SDRule-L: Managing SemanticallyRich Business Decision Processes(short paper)Yan Tang Demey and Christophe DebruyneVrije Universiteit Brussel STARLab2013-08-27 @ EC-Web 2013

2013-08-27 @ EC-Web 2013| page 1

Outline

Introduction and Problem

SDRule-L

Approach

Implementation

Limitations, Future Work and Conclusions

2013-08-27 @ EC-Web 2013| page 2

Introduction and Problem

Introduction

I Semantic Decision Support Systems (SDSS)→ requires propercapturing of business semantics to

I Fact-based Modeling (FBM) & Object Role Modeling (ORM)I Fact-oriented modeling languageI Facts vs. Fact-types

Example

Fact: ”Christophe” knows ”Yan”Fact-types: [Person] knows (/known by) [Person]

[Person] having (/ of) [Name]

I Important in FBM and ORM are ...I grounding in natural languageI verbalizationI graphical notation

... for the involvement of non-technical domain experts

2013-08-27 @ EC-Web 2013| page 4

Introduction

I However, FBM and ORM have ...I limited support for modeling and reasoning over dynamic rules

such as the sequence of events

I SDRule-L extends ORM to capture such rules

I How can we reason over such rules to support decision making?

2013-08-27 @ EC-Web 2013| page 5

SDRule-L

SDRule-L

SDRule-L extends ORM by introducing constraints, operators andcorresponding graphical notations such as:

I Sequence constraints;I Cluster constraints;I Implication- and negation constraints;I Skipper.

2013-08-27 @ EC-Web 2013| page 7

Sequence Constraints

Relations between events:

Formalization of Objectification: Before going into the details of SDRule-L con-straints and operators, it is necessary to explain objectification and the formalization.

Objectification is a way of treating a role pair as an object [3]. Graphically, it is represented as shown in Fig. 1 (O1), which is a minimum constrained fact type with an objectification. A and B are two object types, r1 and r2 are the roles that A and B can play with, and C is an objectified role pair r1/r2. The bar on r1/r2 is a uniqueness constraint, meaning that the populations of A and B with regard to r1/r2 are unique.

O1 O2

Fig. 1. Example of objectification and its equivalent owl-compatible model

The objectification from Fig. 1 (O1) can be mapped to Fig. 1 (O2) without losing any semantics. In Fig. 1 (O2), the objectified role pair r1/r2 is treated as a new object type  C.  Two  new  roles  r1’  and  r2’  are  introduced  for  the  issues  of  formalization  and  implementation. Two mandatory constraints (graphically represented as dots) are applied between C and A, and between C and B.  The  constraints  on  roles  r1’  and  r2’  ensures 1:1 population between C and A, and between C and B. The circled bar in Fig. 1 (O2) is an external uniqueness, which is a direct adaptation from the unique-ness constraint on r1/r2 from Fig. 1 (O1). We use 𝒮𝒪ℐ𝒬(𝐷) – a DL dialect – to for-malize Fig. 1 (O2) as follows:

∃𝑟1. ⊤ ⊑ 𝐴 𝑟1 ≡ 𝑟1 ∃𝑟2. ⊤ ⊑ 𝐵 𝑟2 ≡ 𝑟2′ ∃𝑟1 . ⊤ ⊑ 𝐶 ∃𝑟2 . ⊤ ⊑ 𝐶 𝐶 ≡≤ 1𝑟1 . ⊤ ⊓ ∃𝑟1 . ⊤ ⊔≤ 1𝑟2 . ⊤ ⊓ ∃𝑟2 . ⊤

In what follows, we will use objectification to objectify roles.

Sequence is a common constraint for an event. In SDRule-L, two events can have the relations as indicated in Table 1.

Table 1. SDRule-L Sequence (𝐸 : event on the right of the connector;𝐸 : event on the left)

ID Name Graphical Notation Verbalization 1 Succession 𝐸 is before 𝐸

2 Continuation 𝐸 is exactly before 𝐸

3 Overlap 𝐸 and 𝐸 overlap

4 Trigger 𝐸 triggers 𝐸

5 Terminator 𝐸 is terminated by 𝐸

6 Coincidence 𝐸 and 𝐸 are in parallel Allow us to use  𝐸 for denoting an event. An event contains two basic time indica-

tors: begin time stamp (which we indicate as  𝑇 ) and end time stamp (indicated as  𝑇 ). 𝐸 is a valid event iff  𝑇 ≥ 𝑇 .  We  use  a  dot  “.”  to  indicate  the  holder  of  an  element.  For example, for an event  𝐸 , its begin time stamp is denoted by  𝐸 . 𝑇 . Given two

A r1/r2

C

B A r1/r1'

C B r2'/r2

>>

_ _

-_

>>

>>

|=|

Events have time stamps denoting the beginning and end of an event.

2013-08-27 @ EC-Web 2013| page 8

Sequence Constraints

Example:

Device

opening

opened by

Curtain

sending

sent by

Message

>>

Device opening Curtain TRIGGERS Device sending Message.

2013-08-27 @ EC-Web 2013| page 9

Cluster Constraints

I Treating a set offact types as anobject

I Modalities:Possible→ ♦Mandatory→ �

Device

receiving

received by

Signal

opening

opened by

Curtain

sending

sent by

Message

Opening Curtain

Listen and React�

Sending Message♦

2013-08-27 @ EC-Web 2013| page 10

Implication and Negation

Device

opening

opened by

Curtain

sending

sent by

Message

Device

opening

opened by

Curtain

sending

sent by

Message

¬

¬

IF Device opening Curtain,THEN Device sending Message

IF Device NOT opening Curtain,THEN Device NOT sending Message

2013-08-27 @ EC-Web 2013| page 11

Skipper

Used to ignored rules in a particular commitment.

Device

sending

sent by

Message¶

EACH Device sending AT LEAST ONE Message (SKIPPED)

2013-08-27 @ EC-Web 2013| page 12

Approach

Approach

1) Parse SDRule-L Markup file

2) Translate fact-types into Description Logics

3) Process constraints in SDRule-L Commitment

I Static constraints translated into DL;I Dynamic rules and higher-order constraints

I Adopt approach of Tao et al. 2010I Translate constraints into queries checking existence

counterexamples

4) Notify user of problems

2013-08-27 @ EC-Web 2013| page 14

Translating fact-types into DL

Person

“Employment”

works for

employs

Company

⇓

Person

with

of

Employmentof

with

Company

2013-08-27 @ EC-Web 2013| page 15

Translating fact-types into DL

A

“C”

r1

r2

B A

r1

r1’

C

r2’

r2

B

Translation into DL:

I ∃r1.> v A and ∃r1′.> v C and r1− ≡ r1I ∃r2.> v B and ∃r2′.> v C and r2− ≡ r2I C ≡ ≤ 1r1′.> u ∃r1′.> u ≤ 1r2.> u ∃r2′.>

2013-08-27 @ EC-Web 2013| page 16

Process SDRule-L Constraints

Events – roles involved with sequence constraints – are ”cast” asbeing of the type ”event” and have a start- and end-time.

A

“C”

r1

r2

B

...

>>

Event

with

ofwith

of

Timestamp

2013-08-27 @ EC-Web 2013| page 17

Process SDRule-L Constraints

Create SPARQL ASK queries checking existence of counterexamples

Device

opening

opened by

Curtain

sending

sent by

Message

>>

Test general timestamp correctness

ASK {?x <.../#start> ?t1.?x <.../#end> ?t2.FILTER(?t1 > ?t2) }

Test sequence constraint

ASK {{ ?a <.../#sending_r1> ?rc1. ?rc1 <.../#start> ?t1.

?a <.../#opening_r1> ?rc2. ?rc2 <.../#end> ?t2.FILTER(?t1 >= ?t2) }

UNION{ OPTIONAL{ ?a <.../#sending_r1> ?rc1. }

?a <.../#opening_r1> ?rc2.FILTER(!BOUND(?rc1)) } }

2013-08-27 @ EC-Web 2013| page 18

Process SDRule-L Constraints

Create SPARQL ASK queries checking existence of counterexamples

Device

receiving

received by

Signal

opening

opened by

Curtain

sending

sent by

Message

Opening Curtain

Listen and React�

Sending Message♦

Test cluster constraint

ASK { ?a a <...#Device>.OPTIONAL {?a <.../rec_r1> ?x1. }OPTIONAL {?a <.../ope_r1> ?x2. }FILTER(!BOUND(?x1) ||

!BOUND(?x2)) }

ASK {?a <.../sen_r1> ?b.OPTIONAL { ?a <.../rec_r1> ?c. }FILTER(!BOUND(?c)) }

2013-08-27 @ EC-Web 2013| page 19

Implementation

Implementation

I SDRule-L Markup ParserI Jena and PelletI Available as a module, part of a ”suite”

2013-08-27 @ EC-Web 2013| page 21

Limitations, Future Work andConclusions

Limitations and Future Work

Ambiguous paths→ similar limitation arises in the graphical notationof external uniqueness constraints in ORM.

Sequence on unconnected

fact types Possible path 1 Possible path 2

Fig. 9. An example of a sequence applied on unconnected fact types and two possible paths of connections

5 Conclusion

In this paper, we have discussed the most recent results concerning SDRule-L, which is a semantic decision support language. In particular, we have presented constraints of sequence, cluster and implication, and operators of negation and skipper. We have shown a method of mapping dynamic rules into a combination of static rules and queries for detecting model anomalies. This method is further implemented in the SDRule-L reasoning engine.

Acknowledgements: Our use case and experimental data from this paper are taken from the SBO OSCB project.

References

1. Tang, Y., Meersman, R.: SDRule Markup Language: Towards Modeling and

Interchanging Ontological Commitments for Semantic Decision Making. In Giurca, A., Gasevic, D., Taveter, K., eds. : Handbook of Research on Emerging Rule-Based Languages and Technologies: Open Solutions and Approaches Sec. I. Chapter V. IGI Publishing, USA (2008)

2. FBM: What is Fact Based Modeling? In: Fact Based Modeling Official Website. Available at: http://www.factbasedmodeling.org/

3. Halpin, T., Morgan, T.: Information Modeling and Relational Databases 2nd edn. Morgan Kaufmann (2008)

4. Calì, A., Gottlob, G., Pieris, A.: The Return of the Entity-Relationship Model: Ontological Query Answering. In : Semantic Search over the Web. Springer Berlin Heidelberg, Germany (2012) 255-281

5. Wang, X., Chan, C.: Ontology Modeling Using UML. In Konstantas, D., Léonard, M., Pigneur, Y., Patel, S., eds. : 7th International Conference on Object Oriented Information  Systems  Conference  (OOIS’2001),  Geneva,  vol.  LNCS  2817,  pp.59-68 (2001)

6. Prater, J., Mueller, R., Beauregard, B.: An ontological approach to oracle BPM. In

A ra/rb

B

rc/reEC

ra/rc

rc/rdD

ra/rd>>

A ra/rb

B

rc/reEC

ra/rc

rc/rdD

ra/rd>>

A ra/rb

B

rc/reEC

ra/rc

rc/rdD

ra/rd>>

We currently only support sequence constraints with same ”startingpoint”. Similar limitation for Implication, and Cluster constraints.

2013-08-27 @ EC-Web 2013| page 23

Conclusions

1. Formulated the problem of modeling and reasoning aboutaspects not supported by popular fact-based modelinglanguages

2. Extended ORM with SDRule-L3. Translate SDRule-L into DL4. Check constraint violation by looking for counterexamples5. Implemented ideas, part of IDE

2013-08-27 @ EC-Web 2013| page 24