIVE seti, iyo yakaiswa munharaunda mune makumi mashanu mimwe mitauro vachishandisa nyanzvi vashanduri. Alexa's natural language kunzwisisa (NLU) tekinoroji inotanga kushandura kutaura kuita zvinyorwa, yobva yashandisa akawanda maNLU modhi kune zvinyorwa zvinoongorora kuvepo kwemazwi akakosha kuona kukosha kwemubvunzo wemushandisi.

Chimwe chezvinangwa zvekugadzira uye kushambadza seti ndeyekugadzirisa vabatsiri vezwi kuti vagadzirise ruzivo mumitauro yakati wandei kamwechete, pamwe nekukurudzira vagadziri vebato rechitatu kugadzira maapplication nemasevhisi anowedzera kugona kwevabatsiri vezwi. Kukwezva kutarisisa kwevagadziri, Amazon yakatanga makwikwi ekugadzira yakanakisa generic modhi vachishandisa yakadhindwa dataset.

Parizvino, vabatsiri vezwi vanotsigira mitauro mishoma uye vanoshandisa mhando dzemashini dzekudzidza dzakasungirirwa kune mumwe mutauro. Chirongwa cheMASSIVE chine donzvo rekubvisa chikanganiso ichi nekugadzira mhando dzepasirese uye masisitimu ekudzidza muchina anokwanisa kupatsanura nekugadzirisa ruzivo mumitauro yakasiyana.

Source: opennet.ru

Voeg