A Pattern-based Approach for Semantic Retrieval of Information

Resources in Enterprises

Application Within STMicroelectronics

Sara Bouzid

, Corine Cauvet

, Claudia Frydman

and Jacques Pinaton

Laboratory for Information Sciences and Systems (LSIS), Aix-Marseille University, Marseilles, France

STMicroelectronics, Rousset, France

Keywords: Resource Retrieval, Semantic Search, Alignment Pattern, Goal-Oriented Mechanisms, Business Needs.

Abstract: Information-resource retrieval in enterprises is becoming a major concern nowadays because of the im-

portance of business information in supporting the satisfaction of business objectives. To enhance resource

retrieval in enterprises, this paper argues that it is necessary to coherently include the user need in the search

process, in particular when this need is business-context dependent. A pattern-based approach is proposed

for this purpose. The approach captures the business needs in a company using goal-oriented mechanisms

and integrates them in a keyword search using alignment patterns. These patterns are used to both guide the

search process and semantically fill the gap between the low-level description of information resources and

the high-level needs of business actors. The approach has been applied for resource retrieval in the context

of the manufacturing-process control within the STMicroelectronics Company.

1 INTRODUCTION

With the increase of information resources in enter-

prises, retrieving business information has become a

daily challenge for the actors that have informational

needs for achieving their business tasks.

The com-

plexity of the information systems in enterprises and

the permanent evolution of the business needs make

difficult the retrieval of the heterogeneous infor-

mation resources that we can find in a company.

Also, the growing use of commercial software plat-

forms in enterprises for the extraction of data and the

creation of business information does not facilitate

the access to the resources produced, because exist-

ing commercial platforms do not often support the

semantic management of information.

Indeed, in many recent research works, semantic

techniques proved their ability to enhance infor-

mation retrieval on the web and document retrieval

(Chu, 2003). Semantic techniques rely on the use of

ontologies and semantic web technologies, most of

the time with a keyword search. Most of these tech-

niques focus on the “what” of the user need and

rarely on the “why” aspect. However, the users’

needs in a company may be difficult to express with

only keywords because such needs are generally

complex and business-context dependent.

Basically, business information resources in en-

terprises may be related to the design process of

products, the process control of a manufacturing

activity, the service delivery of a company or either

any information devoted to marketing purposes. To

date, most semantic retrieval approaches are devoted

to information retrieval on the web (Nunes, 2006).

In addition, existing approaches do not consider the

complex business problem of the user in the search

process. The main used techniques to address a user

need are commonly solution oriented. As a conse-

quence, there is a lack of approaches suitable for the

semantic retrieval of information resources used for

business purposes in enterprises. The aim of this

work is to enhance resource retrieval in enterprises

considering, on one hand, the complex business

needs of the users, and on the other hand, the dis-

tance that exists between a business need and the

available information resources in a company.

We propose in this paper a pattern-based ap-

proach aiming at enhancing resource retrieval in

enterprises by filling the gap between the low-level

description of heterogeneous information resources

and the high level needs of business actors. This

approach relies on goal-oriented mechanisms carried

by “alignment patterns” that we integrated in a key-

word search.

193

Bouzid S., Cauvet C., Frydman C. and Pinaton J..

A Pattern-based Approach for Semantic Retrieval of Information Resources in Enterprises - Application Within STMicroelectronics.

DOI: 10.5220/0004886101930200

In Proceedings of the 16th International Conference on Enterprise Information Systems (ICEIS-2014), pages 193-200

ISBN: 978-989-758-027-7

 2014 SCITEPRESS (Science and Technology Publications, Lda.)

The rest of the paper is organized as follows:

section 2 tackles some recent works related to se-

mantic resource retrieval in enterprises. Section 3

gives an overview of the pattern-based approach.

Section 4 outlines the ontology structure. Section 5

presents the meta-model of the alignment patterns.

Section 6 presents the pattern-based search and fi-

nally an example of implementation of the approach

within STMicroelectronics is shown in Section 7.

2 RELATED WORKS

We focus here on the semantic retrieval approaches

that include the business need of the user or its busi-

ness context in the search process. Several tech-

niques can be used to include the knowledge domain

related to the user need in a search process (Yan

2010); (Nunes, 2006); (Chu, 2003); (Belkin and

Croft, 1987).

In (Li et al., 2007), the authors proposed a

framework called EO-Search to semantically tag

heterogeneous engineering resources (e.g., CAD

drawings, design manuals, data sheets, etc.) and

improve their retrieval during the design process in

the automotive industry. The user queries consist of

a list of keywords that are processed with a semantic

disambiguation technique supported with a business

ontology and a lexicon of terms. The authors in (Yao

et al. 2009) proposed a retrieval framework for the

retrieval of resources used in the processes of prod-

uct design and manufacturing. The system maps the

user queries (i.e. business keywords) with the con-

cepts of two business ontologies developed by the

authors. An intention feedback mechanism is pro-

posed to the user at the results’ display to refine

them. In (Zaher et al., 2006), the authors proposed

the HyperTopic approach for the collaborative de-

scription and retrieval of heterogeneous resources

related to software projects conducted in two French

companies. The concept of “point of view” was

mainly used to link the resources to the user need.

The retrieval process was based on navigation along

the concepts of the resulting knowledge model

(Guittard et al., 2005).

Overall, most of the existing works that tackles

resource retrieval in enterprises focused on enhanc-

ing the keyword search with more concepts of the

involved knowledge domain captured in ontologies.

The user need is generally expressed with business

key words or with simple goals. However, infor-

mation resources in enterprises mainly address com-

plex business needs. Up to now, the complex busi-

ness needs of the users are not well included, neither

in the resource description, nor in the search pro-

cesses.

We propose with our approach to deal with the

complex business needs of the users in the resource

search process using goal-oriented mechanisms.

3 APPROACH OVERVIEW

The pattern-based approach aims at providing a

business-need-oriented description and retrieval to

information resources in the companies using align-

ment patterns. We define an alignment pattern as an

artifact of business need that can be simple or com-

plex. Each artifact constitutes a part of a description

process of a resource that carries business infor-

mation. In the alignment patterns, the business needs

are expressed with goals and the patterns are implic-

itly linked to each other through goal decomposi-

tions.

Figure 1: Role of alignment patterns in resource description and retrieval.

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

194

Also the goal decomposition may vary according to

a context related to the user business activity. The

pattern-based approach is supported with a business

ontology that provides the necessary semantics for

information-resource description and retrieval. This

ontology is goal oriented.

Figure 1 depicts the role of the alignment pat-

terns in filling the gap between business needs and

the information resources that we can find in a com-

pany.

The alignment patterns carry the expression of

business goals up to their satisfaction in a given

business context. They provide another level of re-

source description close to the business needs of the

users. Furthermore, we consider that these resources

must have meta-data or semantic descriptors that

provide their low level of description (business us-

age, business concepts, etc.), so to make the link

with the alignment patterns. The business ontology

is mainly used to consolidate this link. As a result,

the alignment patterns enable somehow to “align”

the low-level description of business information

resources in a company with the high-level needs of

the business actors.

In the proposed approach, the alignment patterns

are integrated in a keyword search system, to pro-

vide a business-need-focused search of resources to

the users. These patterns are progressively created,

stored and reused during each search process re-

quired by the user. On the whole, the alignment

patterns meet two purposes:

i. they constitute a source of knowledge related to

business needs and how they must be addressed

with resources, so to be shared and reused

ii. they guide the resource retrieval process, to in-

crease the relevance of the search results to the

user need.

4 THE BUSINESS ONTOLOGY

The business ontology proposed in this work gets a

standard and abstract description of the knowledge

domain enhanced with a goal orientation. This goal

orientation enables to support the creation of the

alignment patterns and the semantic retrieval of

resources.

Because this work was conducted with the

STMicroelectronics Company, we developed a

manufacturing process ontology (Bouzid et al.,

2013b). The scope of this ontology relies on four

views of description:

‐ Organization View: any business activity (func-

tion) related to the manufacturing process. We

included in the organization view the business

profiles that are related to the manufacturing ac-

tivity (e.g. process engineering, maintenance,

etc.)

‐ Function View: describes the manufacturing

objectives of the company.

‐ Data View: gathers the data types involved in the

manufacturing process (i.e. data about equip-

ment, data about products, etc.)

‐ Control View: gathers a process control descrip-

tion of the manufacturing process. The process

control regroups the set of techniques and meth-

ods used to control a manufacturing process to

ensure its well achievement. Note that because of

the vocabulary heterogeneity related to this view,

we also proposed a process control dictionary

that rationalizes the process control terminolo-

gies, but this part of work is out of scope of this

paper (Bouzid et al., 2013c).

The proposed views refer in fact to the descriptive

levels of the ARIS architecture (Ferdian, 2001).

They were chosen as a starting point for the domain

description because the ARIS approach seeks reduc-

ing the complexity of modeling industrial business

processes using these levels

5 META-MODEL OF

ALIGNMENT PATTERNS

The pattern notion was originally developed by

Christopher Alexander (Alexander, 1979) to respond

to recurrent problems in the architectural domain.

This notion was reused and readapted by Gamma et

al., (1995) in software engineering. In this paper,

patterns are used to respond to recurrent business

needs whose solutions are expressed with goals.

Figure 2 depicts the meta-model of an alignment

pattern. An alignment pattern is generally composed

of a goal, a context and a solution.

A goal expresses a business need, which can be

complex or atomic. A complex goal can be refined

into sub goals, whereas an atomic goal cannot be

refined.

The context of a need requires the business pro-

file or activity of the user. The business profiles and

activities in a company are generally identified in

the business ontology that describes its core business

process.

The solution of an alignment pattern can be

business goals or one or several semantic descriptors

of resources, where a semantic descriptor refers to a

resource. In fact, the solution for a pattern differs

APattern-basedApproachforSemanticRetrievalofInformationResourcesinEnterprises-ApplicationWithin

STMicroelectronics

195

Figure 2: Meta-model of an alignment pattern.

according to the type of goal that composes it (i.e.

complex or atomic). A solution of a pattern can be a

set of goals when the goal of the pattern is complex.

In this way, each pattern solution to a complex goal

provides an “And” decomposition that enables to

realize it.

Figure 3 shows a general example of an align-

ment pattern. Variability reduction is a complex goal

in the example, its solution is the sub goals Lot con-

trol and Equipment control that contribute to its

realization. The link between a pattern solution and

another pattern is not defined in the content of the

patterns, but rather deduced during the goal decom-

position of the search process.

Figure 3: Example of an alignment pattern.

Each complex goal defined in a pattern can have

alternatives of decomposition according to different

contexts, because a same goal can be associated to

different contexts for its achievement (Figure 2).

The alternatives express in fact the “Or” decomposi-

tions, but in practice, the alternatives of decomposi-

tion are carried with different patterns.

Thus, two patterns P1 and P2 can have a same

complex goal G1 in their definition but different

“And” decompositions in their solution, depending

on the context of achieving the goal G1.

Goals that cannot be refined constitute the low

level of the hierarchical structure of goals and are

known as atomic goals. Thus, when an alignment

pattern has an atomic goal, its solution is the set of

semantic descriptors that satisfy this atomic goal

according to the specified context.

Finally, during the search process, the alignment

patterns are dynamically linked through the goal

decomposition mechanism, which guides the re-

source retrieval for every user need.

6 THE PATTERN-BASED

6.1 Context of Application

The basic problem facing any user searching for a

resource in a business context is how to well capture

his need and how to retrieve the resources relevant

to his business objectives in a company. We consid-

er that using semantic-based solutions to support

information-resource retrieval in such contexts must

consider not only the semantics of the resources but

also the semantic alignment between the resources

and the complex business needs of the users. This

ascertainment has been deduced through the study of

the information system of the STMicroelectronics

Company, a semi-conductor manufacturer. We stud-

ied in this company a set of business information

resources used to support the control of the manu-

facturing process. These resources must be shared

and retrieved by company engineers of different

business activities and profiles.

Basically, the STMicroelectronics’ products are

manufactured in lots where each lot is a silicon wa-

fer used as support for the construction of electronic

chips. The manufacturing process spread over sever-

al work areas (Photolithography, Etching, defectivi-

ty, etc) where each step and operation inside it can

be remade several times. Controlling this complex

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

196

manufacturing process consists in monitoring and

analysing manufacturing information extracted and

processed with specific data management systems.

In this context, the resulting resources carry infor-

mation about the manufacturing activity.

A first part of this research work has been al-

ready done within the company to create semantic

descriptors to information resources related to the

context of work, using a bottom-up description ap-

proach. More details about this work can be found in

(Bouzid et al., 2013b); (Bouzid et al., 2013c).

In a second part of this work, the pattern-based

search is proposed to put the business needs of the

users (i.e. business actors) in the heart of the re-

source retrieval process. The users formulate their

query using business keywords, a goal and a context,

and the system performs these queries and enriches

at the same time the resource description through the

alignment patterns. In fact, when a user formulates

his query, two situations can happen:

a. the alignment pattern(s) do(es) not yet exist, the

system builds pattern-based solutions starting

from the user query

b. the alignment patterns that meet the user need

exist, the system takes the solutions of the pat-

terns that better meet the user query

The interesting asset of the proposed pattern-based

search is that business need artefacts are progres-

sively captured in alignment patterns and stored in a

pattern base for reuse.

6.2 Creation of Alignment Patterns

during the Search Process

Three goal-oriented mechanisms are used to capture

and encapsulate business-need artifacts in alignment

patterns: goal decomposition, goal-sibling decompo-

sition and goal abstraction. These mechanisms are

complementary and are repeatedly used as much

time as they are needed for the construction of the

necessary alignment patterns that respond to a given

high-level business need

. When some patterns al-

ready exist and others must be created during the

search process, the created ones are dynamically

linked to the existing patterns following the goal

decomposition relations. The use of these goal-

oriented mechanisms enables to progressively fill

the gap between the business needs of the users and

the information resources during each search pro-

cess.

6.2.1 Goal Decomposition

The standard goal decomposition technique

(Lamsweerde, 2001) is used here to refine complex

business goals. The goal decomposition is expressed

when, considering a complex goal 1, all the sub-

goals of 1 must be achieved in order to achieve

this goal. We can also say from a bottom up view

that the goals SG1.1 and SG1.2 contribute to the

realization of the goal CG1. In addition, 1 can

have alternatives of decomposition according to the

given business context.

Figure 4: Examples of goal decomposition.

Figure 4 shows an example of a goal decomposi-

tion with two alternatives. Variability reduction is a

goal that can be decomposed into Lot control and

Equipment control in the business context process

engineering, or into Defect control and Equipment

control in the business context Defectivity.

In general, three scenarios of decompositions can

be identified in a same context as follows:

- a complex goal CG is refined into complex sub

goals SG



. Each sub goal SG



must be refined in

turn until achieving atomic goals

- a complex goal CG is refined into complex goals



and into atomic goals AG



. The complex

sub goalsSG



must be refined in turn until

achieving atomic goals

- a complex goal  is refined into atomic goals





only. There is no more refinement to do in

this case

6.2.2 Goal-sibling Decomposition

The goal-sibling decomposition is a special case of

the goal decomposition. The decomposition of goals

is identified starting from a sibling goal. Considering

two goals CG1.1 and CG1.2 (Figure 5 ), if CG1.1 is

sibling of CG1.2, and CG1.2 is used in the same

context as CG1.1, thus CG1.2 will be also captured

in an alignment pattern and decomposed in turn if it

is a complex goal.

APattern-basedApproachforSemanticRetrievalofInformationResourcesinEnterprises-ApplicationWithin

STMicroelectronics

197

Note that the goal decomposition scenarios men-

tioned before are the same for the sibling-goal de-

composition.

Figure 5: Example of goal-sibling decomposition.

6.2.3 Goal Abstraction

The goal abstraction is expressed when, considering

two goals SG1.1 and SG1.2, they contribute to the

realization of a goal CG1. The abstraction of goals

can also be associated to a given context.

In Figure 6, the goals Cycle time optimization

and Procedure optimization contribute to the realiza-

tion of the same goal Cost optimization in the same

context.

Figure 6: Example of goal abstraction.

6.2.4 Example

We consider the following user query as example

from the STMicroelectronics’ context:

Keywords: {‘lot’, ‘out of control’}

Goal: {‘cycle time optimization’}

Context: {‘process engineering’}

The manufacturing process ontology of the company

contains a description of its core business activity

according to four views of description and each view

describes a business aspect of the process, including

the business objectives of the company related to

each business activity (Bouzid et al., 2013c). Thus,

in this ontology, the goal “Cycle time optimization”

specified in the user query is in fact a complex goal.

It will be then refined following its relations with

other goals defined in the business ontology (Figure

7) and according to the specified business activity

“Process engineering”. Starting from this goal, a

pattern P1 will be created (or identified in the pattern

base). Figure 7 shows the obtained patterns for this

example (to simplify the example, the patterns’ iden-

tifiers like P1, P2, etc. are used to refer to the pattern

solutions instead of using goals).

The pattern P1 with its goal and context is used

as starting point to identify or create the patterns

linked to P1 through the goal-oriented mechanisms.

For example, the pattern P7 will be created through

the goal abstraction, the patterns P8 and P3 through

the goal-sibling decomposition and the patterns P4

and P5 through the goal decomposition. The result-

ing resources according to this example would be:

R1, R2, R3, R7, R4, R6.

Figure 7: Example of resulting pattern architecture after query processing.

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

198

Finally, each new pattern resulting from a search

process is stored in a dedicated base for reuse.

6.3 Combining the Alignment Patterns

with a Keyword Search for

Resource Retrieval

As mentioned before, the alignment patterns are

integrated in a keyword search to enhance the re-

source retrieval process in business contexts.

Figure 8 outlines the search process after the re-

quired alignment patterns are created (or found in

the pattern base). The alignment patterns with their

solutions can be reused in several queries to guide

the refinement of the user need until reaching the

atomic goals and then the resources that meet them.

The keyword search is processed separately from

the pattern-based search. The keyword search con-

sists here in matching the keywords of the user que-

ry with the concepts of the semantic descriptors of

resources.

Figure 8: Processing of the user query by combining a

keyword search and alignment patterns.

A standard string matching is applied, and this

process is semantically supported with the business

ontology and a process control dictionary. Indeed,

each information resource has a semantic descriptor

that provides its basic usage. In our context, the

semantic descriptors provide a manufacturing de-

scription and a control description of each resource

used to control a manufacturing process. These de-

scriptors were created with a semantic mapping

technique and capitalized in a dedicated base in a

previous work (Bouzid et al., 2013a).

At the end of the search processing, the results

obtained with the alignment patterns and with the

keyword search are merged, filtered and ranked, so

to only keep the corresponding resources to the user

need. The ranking is also applied in order to display

the results in a suitable way to the user.

Table 1 shows four categories identified for

ranking the resources according to their matching

with the user query. The display of the results to the

user is done in the ascending order of these catego-

ries.

Table 1: Categories of ranking.

Goal Context Keyword(s) Ranking categories

match(



) match(



) match(



)

match(



) match(



) match(



)

match(



) match(



) match(



)

match(



) match(



) match(



)

match(



) match(



) match(



)

According to the example in we consider that we

have obtained the following resources with the key-

words’ matching: R1, R2, R5, R6, R7. Knowing that

5 resources were obtained with the alignment pat-

terns, the resources that better satisfy the user need

after ranking in this example would be: R1, R2, R6,

R5, R7, R3, R4 and will be displayed in this order.

7 IMPLEMENTATION

The general retrieval approach was implemented

within STMicroelectronics with a web prototype.

Following the Information Technology policy of the

company, the Php and C technologies were used for

the programming logic and the XML technology

was used to implement the business ontology.

An example of a pattern-based search is present-

ed in the appendix. Basically, the system takes as

user query one or a set of keywords, a goal and a

context. A keyword here can be any concept of the

manufacturing process description of the company.

When the user specifies key concepts, the system

automatically displays meta-information about each

concept, so to help the user in choosing the right

concepts to express his need. The semantics of the

manufacturing process ontology is also used to assist

the user in defining his business need with goals.

The system processes the goal refinement task with

the alignment patterns regardless of the level of

abstraction of the selected goal by the user.

The pattern-based approach is still under test to-

day within the STMicroelectronics Company. The

first experimentations were done using a set of sce-

narios of needs defined with their solutions by the

business experts of the company. 12 scenarios were

tested on the keyword search alone and on the pat-

tern-based search (i.e. the combination of the key-

word search and the alignment patterns). The ap-

proach showed good results on the latter with 96%

APattern-basedApproachforSemanticRetrievalofInformationResourcesinEnterprises-ApplicationWithin

STMicroelectronics

199

of precision and 87% of recall comparing to the

keyword search alone (78% of precision and 85% of

recall).

8 CONCLUSION

This paper presented the pattern-based approach,

devoted to enhance resource retrieval in enterprises.

The proposed approach emphasizes a business-need

focused search based on alignment patterns in a

standard keyword search. The alignment patterns are

created dynamically when a user searches for infor-

mation resources in a business context. These pat-

terns progressively capture business need artefacts

and support their reuse for addressing recurrent

needs. This original search approach presents two

main interesting assets: (i) it enriches resource de-

scription with a high-level business semantics main-

taining in this way the link between the resources

and the needs of business actors in a company, and

(ii) it progressively capitalizes on business know-

how by storing the semantics of the information

resources used in an activity domain.

In future work, more experimentations of the ap-

proach are planned with the business experts of the

STMicroelectronics Company. A specific string-

similarity measure is also planned to be tested on the

matching between the keywords and the resource

descriptors. This technique was previously used in

our work where it proved its efficiency (Bouzid et

al., 2013a). We consider that it would be worth to

reuse it to foster the keyword matching during the

search process.

REFERENCES

Alexander, C., 1979. The Timeless Way of Building, New

York, USA: Oxford University.

Belkin, N.J. & Croft, W.B., 1987. Retrieval Techniques.

In M. E. Williams, ed. Annual Review of Information

Science and Technoloy (ARIST). Elsevier Science

Publishers B.V., pp. 109 – 145.

Bouzid, S. et al., 2013a. A Bottom up Search Technique of

Manufacturing Indicators. In 24th Annual SEMI

Advanced Semiconductor Manufacturing Conference

(ASMC 2013). Saratoga Springs, New York: IEEE.

Bouzid, S. et al., 2013b. A Semantic Mapping Approach

to Retrieve Manufacturing Information Resources. In

The IEEE IFAC-MIM Conference. Saint-Petersburg,

Russia: IEEE Xplore.

Bouzid, S. et al., 2013c. A Semantic Support to Improve

the Collaborative Control of Manufacturing Processes

in Industries. In 17th IEEE International Conference

on Computer Supported Cooperative Work in Design

(CSCWD 2013). Whistler, BC, Canada: IEEE.

Chu, H., 2003. Information Representation and Retrieval :

An Overview. In I. Information Today, ed.

Information Representation and Retrieval: An

Overview. pp. 1–25.

Ferdian, 2001. A Comparison of Event-driven Process

Chains and UML Activity Diagram for Denoting

Business Processes,

Gamma, E. et al., 1995. Design patterns: Elements of

Reusable Object-Oriented Software, Addison-Wesley.

Guittard, C., Zaher, L.H. & Cahier, J., 2005.

Experimentation of a socially constructed “ Topic Map

” by the OSS community. In In proceedings of the

IJCAI-05 workshop on Knowledge Management and

Ontology Management (KMOM). Edimbourg.

Lamsweerde, A. Van, 2001. Goal-Oriented Requirements

Engineering : A Guided Tour. In Proceedings Fifth

IEEE International Symposium on requirements

Engineering. Toronto: IEEE, pp. 249–263.

Li, Z., Raskin, V. & Ramani, K., 2007. A Methodology of

Engineering Ontology Development for Information

Retrieval. In International Conference on Engineering

Design, ICED’07. Paris, France, pp. 1–12.

Nunes, S., 2006. State of the Art in Web Information

Retrieval, Porto, Portugal.

Yan, W., 2010. Component Retrieval Based on Ontology

and Graph. Journal of Information & Computational

Science, 4, pp.893–900.

Yao, Y., Lin, L. & Dong, J., 2009. Research on Ontology-

Based Multi-source Engineering Information Retrieval

in Integrated Environment of Enterprise. In

International Conference on Interoperability for

Enterprise Software and Applications. China: Ieee, pp.

277–282.

Zaher, L. H. et al., 2006. The Agoræ / Hypertopic

approach. In International Workshop IKHS - Indexing

and knowledge in Human Sciences (Sdc). Nantes,

France.

APPENDIX

ICEIS2014-16thInternationalConferenceonEnterpriseInformationSystems

200