資源描述：
《natural language understanding and prediction technologies》由會員上傳分享，免費在線閱讀，更多相關(guān)內(nèi)容在學(xué)術(shù)論文-天天文庫。
1、NaturalLanguageUnderstandingandPredictionTechnologiesNicolaeDutaCloudML@Microsoft1IJCAI2015TutorialOutline?Voiceandlanguagetechnologies:history,examplesandtechnologicalchallenges?ShortintrotoASR:modeling,architecture,analytics?Languageprediction(akamodeling)?NaturalLangua
2、geUnderstanding?Supervisedlearningapproaches:training&annotationissues?Semi-supervisedlearningapproaches?Parsers&hybridmodels,multilingualmodels?Client-serverarchitectures,dialog&semanticequations?Humaninteractionwithvoice&languagetechnologies?Semanticweb-search?Disclosur
3、e2IJCAI2015TutorialDeployedlanguagetechnologiesMostapplicationsthattranslatesomesignalintotextemployaBayesianapproach:argmaxP(sentence
4、signal)?sentenceargmaxP(signal
5、sentence)?P(sentence)sentenceApplications?Speechrecognition?Opticalcharacterrecognition?Handwritingrecogni
6、tion?Machinetranslation?Spellingcorrection?Word/sentenceautocompletion3IJCAI2015TutorialTechnologiesbasedonvoiceinput?Technologiesthatusespokeninputforrequestinginformation,webnavigationorcommandexecution–DAsystems:Nuance(bNuance+PhoneticSystems),BBN/Nortel,TellMe/Microso
7、ft,Jingle,Google,AT&T,IBM(mid1990s)–Dictation/speechtotextsystems:Dragon(mid1990s)–TVclosecaptioningBBN/NHK(early2000s)–Automatedattendant&Callrouting:AT&T,BBN,Nuance,IBM(early2000s)–Form-fillingdirecteddialog(flightreservations)(early2000s)–Personalassistants/Fullwebsear
8、ch:Siri/Apple,DragonGo,GoogleVoice,Vlingo/SVoice,MicrosoftCortana(from2008)–Manydedicatedsystems:–TVcontrol+music/videomanagement:DragonTV,Xboxone–Onlinebanking&Stockpricesearch–Productreviews&FAQsearch–Medicalfactextractionfrommedicalreports4IJCAI2015TutorialTechnologies
9、basedonvoiceinput:history?Architecture:Speechrecognizer+NLU+Dialogmanager–Oldersystems:centralized,deployedinthecustomer’sprocessingcenters–Newsystems:client-server,serverdeployedinthemanufacturer’sprocessingcenter,clientapponuser’s(mobile)device?NLUapproaches:–Handwritte
10、ngrammarrules(top-down):STUDENT,ELIZA–Contextindependentgrammarsfromtrainingtext:Tina(MIT)–Super
