I-algorithm ngemuva komculo

Hambisa ngapha, American Idol.

Indaba enkulu elandelayo yempumelelo embonini yomculo ngeke itholakale emiqhudelwaneni yamathalente ephrofayili ephezulu. Kunalokho, izokhonjwa kumasethi wedatha ngama-algorithms ayinkimbinkimbi aklanyelwe ukwembula ukusetshenziswa namathrendi ebhizinisi.

Uma ubheka phezulu, le ndlela izwakala yomile futhi ingenawo umzwelo ukwedlula ukugxeka kukaSimon Cowell, kodwa empeleni iyindlela umphakathi okhetha ngayo “into enkulu elandelayo.” Ngaso sonke isikhathi uma umphakathi uchofoza izixhumanisi ze-YouTube, uthumela izithombe zekhonsathi ku-Twitter, noma uxoxa ngamabhendi ku-Facebook, unikela kuqoqo lolwazi olubizwa ngokuthi idatha enkulu. Igama libhekisela eqoqweni lamasethi edatha amakhulu futhi aqukethe ukuhlobana okuyinkimbinkimbi. Cabanga ngesakhiwo sezinkundla zokuxhumana. Ziqukethe izigidi zamaphrofayela omsebenzisi ngamunye axhunywe ndawonye ngobungane, ‘okuthandwayo’, ubulungu beqembu, njalo njalo.Eqinisweni, idatha enkulu ibukisa ngesakhiwo salezi zinkundla.

Embonini yomculo, idatha enkulu ikhiqizwa imisebenzi efana nokuthengiswa ku-inthanethi, ukulanda, nokuxhumana okwenziwa ngezinhlelo zokusebenza noma izindawo zenkundla yezokuxhumana. Amamethrikhi alinganisiwe ahlanganisa “inani lezikhathi izingoma ezidlalwa ngazo noma ezeqiwa ngazo, kanye nezinga lokudonseka ezikuthola ezinkundleni zokuxhumana ngokusekelwe ezenzweni ezifana nokuthandwa kwe-Facebook nama-tweets.” Amathuluzi okuhlaziya anquma ukuthandwa kukonke kwamakhasi abalandeli futhi abhalise ukuphawula okuhle noma okungekuhle mayelana namaciko. Ndawonye, ​​lolu lwazi luhlonza amathrendi amanje, luhlola i-digital pulse yabaculi, futhi luholela ekuthengisweni ngama-single, okuthengiswayo, amathikithi ekhonsathi, ngisho nokubhaliselwe kumasevisi okusakaza umculo.

Mayelana nokuvundulula ithalente elisha, idatha enkulu idlala indima ebalulekile ekukhiqizeni intshisekelo kumalebula amarekhodi amakhulu. Ezimweni eziningi, izinkampani zibala ukubukwa kwekhasi lomculi, 'okuthandwayo', nabalandeli. Khona-ke, izinombolo zingaqhathaniswa kalula nabanye abaculi abakulolu hlobo lomculo. Uma isenzo sesikhiqize abalandeli abayizinkulungwane eziyikhulu kanye ne-Facebook noma i-Twitter, abaphathi bamathalenta bayaqaphela futhi baqale ukukhulisa isithakazelo embonini yomculo ngokwayo.

Idatha enkulu ekhetha i-Top 40 elandelayo enkulu

Ikhono lokuhlonza izitayela zamanje nokubikezela i-megastar elandelayo iza nemivuzo emikhulu yezezimali yawo wonke umuntu ohilelekile. Ngokwesibonelo, ososayensi bedatha bafunde umthelela wenkundla yezokuxhumana ku-albhamu ye-iTunes futhi balandelela ukuthengiswa ngokuqhathanisa ama-metrics omuntu nemali engenayo yomunye. Baphethe ngokuthi imisebenzi yenkundla yezokuxhumana ihambisana nokwanda kokuthengiswa kwama-albhamu namathrekhi. Ngokuqondile, ukubukwa kwe-YouTube kunomthelela omkhulu ekuthengisweni; okutholakele okwenze amalebula amarekhodi amaningi ukuthi alayishe amavidiyo omculo webhajethi enkulu endaweni yesikhulumi ukuze aphromothe abangashadile. Ngaphambi kokuchitha izigidi ekukhiqizweni kwamavidiyo, ukuhlaziya kusetshenziswa ukukhomba ukuthi yiziphi izingoma okungenzeka zibe yingoma ethandwa kakhulu ngokusekelwe emisebenzini eku-inthanethi yezithameli eziqondiwe. Ukunemba kwalezi zibikezelo kuhlotshaniswa nekhwalithi yokuhlaziywa kwedatha enkulu.

Osomabhizinisi abangaphakathi embonini yomculo manje bazama izindlela ezintsha zokuthuthukisa ama-algorithms avuna ulwazi ngokusebenza kahle nangokunemba okukhulu. Esinye sezibonelo eziphawuleka kakhulu ibhizinisi elihlanganyelwe phakathi kwe-EMI Music ne-Data Science yaseLondon ebizwa ngokuthi i-EMI Million Interview Dataset. Ichazwa “njengenye yedathasethi yokwaziswa yomculo ecebe kakhulu nenkulu kunawo wonke ake atholakala - isethi yedatha enkulu, eyingqayizivele, ecebile, yekhwalithi ephezulu ehlanganiswe ocwaningweni lomhlaba wonke oluqukethe izinto ezithakaselwayo, izimo zengqondo, ukuziphatha, ukujwayela, nokwazisa umculo njengoba kuvezwa abalandeli bomculo.”

UDavid Boyle, iPhini Likamongameli Eliphezulu le-Insight kwa-EMI Music, uyachaza, “(Kukhona) kuhlanganisa izingxoxo eziyisigidi ezidingida izihloko ezifana nezinga lothando lohlobo oluthile lomculo nohlobo oluncane, izindlela ezithandwayo zokutholwa komculo, abaculi bomculo abayintandokazi, imicabango ngokukhohlisa komculo, ukusakazwa komculo, amafomethi omculo, kanye nezibalo zabalandeli.”

Umgomo wephrojekthi uwukukhipha leli qoqo lolwazi emphakathini nokuthuthukisa ikhwalithi yebhizinisi embonini yomculo.

“Sibe nempumelelo enkulu sisebenzisa idatha ukuze sisize thina namaciko ethu ukuthi siqonde abathengi, futhi siyajabula ukwabelana ngedatha yethu ukuze sisize abanye benze okufanayo,” kusho uBoyle.

Ngo-2012, i-EMI Music and Data Science yaseLondon yathatha iphrojekthi isinyathelo esisodwa ngokuqhubekayo ngokusingatha i-Music Data Science Hackathon. I-EMC, umholi womhlaba kusayensi yedatha kanye nezixazululo ezinkulu zedatha, ijoyine leli bhizinisi futhi yanikeza ingqalasizinda ye-IT. Phakathi nenkathi yamahora angu-24, ososayensi bedatha abangu-175 bathuthukise amafomula angu-1,300 nezindlela zokuxazulula le nkinga ukuze baphendule lo mbuzo: “Ungabikezela ukuthi umlaleli uzoyithanda yini ingoma entsha?” Imiphumela yabonisa amandla obuhlakani obuqoqiwe futhi ababambiqhaza bakha amafomula achazwa njengasezingeni lomhlaba.

"Imininingwane evezwe kulokhu kusikisela kwe-hackathon emandleni namandla I-Big Data enawo - kokubili ukutholwa kobuhlakani kanye nenani elikhulayo lebhizinisi lezinhlangano zazo zonke izinhlobo," kusho uChris Roche, uMqondisi Wesifunda we-EMC Greenplum.

Kodwa ubakhokhela kanjani abaculi?

Ngemuva kokuthi imboni inqume ukuthi ingoma inamandla futhi iyikhiphe njengeyodwa, ibala kanjani imali yokukhokha uma ingoma idlalwa ezinkundleni zokuxhumana noma ezingosini zokusakaza? Njengamanje, "amalebula amarekhodi abo bonke osayizi abhekene nenkinga ekhulayo yokuhlanganisa imininingwane evela ezinkampanini ezisakazwayo ezifana ne-Spotify, Deezer, ne-YouTube, kodwa abe nabantu abambalwa kunangaphambili ukwenza kanjalo."

Enye yezinselele eziyinhloko ezivela kumbono wokuphathwa kolwazi ukuthi izinhlelo eziningi zokuphatha isizindalwazi azizange zenzelwe ukuphatha amasethi edatha amakhulu futhi ayinkimbinkimbi njengedatha enkulu. Isibonelo, usayizi wamafayela edatha yedijithali akhiqizwa abasabalalisi bomculo angaphezulu kakhulu kwalokho okungenziwa izinhlelo ezifana ne-Excel. Lokhu kudala izinkinga ezihlanganisa idatha engekho kanye namalebula wefayela angahambisani nesofthiwe yokubala.

Ezimweni eziningi, zonke lezi zinkinga zilungiswa ngababali bezimali, zengeza isikhathi esengeziwe nomsebenzi kumthwalo osuvele unzima. Ezimweni eziningi, iphesenti elikhulu le-overhead yelebula liboshelwe emnyangweni wokubalwa kwezimali.

Ukuze kuliwe nalezi zinkinga, osomabhizinisi bathuthukisa izinkundla zezobunhloli zebhizinisi ezinamandla okuhlela nokuhlaziya idatha enkulu. Esinye sezibonelo ezinhle kakhulu inkampani yase-Austrian i-Rebeat, echaza izinsizakalo zayo ngokuthi "i-royalty accounting ngokuchofoza kathathu." Yasungulwa ngo-2006, ikhule ngokushesha yaba isabalalisi sedijithali ehamba phambili eYurophu futhi inikeza ukufinyelela ezinsizeni zedijithali ezingama-300 emhlabeni jikelele. Empeleni, i-Rebeat iqondisa izinqubo zokubala futhi isingathe umsebenzi ongasemuva, njengokufanisa izinkambu zedatha kusofthiwe yokubala, ukuze umnyango wokubalwa kwezimali ukhululekile ukuphatha ibhajethi. Baphinde banikeze ingqalasizinda yokuphatha izinkokhelo zezindlu zasebukhosini ngokuhambisana nezivumelwano zenkontileka, izivumelwano eziqondile nezitolo zomculo wedijithali, ukukhiqiza amagrafu okulandelela ukuthengisa, futhi okubaluleke kakhulu, ukuthekelisa idatha kumafayela e-CSV.

Yebo, isevisi iza nenani. I-Forbes ibike ukuthi amalebula okurekhoda kufanele asebenzise i-Rebeat njengomsabalalisi ukuze akwazi ukufinyelela idatha yenkampani, ebiza ikhomishini yokuthengisa engu-15% kanye nemali engaguquki engu-$649 unyaka ngamunye. Izilinganiso ziphakamisa, nokho, ukuthi ezimeni eziningi imbondela yokubala yelebula ivamise kubiza kakhulu, okusho ukuthi ukusayina ne-Rebeat kungase kugcine kukonga imali.