《深度自然语言处理》课程教学课件(Natural language processing with deep learning)16 Natural Language Generation

西安交通大学Natural languageprocessingwith deeplearningXIANHAOTONGUNIVERSITYNatural Language Generation交通大学ChenLicli@xjtu.edu.cn2023
Chen Li cli@xjtu.edu.cn 2023 Natural Language Generation Natural language processing with deep learning

Outlines1. What is NLG?2. Formalizing NLG3.Decoding from NLG models1Training NLG models5.Evaluating NLG Systems6. Ethical Considerations
Outlines 1. What is NLG? 2. Formalizing NLG 3. Decoding from NLG models 4. Training NLG models 5. Evaluating NLG Systems 6. Ethical Considerations

NLGWhat is natural language generation?Natural language generation (NLG) is asub-field of natural language processingFocused on building systems thatautomatically produce coherent anduseful writtenorspokentextforhumanconsumptionNLG systems are already changing theworld we live in
NLG What is natural language generation? • Natural language generation (NLG) is a sub-field of natural language processing • Focused on building systems that automatically produce coherent and useful written or spoken text for human consumption • NLG systems are already changing the world we live in

NLGMachineTranslationEG文ax1FrenchEnglishXJ'aimange avecIatewithmylawyertodaymonavocataujourd'hui口D1
NLG Machine Translation

NLGDialogue Systems9.41AMWhatcanIhelpyou with?ama:
NLG Dialogue Systems

NLGSummarizationDocumentSummarizationE-mailSummarization1-5.0f5re-thinkingcom.cy-1minread,122.wordsoInboxxTL:DR: Anyone should be able to buya.cy domain regardless of location in a quick anefficiertwayEmintesd, 122,wordsSep 8, 2019; 11:53 AMArgyrouArgyriscatgyrou,aggmeil.comto.me *Hey.Cyprus country codeTLDregistraraitay operaled bytheUnversityorfCyprusis theONLYwaytoregisteracom.CydomaininCyprus.Wearetakingabout abureaucraticprocessIssildon'tgetitwhywecan'tfreelyregister.cynames,Rightnowyoucan'tbuy.cydomainsonlygamdyanda listof other ywtiatexetueelesa.cy domain extensioes.Releasing.cy will help the sales and promotion of our nationnl country code top levol donItwilbeanewdomain introfree.toregistor.Aeyoneshouidbeabletobuyacydomain tegardiessofjocation,in aquicandeffidentwaynicCyshouldprovidethihttp://mogren.one/lic/nttnss//cnroneestpre/cetal/emesummariation
NLG Summarization https://chrome.google.com/webstore/detail/gmail-summarization/ Document Summarization E-mail Summarization http://mogren.one/lic/

NLGData-to-TextGenerationTable Title: Robert Craig (American football)Section Title: National Football League statisticsTable Description:NoneRUSHINGRECEIVINGATTYDSLNGTDNO.YDSLNGTDYEARTEAMAVGAVG1983SF176725718484272344.18.931984SF1556494.2284716759.564SF99273198521410504.96210161162577.71986SF2048304.181624480SF2153.82537.53519878156649211988SF3104.8469765347.02211502271SF3.9276494739.7441989105411990SF1414393.1261252018.0310172001991RAI1625903.61511368.04227.5221992MIN1054164.02116401993MIN381193.111198.9311116956617199181894.1715649118.773Totals-Target Text: Craig finished his eleven NFL seasons with 8,189 rushing yards and 566 receptions for 4,911 receiving yards.(Parikhetal.,EMNLP2020)The Atlanta Hawks defeated the Miami Heat,103 -95,at Philips Arena on WednesdayWINRBLOSSPTSFG_PCTAS...Atlanta was in desperate need of a win andTEAMthey were able to take care of a shorthandedHeat1712103494727Miami team hereDefense was key for1595433320Hawksthe Hawks , as they held the Heat to 42percent shooting and forced them to commit16 turnoversAtlanta also dominated in thePTASRBFGFGACITYpaint, winning the rebounding battle,47.-PLAYER-34,and outscoring them in the paint 58-26.The Hawks shot 49 percent from the5227Tyler Johnson816Miamifield and assisted on 27 of their 43 madeo4172311Dwight HowardAtlantabaskets.This was a near wire -to -wire2921812Paul MillsapAtlantawin for the Hawks,as Miami held just one2m421817Goran DragicMiamilead in the first five minutes., Miami (7-2719715Wayne EllingtonMiami15)are as beat-up as anyone right now417815Dennis SchroderAtlantaand it 's taking a toll on the heavily used55113.8Rodney McGruderMiamistarters .Hassan Whiteside really struggled5551011Thabo SefoloshaAtlantain this game, as he amassed eight points,35399KyleKorverAtlanta12 rebounds and one blocks on 4 - of -12shooting ...(WisemanandRush.,EMNLP2017)
NLG Data-to-Text Generation (Wiseman and Rush., EMNLP 2017) (Parikh et al., EMNLP 2020)

NLGVisualDescriptionhotdog stand on busy streetbowls are food in trianpulur shape are sitting on tableman in whitetshit isholding umbrellaand ice cream carttable filled with many plates of various breakfast foodstabletoppedwith lots ofdifferent typesofdonutsman in white shirt is pushing his cart down streetSTOPmaningraduationrobesridingbicycleonemanandtwowomensittingin livingroomCyclist giving thumbs upposes withhisbicycle by rightmanandwomanareplayingwil gamewhile womanofwaysignatparksits on couch withwineglass in her handman ridingmotorcycleon streetgroupofpeople.sitting on couchwith.theirlaptops(Karpathy&Li.,CVPR2015)
NLG (Karpathy & Li., CVPR 2015) Visual Description

NLGCreativeGenerationPoetryStories&NarrativesVocabularybigbird'sbirthday celebrationPeorstyieEncouraiget wortismommsStyleCurse wordsrepetitionalterationStoryord lengehcookiemonstereats?迪??Outlinerollerskatingrinksentimenttopicakwordismanosylablewordsconcrete wordsbig birthday cake-常PlotdynamicsloveGenerateRe-generate with same rhyme wordsP'= paragraphiPoemOutline-conditioned Story Generation合My lovely lady sweet and sweet temptation,It is Big Bird's birthday,and he goes to the rollerThe lucky woman on the wedding night,I reallyneedafriendof consoiationskating rink with his friends.Alonelypartof youand metonightBackat Sesame Street,Maria and Susan take out thebig(a) Poem generated with default style setingsPIbirthday cakeand leave iton atableVocabularyResetStylEncourgge wordmommaCookieMonstersees the cake,but insteadofeating itStylerepectioncurse wordsaliterationwordlengthandspoilingthepartyheeatsachairandotherthingsall?P物福overSesameStreetsentimenttopical wordsmonosyilabie wordsconcreteBigBirdandtheotherskatersreturntoSesameStreetbandareshockedatwhatCookieMonsterate.thoughtheloveGenerateRe-generate with same rhyme wordscake is safeP2/PoemGinaand Count Von Count presentsthe caketo Big Bird食面青吉Thanks toryourteedbackIthas548candleseventhoughBigBird is6yearsoldMymemy littleloveandsweettemotation,Attheend,whenGinaannouncesthesponsors,CookieThe lucky lady on a wedding nightShesingsthesweetestsongofconsolationMonstereatsthemalongwithhiscake.Alovelydreamofyouandme.tonight(b) Poem generated with user adjusted style settings(Ghazvininejadetal..AcL2017)(Rashkinetal.,EMNLP2020)
NLG Creative Generation (Ghazvininejad et al., ACL 2017) (Rashkin et al., EMNLP 2020) Stories & Narratives Poetry

NLGWhat isnatural languagegeneration?Anytaskinvolvingtextproductionforhumanconsumption requires natural language generationDeep Learning is powering next-gen NLG systems!
NLG What is natural language generation? Any task involving text production for human consumption requires natural language generation Deep Learning is powering next-gen NLG systems!
按次数下载不扣除下载券;
注册用户24小时内重复下载只扣除一次;
顺序:VIP每日次数-->可用次数-->下载券;
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)14 Question Answering.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)15 Machine translation.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)10 information extraction.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)11 coreference resolution.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)12 sentiment analysis.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)09 Language Model & Distributed Representation(6/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)08 Language Model & Distributed Representation(5/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)07 Language Model & Distributed Representation(4/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)05 Language Model & Distributed Representation(2/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)06 Language Model & Distributed Representation(3/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)04 Language Model & Distributed Representation(1/6).pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)03 Fundamental Tasks of NLP.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)01 About the course.pdf
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)02 What is NLP, why NLP and How NLP.pdf
- 佛山大学(佛山科学技术学院):2022年版计算机科学与技术专业理论课程教学大纲汇编.pdf
- 佛山大学(佛山科学技术学院):2022年版物联网实践课程教学大纲汇编.pdf
- 佛山大学(佛山科学技术学院):2022年版智能科学与技术专业理论课程教学大纲汇编.pdf
- 佛山大学(佛山科学技术学院):2022年版物联网实验课程教学大纲汇编.pdf
- 《物联网导论》课程教学资源(PPT课件)第16章 SLAM空间智能计算.pptx
- 《物联网导论》课程教学资源(PPT课件)第15章 低功耗广域网.pptx
- 《深度自然语言处理》课程教学课件(Natural language processing with deep learning)17 Deep leanring Programing framework.pdf
- 全国信息安全标准化技术委员会:大数据安全标准化白皮书(2018 版).pdf
- 沈阳师范大学:《大学计算机基础》课程教学大纲 Fundamentals of University Computer A.pdf
- 沈阳师范大学:《大学计算机基础》课程授课教案(讲义,共五章,任课教师:刘冰).pdf
- 《大学计算机基础》课程教学资源(教案讲义,共五章,沈阳师范大学:刘冰).pdf
- 《大学计算机基础》课程教学大纲 Fundamentals of University Computer A.pdf
- 《大学计算机基础》课程教学资源(PPT课件,完整讲稿,共五章).pptx
- 《数据库技术与应用》课程教学资源(授课教案)第1章 数据库基础、第2章 数据库和表(沈阳师范大学:安晓飞).pdf
- 沈阳师范大学:《大学计算机基础》课程教学资源(PPT课件,完整讲稿,共五章).pptx
- 沈阳师范大学:《数据库原理》课程教学大纲 DataBase Principle.pdf
- 沈阳师范大学:《数据库原理》课程授课教案(讲义,共十章,主讲:马佳琳).pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第10章 数据库恢复技术.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第11章 并发控制.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第1章 概论(主讲:马佳琳).pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第2章 关系数据库.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第3章 关系数据库标准语言SQL.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第4章 数据库安全性.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第5章 数据库完整性.pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第6章 关系数据理论(1/4).pdf
- 沈阳师范大学:《数据库原理》课程教学课件(讲稿)第6章 关系数据理论(2/4).pdf
