Hello, Habr! A yau za mu nuna muku yadda ake amfani da Azure don magance matsalolin da galibi ke buƙatar sa hannun ɗan adam. Wakilai suna ciyar da lokaci mai yawa don amsa tambayoyi iri ɗaya, sarrafa kiran waya da saƙonnin rubutu. Chatbots suna sarrafa sadarwa ta atomatik da ganewa kuma suna rage nauyi akan mutane. Hakanan ana amfani da Bots a cikin Azure DevOps, inda suke ba da izini, alal misali, don amincewa da sakewa, sarrafa abubuwan gini - duba, farawa da tsayawa - kai tsaye daga Slack ko Ƙungiyoyin Microsoft. A zahiri, chatbot yana da ɗan tunowa da CLI, mai mu'amala ne kawai, kuma yana bawa mai haɓaka damar ci gaba da kasancewa cikin mahallin tattaunawar.
A cikin wannan labarin, za mu yi magana game da kayan aikin ƙirƙira chatbots, nuna yadda za a iya inganta su tare da ayyukan fahimi, da bayyana yadda ake haɓaka haɓakawa tare da shirye-shiryen da aka yi a Azure.
Chatbots da sabis na fahimi: menene kamanceceniya kuma menene bambance-bambance?
Don ƙirƙirar bots a cikin Microsoft Azure, kuna amfani da Sabis na Bot na Azure da Tsarin Bot. Tare suna wakiltar saitin software don ginawa, gwaji, ƙaddamarwa da gudanar da bots, wanda ke ba ku damar ƙirƙira daga shirye-shiryen da aka ƙera duka tsarin sadarwa mai sauƙi da ci gaba tare da tallafin magana, ƙwarewar harshe na yanayi da sauran damar.
Bari mu ɗauka cewa kana buƙatar aiwatar da bot mai sauƙi dangane da sabis na Q&A na kamfani ko, akasin haka, ƙirƙirar bot mai aiki tare da hadaddun tsarin sadarwa mai rassa. Don yin wannan, zaka iya amfani da kayan aikin da dama, waɗanda aka kasu kashi uku:
- Sabis don saurin haɓaka hanyoyin sadarwa (bots).
- Shirye-shiryen fahimi AI sabis don lokuta daban-daban na amfani (bayanin ƙima, fahimtar magana, tushen ilimi da bincike).
- Ayyuka don ƙirƙira da horar da samfuran AI.
Yawanci, mutane da hankali suna rikitar da "bots" da "sabis na fahimi" saboda dukkanin ra'ayoyin sun dogara ne akan ka'idar sadarwa, kuma yanayin amfani da bots da ayyuka ya ƙunshi tattaunawa. Amma chatbots suna aiki tare da mahimman kalmomi da abubuwan jan hankali, kuma sabis na fahimi yana aiki tare da buƙatun sabani waɗanda yawanci mutane ke sarrafa su:
Sabis na fahimi wata hanya ce don sadarwa tare da mai amfani, yana taimakawa canza buƙatun sabani zuwa bayyananniyar umarni da aika shi ga bot.
Don haka, chatbots aikace-aikace ne don yin aiki tare da buƙatun, kuma sabis na fahimi kayan aiki ne don tantance buƙatun da aka ƙaddamar daban, amma wanda chatbot zai iya shiga, ya zama “masu hankali.”
Ƙirƙirar chatbots
Shawarar ƙira da aka ba da shawarar don bot a Azure shine kamar haka:
Don ƙira da haɓaka bots a cikin Azure, yi amfani . Akwai akan GitHub , iyawar tsarin yana canzawa, don haka wajibi ne a yi la'akari da sigar SDK da ake amfani da su a cikin bots.
Tsarin yana ba da zaɓuɓɓuka da yawa don ƙirƙirar bots: ta amfani da lambar gargajiya, kayan aikin layin umarni ko taswirar ruwa. Zaɓin na ƙarshe yana hango maganganu; don wannan zaka iya amfani da mai sarrafa . An gina shi akan Bot Framework SDK azaman kayan haɓaka na gani wanda ƙungiyoyin ladabtarwa zasu iya amfani da su don ƙirƙirar bots.
Bot Framework Composer yana ba ku damar amfani da tubalan don ƙirƙirar tsarin tattaunawa wanda bot ɗin zai yi aiki da shi. Bugu da ƙari, zaku iya ƙirƙirar abubuwan jan hankali, wato, kalmomin shiga waɗanda bot ɗin zai amsa yayin tattaunawar. Misali, kalmomin “operator”, “sata” ko “tsaya” da “isa”.
A cikin Mawaƙin Tsarin Tsarin Bot, zaku iya ƙirƙirar tsarin maganganu masu rikitarwa ta amfani da su . Tattaunawa na iya amfani da duka sabis na fahimi da katunan taron (Katunan Adaɗi):
Bayan ƙirƙira, zaku iya tura chatbot ɗin a cikin biyan kuɗi, kuma rubutun da aka shirya ta atomatik zai ƙirƙiri duk abubuwan da ake buƙata: sabis na fahimi, shirin aikace-aikacen, Bayanan aikace-aikacen, bayanan bayanai, da sauransu.
Mai QnA
Don ƙirƙirar bots masu sauƙi dangane da bayanan Q&A na kamfani, zaku iya amfani da sabis na fahimi na QnA Maker. An aiwatar da shi azaman mayen gidan yanar gizo mai sauƙi, yana ba ku damar shigar da hanyar haɗi zuwa tushen ilimin kamfanoni (FAQ Urls) ko amfani da bayanan daftarin aiki a cikin * .doc ko * .pdf tsarin a matsayin tushe. Bayan ƙirƙirar fihirisar, bot ɗin zai zaɓi mafi dacewa amsoshin tambayoyin mai amfani ta atomatik.
Yin amfani da QnAMAker, zaku iya ƙirƙirar sarƙoƙi na fayyace tambayoyi tare da ƙirƙirar maɓalli ta atomatik, haɓaka tushen ilimi tare da metadata, da ƙara horar da sabis yayin amfani.
Ana iya amfani da sabis ɗin azaman chatbot wanda ke aiwatar da wannan aikin guda ɗaya kawai, ko kuma a matsayin wani ɓangare na hadadden chatbot wanda ke amfani da shi, gwargwadon buƙatun, wasu ayyukan AI ko abubuwan da ke cikin Bot Framework.
Yin aiki tare da sauran sabis na fahimi
Akwai hidimomi daban-daban na fahimi akan dandalin Azure. A fasaha, waɗannan ayyukan gidan yanar gizo ne masu zaman kansu waɗanda za a iya kira daga lamba. Don amsawa, sabis ɗin yana aika json na wani tsari, wanda za'a iya amfani dashi a cikin chatbot.
Mafi yawan amfani da chatbots sune:
- Gane rubutu.
- Gane nau'ikan hoton Sabis na hangen nesa mai haɓakawa (harka samarwa: sanin ko ma'aikaci yana sanye da hula mai wuya, tabarau ko abin rufe fuska).
- Gane fuska (kyakkyawan yanayin amfani shine bincika ko mutumin da ake binciken ya buga fuskarsa, ko, a ce, hoton kare ko hoton mutumin da ke da bambancin jinsi).
- Gane magana.
- Binciken hoto.
- Fassara (dukkanmu mun tuna nawa amo nawa fassarar lokaci guda a Skype ya haifar).
- Duban haruffa da shawarwari don gyara kurakurai.
LUIS
Hakanan, don ƙirƙirar bots kuna iya buƙata (Language Understanding Intelligent Service). Makasudin sabis:
- Ƙayyade ko bayanin mai amfani yana da ma'ana kuma ko amsawar bot ya zama dole.
- Rage ƙoƙarin kwafin magana mai amfani (rubutu) zuwa umarni da aka fahimta ga bot.
- Yi tsinkaya maƙasudai/manufofin mai amfani na gaskiya kuma cire mahimman bayanai daga jimloli a cikin tattaunawa.
- Bada izini ga mai haɓakawa ya ƙaddamar da bot ta amfani da ƴan misalan ƙayyadaddun ma'ana da ƙarin horo na bot yayin aiki.
- Ba da damar mai haɓakawa don amfani da hangen nesa don tantance ingancin rubutun umarni.
- Taimaka wajen haɓaka haɓakawa a gano ainihin manufa.
A zahiri, babban burin LUIS shine fahimtar tare da takamaiman yuwuwar abin da mai amfani ke nufi da canza buƙatun halitta zuwa umarni mai jituwa. Don gane ƙimar tambaya, LUIS tana amfani da saitin abubuwan niyya (ma'ana, niyya) da ƙungiyoyi (ko dai waɗanda masu haɓakawa suka tsara su, ko kuma waɗanda aka riga aka tsara su "yankin" - wasu shirye-shiryen dakunan karatu na daidaitattun jimlolin da Microsoft ta shirya).
Misali mai sauƙi: kuna da bot wanda ke ba ku hasashen yanayi. A gare shi, manufar za ta zama fassarar buƙatun halitta zuwa "aiki" - buƙatun hasashen yanayi, kuma ƙungiyoyin za su kasance lokaci da wuri. Anan akwai zane na yadda niyyar CheckWeather ke aiki don irin wannan bot.
Niyya
Essence
Misalin tambaya ta halitta
Duba Yanayi
{"nau'in": "wuri", "hukunta": "moscow"}
{"nau'in": "builtin.datetimeV2.date", "haɓaka": "gaba","ƙuduri":"2020-05-30"}
Yaya yanayin zai kasance gobe a Moscow?
Duba Yanayi
{"type": "date_range", "entity": "this weekend" }
Nuna mani hasashen wannan karshen mako
Don haɗa QnA Maker da LUIS za ku iya amfani da su .
Lokacin da kuke aiki tare da QnA Maker kuma kuna karɓar buƙatu daga mai amfani, tsarin yana ƙayyade adadin yuwuwar amsa daga QnA tayi daidai da buƙatar. Idan yuwuwar ta yi girma, ana ba mai amfani kawai amsa daga tushen ilimin kamfani; idan yana da ƙasa, ana iya aika buƙatar zuwa LUIS don ƙarin haske. Yin amfani da Dispatcher yana ba ku damar tsara wannan dabaru, amma don ƙayyade wannan gefen rarrabuwar buƙatun da sauri da rarraba su.
Gwaji da buga bot
Ana amfani da wani aikace-aikacen gida don gwaji, . Yin amfani da abin koyi, zaku iya sadarwa tare da bot kuma duba saƙonnin da yake aikawa da karɓa. Mai kwaikwayon yana nuna saƙonni kamar yadda zasu bayyana a cikin mahaɗin taɗi na yanar gizo kuma suna yin rajistar buƙatun JSON da martani lokacin aika saƙon bot.
An gabatar da misalin yin amfani da kwaikwayi a cikin wannan demo, wanda ke nuna ƙirƙirar mataimaki mai kama da BMW. Bidiyon kuma yayi magana game da sababbin masu haɓakawa don ƙirƙirar taɗi - samfuri:
Hakanan zaka iya amfani da samfuri lokacin ƙirƙirar bot ɗin ku.
Samfuran suna ba ku damar rubuta daidaitattun ayyukan bot sabo, amma don ƙara lambar da aka yi shirye-shirye azaman “ƙwarewa”. Misali zai iya zama aiki tare da kalanda, yin alƙawura, da sauransu. Lambar fasahar da aka shirya na github.
Gwajin ya yi nasara, bot yana shirye, kuma yanzu yana buƙatar bugawa kuma an haɗa tashoshi. Ana yin bugu ta hanyar amfani da Azure, kuma ana iya amfani da manzo ko shafukan sada zumunta azaman tashoshi. Idan baku da tashar da ake buƙata don shigar da bayanai, zaku iya nemo ta a cikin al'ummar da ta dace akan GitHab.
Hakanan, don ƙirƙirar cikakken chatbot azaman hanyar sadarwa don sadarwa tare da mai amfani da sabis na fahimi, ba shakka, kuna buƙatar ƙarin sabis na Azure, kamar bayanan bayanai, marasa sabar (Ayyukan Azure), da sabis na LogicApp da, mai yiwuwa. , Grid Event.
Ƙimar da Bincike
Don ƙididdige hulɗar mai amfani, zaku iya amfani da duka ginanniyar nazari na Sabis na Bot na Azure da sabis na Fahimtar Aikace-aikace na musamman.
A sakamakon haka, zaku iya tattara bayanai bisa ma'auni masu zuwa:
- Masu amfani nawa ne suka sami damar shiga bot daga tashoshi daban-daban a lokacin da aka zaɓa.
- Masu amfani nawa da suka aika sako daya suka dawo daga baya suka aika wani.
- Ayyuka nawa ne aka aika da karɓa ta amfani da kowane tashoshi yayin ƙayyadaddun tazarar lokaci.
Yin amfani da Haɓaka Aikace-aikacen, zaku iya saka idanu akan kowane aikace-aikacen a cikin Azure kuma, musamman, taɗi, samun ƙarin bayanai game da halayen mai amfani, lodi da halayen chatbot. Ya kamata a lura cewa sabis ɗin Insights na Aikace-aikacen yana da nasa keɓancewa a cikin tashar Azure.
Hakanan zaka iya amfani da bayanan da aka tattara ta wannan sabis ɗin don ƙirƙirar ƙarin abubuwan gani da rahotannin nazari a cikin PowerBI. Ana iya ɗaukar misalin irin wannan rahoto da samfuri na PowerBI .
Na gode duka saboda kulawar ku! A cikin wannan labarin mun yi amfani da shi daga gidan yanar gizon Microsoft Azure m Anna Fenyushina "Lokacin da mutane ba su da lokaci. Yadda ake amfani da 100% na chatbots da sabis na fahimi don sarrafa ayyukan yau da kullun", inda muka nuna a sarari menene chatbots a cikin Azure da menene yanayin amfani da su, kuma mun nuna yadda ake ƙirƙirar bot a cikin QnA Maker a cikin mintuna 15 da kuma yadda An ƙaddamar da tsarin tambaya a cikin LUIS.
Mun yi wannan webinar a matsayin wani ɓangare na marathon kan layi don masu haɓaka Dev Bootcamp. Ya kasance game da samfuran da ke haɓaka haɓakawa da sauƙaƙe wasu ayyukan yau da kullun daga ma'aikatan kamfani ta amfani da kayan aikin sarrafa kansa da kuma shirye-shiryen da aka riga aka tsara na Azure. Ana samun rikodin sauran gidajen yanar gizon da aka haɗa a cikin marathon a hanyoyin haɗin yanar gizo masu zuwa:
-
Muna magana game da juyin halitta da sabbin damar sabis na Azure DevOps. Muna yin nazari dalla-dalla dalla-dalla duk manyan abubuwan da ke tattare da shi da kuma yadda sabis ɗin ke taimakawa ƙungiyar haɓaka software na masana'antu. Mai gabatarwa: Vladimir Gusarov Microsoft MVP.
Mun yi la'akari da yadda ake aiwatar da tsarin "majalisar a matsayin lambar" ga kowane yarukan shirye-shirye ta amfani da bututun YAML masu matakai da yawa. Mai gabatarwa: Vladimir Gusarov Microsoft MVP.
Mun tattauna yadda ake haɓaka dabarun adana lambar tushen software da kuma yadda ake amfani da tsarin ajiyar sigar ta amfani da Azure Repos. Mai gabatarwa: Vladimir Gusarov Microsoft MVP.
Muna gabatar da buɗaɗɗen dabarun sarrafa tushen tushe a cikin haɓakawa da aikace-aikacen aikace-aikacen kayan aikin WhiteSource. Muna magana ne game da neman lahani a cikin abubuwan buɗe tushen tushe da kiyaye tsabtar lasisi ta amfani da WhiteSource. Masu gabatarwa: Diana Lisbaron, Daraktan Tallace-tallace ta WhiteSource, da Daria Oreshkina, Sarrafa Yanar Gizo, mai rarraba hukuma na WhiteSource a Rasha.
source: www.habr.com