“towards building a cognitive system to fight for national college admission challenge”
TRANSCRIPT
Towards Building a Cognitive System toFight for National College Admission Challenge
Yansong FengJoint work with
Kun Xu, Songfang Huang, Dongyan Zhao
Peking UniversityIBM China Research Lab
December 1, 2016
Feng et al. (PKU) Question Answering December 1, 2016 1 / 25
Pass the Exam: A New AI Challenge
The Todai Robot ProjectNational Institute of Informatics andcollaboratorsNational Center Test for UniversityAdmissions ( 2016)Entrance Exam of University ofTokyo (2021)
Japanese, Social Science, Math,Physics
The Project Aristo and EuclidThe Allen Institute for ArtificialIntelligenceElementary School:
Science andMath
High School:
Geometry
Feng et al. (PKU) Question Answering December 1, 2016 2 / 25
Pass the Exam: A New AI Challenge
The Todai Robot ProjectNational Institute of Informatics andcollaboratorsNational Center Test for UniversityAdmissions ( 2016)Entrance Exam of University ofTokyo (2021)Japanese, Social Science, Math,Physics
The Project Aristo and EuclidThe Allen Institute for ArtificialIntelligenceElementary School:
Science andMath
High School:
Geometry
Feng et al. (PKU) Question Answering December 1, 2016 2 / 25
Pass the Exam: A New AI Challenge
The Todai Robot ProjectNational Institute of Informatics andcollaboratorsNational Center Test for UniversityAdmissions ( 2016)Entrance Exam of University ofTokyo (2021)Japanese, Social Science, Math,Physics
The Project Aristo and EuclidThe Allen Institute for ArtificialIntelligenceElementary School:
Science andMath
High School:
Geometry
Feng et al. (PKU) Question Answering December 1, 2016 2 / 25
Pass the Exam: A New AI Challenge
The Todai Robot ProjectNational Institute of Informatics andcollaboratorsNational Center Test for UniversityAdmissions ( 2016)Entrance Exam of University ofTokyo (2021)Japanese, Social Science, Math,Physics
The Project Aristo and EuclidThe Allen Institute for ArtificialIntelligenceElementary School: Science andMathHigh School: Geometry
Feng et al. (PKU) Question Answering December 1, 2016 2 / 25
The GaoKao Challenge
Gaokao in ChinaNational College Entrance ExaminationChinese, Math, English, History, Geography, Politics, Physics,Chemistry, Biologyover 9,400,000 students in 2016
The China Gaokao ChallengePrompt research in Artificial IntelligenceTeam: national research institutes, universities and companiesReal National College Entrance ExaminationsChinese, History, Math,
Feng et al. (PKU) Question Answering December 1, 2016 3 / 25
The GaoKao Challenge
Gaokao in ChinaNational College Entrance ExaminationChinese, Math, English, History, Geography, Politics, Physics,Chemistry, Biologyover 9,400,000 students in 2016
The China Gaokao ChallengePrompt research in Artificial IntelligenceTeam: national research institutes, universities and companiesReal National College Entrance ExaminationsChinese, History, Math, Geography
Feng et al. (PKU) Question Answering December 1, 2016 3 / 25
The GaoKao Challenge
Gaokao in ChinaNational College Entrance ExaminationChinese, Math, English, History, Geography, Politics, Physics,Chemistry, Biologyover 9,400,000 students in 2016
The China Gaokao ChallengePrompt research in Artificial IntelligenceTeam: national research institutes, universities and companiesReal National College Entrance ExaminationsChinese, History, Math, Geography
Feng et al. (PKU) Question Answering December 1, 2016 3 / 25
Is that Difficult?
Which is the correct ranking of provinces according to theiraverage altitutes, from highest to lowest?
Xiang, Liao, Ning
→ Hunan, Liaoning, Ningxia
Tai, Lu, Su
→Taiwan, Shandong, Jiangsu
Qing, Yue, Jin
→Qinghua, Guangdong, Shanxi
Gui, Gan, Yu
→ Guangxi, Gansu, Henan
1 Knowledge:short names of provincesaverage altitudes of provinces
2 Reasoningrelative comparisons of provinces’ altituderanking
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 4 / 25
Is that Difficult?
Which is the correct ranking of provinces according to theiraverage altitutes, from highest to lowest?
Xiang, Liao, Ning → Hunan, Liaoning, NingxiaTai, Lu, Su →Taiwan, Shandong, JiangsuQing, Yue, Jin →Qinghua, Guangdong, ShanxiGui, Gan, Yu → Guangxi, Gansu, Henan
1 Knowledge:short names of provincesaverage altitudes of provinces
2 Reasoningrelative comparisons of provinces’ altituderanking
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 4 / 25
Is that Difficult?
Which is the correct ranking of provinces according to theiraverage altitutes, from highest to lowest?
Xiang, Liao, Ning → Hunan, Liaoning, NingxiaTai, Lu, Su →Taiwan, Shandong, JiangsuQing, Yue, Jin →Qinghua, Guangdong, ShanxiGui, Gan, Yu → Guangxi, Gansu, Henan
1 Knowledge:short names of provincesaverage altitudes of provinces
2 Reasoningrelative comparisons of provinces’ altituderanking
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 4 / 25
Is that Difficult?
Which is the correct ranking of provinces according to theiraverage altitutes, from highest to lowest?
Xiang, Liao, Ning → Hunan, Liaoning, NingxiaTai, Lu, Su →Taiwan, Shandong, JiangsuQing, Yue, Jin →Qinghua, Guangdong, ShanxiGui, Gan, Yu → Guangxi, Gansu, Henan
1 Knowledge:short names of provincesaverage altitudes of provinces
2 Reasoningrelative comparisons of provinces’ altituderanking
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 4 / 25
What about this one?
Missouri River Valley is an important agricultural area of theUnited States. Aerial figure 1 shows the winter of the MissouriRiver, where the white part is snow.1. Why is the farmland on the peninsula shaped as circular?
their watering approachrugged terraintheir farming approachlack of farmland
Feng et al. (PKU) Question Answering December 1, 2016 5 / 25
What about this one?
2. Could you have a guess what is the main crop in this area?Winter wheatCornRicePotato
1 Knowledge:agricultureclimatelongitude and latituderead maps and images
2 Reasoningrelationship among those factorsfind analogiescommon sense knowledge
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 6 / 25
What about this one?
2. Could you have a guess what is the main crop in this area?Winter wheatCornRicePotato
1 Knowledge:agricultureclimatelongitude and latituderead maps and images
2 Reasoningrelationship among those factorsfind analogiescommon sense knowledge
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 6 / 25
What about this one?
2. Could you have a guess what is the main crop in this area?Winter wheatCornRicePotato
1 Knowledge:agricultureclimatelongitude and latituderead maps and images
2 Reasoningrelationship among those factorsfind analogiescommon sense knowledge
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 6 / 25
What about this one?
2. Could you have a guess what is the main crop in this area?Winter wheatCornRicePotato
1 Knowledge:agricultureclimatelongitude and latituderead maps and images
2 Reasoningrelationship among those factorsfind analogiescommon sense knowledge
Not very challenging?
Feng et al. (PKU) Question Answering December 1, 2016 6 / 25
What We Need?
1 Solid knowledge:every aspects about the Syllabus
2 Math3 Reasoning:
logical inferenceuse common sense knowledgetextual entailment
A practical starting point: Answering Factoid Questions withKnowledge
Information Retrieval Based Question AnsweringAnswering Factoid Questions with Structured Knowledge BaseAnswering Factoid Questions with both Structured andUnstructured KBs
Feng et al. (PKU) Question Answering December 1, 2016 7 / 25
What We Need?
1 Solid knowledge:every aspects about the Syllabus
2 Math3 Reasoning:
logical inferenceuse common sense knowledgetextual entailment
A practical starting point: Answering Factoid Questions withKnowledge
Information Retrieval Based Question AnsweringAnswering Factoid Questions with Structured Knowledge BaseAnswering Factoid Questions with both Structured andUnstructured KBs
Feng et al. (PKU) Question Answering December 1, 2016 7 / 25
What We Need?
1 Solid knowledge:every aspects about the Syllabus
2 Math3 Reasoning:
logical inferenceuse common sense knowledgetextual entailment
A practical starting point: Answering Factoid Questions withKnowledge
Information Retrieval Based Question AnsweringAnswering Factoid Questions with Structured Knowledge BaseAnswering Factoid Questions with both Structured andUnstructured KBs
Feng et al. (PKU) Question Answering December 1, 2016 7 / 25
What We Need?
1 Solid knowledge:every aspects about the Syllabus (Knowledge Bases)
2 Math3 Reasoning:
logical inferenceuse common sense knowledge (a little bit)textual entailment
A practical starting point: Answering Factoid Questions withKnowledge
Information Retrieval Based Question AnsweringAnswering Factoid Questions with Structured Knowledge BaseAnswering Factoid Questions with both Structured andUnstructured KBs
Feng et al. (PKU) Question Answering December 1, 2016 7 / 25
The Task
What else did the director of the movie Interstellar direct ?
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Inception, The Dark Knight Rises
The Dark KnightBatman Begins
…..
Feng et al. (PKU) Question Answering December 1, 2016 8 / 25
The Task
What else did the director of the movie Interstellar direct ?
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Inception, The Dark Knight Rises
The Dark KnightBatman Begins
…..
���������� ��
� ���������
Feng et al. (PKU) Question Answering December 1, 2016 8 / 25
Question Answering over Structured Knowledge Bases
GoalAnswer Natural Language Questions against Structured KnowledgeBases
Feng et al. (PKU) Question Answering December 1, 2016 9 / 25
Related Work
Information Retrieval CommunityNatural Language Processing Community
Semantic Parsing BasedPCCG style: (Zettlemoyer and Collins, 2005, Cai and Yates, 2013;Kwiatkowski et al. 2013, Reddy et al., 2014)Syntactic parsing style: (Clarke et al., 2010, Liang et al., 2011,Berant et al. 2013, Berant and Liang, 2014, Xu et al., 2014)
Information Extraction Based(Yao and van Durme 2014, Bao et al., 2014, Yih et al., 2015, Donget al., 2015)Deep Learning, End2End style(Bordes et al., 2014a, 2014b, Yang et al., 2014, Bordes et al.,2015, Zhang et al., 2016 )
Feng et al. (PKU) Question Answering December 1, 2016 10 / 25
Related Work
Information Retrieval CommunityNatural Language Processing Community
Semantic Parsing BasedPCCG style: (Zettlemoyer and Collins, 2005, Cai and Yates, 2013;Kwiatkowski et al. 2013, Reddy et al., 2014)Syntactic parsing style: (Clarke et al., 2010, Liang et al., 2011,Berant et al. 2013, Berant and Liang, 2014, Xu et al., 2014)
Information Extraction Based(Yao and van Durme 2014, Bao et al., 2014, Yih et al., 2015, Donget al., 2015)Deep Learning, End2End style(Bordes et al., 2014a, 2014b, Yang et al., 2014, Bordes et al.,2015, Zhang et al., 2016 )
Feng et al. (PKU) Question Answering December 1, 2016 10 / 25
Semantic Parsing Based Methods
Challenges:1 Convert questions into proper meaning representations2 Ground the meaning representation into a database query
Previously:1 Search space is huge2 Difficult to adapt to other KBs
Feng et al. (PKU) Question Answering December 1, 2016 11 / 25
Semantic Parsing Based Methods
Challenges:1 Convert questions into proper meaning representations2 Ground the meaning representation into a database query
Previously:1 Search space is huge2 Difficult to adapt to other KBs
Feng et al. (PKU) Question Answering December 1, 2016 11 / 25
Motivation
1 Meaning representation should be KB-independent
[what] did �the director of] [the movie] [Interstellar]else �movies] [direct]
2 Separation of meaning representation and instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
ns:Interstellar dbo:type dbo:filmdbp:director ?x[ ]
select ?y
?y
?y
ns:Interstellar
dbp:director ?x
dbo:filmdbo:type
Feng et al. (PKU) Question Answering December 1, 2016 12 / 25
Motivation
1 Meaning representation should be KB-independent
[what] did �the director of] [the movie] [Interstellar]else �movies] [direct]
2 Separation of meaning representation and instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
ns:Interstellar dbo:type dbo:filmdbp:director ?x[ ]
select ?y
?y
?y
ns:Interstellar
dbp:director ?x
dbo:filmdbo:type
Feng et al. (PKU) Question Answering December 1, 2016 12 / 25
Framework
what else movies did the director of the movie Interstellar direct
Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Feng et al. (PKU) Question Answering December 1, 2016 13 / 25
Framework
what else movies did the director of the movie Interstellar direct
Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Feng et al. (PKU) Question Answering December 1, 2016 13 / 25
Framework
what else movies did the director of the movie Interstellar direct
Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Feng et al. (PKU) Question Answering December 1, 2016 13 / 25
Framework
what else movies did the director of the movie Interstellar direct
Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Feng et al. (PKU) Question Answering December 1, 2016 13 / 25
Framework
what else movies did the director of the movie Interstellar direct
Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Instantiation
fb:m.0fkf28 fb:object.type fb:film.film
fb:m.0fkf28 fb:film.film.directed_by ?x[ ]select ?y
?x fb:film.director.fim ?y?y fb:object.type fb:film.film
Feng et al. (PKU) Question Answering December 1, 2016 13 / 25
Phrase Dependency Graph
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
NodeA phrase with a semantic label l ∈ {entity, category, variable, relation}
EdgeA predicate-argument dependency between phrasesunary predicatebinary predicate
Feng et al. (PKU) Question Answering December 1, 2016 14 / 25
Structure Prediction
Input: a natural language questionOutput: a phrase dependency graph
A pipeline framework to predict the structure1 Phrase Detection
what did the director of the movie Interstellar
variable relation category entity
else movies direct
category relation
2 Phrase Dependency Parsing
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
Feng et al. (PKU) Question Answering December 1, 2016 15 / 25
Instantiation
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
1 Converting Phrase Dependency Graph into Structured Queries2 Instantiating Structured Query against KB
Feng et al. (PKU) Question Answering December 1, 2016 16 / 25
Applying Rules
[what] did �the director of] [the movie] [Interstellar]variable relation category entity
else �movies] [direct]category relation
variable category variablecategory
rule#1 rule#1 variablerelationrelationentityrule#8
?y type movies type ?x?x ?y
moviesInterstellar Interstellardirect
the director of
[ ]select ?y
type moviesInterstellar
?xInterstellar the director of?x ?ydirect
?y type movies
Feng et al. (PKU) Question Answering December 1, 2016 17 / 25
Probabilistic Model
[ ]select ?y
type moviesInterstellar
?xInterstellar the director of?x ?ydirect
?y type movies
Qind Qd
ns:Interstellar dbo:type dbo:filmdbp:director ?x[ ]
select ?y
?y
?y
ns:Interstellar
dbp:director ?x
dbo:filmdbo:type
Q∗d = arg max P(Qd |Qind)
P(Qd |Qind) =n∏
i=1
P(sdi |sindi )P(odi |oindi )P(pdi |pindi )
Feng et al. (PKU) Question Answering December 1, 2016 18 / 25
P(sdi |sindi )P(odi |oindi )Freebase Search API ⇒ wikipedia ID ⇒ DBpedia Entity
P(pd |pind)We construct the co-occurrence matrix from the patty relationphrase dataset which includes 1,631,530 relation phrases
Feng et al. (PKU) Question Answering December 1, 2016 19 / 25
Results on QALDs
Question Answering over Linked Data
Processed Right Partial Recall Precision F-12014 40 34 6 0.71 0.72 0.722015 42 26 7 0.72 0.74 0.73
Nice for longer/complex sentencesEfficient: around 0.33 sec per sentence
Consistent performances0.76 of F-1 on Free9170.41 of F-1 on WebQuestions
Feng et al. (PKU) Question Answering December 1, 2016 20 / 25
Results on QALDs
Question Answering over Linked Data
Processed Right Partial Recall Precision F-12014 40 34 6 0.71 0.72 0.722015 42 26 7 0.72 0.74 0.73
First Place in CLEF QALD 4 and 5
Nice for longer/complex sentencesEfficient: around 0.33 sec per sentence
Consistent performances0.76 of F-1 on Free9170.41 of F-1 on WebQuestions
Feng et al. (PKU) Question Answering December 1, 2016 20 / 25
Results on QALDs
Question Answering over Linked Data
Processed Right Partial Recall Precision F-12014 40 34 6 0.71 0.72 0.722015 42 26 7 0.72 0.74 0.73
Nice for longer/complex sentencesEfficient: around 0.33 sec per sentence
Consistent performances0.76 of F-1 on Free9170.41 of F-1 on WebQuestions
Feng et al. (PKU) Question Answering December 1, 2016 20 / 25
Results on QALDs
Question Answering over Linked Data
Processed Right Partial Recall Precision F-12014 40 34 6 0.71 0.72 0.722015 42 26 7 0.72 0.74 0.73
Nice for longer/complex sentencesEfficient: around 0.33 sec per sentenceConsistent performances0.76 of F-1 on Free9170.41 of F-1 on WebQuestions
Feng et al. (PKU) Question Answering December 1, 2016 20 / 25
Further Improvement
Semantic interpretation for superlatives
Nile is the longest river in the world.
Keys:the target: Nilethe comparison set: all rivers in the worldthe comparison dimension: the length of a river
→/geography/river/lengththe ranking order: descending
Feng et al. (PKU) Question Answering December 1, 2016 21 / 25
A Little More Extraction
For simple sentences:
Who does michael keaton play in cars
Who
michael keaton cars
Michael Keaton
The Merry GentlemanPenthouse North
cvt1Cars
cvt2
Chick Hicks
film
starring
starring
filmdirect direct by
ctv3spouse_s
Caroline McWilliams
spousespouse_s
spouse6/5/1982 from
1/29/1990
to
Marriage
type of unioncharacter
portrayer
film
character
starring
charactercharacter
Question:
Star Graph:
Freebase Graph:
cvt3
film
actor
Leona Elizabeth Loftus George A. Douglas
parentschild
Feng et al. (PKU) Question Answering December 1, 2016 22 / 25
A Little More Extraction
For simple sentences:
Join Entitly Linking and Relation Extractiongives 0.49 (+0.06) F1 on WebQuestions.
Feng et al. (PKU) Question Answering December 1, 2016 22 / 25
With Hybrid Knowledge Base Resources
a little bit complex...Where should a visitor see in Germany ?
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
a little bit complex...Where should a visitor see in Germany ?What is the most popular crop during 1900s in USA ?
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
a little bit complex...Where should a visitor see in Germany ?What is the most popular crop during 1900s in USA ?Who did Shaq first play for ?
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
Either subjective, or hard to map against existing Structured KBs
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
Either subjective, or hard to map against existing Structured KBsUsing both structured knowledge bases and texts, e.g.,Wikipedia or existing community QA archives.
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
who did shaq first play for
KB-QA
Entity Linking Relation Extraction
Joint Inference
shaq: m.012xdfshaq: m.05n7bpshaq: m.06_ttvh
sports.pro_athlete.teams..sports.sports_team_roster.teambasketball.player.statistics..basketball.player_stats.team
……
m.012xdf sports.pro_athlete.teams..sports.sports_team_roster.team
Los Angeles Lakers,Boston Celtics,Orlando Magic,
Miami Heat
Freebase
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
Answer Refinement
Los Angeles Lakers,Boston Celtics,Orlando Magic,
Miami Heat
Freebase
Shaquille O'Neal
O'Neal signedas a free agent with the Los Angeles Lakers
Shaquille O'Neal
O'Neal played for the Boston Celtics in the 2010-11 season before retiring
Shaquille O'Neal
O'Neal was drafted
in the 1992 NBA draftby the Orlando Magic with the first overall pick
Los Angeles Lakers Boston Celtics Orlando Magic
O’Neal was drafted by the OrlandoMagic with the first overall pick in
the 1992 NBA draft
O’Neal played for the Boston Celtics in the 2010-11 season before retiring
O’Neal signed as a free agent with the Los Angeles Lakers
Refinement Model
+- -
Orlando Magic
Wikipedia Dump
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
With Hybrid Knowledge Base Resources
TextualRelation Extraction
Triple SolverTriple Solver
who is the front man of the band that wrote Coffee & TV
Question Decomposition< ans, is the front man of, var1 >
< var1 , is a , band >< var1 , wrote , Coffee & TV >
< var1 , wrote, Coffee & TV >
Triple Solver
Entity Linking
Multi-Channel Neural NetworkParaphrase Model
Wikipedia Dump
TextualKB
KB-basedRelation Extraction
DBpediaFreebase
DBpedia Lookup
Coffee & TV
Bitter_Coffee_(Iranian_video_series)
Irish_Coffee_(TV_series)Coffee & TV
influencedByassociatedMusicalArtist
associatedBandwriter
front man ofis written by
lead vocalist of
Open Information Extractor
wrote
is the front man of
Joint Inference
Damon Albarn
Feng et al. (PKU) Question Answering December 1, 2016 23 / 25
Conclusion
At this point...still a lot to do for GaoKaobut, a flexible QA frameworkWith multiple resources, e.g., structured knowledge bases,Wikipedia, text books, exercises, even news papers, etc.Our collaborations with IBM China Research Lab
contribute to the Watson Competitionscontribute to a Multi-Modal QA system (with Vision China)
The KeysFrom natural languages to knowledge basesInference over structured knowledgeAnswer with common-sense knowledgeUnderstand various images, tables, figures, diagrams...
Feng et al. (PKU) Question Answering December 1, 2016 24 / 25
Conclusion
At this point...still a lot to do for GaoKaobut, a flexible QA frameworkWith multiple resources, e.g., structured knowledge bases,Wikipedia, text books, exercises, even news papers, etc.Our collaborations with IBM China Research Lab
contribute to the Watson Competitionscontribute to a Multi-Modal QA system (with Vision China)
The KeysFrom natural languages to knowledge bases → on the wayInference over structured knowledgeAnswer with common-sense knowledgeUnderstand various images, tables, figures, diagrams...
Feng et al. (PKU) Question Answering December 1, 2016 24 / 25
Conclusion
At this point...still a lot to do for GaoKaobut, a flexible QA frameworkWith multiple resources, e.g., structured knowledge bases,Wikipedia, text books, exercises, even news papers, etc.Our collaborations with IBM China Research Lab
contribute to the Watson Competitionscontribute to a Multi-Modal QA system (with Vision China)
The KeysFrom natural languages to knowledge bases → on the wayInference over structured knowledge → challengingAnswer with common-sense knowledgeUnderstand various images, tables, figures, diagrams...
Feng et al. (PKU) Question Answering December 1, 2016 24 / 25
Conclusion
At this point...still a lot to do for GaoKaobut, a flexible QA frameworkWith multiple resources, e.g., structured knowledge bases,Wikipedia, text books, exercises, even news papers, etc.Our collaborations with IBM China Research Lab
contribute to the Watson Competitionscontribute to a Multi-Modal QA system (with Vision China)
The KeysFrom natural languages to knowledge bases → on the wayInference over structured knowledge → challengingAnswer with common-sense knowledge → still missingUnderstand various images, tables, figures, diagrams... → stillmissing
Feng et al. (PKU) Question Answering December 1, 2016 24 / 25