Efficiently answering top-k typicality queries on large databases
FUZZY COMPUTATIONAL ONTOLOGIES45.55.222.147/wp-content/uploads/2016/10/smp2015_lia… · ·...
Transcript of FUZZY COMPUTATIONAL ONTOLOGIES45.55.222.147/wp-content/uploads/2016/10/smp2015_lia… · ·...
FUZZY COMPUTATIONAL ONTOLOGIES
Ho-fung Leung
The Chinese University of Hong Kong
SOCIAL MEDIA
‘Social Media is a group of Internet-based applications that build on the ideological and technological foundations of Web 2.0, and that allow the creation and exchange of User Generated Content.’
-- A. M. Kaplan and M. Haenlein, 2010. Users of the world, unite! The challenges and opportunities of Social Media, Business Horizons, Volume 53, Issue 1, Pages 59-68.
"Conversationprism" by Brian Solis and JESS3 -http://www.theconversationprism.com/. Licensed under CC BY 2.5 via Commons -https://commons.wikimedia.org/wiki/File:Conversationprism.jpeg#/media/File:Conversationprism.jpeg
SOCIAL MEDIA
Baidu Tieba Facebook Google+Hong Kong Discuss
Forum
Hong Kong Golden Forum Instagram Podcast Wikipedia
Taifeng Luntan Tianya Club Twitter
and so on…
SOCIAL MEDIA
In many situations, people wish to formally represent and categorise the User Generated Content so as to analyse them.
A way to do this is by using computational ontologies (or simply ontologies in this talk).
ONTOLOGIES
‘A body of formally represented knowledge is based on a conceptualization: the , , and other entities that are assumed to exist in some area of interest and the
that hold among them.’
‘An ontology is an explicit specification of a conceptualization.’
— Thomas R. Gruber (1995)
"Ontology Bronco" by Ⓔcw.ⓣechnoid.ⓓweeb - I vectorized .. Licensed under CC BY-SA 2.5 via Wikipedia -https://en.wikipedia.org/wiki/File:Ontology_Bronco.svg#/media/File:Ontology_Bronco.svg
GRUBER, T., 1995. Toward principles for the design of ontologies used for knowledge sharing? Int. J. Hum-Comput. Int., 43(5-6), Elsevier, 907–928.
ONTOLOGIES
‘A body of formally represented knowledge is based on a conceptualization: the , , and other entities that are assumed to exist in some area of interest and the
that hold among them.’
‘An ontology is an explicit specification of a conceptualization.’
— Thomas R. Gruber (1995)
Review
Positive Review
…
…
Negative Review
Honest Negative Review
Biased Negative Review
…
GRUBER, T., 1995. Toward principles for the design of ontologies used for knowledge sharing? Int. J. Hum-Comput. Int., 43(5-6), Elsevier, 907–928.
A MODEL OF FUZZY ONTOLOGIES
Fuzziness of Concepts.
Typicality of Objects. Positive
Review
A MODEL OF FUZZY ONTOLOGIES
Fuzziness of Concepts.
Typicality of Objects.
A MODEL OF FUZZY ONTOLOGIES
Fuzziness of Concepts.
Typicality of Objects.
A MODEL OF FUZZY ONTOLOGIES
O = (C, P, I, R)
A set of Concepts
A set of Properties
A set of Instances
A set of Rules that specifies the relations
between concepts and
properties
A MODEL OF FUZZY ONTOLOGIES
O = (C, P, I, R)
A set of
A MODEL OF FUZZY ONTOLOGIES
Studies of ‘Concepts’ are found in the realm of Cognitive Psychology.
Classical View
(Common until 1970’s)A Concept is defined by a set of singly necessary and jointly sufficient features(properties).
4 sides
sides all equal in length
all angles measure
90°
A MODEL OF FUZZY ONTOLOGIES
Studies of ‘Concepts’ are found in the realm of Cognitive Psychology.
Prototype View
(Typicality as degree of ‘goodness’) A Concept is represented by an abstract prototype with all the salient properties.
abstract prototype of CAT
A MODEL OF FUZZY ONTOLOGIES
Studies of ‘Concepts’ are found in the realm of Cognitive Psychology.
Prototype View
(Typicality vs. Membership)
A MODEL OF FUZZY ONTOLOGIES
O = (C, P, I, R)
A set of
A MODEL OF FUZZY ONTOLOGIES
O = (C, P, I, R)
A set of
A MODEL OF FUZZY ONTOLOGIES
A Concept is described by a Characteristic Vector.
1 1 2 2( : , : , , : )n nc c w c w c wr
KPositve Review
A MODEL OF FUZZY ONTOLOGIES
A Concept is described by a Characteristic Vector .
An Object is described by a Property Vector .
1 1 2 2( : , : , , : )n nc c w c w c wr
KPositve Review
1 1 2 2( : , : , , : )n np p v p v p vr
KReview-1
cr
pr
A MODEL OF FUZZY ONTOLOGIES
A Concept x is a subconcept of y if and only if for all i .
cr
Review
, ,x i y ic c
cr
Positve Review
, ,i ic cPositive Review Review
ôPositive Review Review
with less, or more basic, defining properties
with more, or stricter, defining properties
A MODEL OF FUZZY ONTOLOGIES
1 1 2 2( : , : , , : )n nc c w c w c wr
KPositve Review
1 1 2 2( : , : , , : )n np p v p v p vr
KReview-1
Likeliness of an Individual in a Concept
( ) [0,1]λ Positve Review Review-1
A MODEL OF FUZZY ONTOLOGIES
1 1 2 2( : , : , , : )n nc c w c w c wr
KPositve Review
1 1 2 2( : , : , , : )n np p v p v p vr
KReview-1
Likeliness of an Individual in a Concept
( ) [0,1]λ Positve Review Review-1
0.1
0.2
0.2
0.8
0.6
A MODEL OF FUZZY ONTOLOGIES
λx(a)=1 ↔ cx,i>0 → pa,i=1 for all i.
λx(a)=0 ↔ cx,i>0 and pa,i=0 for some i.
λx(a)>λx(b) if for some j, cx,j>0 and pa,j>pb,j, and pa,i=pb,i for all
i≠j.
λy(a)>λx(a) if for some j, cx,j≥cy,j>0 and 1>pa,j>0, and cx,i=cy,i and
pa,i>0 for all i≠j.
λy(a)=λx(a) if for some j, cx,j≥cy,j>0 and pa,j=1, and cx,i=cy,i and
pa,i>0 for all i≠j.
A MODEL OF FUZZY ONTOLOGIES
λx(a)=1 ↔ cx,i>0 → pa,i=1 for all i.
λx(a)=0 ↔ cx,i>0 and pa,i=0 for some i.
λx(a)>λx(b) if for some j, cx,j>0 and pa,j>pb,j, and pa,i=pb,i for all
i≠j.
λy(a)>λx(a) if for some j, cx,j≥cy,j>0 and 1>pa,j>0, and cx,i=cy,i and
pa,i>0 for all i≠j.
λy(a)=λx(a) if for some j, cx,j≥cy,j>0 and pa,j=1, and cx,i=cy,i and
pa,i>0 for all i≠j.
A MODEL OF FUZZY ONTOLOGIES
Cat 1 1 2 2( : , : , , : )n nt t w t w t wr
K
A Prototype is described by a Prototype Vector.A Prototype vector is the weighted average of the Property Vectors of the individuals in the Concept.
A MODEL OF FUZZY ONTOLOGIES
Cat 1 1 2 2( : , : , , : )n nt t w t w t wr
K
A Prototype is described by a Prototype Vector.A Prototype vector is the weighted average of the Property Vectors of the individuals in the Concept.
Cat-1 1 1 2 2( : , : , , : )n np p v p v p vr
K
A MODEL OF FUZZY ONTOLOGIES
Cat 1 1 2 2( : , : , , : )n nt t w t w t wr
K
A Prototype is described by a Prototype Vector.A Prototype vector is the weighted average of the Property Vectors of the individuals in the Concept.
Typicality of an Individual in a Concept
Cat Cat-1( ) [0,1]τ
Cat-1 1 1 2 2( : , : , , : )n np p v p v p vr
K
A MODEL OF FUZZY ONTOLOGIES
Likeliness the extent to which an individual object is an instance of a concept.
Typicality how typical or how representative an individual is to a concept.
A MODEL OF FUZZY ONTOLOGIES
Likeliness the extent to which an individual object is an instance of a concept.
Typicality how typical or how representative an individual is to a concept.
MULTI-PROTOTYPE CONCEPTS
In some situations there are more than one prototypes for a concept.
For example, a video is influential if it has
Property 1: many people like it;
Property 2: …
Or
Property 1’: many people discuss about it;
Property 2’: …
MULTI-PROTOTYPE CONCEPTS
In some situations there are more than one prototypes for a concept.
For example, a video is influential if it has
Property 1: many people like it;
Property 2: …
Or
Property 1’: many people discuss about it;
Property 2’: …
PROPERTY PRIORITY
In some other applications, properties would have priority, in addition to importance.
For example, a good review on a camera has many properties.
Property 1: the review is about camera
Property 2: the review is written by anexpert
Property 3: …
Property 1 should have a higher priority than all other properties.
PROPERTY PRIORITY
In some other applications, properties would have priority, in addition to importance.
For example, a good review on a camera has many properties.
Property 1: the review is about camera
Property 2: the review is written by anexpert
Property 3: …
Property 1 should have a higher priority than all other properties.
CONTEXT AWARENESS
Typicality of a user generated content is often different in different contexts, as the perspectives are different.
ONTOLOGY
Ontology is …………………
A typical good article!
Not a typical good article!
A MODEL OF FUZZY ONTOLOGIES
Models of ontologies and categorisation
Traditional (crisp) categories and ontology models
Fuzzy categories and ontologies
Categories and ontology modelling with typicality of objects
A MODEL OF FUZZY ONTOLOGIES
Traditional Models
Concept = crisp set of objects
Complex concepts are constructed by operators in Description Logic.
A ≡ B⊓C
D⊑∃R.E
…
Unable to express fuzziness of concepts (‘positive review’, ‘important remarks’, etc.)
A MODEL OF FUZZY ONTOLOGIES
Fuzzy Description Logic & Ontologies
Fuzzy description logics
Fuzzy ontologies have been proposed for medical document retrieval, multilingual information retrieval, etc.
A MODEL OF FUZZY ONTOLOGIES
Formal prototypes and derived typicality values
Typicality versus likeliness
Prioritised properties
Multi-prototype concepts
Contextual effects
POTENTIAL APPLICATIONS
Ranking the user generated contents by typicality.
Presenting potentially useful results.
APPLICATION 1: WEB OF THINGS RECOMMENDATIONS
We design and develop a novel recommendation method underpinned
by the principle of object typicality to address the issues related to
Web of Things (WoT) recommendations. Since the proposed method
is more effective and efficient than other baseline methods given
sparse training data. It also significantly outperforms state-of-the-art
recommendation methods in terms of Mean Absolute Error (MAE).
The business implication of our research is that the proposed
recommendation method can enhance the situation awareness of
Web of Things (WoT) applications which facilitate the reuse of
enterprise resources and the interoperability among enterprises.
CAI, Y., LAU, R. Y. K., LIAO, S. S. Y., LI, C. P., LEUNG, H. F. and MA, L. C. K. 2014. Object Typicality for Effective Web ofThings Recommendations. Decis. Support Syst., 63. Elsevier, 52-63.
APPLICATION 2: TYPICALITY QUERY ANSWERING
We apply the idea of typicality analysis from cognitive psychology to query answering, and propose a novel method to answer typicality queries effectively based on theories in cognitive psychology. The proposed method adopts multi-prototype concept modelling and basic level category detection. By a systematic empirical evaluation using real data sets, we verify the accuracy and the effectiveness of our method on answering typicality queries.
CAI, Y., ZHAO, H. K., HAN, H., LAU, R. Y. K., LEUNG, H. F. and MIN, H. Q., 2012. Answering Typicality Query based on Automatically Prototype Construction. In: Proceedings of the 2012 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT), Macau, China, 4-7 December 2012. 362 - 366.
APPLICATION 3: TYPICALITY-BASED CF RECOMMENDATION
We propose a novel typicality-based collaborative filtering
recommendation method named TyCo, which finds ‘neighbours’ of
users based on user typicality degrees in user groups. TyCo
outperforms many CF recommendation methods on recommendation
accuracy (in terms of MAE), especially with sparse training data. It has
lower time cost than other CF methods. Further, it can obtain more
accurate predictions with less number of big-error predictions.
CAI, Y., LEUNG, H. F., LI, Q., MIN, H. Q., HAN, H., TANG, J. and LI, J. Z., 2014. Typicality-based Collaborative Filtering Recommendation. IEEE Trans. Knowl. Data Eng., 26(3). IEEE Computer Society, 766-779.
CAI, Y., LEUNG, H. F., LI, Q., TANG, J. and LI, J. Z., 2010. TyCo: Towards Typicality-based Collaborative Filtering Recommendation. In: Proceedings of the 22nd IEEE International Conference on Tools with Artificial Intelligence, Volume 2, Arras, France, 27-29 October 2010. Los Alamitos: IEEE Computer Society, 97-104.
CAI, Y., LEUNG, H. F., LI, Q., TANG, J. and LI, J. Z., 2010. Recommendation Based On Object Typicality. In: Proceedings of the 19th ACM International Conference on Information and Knowledge Management, Toronto, Canada, 26-30 October 2010.
We apply the idea of typicality analysis to database query answering,
and study the novel problem of answering top-k typicality queries.
Three types of top-k typicality queries are formulated. We develop a
series of approximation methods for various situations. Typicality
queries can be answered efficiently with quality guarantees. An
extensive performance study using two real data sets and a series of
synthetic data sets clearly shows that top-k typicality queries are
meaningful and our methods are practical.
APPLICATION 4: TOP-k TYPICALITY QUERIES
HUA, M., PEI, J., FU, A. W. C., LIN, X. M. and LEUNG, H. F., 2009. Top-k Typicality Queries and Efficient Query Answering Methods on Large Databases. VLDB J., 18(3), Berlin: Springer, 809-835.
HUA, M., PEI, J., FU, A. W. C., LIN, X. M. and LEUNG, H. F., 2007. Efficiently Answering Top-k Typicality Queries on Large Databases. In: Proceedings of the 33rd International Conference on Very Large Data Bases, Vienna, Austria, 23-28 September 2007. VLDB Endowment, 890-901.
RELATED PUBLICATIONS
Publication supported byNFAPST
国家科学技术学术著作出版基金
CAI, Y., AU YEUNG, C. M. and LEUNG, H. F., 2012. Fuzzy Computational Ontologies in Contexts. Higher Education Press / Springer.
RELATED PUBLICATIONS
APPLICATIONS
CAI, Y., LAU, R. Y. K., LIAO, S. S. Y., LI, C. P., LEUNG, H. F. and MA, L. C. K. 2014. Object Typicality for Effective Web of Things Recommendations. Decis. Support Syst., 63. Elsevier, 52-63.
CAI, Y., LEUNG, H. F., LI, Q., MIN, H. Q., HAN, H., TANG, J. and LI, J. Z., 2014. Typicality-based Collaborative Filtering Recommendation. IEEE Trans. Knowl. Data Eng., 26(3). IEEE Computer Society, 766-779.
CAI, Y., ZHAO, H. K., HAN, H., LAU, R. Y. K., LEUNG, H. F. and MIN, H. Q., 2012. Answering Typicality Query based on Automatically Prototype Construction. In: Proceedings of the 2012 IEEE/WIC/ACM International Conference on Web Intelligence and Intelligent Agent Technology (WI-IAT), Macau, China, 4-7 December 2012. 362 - 366.
CAI, Y., LEUNG, H. F., LI, Q., TANG, J. and LI, J. Z., 2010. TyCo: Towards Typicality-based Collaborative Filtering Recommendation. In: Proceedings of the 22nd IEEE International Conference on Tools with Artificial Intelligence, Volume 2, Arras, France, 27-29 October 2010. Los Alamitos: IEEE Computer Society, 97-104.
CAI, Y., LEUNG, H. F., LI, Q., TANG, J. and LI, J. Z., 2010. Recommendation Based On Object Typicality. In: Proceedings of the 19th ACM International Conference on Information and Knowledge Management, Toronto, Canada, 26-30 October 2010.
RELATED PUBLICATIONS
APPLICATIONS (Continued)
HUA, M., PEI, J., FU, A. W. C., LIN, X. M. and LEUNG, H. F., 2009. Top-k Typicality Queries and Efficient Query Answering Methods on Large Databases. VLDB J., 18(3), Berlin: Springer, 809-835.
HUA, M., PEI, J., FU, A. W. C., LIN, X. M. and LEUNG, H. F., 2007. Efficiently Answering Top-kTypicality Queries on Large Databases. In: Proceedings of the 33rd International Conference on Very Large Data Bases, Vienna, Austria, 23-28 September 2007. VLDB Endowment, 890-901.
RELATED PUBLICATIONS
THEORY
CAI, Y. and LEUNG, H. F., 2011. Formalizing Object Membership in Fuzzy Ontology with Property Importance and Property Priority. In: Proceedings of the 2011 IEEE International Conference on Fuzzy Systems, Taipei, China, 27-30 June, 2011. IEEE, 1719-1726.
CAI, Y. and LEUNG, H. F., 2010. A Fuzzy Description Logic with Automatic Object Membership Measurement. In: Y. X. BI and M.-A. WILLIAMS, Eds., Knowledge Science, Engineering and Management, 4th International Conference, KSEM 2010, Lecture Notes in Artificial Intelligence, Volume 6291, Belfast, Northern Ireland, UK, 1-3 September 2010. Springer, 76-87.
CAI, Y. and LEUNG, H. F., 2008. Formalizing Object Typicality in Context-aware Ontology. In:Proceedings of the 20th IEEE International Conference on Tools with Artificial Intelligence, Dayton, Ohio, U.S.A., 3-5 November 2008. Los Alamitos: IEEE Computer Society, 233-240.
CAI, Y. and LEUNG, H. F., 2008. A Formal Model of Fuzzy Ontology with Property Hierarchy and Object Membership. In: Q. LI, S. SPACCAPIETRA, E. YU and A. OLIVÉ, Eds., Conceptual Modeling - ER 2008, Lecture Notes in Computer Science, Volume 5231, Barcelona, Catalonia, Spain, 20-23 October 2008. Springer Berlin / Heidelberg, 69-82.
CAI, Y., LEUNG, H. F. and FU, A. W. C., 2008. Multi-Prototype Concept and Object Typicality in Ontology. In: Proceedings of the 21st International Florida Artificial Intelligence Research Society Conference, Coconut Grove, Florida, USA, 15-17 May, 2008. Menlo Park, Calif.: AAAI Press.
RELATED PUBLICATIONS
THEORY (Continued)
AU YEUNG, C. M. and LEUNG, H. F., 2010. A Formal Model of Ontology for Handling Fuzzy Membership and Typicality of Instances. Comput. J., 53(3). Oxford: Oxford University Press, 316-341.
AU YEUNG, C. M. and LEUNG, H. F., 2006. Ontology with Likeliness and Typicality of Objects in Concepts. In: D. W. EMBLEY, A. OLIVÉ and S. RAM, eds., Conceptual Modeling - ER 2006, Lecture Notes in Computer Science, Volume 4215, Tucson, Arizona, USA, 6-9 November 2006. Springer Berlin / Heidelberg, 98-111.
AU YEUNG, C. M. and LEUNG, H. F., 2006. Formalizing Typicality of Objects and Context-sensitivity in Ontologies. In: P. STONE and G. WEISS, eds., Proceedings of the 5th International Joint Conference on Autonomous Agents and Multiagent Systems, Hakodate, Japan, 8-12 May 2006. New York: ACM Press, 946-948.
AU YEUNG, C. M. and LEUNG, H. F., 2006. Formalizing Concepts in Description Logics Using a Cognitive Approach. In: D. LUKOSE and Z. SHI, eds., Proceedings of the 8th Pacific Rim International Workshop on Multi-Agents, PRIMA 2005, Kuala Lumpur, Malaysia, 26-28 September 2005.
RESEARCHERS
Yi Cai
ProfessorSouth China University
of Technology
Ho-fung Leung
ProfessorThe Chinese University
of Hong Kong
Ching-man Au Yeung
Co-founder & DirectorAxon Labs Limited
ACKNOWLEDGEMENT
I would like to thank the organiser of the Fourth National Conference of Social Media Processing for the kind support.