Sawubona, Habr! Namuhla sizokukhombisa ukuthi ungayisebenzisa kanjani i-Azure ukuxazulula izinkinga ezidinga ukungenelela komuntu. Ama-ejenti achitha isikhathi esiningi ephendula imibuzo efanayo, ephatha izingcingo kanye nemiyalezo. Ama-Chatbots enza ukuxhumana nokuqashelwa ngokuzenzakalela futhi anciphise umthwalo kubantu. Amabhothi aphinde asetshenziswe ku-Azure DevOps, lapho avumela khona, ngokwesibonelo, ukugunyaza ukukhishwa, ukuphatha ukwakha - buka, qala futhi ume - ngqo kusuka ku-Slack noma i-Microsoft Teams. Empeleni, i-chatbot ikhumbuza i-CLI ngandlela thile, iyasebenzisana kuphela, futhi ivumela unjiniyela ukuthi ahlale kumongo wengxoxo.
Kulesi sihloko, sizokhuluma ngamathuluzi okudala ama-chatbots, sibonise ukuthi angathuthukiswa kanjani ngezinsizakalo zokuqonda kwengqondo, futhi sichaze indlela yokusheshisa intuthuko ngezinsizakalo ezenziwe ngomumo e-Azure.
Ama-Chatbots kanye nezinsizakalo zokuqonda: yikuphi ukufana futhi yini umehluko?
Ukwakha ama-bots ku-Microsoft Azure, usebenzisa i-Azure Bot Service kanye ne-Bot Framework. Ngokuhlangene bamele isethi yesofthiwe yokwakha, ukuhlola, ukuthumela nokuphatha ama-bots, okukuvumela ukuthi udale kusuka kumamojula enziwe ngomumo womabili amasistimu okuxhumana alula futhi athuthukile ngokusekelwa kwenkulumo, ukuqashelwa kolimi lwemvelo namanye amakhono.
Ake sicabange ukuthi udinga ukusebenzisa i-bot elula esekelwe kusevisi ye-Q&A yebhizinisi noma, ngokuphambene, udale i-bot esebenzayo enohlelo lokuxhumana oluyinkimbinkimbi, olunamagatsha. Ukuze wenze lokhu, ungasebenzisa amathuluzi amaningi, ahlukaniswe ngamaqembu amathathu:
- Izinsizakalo zokuthuthukiswa okusheshayo kwezokuxhumana zengxoxo (ama-bots).
- Izinsizakalo ze-AI zengqondo esele zenziwe ezimweni ezihlukene zokusetshenziswa (ukubonwa kwephethini, ukunakwa kwenkulumo, isisekelo solwazi nosesho).
- Izinsizakalo zokudala nokuqeqesha amamodeli e-AI.
Imvamisa, abantu badida “amabhothi” kanye “nezinsizakalo zokuqonda” ngoba yomibili imiqondo isekelwe kumgomo wokuxhumana, futhi ukusetshenziswa kwe-bots namasevisi kuhilela izingxoxo. Kodwa ama-chatbot asebenza ngamagama angukhiye nezibangeli, futhi izinsizakalo zokuqonda zisebenza nezicelo ezingafanele ezivame ukucutshungulwa abantu:
Izinsizakalo zokuqonda zingenye indlela yokuxhumana nomsebenzisi, okusiza ukuguqula isicelo esinqunyiwe sibe umyalo ocacile futhi usidlulisele ku-bot.
Ngakho-ke, ama-chatbot ayizicelo zokusebenza ngezicelo, futhi izinsiza zokuqonda zingamathuluzi okuhlaziya okuhlakaniphile kwezicelo ezethulwa ngokuhlukana, kodwa ezingafinyelelwa yi-chatbot, zibe “ezihlakaniphile.”
Ukudala ama-chatbots
Umdwebo onconyiwe we-bot e-Azure umi kanje:
Ukuklama nokuthuthukisa ama-bots e-Azure, sebenzisa . Itholakala ku-GitHub , amandla ohlaka ashintsha, ngakho-ke kuyadingeka ukucabangela inguqulo ye-SDK esetshenziswa kuma-bots.
Uhlaka luhlinzeka ngezinketho eziningana zokudala ama-bots: ukusebenzisa ikhodi yakudala, amathuluzi omugqa womyalo noma ama-flowchart. Inketho yokugcina ibona ngeso lengqondo izingxoxo; kulokhu ungasebenzisa umphathi . Yakhelwe ku-Bot Framework SDK njengethuluzi lokuthuthukisa elibonakalayo amaqembu aqondisa izigwegwe angalisebenzisa ukuze enze ama-bots.
I-Bot Framework Composer ikuvumela ukuthi usebenzise amabhulokhi ukuze udale isakhiwo sengxoxo lapho i-bot izosebenza khona. Ukwengeza, ungakha izimbangela, okungukuthi, amagama angukhiye lapho i-bot izosabela khona phakathi nenkhulumomphendvulwano. Isibonelo, amagama athi “opharetha”, “ukweba” noma “yeka” kanye “nokwanele”.
Ku-Bot Framework Composer, ungakha amasistimu engxoxo ayinkimbinkimbi usebenzisa . Izingxoxo zingasebenzisa kokubili amasevisi okuqonda kanye namakhadi omcimbi (Amakhadi Aguquguqukayo):
Ngemva kokudala, ungasebenzisa i-chatbot ekubhaliseni, futhi iskripthi esilungiselelwe ngokuzenzakalelayo sizodala zonke izinsiza ezidingekayo: amasevisi okuqonda, uhlelo lohlelo lokusebenza, imininingwane yohlelo lokusebenza, isizindalwazi, njalonjalo.
Umenzi we-QnA
Ukuze udale ama-bots alula asuselwe kuzizindalwazi ze-Q&A zebhizinisi, ungasebenzisa isevisi yengqondo ye-QnA Maker. Isetshenziswa njengesilekeleli sewebhu esilula, ikuvumela ukuthi ufake isixhumanisi kusizinda solwazi lwenkampani (FAQ Urls) noma usebenzise isizindalwazi sedokhumenti *.doc noma *.pdf format njengesisekelo. Ngemva kokudala inkomba, i-bot izokhetha ngokuzenzakalelayo izimpendulo ezifanele kakhulu zemibuzo yomsebenzisi.
Usebenzisa i-QnAMaker, ungase futhi udale uchungechunge lwemibuzo ecacisayo ngokudala izinkinobho ezizenzakalelayo, wengeze isisekelo solwazi ngemethadatha, futhi uqhubeke uqeqesha isevisi ngesikhathi sokusetshenziswa.
Isevisi ingasetshenziswa njenge-chatbot esebenzisa lo msebenzi owodwa kuphela, noma njengengxenye ye-chatbot eyinkimbinkimbi esebenzisa, kuye ngesicelo, amanye amasevisi e-AI noma izici ze-Bot Framework.
Ukusebenza namanye amasevisi okuqonda
Kunezinsizakalo eziningi ezahlukene zokuqonda endaweni yesikhulumi se-Azure. Ngobuchwepheshe, lezi yizinsizakalo zewebhu ezizimele ezingabizwa ngekhodi. Ukuphendula, isevisi ithumela i-json yefomethi ethile, engasetshenziswa ku-chatbot.
Ukusetshenziswa okuvame kakhulu kwama-chatbots yilezi:
- Ukubona umbhalo.
- Ukuqashelwa kwezigaba zezithombe Zesevisi Yokubona Ngokwezifiso ezichazwe unjiniyela (icala lokukhiqiza: ukunakwa kokuthi isisebenzi sigqoke isigqoko esiqinile, izibuko noma imaski).
- Ukuqashelwa kobuso (ukusetshenziswa okuhle kakhulu ukuhlola ukuthi umuntu ohlolwayo uthumele ubuso bakhe yini, noma, yithi, isithombe senja noma isithombe somuntu wobulili obuhlukile).
- Ukuqashelwa kwenkulumo.
- Ukuhlaziywa kwesithombe.
- Ukuhumusha (sonke siyakhumbula ukuthi ungakanani umsindo owenziwe ukuhumusha ngesikhathi esisodwa ku-Skype).
- Ukuhlola ukupela neziphakamiso zokulungisa amaphutha.
LUIS
Futhi, ukuze udale ama-bots ungase udinge (Isevisi Yokuqondisisa Ngolimi). Izinjongo zesevisi:
- Nquma ukuthi ingabe isitatimende somsebenzisi sinengqondo nokuthi impendulo ye-bot iyadingeka yini.
- Yehlisa imizamo yokuloba inkulumo yomsebenzisi (umbhalo) ibe imiyalo eqondakalayo ku-bot.
- Qagela imigomo/izinhloso zangempela zomsebenzisi futhi ukhiphe imininingwane ebalulekile emishwaneni engxoxweni.
- Vumela unjiniyela ukuthi aqalise i-bot esebenzisa izibonelo ezimbalwa zokubonwa kwencazelo kanye nokuqeqeshwa okungeziwe okulandelayo kwe-bot phakathi nokusebenza.
- Nika amandla unjiniyela ukuthi asebenzise ukubona ngamehlo engqondo ukuze ahlole ikhwalithi yokulotshwa komyalo.
- Siza ekuthuthukisweni okukhulayo ekubonweni kwangempela kwethagethi.
Eqinisweni, umgomo oyinhloko we-LUIS ukuqonda ngamathuba athile ukuthi umsebenzisi wayesho ukuthini futhi uguqule isicelo semvelo sibe umyalo ovumelanayo. Ukubona amanani emibuzo, i-LUIS isebenzisa isethi yezinhloso (izincazelo, izinhloso) kanye nezinhlangano (okungaba okulungiselelwe kusengaphambili onjiniyela, noma “izizinda” ezithathwe futhi zakhiwe kusengaphambili - eminye imitapo yolwazi eseyenziwe ngomumo yemishwana evamile elungiselelwe i-Microsoft).
Isibonelo esilula: une-bot ekunikeza isibikezelo sezulu. Kuye, inhloso kuyoba ukuhunyushwa kwesicelo semvelo "esenzweni" - isicelo sesibikezelo sezulu, futhi izinhlangano zizoba isikhathi nendawo. Nawu umdwebo wokuthi inhloso ye-CheckWeather isebenza kanjani ku-bot enjalo.
Inhloso
Essence
Isibonelo sombuzo wemvelo
I-CheckWeather
{"type": "location", "entity": "moscow"}
{"type": "builtin.datetimeV2.date", "entity": "future","resolution":"2020-05-30"}
Siyoba njani isimo sezulu kusasa eMoscow?
I-CheckWeather
{ "type": "date_range", "entity": "kule mpelasonto" }
Ngibonise isibikezelo sale mpelasonto
Ukuhlanganisa i-QnA Maker ne-LUIS ungasebenzisa .
Uma usebenza ne-QnA Maker futhi uthola isicelo esivela kumsebenzisi, isistimu inquma ukuthi yiliphi iphesenti lamathuba empendulo evela ku-QnA efana nesicelo. Uma amathuba aphezulu, umsebenzisi umane anikezwe impendulo evela kusisekelo solwazi lwebhizinisi; uma liphansi, isicelo singathunyelwa ku-LUIS ukuze sicaciselwe. Ukusebenzisa i-Dispatcher kukuvumela ukuthi ungahleli le logic, kodwa ukucacisa ngokuzenzakalelayo lo mkhawulo wokuhlukaniswa kwezicelo futhi usabalalise ngokushesha.
Ukuhlola nokushicilela i-bot
Olunye uhlelo lokusebenza lwasendaweni lusetshenziselwa ukuhlola, . Usebenzisa i-emulator, ungakwazi ukuxhumana ne-bot futhi uhlole imilayezo eyithumelayo neyitholayo. I-emulator ibonisa imilayezo ngendlela ezovela ngayo kusixhumi esibonakalayo sengxoxo yewebhu futhi ibhala izicelo nezimpendulo ze-JSON lapho ithumela umlayezo i-bot.
Isibonelo sokusebenzisa i-emulator sivezwa kule demo, ebonisa ukudalwa komsizi obonakalayo we-BMW. Ividiyo iphinde ikhulume ngama-accelerator amasha okudala ama-chatbots - izifanekiso:
Ungasebenzisa futhi izifanekiso lapho udala ama-chatbots akho.
Izifanekiso zikuvumela ukuthi ungabhali imisebenzi ye-bot evamile kabusha, kodwa wengeze ikhodi eseyenziwe ngomumo “njengekhono”. Isibonelo kungaba ukusebenza ngekhalenda, ukwenza ama-aphoyintimenti, njll. Ikhodi yamakhono asevele enziwe ku github.
Ukuhlola kube yimpumelelo, i-bot isilungile, futhi manje idinga ukushicilelwa futhi kuxhunywe iziteshi. Ukushicilela kwenziwa kusetshenziswa i-Azure, futhi izithunywa noma izinkundla zokuxhumana zingasetshenziswa njengeziteshi. Uma ungenaso isiteshi esidingekayo sokufaka idatha, ungayisesha emphakathini ohambisanayo ku-GitHab.
Futhi, ukuze wakhe i-chatbot egcwele ngokugcwele njengesixhumi esibonakalayo sokuxhumana nomsebenzisi nezinsizakalo zokuqonda kwengqondo, uzodinga izinsiza ezengeziwe ze-Azure, ezifana nedathabheyisi, i-serverless (Imisebenzi ye-Azure), kanye nezinsizakalo ze-LogicApp futhi, ngokunokwenzeka. , Igridi Yomcimbi.
Ukuhlola nokuhlaziya
Ukuze uhlole ukusebenzisana komsebenzisi, ungasebenzisa kokubili ukuhlaziya okwakhelwe ngaphakathi kwe-Azure Bot Service kanye nesevisi ye-Application Insights ekhethekile.
Ngenxa yalokho, ungakwazi ukuqoqa ulwazi ngokusekelwe kulezi zindlela ezilandelayo:
- Bangaki abasebenzisi abafinyelele i-bot eziteshini ezihlukahlukene phakathi nenkathi yesikhathi ekhethiwe.
- Bangaki abasebenzisi abathumele umlayezo owodwa ababuye kamuva futhi bathumela omunye.
- Zingaki izenzo ezithunyelwe futhi zamukelwe kusetshenziswa isiteshi ngasinye phakathi nesikhawu sesikhathi esishiwo.
Usebenzisa imininingwane yohlelo lokusebenza, ungaqapha noma yiluphi uhlelo lokusebenza ku-Azure futhi, ikakhulukazi, ama-chatbots, uthole idatha eyengeziwe mayelana nokuziphatha komsebenzisi, imithwalo kanye nokusabela kwe-chatbot. Kufanele kuqashelwe ukuthi insizakalo ye-Application Insights inokuxhumana kwayo kuphothali ye-Azure.
Ungasebenzisa futhi idatha eqoqwe ngale sevisi ukuze udale ukubonwa okwengeziwe nemibiko yokuhlaziya ku-PowerBI. Isibonelo sombiko onjalo nesifanekiso se-PowerBI singathathwa .
Siyabonga nonke ngokunaka kwenu! Kulesi sihloko sisebenzise kusukela kuwebhu ye-Microsoft Azure umakhi u-Anna Fenyushina “Lapho abantu bengenaso isikhathi. Ungawasebenzisa kanjani ngo-100% ama-chatbots kanye nezinsizakalo zokuqonda ukuze uguqule izinqubo zenjwayelo”, lapho sibonise khona ngokucacile ukuthi yiziphi izingxoxo eziku-Azure nokuthi yiziphi izimo zokusetshenziswa kwazo, futhi sabonisa nendlela yokudala i-bot ku-QnA Maker ngemizuzu engu-15 nokuthi isakhiwo sombuzo sichazwa nge-LUIS.
Senze le webinar njengengxenye yomjaho we-inthanethi wabathuthukisi be-Dev Bootcamp. Kwakumayelana nemikhiqizo esheshisa intuthuko futhi ekhulula omunye wemisebenzi evamile evela kubasebenzi benkampani abasebenzisa amathuluzi ezishintshayo kanye namamojula e-Azure aselungiselelwe kusengaphambili enziwe. Ukuqoshwa kwamanye ama-webinars afakwe emjahweni kuyatholakala kulezi zixhumanisi ezilandelayo:
-
Sikhuluma ngokuziphendukela kwemvelo namandla amasha esevisi ye-Azure DevOps. Sihlaziya ngokuningiliziwe zonke izingxenye zayo ezinhlanu eziyinhloko nokuthi isevisi isiza kanjani inhlangano yokuthuthukiswa kwesofthiwe yezimboni. Umethuli: UVladimir Gusarov Microsoft MVP.
Sicabanga ukuthi singayisebenzisa kanjani indlela "yokuhlanganiswa njengekhodi" kunoma yiziphi izilimi zokuhlela kusetshenziswa amapayipi e-YAML anezigaba eziningi. Umethuli: UVladimir Gusarov Microsoft MVP.
Sidingida ukuthi singawathuthukisa kanjani amasu okugcina ikhodi yomthombo wesoftware nokuthi ungasebenzisa kanjani izinhlelo zokulondoloza inguqulo usebenzisa i-Azure Repos. Umethuli: UVladimir Gusarov Microsoft MVP.
Sethula amasu okuphatha umthombo ovulekile ekuthuthukisweni nasekusetshenzisweni okungokoqobo kwethuluzi le-WhiteSource. Sikhuluma ngokusesha ubungozi ezingxenyeni zomthombo ovulekile kanye nokugcina ubumsulwa belayisense sisebenzisa i-WhiteSource. Abethuli: Diana Lisbaron, Umqondisi Wezokuthengisa WhiteSource, kanye no-Daria Oreshkina, Ukulawulwa Kwewebhu, umsabalalisi osemthethweni we-WhiteSource eRussia.
Source: www.habr.com