Molo, Habr! Namhlanje siza kukubonisa indlela yokusebenzisa iAzure ukusombulula iingxaki ezifuna ungenelelo lomntu. Iiarhente zichitha ixesha elininzi ziphendula imibuzo efanayo, zibamba iifowuni kunye nemiyalezo. IiChatbots zizenza ngokuzenzekelayo unxibelelwano kunye nokuqatshelwa kunye nokunciphisa umthwalo ebantwini. Iibhothi zikwasetyenziswa kwi-Azure DevOps, apho zivumela, umzekelo, ukuvuma ukukhutshwa, ukulawula ukwakha - ukujonga, ukuqala nokuyeka - ngokuthe ngqo kwi-Slack okanye iMicrosoft Teams. Ngokwenene, i-chatbot ikhumbuza i-CLI, isebenzisana kuphela, kwaye ivumela umphuhlisi ukuba ahlale kumxholo wengxoxo yengxoxo.
Kweli nqaku, siza kuthetha ngezixhobo zokwenza ii-chatbots, bonisa ukuba zingaphuculwa njani ngeenkonzo zokuqonda, kwaye sichaze indlela yokukhawulezisa uphuhliso ngeenkonzo esele zenziwe e-Azure.
Ii-Chatbots kunye neenkonzo zokuqonda: zeziphi izinto ezifanayo kwaye yintoni umahluko?
Ukwenza i-bots kwiMicrosoft Azure, usebenzisa iNkonzo yeBot yeAzure kunye neSakhelo seBot. Ngokudibeneyo bamele isethi yesofthiwe yokwakha, ukuvavanya, ukuhambisa kunye nokulawula i-bots, ekuvumela ukuba udale kwiimodyuli ezilungeleyo zombini iinkqubo zonxibelelwano ezilula kunye neziphambili kunye nenkxaso yentetho, ukuqaphela ulwimi lwendalo kunye nezinye izakhono.
Makhe sicinge ukuba kufuneka uphumeze i-bot elula esekelwe kwinkonzo ye-Q & A yenkampani okanye, ngokuchaseneyo, udale i-bot esebenzayo kunye nenkqubo yonxibelelwano edibeneyo. Ukwenza oku, ungasebenzisa inani lezixhobo, ezahlulwe ngokwamaqela amathathu:
- Iinkonzo zophuhliso olukhawulezayo lwe-dialog interfaces (bots).
- Iinkonzo ze-AI zokuqonda ezisele zenziwe kwiimeko ezahlukeneyo zokusetyenziswa (ukuqatshelwa kwephethini, ukuqondwa kwentetho, isiseko solwazi kunye nokukhangela).
- Iinkonzo zokudala kunye noqeqesho lweemodeli ze-AI.
Ngokuqhelekileyo, abantu badibanisa nge-intuitively "i-bots" kunye "neenkonzo zokuqonda" kuba zombini iingcamango zisekelwe kumgaqo wonxibelelwano, kwaye imeko yokusetyenziswa kwe-bots kunye neenkonzo ibandakanya iingxoxo. Kodwa ii-chatbots zisebenza ngamagama angundoqo kunye nezinto ezibangelayo, kwaye iinkonzo zokuqonda zisebenza kunye nezicelo ezingafunekiyo ezihlala zisenziwa ngabantu:
Iinkonzo zokuqonda enye indlela yokunxibelelana nomsebenzisi, inceda ukuguqula isicelo esingenasizathu kumyalelo ocacileyo kwaye uyidlulisele kwi-bot.
Ke, ii-chatbots zizicelo zokusebenza kunye nezicelo, kwaye iinkonzo zokuqonda zizixhobo zokuhlalutya okukrelekrele kwezicelo eziqaliswe ngokwahlukeneyo, kodwa apho i-chatbot inokufikelela kuyo, ibe "nengqondo."
Ukudala ii-chatbots
Umzobo ocetyiswayo woyilo lwe-bot e-Azure umi ngolu hlobo lulandelayo:
Ukuyila kunye nokuphuhlisa i-bots kwi-Azure, sebenzisa . Ifumaneka kwi-GitHub , amandla otshintsho lwesakhelo, ngoko ke kuyimfuneko ukuqwalasela inguqu ye-SDK esetyenziswa kwi-bots.
Isakhelo sibonelela ngeendlela ezininzi zokudala ii-bots: usebenzisa ikhowudi yakudala, izixhobo zomgca womyalelo okanye ii-flowcharts. Ukhetho lokugqibela lubonisa iincoko zababini, koku ungasebenzisa umphathi . Yakhelwe kwi-SDK ye-Bot Framework njengesixhobo sophuhliso olubonakalayo apho amaqela anqamlekileyo angasebenzisa ukwenza i-bots.
Umqambi we-Framework ye-Bot ikuvumela ukuba usebenzise iibhloko ukwenza ulwakhiwo lwencoko apho i-bot iza kusebenza khona. Ukongeza, unokwenza izinto ezinokubangela, oko kukuthi, amagama angundoqo apho i-bot iya kusabela ngayo ngexesha lengxoxo. Umzekelo, amagama athi "umqhubi", "ubusela" okanye "yeka" kunye "nokwaneleyo".
KwiBot Framework Composer, ungenza inkqubo yencoko yababini enzima usebenzisa . Iingxoxo zinokusebenzisa zombini iinkonzo zokuqonda kunye namakhadi esiganeko (Amakhadi e-Adaptive):
Emva kokudala, unokusebenzisa i-chatbot ekubhaliseni, kwaye iskripthi esilungiselelwe ngokuzenzekelayo siya kudala zonke izibonelelo eziyimfuneko: iinkonzo zokuqonda, isicwangciso seSicelo, i-Application Insights, i-database, njalo njalo.
UMenzi weQnA
Ukudala i-bots elula esekelwe kwi-database ye-Q & A yenkampani, ungasebenzisa inkonzo yengqondo ye-QnA Maker. Iphunyezwe njengewizadi yewebhu elula, ikuvumela ukuba ufake ikhonkco kwisiseko solwazi lwenkampani (FAQ Urls) okanye usebenzise idatabase yoxwebhu kwi *.doc okanye *.pdf ifomathi njengesiseko. Emva kokudala isalathisi, i-bot iya kukhetha ngokuzenzekelayo iimpendulo ezifanelekileyo kakhulu kwimibuzo yomsebenzisi.
Ukusebenzisa i-QnAMaker, unokwenza kwakhona imixokelelwane yokucacisa imibuzo ngokudala ngokuzenzekelayo amaqhosha, ukongeza isiseko solwazi kunye nemethadatha, kwaye uqeqeshe ngakumbi inkonzo ngexesha lokusetyenziswa.
Inkonzo ingasetyenziswa njenge-chatbot eyenza lo msebenzi omnye kuphela, okanye njengenxalenye ye-chatbot enzima esebenzisa, ngokuxhomekeke kwisicelo, ezinye iinkonzo ze-AI okanye izinto ze-Bot Framework.
Ukusebenza nezinye iinkonzo zengqondo
Kukho iinkonzo ezininzi ezahlukeneyo zengqondo kwiqonga le-Azure. Ngokobuchwephesha, ezi ziinkonzo zewebhu ezizimeleyo ezinokubizwa kwikhowudi. Ukuphendula, inkonzo ithumela i-json yefomathi ethile, enokusetyenziswa kwi-chatbot.
Olona setyenziso luqhelekileyo lwee-chatbots zezi:
- Ukunakana okubhaliweyo.
- Ukuqondwa komphuhlisi-ochaziweyo kwiindidi zemifanekiso yeNkonzo yeSiko loMbono (imeko yemveliso: ukuqondwa kokuba umqeshwa unxibe umnqwazi oqinileyo, izibuko okanye imaski).
- Ukuqondwa kobuso (imeko yokusetyenziswa egqwesileyo kukujonga ukuba umntu ophononongwayo uthumele ubuso bakhe, okanye, masithi, ifoto yenja okanye ifoto yomntu wesini esahlukileyo).
- Ukuqondwa kwentetho.
- Uhlalutyo lomfanekiso.
- Uguqulo (sonke siyakhumbula ukuba ingakanani ingxolo eyabangelwa kukuguqulelwa ngaxeshanye kwiSkype).
- Ukujonga upelo kunye neengcebiso zokulungisa iimpazamo.
LUIS
Kwakhona, ukwenza i-bots unokufuna (ULwimi lokuQonda iNkonzo yobuNgcali). Iinjongo zenkonzo:
- Qinisekisa ukuba ingxelo yomsebenzisi iyavakala kwaye ingaba impendulo yebhot iyimfuneko.
- Nciphisa iinzame zokukhuphela intetho yomsebenzisi (isicatshulwa) kwimiyalelo eqondakalayo kwi-bot.
- Qikelela iinjongo zabasebenzisi bokwenyani/iinjongo kwaye ukhuphe iimbono eziphambili kumabinzana akwingxoxo.
- Vumela umphuhlisi ukuba aqalise i-bot esebenzisa imizekelo embalwa yokuqondwa kwentsingiselo kunye noqeqesho olongezelelweyo olulandelayo lwebhot ngexesha lokusebenza.
- Yenza umphuhlisi asebenzise umboniso ukuvavanya umgangatho woshicilelo lomyalelo.
- Ukuncedisa kuphuculo olongezelelekileyo ekunakaneni okujoliswe kuko okuyinyaniso.
Ngapha koko, eyona njongo iphambili ye-LUIS kukuqonda ngokunokwenzeka okuthile ukuba umsebenzisi wayethetha ntoni kwaye uguqule isicelo sendalo sibe ngumyalelo ohambelanayo. Ukuqaphela amaxabiso emibuzo, i-LUIS isebenzisa isethi yeenjongo (iintsingiselo, iinjongo) kunye nemibutho (inokuthi iqwalaselwe kwangaphambili ngabaphuhlisi, okanye ithathwe kwaye yenziwe kwangaphambili "iindawo" - ezinye iilayibrari esele zenziwe amabinzana aqhelekileyo alungiswe nguMicrosoft).
Umzekelo olula: unebhot ekunika uqikelelo lwemozulu. Kuye, injongo iya kuba kukuguqulelwa kwesicelo sendalo "kwisenzo" - isicelo sesimo sezulu, kwaye amaqumrhu aya kuba lixesha kunye nendawo. Nanku umzobo wendlela injongo yeCheckWeather esebenza ngayo kwibhot enjalo.
Injongo
Isincoko
Umzekelo wombuzo wendalo
CheckWeather
{"type": "location", "entity": "moscow"}
{"type": "builtin.datetimeV2.date", "entity": "future","resolution":"2020-05-30"}
Imozulu iya kuba njani ngomso eMoscow?
CheckWeather
{ "type": "date_range", "entity": "kulempelaveki"}
Ndibonise uqikelelo lwale mpelaveki
Ukudibanisa uMenzi we-QnA kunye ne-LUIS ongayisebenzisa .
Xa usebenza noMenzi we-QnA kwaye ufumana isicelo kumsebenzisi, inkqubo imisela ukuba yeyiphi ipesenti yokunokwenzeka impendulo evela kwi-QnA ehambelana nesicelo. Ukuba amathuba angaphezulu, umsebenzisi unikwa nje impendulo evela kwisiseko solwazi; ukuba iphantsi, isicelo sinokuthunyelwa kuLUIS ukuze sicaciselwe. Ukusebenzisa i-Dispatcher ikuvumela ukuba ungacwangciswanga le ngqiqo, kodwa ukugqiba ngokuzenzekelayo lo mda wokwahlulwa kwezicelo kwaye usasaze ngokukhawuleza.
Ukuvavanya kunye nokupapasha ibhot
Esinye isicelo sasekhaya sisetyenziselwa uvavanyo, . Ukusebenzisa i-emulator, unokunxibelelana ne-bot kwaye ujonge imiyalezo eyithumelayo kwaye iyifumana. I-emulator ibonisa imiyalezo njengoko iya kubonakala kwi-intanethi yengxoxo yewebhu kwaye ifake izicelo ze-JSON kunye neempendulo xa uthumela umyalezo kwi-bot.
Umzekelo wokusebenzisa i-emulator uboniswa kule demo, ebonisa ukudalwa komncedisi we-BMW. Ividiyo ikwathetha malunga nee-accelerator ezintsha zokwenza ii-chatbots-itemplates:
Unokusebenzisa iitemplates xa usenza ii-chatbots zakho.
Iitemplates zikuvumela ukuba ungabhali imisebenzi ye-bot esemgangathweni ngokutsha, kodwa ukongeza ikhowudi esele yenziwe “njengesakhono”. Umzekelo unokusebenza ngekhalenda, ukwenza amadinga, njl. njl. Ikhowudi yezakhono esele zenziwe kwi github.
Uvavanyo luphumelele, i-bot ilungile, kwaye ngoku kufuneka ipapashwe kwaye iziteshi zixhunyiwe. Upapasho lwenziwa kusetyenziswa i-Azure, kwaye abathunywa ngoko nangoko okanye iinethiwekhi zentlalo zinokusetyenziswa njengetshaneli. Ukuba awunayo itshaneli efunekayo yokufaka idatha, ungayikhangela kuluntu oluhambelanayo kwiGitHab.
Kwakhona, ukwenza i-chatbot egcwele ngokupheleleyo njengojongano lokunxibelelana nomsebenzisi kunye neenkonzo zokuqonda, ngokuqinisekileyo uya kufuna iinkonzo ezongezelelweyo ze-Azure, ezinje ngogcino-lwazi, i-serverless (i-Azure Functions), kunye neenkonzo ze-LogicApp kwaye, mhlawumbi. , Igridi yoMnyhadala.
UVavanyo kunye noHlalutyo
Ukuvavanya ukusebenzisana komsebenzisi, ungasebenzisa zombini uhlalutyo olwakhelwe ngaphakathi lweNkonzo ye-Azure Bot kunye nenkonzo ekhethekileyo ye-Application Insights.
Ngenxa yoko, unokuqokelela ulwazi ngokusekwe kwezi ndlela zilandelayo:
- Bangaphi abasebenzisi abafikelele kwi-bot kwiitshaneli ezahlukeneyo ngexesha elikhethiweyo.
- Bangaphi abasebenzisi abathumele umyalezo omnye ababuyileyo kamva baza bathumela omnye.
- Zingaphi izenzo ezithunyelwe kwaye zamkelwa kusetyenziswa itshaneli nganye ngexesha lekhefu elichaziweyo.
Ukusebenzisa i-Application Insights, unokubeka iliso nasiphi na isicelo kwi-Azure kwaye, ngokukodwa, ii-chatbots, ukufumana idatha eyongezelelweyo malunga nokuziphatha komsebenzisi, imithwalo kunye neempendulo ze-chatbot. Kufuneka kuqatshelwe ukuba inkonzo ye-Application Insights ine-interface yayo kwi-portal ye-Azure.
Ungasebenzisa kwakhona idatha eqokelelwe ngale nkonzo ukwenza imiboniso eyongezelelweyo kunye neengxelo zohlalutyo kwi-PowerBI. Umzekelo wengxelo enjalo kunye netemplate ye-PowerBI ingathathwa .
Enkosi nonke ngengqalelo yenu! Kweli nqaku sisebenzise ukusuka kwiwebhu yewebhu nguMicrosoft Azure umakhi uAnna Fenyushina “Xa abantu bengenaxesha. Uzisebenzisa njani i-100% ii-chatbots kunye neenkonzo zokuqonda ukwenza iinkqubo zesiqhelo ”, apho sibonise ngokucacileyo ukuba zeziphi ii-chatbots ezikwi-Azure kwaye zeziphi iimeko zokusetyenziswa kwazo, kwaye sibonise nendlela yokwenza i-bot kwi-QnA Maker kwimizuzu eli-15 kunye nendlela Ulwakhiwo lombuzo lucaciswa kwi-LUIS.
Senze le webinar njengenxalenye yomdyarho we-intanethi wabaphuhlisi beDev Bootcamp. Yayimalunga neemveliso ezikhawulezisa uphuhliso kunye nokunciphisa ezinye zemisebenzi yesiqhelo evela kubasebenzi benkampani usebenzisa izixhobo ezizenzekelayo kunye neemodyuli ze-Azure ezilungiselelwe kwangaphambili. Ukurekhodwa kwezinye iiwebinars ezibandakanyiweyo kumdyarho ziyafumaneka kula makhonkco alandelayo:
-
Sithetha ngendaleko kunye nobuchule obutsha benkonzo ye-Azure DevOps. Sihlalutya ngokweenkcukacha zontlanu zamacandelo ayo aphambili kunye nendlela inkonzo inceda ngayo umbutho wophuhliso lwesoftware yeshishini. Umbonisi: Vladimir Gusarov Microsoft MVP.
Sijonga indlela yokuphumeza indlela "yokudibana njengekhowudi" kuzo naziphi na iilwimi zenkqubo kusetyenziswa imibhobho ye-YAML enamanqanaba amaninzi. Umbonisi: Vladimir Gusarov Microsoft MVP.
Sixoxa ngendlela yokuphuhlisa amaqhinga okugcina ikhowudi yomthombo wesoftware kunye nendlela yokusebenzisa iinkqubo zokugcina iinguqulelo usebenzisa iAzure Repos. Umbonisi: Vladimir Gusarov Microsoft MVP.
Sazisa iindlela zolawulo lomthombo ovulekileyo kuphuhliso kunye nokusetyenziswa okusebenzayo kwesixhobo seWhiteSource. Sithetha malunga nokukhangela ubuthathaka kumthombo ovulekileyo kunye nokugcina ukucoceka kwelayisensi kusetyenziswa i-WhiteSource. Ababonisi: UDiana Lisbaron, uMlawuli weSales WhiteSource, kunye noDaria Oreshkina, uLawulo lweWebhu, umthengisi osemthethweni we-WhiteSource eRashiya.
umthombo: www.habr.com