IJCNLP 2017 The Eighth International Joint Conference on Natural Language Processing ...

234 Pages·2017·12.31 MB·English

IJCNLP 2017
The Eighth International Joint Conference
on Natural Language Processing
Proceedings of the IJCNLP 2017, Shared Tasks
November 27 – December 1, 2017
Taipei, Taiwan
(cid:13)c 2017AsianFederationofNaturalLanguageProcessing
978-1-948087-03-2(ProceedingsoftheIJCNLP2017,SharedTasks)
ii
Introduction
The 8th International Joint Conference on Natural Language Processing (IJCNLP 2017) took place in
Taipei,TaiwanfromNovember27toDecember1,2017. ItwasorganizedbytheNationalTaiwanNormal
University and by the Association for Computational Linguistics and Chinese Language Processing
(ACLCLP),anditwashostedbytheAsianFederationofNaturalLanguageProcessing(AFNLP).
ForafirsttimeinthehistoryofIJCNLP,theconferencefeaturedsharedtasks. Wereceivedatotaloften
taskproposals,andafterarigurousreview,weacceptedthefollowingfiveofthem:
• Task 1: Chinese Grammatical Error Diagnosis. Participants were asked to build systems
to automatically detect the errors in Chinese sentences made by Chinese-as-Second-Language
learners, i.e., redundant word, missing word, word selection and word ordering. (Organized by:
GaoqiRao,BaolinZhang,andEndongXun)
• Task 2: Dimensional Sentiment Analysis for Chinese Phrases. Given a word or a phrase,
participants were asked to generate a real-valued score between 1 and 9, indicating the degree of
valence, from most negative to most positive, and for the degree of arousal, from most calm to
mostexcited. (OrganizedbyLiang-ChihYu,Lung-HaoLee,JinWang,andKam-FaiWong)
• Task3: ReviewOpinionDiversification. Participantswereaskedtobuildsystemstorankproduct
reviews based on a summary of opinions in two domains: books and electronics. (Organized by
AnilKumarSingh,JulianMcAuley,AvijitThawani,MayankPanchal,AnubhavGupta,andRajesh
KumarMundotiya)
• Task 4: Customer Feedback Analysis. Participants were asked to train classifiers for the
detectionofmeaningincustomerfeedbackinEnglish,French,Spanish,andJapanese: comment,
request,bug,complaint,meaningless,andundetermined. (OrganizedbyChao-HongLiu,Yasufumi
Moriya,AlbertoPoncelas,andDeclanGroves)
• Task 5: Multi-choice Question Answering in Examinations. Participants were asked to build
systems to choose the correct option for a multi-choice question: for English and Chinese.
(OrganizedbyJunZhao,KangLiu,ShizhuHe,ZhuoyuWei,andShangminGuo)
A total of 40 teams participated in the five tasks (and many more registered to participate, but ended
upnotsubmittingsystems),submittinghundredsofrunsforthedifferenttasksandtheirsubtasks: 5for
task 1, 13 for task 2, 3 for task 3, 12 for task 4, and 7 for task 5. Moreover, most of the participating
teamscontributedasystemdescriptionpaper: 3fortask1,10fortask2,3fortask3,9fortask4,and6
fortask5. Finally,theorganizersofeachtaskpreparedataskdescriptionpaper. Alltheseappearinthe
presentproceedings.
Wethankthesharedtaskparticipants,aswellasthetaskorganizers,foralltheirgreatwork. Wefurther
taketheopportunitytothanktheprogramcommitteeandallreviewersfortheirthoroughreviews.
TheIJCNLP’2017SharedTaskCo-Chairs:
Chao-HongLiu,ADAPTCentre,DublinCityUniversity,Ireland
PreslavNakov,QatarComputingResearchInstitute,HBKU,Qatar
NianwenXue,BrandeisUniversity,USA
iii

Organizing Committee
SharedTaskWorkshopCo-Chairs
Chao-HongLiu,ADAPTCentre,DublinCityUniversity
PreslavNakov,QatarComputingResearchInstitute,HBKU
NianwenXue,BrandeisUniversity
Task1: ChineseGrammaticalErrorDiagnosis(CGED)Organizers
GaoqiRao,BeijingLanguageandCultureUniversity
BaolinZhang,BeijingLanguageandCultureUniversity
EndongXun,BeijingLanguageandCultureUniversity
Task2: DimensionalSentimentAnalysisforChinesePhrases(DSAP)Organizers
Liang-ChihYu,YuanZeUniversity
Lung-HaoLee,NationalTaiwanNormalUniversity
JinWang,YunnanUniversity
Kam-FaiWong,TheChineseUniversityofHongKong
Task3: ReviewOpinionDiversificationOrganizers
AnilKumarSingh,IndianInstituteofTechnology(BHU)Varanasi
JulianMcAuley,UniversityofCalifornia,SanDiego
AvijitThawani,IndianInstituteofTechnology(BHU)Varanasi
MayankPanchal,IndianInstituteofTechnology(BHU)Varanasi
AnubhavGupta,IndianInstituteofTechnology(BHU)Varanasi
RajeshKumarMundotiya,IndianInstituteofTechnology(BHU)
Task4: CustomerFeedbackAnalysisOrganizers
Chao-HongLiu,ADAPTCentre,DublinCityUniversity
YasufumiMoriya,ADAPTCentre,DublinCityUniversity
AlbertoPoncelas,ADAPTCentre,DublinCityUniversity
DeclanGroves,MicrosoftDublin
Task5: Multi-choiceQuestionAnsweringinExaminationsOrganizers
JunZhao,InstituteofAutomation,ChineseAcademyofSciences
KangLiu,InstituteofAutomation,ChineseAcademyofSciences
ShizhuHe,InstituteofAutomation,ChineseAcademyofSciences
ZhuoyuWei,InstituteofAutomation,ChineseAcademyofSciences
ShangminGuo,InstituteofAutomation,ChineseAcademyofSciences
v
Program Committee
Reviewers
AlbertoPoncelas,ADAPTCentre,DublinCityUniversity
AnilKumarSingh,IndianInstituteofTechnology(BHU)Varanasi
AnubhavGupta,IndianInstituteofTechnology(BHU)Varanasi
AvijitThawani,IndianInstituteofTechnology(BHU)Varanasi
BaolinZhang,BeijingLanguageandCultureUniversity
Chao-HongLiu,ADAPTCentre,DublinCityUniversity
EndongXun,BeijingLanguageandCultureUniversity
GaoqiRao,BeijingLanguageandCultureUniversity
HaithemAfli,ADAPTCentre,DublinCityUniversity
JinWang,YunnanUniversity
JulianMcAuley,UniversityofCalifornia,SanDiego
JunZhao,InstituteofAutomation,ChineseAcademyofSciences
Kam-FaiWong,TheChineseUniversityofHongKong
KangLiu,InstituteofAutomation,ChineseAcademyofSciences
Liang-ChihYu,YuanZeUniversity
Lung-HaoLee,NationalTaiwanNormalUniversity
MayankPanchal,IndianInstituteofTechnology(BHU)Varanasi
MonalisaDey,JadavpurUniversity
NianwenXue,BrandeisUniversity
PreslavNakov,QatarComputingResearchInstitute,HBKU
PruthwikMishra,InternationalInstituteofInformationTechnology,Hyderabad
RajeshKumarMundotiya,IndianInstituteofTechnology(BHU)Varanasi
ShangminGuo,InstituteofAutomation,ChineseAcademyofSciences
Shih-HungWu,ChaoyangUniversityofTechnology
YasufumiMoriya,ADAPTCentre,DublinCityUniversity
vi
Invited Talk
Public Health Surveillance Using Twitter: The Case for Biosurveillance and
Pharmacovigilance
AntonioJimenoYepes
IBMResearch,Australia
Abstract
Publichealthsurveillanceusingclinicaldataischallengingduetoissuesrelatedtoaccessinghealth
care data in a homogeneous way and in real-time, which is further affected by privacy concerns.
Yet, it is still relevant to access this data in real-time to model potential disease outbreaks and to
detect post-marketing adverse events of drugs. Social networks such as Twitter provide a large
quantityofinformationthatcanberelevantasanalternativetoclinicaldata. Wehaveresearched
theusageofTwitterinseveraltasksrelatedtopublichealthsurveillance. Inthistalk,Iwillpresent
the work that we have done in IBM Research Australia using Twitter in public health related
problems and the challenges that we have faced using Twitter. Specifically, I will show results
related to the prediction of the prevalence of flu in the USA and related to the identification of
post-marketingadverseeventsofdrugs.
Biography
Dr Antonio Jimeno Yepes is a senior researcher in text analytics in the Biomedical Data Science
team at IBM Research Australia. Before joining IBM, he worked as software engineer at CERN
from2000to2006,thenassoftwareengineerattheEuropeanBioinformaticsInstitute(EBI)from
2006to2010,asapost-doctoralresearcherattheUSANationalLibraryofMedicine(NIH/NLM)
from2010to2012,asaresearcheratNationalICTAustraliafrom2012to2014andasresearcher
attheCISdepartmentattheUniversityofMelbournein2014. HeobtainedhisMastersdegreein
Computer Science in 2001, a master in Intelligent systems in 2008 and his PhD degree related to
biomedicalnaturallanguagesandontologiesin2009fromUniversityJaumeI.
vii

Table of Contents
IJCNLP-2017Task1: ChineseGrammaticalErrorDiagnosis
GaoqiRAO,BaolinZhang,EndongXUNandLung-HaoLee ................................ 1
IJCNLP-2017Task2: DimensionalSentimentAnalysisforChinesePhrases
Liang-ChihYu,Lung-HaoLee,JinWangandKam-FaiWong................................9
IJCNLP-2017Task3: ReviewOpinionDiversification(RevOpiD-2017)
AnilKumarSingh,AvijitThawani,MayankPanchal,AnubhavGuptaandJulianMcAuley.....17
IJCNLP-2017Task4: CustomerFeedbackAnalysis
Chao-HongLiu,YasufumiMoriya,AlbertoPoncelasandDeclanGroves.....................26
IJCNLP-2017Task5: Multi-choiceQuestionAnsweringinExaminations
ShangminGuo,KangLiu,ShizhuHe,CaoLiu,JunZhaoandZhuoyuWei...................34
AlibabaatIJCNLP-2017Task1: EmbeddingGrammaticalFeaturesintoLSTMsforChineseGrammat-
icalErrorDiagnosisTask
Yiyang,PengjunXie,Juntao,Guangweixu,LinlinliandSiluo............................41
THU_NGN at IJCNLP-2017 Task 2: Dimensional Sentiment Analysis for Chinese Phrases with Deep
LSTM
ChuhanWu,FangzhaoWu,YongfengHuang,SixingWuandZhigangYuan..................47
IIIT-HatIJCNLP-2017Task3: ABidirectional-LSTMApproachforReviewOpinionDiversification
PruthwikMishra,PrathyushaDanda,SilpaKannegantiandSoujanyaLanka..................53
Bingo at IJCNLP-2017 Task 4: Augmenting Data using Machine Translation for Cross-linguistic Cus-
tomerFeedbackClassification
HebaElfardy,ManishaSrivastava,WeiXiao,JaredKramerandTarunAgarwal...............59
ADAPT Centre Cone Team at IJCNLP-2017 Task 5: A Similarity-Based Logistic Regression Approach
toMulti-choiceQuestionAnsweringinanExaminationsSharedTask
DariaDzendzik,AlbertoPoncelas,CarlVogelandQunLiu.................................67
YNU-HPCC at IJCNLP-2017 Task 1: Chinese Grammatical Error Diagnosis Using a Bi-directional
LSTM-CRFModel
QuanleiLiao,JinWang,JinnanYangandXuejieZhang....................................73
CVTE at IJCNLP-2017 Task 1: Character Checking System for Chinese Grammatical Error Diagnosis
Task
XianLi,PengWang,SuixueWang,GuanyuJiangandTianyuanYou ........................ 78
LDCCNLP at IJCNLP-2017 Task 2: Dimensional Sentiment Analysis for Chinese Phrases Using Ma-
chineLearning
PengZhongandJingbinWang...........................................................84
CKIPatIJCNLP-2017Task2: NeuralValence-ArousalPredictionforPhrases
Peng-HsuanLi,Wei-YunMaandHsin-YangWang.........................................89
CIAL at IJCNLP-2017 Task 2: An Ensemble Valence-Arousal Analysis System for Chinese Words and
Phrases
Zheng-WenLin,Yung-ChunChang,Chen-AnnWang,Yu-LunHsiehandWen-LianHsu ...... 95
ix
AlibabaatIJCNLP-2017Task2: ABoostedDeepSystemforDimensionalSentimentAnalysisofChinese
Phrases
XinZhou,JianWang,XuXie,ChanglongSunandLuoSi.................................100
NLPSA at IJCNLP-2017 Task 2: Imagine Scenario: Leveraging Supportive Images for Dimensional
SentimentAnalysis
szu-minchen,Zi-YuanChenandLun-WeiKu............................................105
NCYUatIJCNLP-2017Task2: DimensionalSentimentAnalysisforChinesePhrasesusingVectorRep-
resentations
Jui-FengYeh,Jian-ChengTsai,Bo-WeiWuandTai-YouKuang............................112
MainiwayAIatIJCNLP-2017Task2: EnsemblesofDeepArchitecturesforValence-ArousalPrediction
YassineBenajiba,JinSun,YongZhang,ZhiliangWengandOrBiran.......................118
NCTU-NTUT at IJCNLP-2017 Task 2: Deep Phrase Embedding using bi-LSTMs for Valence-Arousal
RatingsPredictionofChinesePhrases
Yen-HsuanLee,Han-YunYeh,Yih-RuWangandYuan-FuLiao............................124
NTOUAatIJCNLP-2017Task2: PredictingSentimentScoresofChineseWordsandPhrases
Chuan-JieLinandHao-TsungChang....................................................130
CYUTatIJCNLP-2017Task3: SystemReportforReviewOpinionDiversification
Shih-HungWu,Su-YuChangandLiang-PuChen.........................................134
JUNLPatIJCNLP-2017Task3: ARankPredictionModelforReviewOpinionDiversification
MonalisaDey,AnupamMondalandDipankarDas........................................138
All-In-1atIJCNLP-2017Task4: ShortTextClassificationwithOneModelforAllLanguages
BarbaraPlank.........................................................................143
SentiNLPatIJCNLP-2017Task4: CustomerFeedbackAnalysisUsingaBi-LSTM-CNNModel
ShuyingLin,HuoshengXie,Liang-ChihYuandK.RobertLai.............................149
IIIT-HatIJCNLP-2017Task4: CustomerFeedback Analysis usingMachineLearningandNeuralNet-
workApproaches
PrathyushaDanda,PruthwikMishra,SilpaKannegantiandSoujanyaLanka.................155
ADAPT at IJCNLP-2017 Task 4: A Multinomial Naive Bayes Classification Approach for Customer
FeedbackAnalysistask
PintuLohar,KoelDuttaChowdhury,HaithemAfli,MohammedHasanuzzamanandAndyWay161
OhioStateatIJCNLP-2017Task4: ExploringNeuralArchitecturesforMultilingualCustomerFeedback
Analysis
DushyantaDhyani.....................................................................170
YNU-HPCCatIJCNLP-2017Task4: Attention-basedBi-directionalGRUModelforCustomerFeedback
AnalysisTaskofEnglish
NanWang,JinWangandXuejieZhang..................................................174
NITMZ-JUatIJCNLP-2017Task4: CustomerFeedbackAnalysis
SomnathBanerjee,ParthaPakray,RiyankaManna,DipankarDasandAlexanderGelbukh....180
IITPatIJCNLP-2017Task4: AutoAnalysisofCustomerFeedbackusingCNNandGRUNetwork
DeepakGupta,PabitraLenka,HarsimranBedi,AsifEkbalandPushpakBhattacharyya.......184
x