A Legacy ERP System Integration Framework

based on Ontology Learning

Chuangtao Ma and Bálint Molnár

Faculty of Informatics, Eötvös Loránd University, Budapest 1117, Hungary

Keywords: Integration Framework, Ontology Learning, Legacy ERP Systems, Data Integration.

Abstract: In the past decades, there are various legacy ERP systems that exist in different departments or sub-

organizations within the enterprise. The majority of the legacy ERP systems are heterogeneous systems, that

may be developed by different software companies under different development framework, which create a

big challenge for organizations to develop and implement centralized and integrated management systems

based on their existing legacy ERP systems to respond the dynamic business environment with agility.

Ontologies are viewed as an effective technology to integrate different data from multiple heterogeneous

sources, the ontology learning methods were proposed to achieve (semi-)automated construction of

ontologies. This paper proposes a general framework for legacy ERP system integration based on ontology

learning to tackle this challenge. Initially, the related literature is reviewed from the perspective of system

integration and ontology learning, then an integration framework based on ontology learning is given, and the

basic workflow and ontology learning process are analysed and illustrated.

1 INTRODUCTION

The terminology of the legacy information system

was proposed by Brodie (Brodie, M.L., 1993) to

describe the class of information systems, which are

developed a few years ago and their technology isn’t

the most modern one, but these systems also run

normally in organizations to provide the management

and decision support currently. Enterprise Resource

Planning (ERP) system are developed as a critical

assistant tool to provide management decision

support and optimal solution for enterprise manager

in the past decades. The efficiency of the management

and decision was improved by ERP systems

significantly that promote the development of the

enterprise. Recently, by the growth and expansion of

the business, an increasing enterprise plan to, or

already have achieved to update and to upgrade their

ERP system. Hence, various legacy ERP systems is

emerging in different organizations and departments.

The majority of the legacy ERP systems are

heterogeneous systems that may be developed by

different software companies under different

development frameworks, so that fact it creates a big

challenge for organizations to implement the

integrated business intelligence system based on

various legacy ERP information systems, and to

achieve collaborative decision making in the process

of management. Because of the development and

great progress in information systems technology

(e.g., service-oriented system architecture, business

intelligence, etc), there is a trend that many

enterprises plan to develop and implement business

intelligence system based on their existing legacy

ERP systems to respond to the dynamic business

environment quickly and effectively. Therefore, the

problems of how to reconcile and integrate these

heterogeneous legacy ERP systems efficiently and

effectively became an urgent task that should be

researched and resolved.

Ontologies are viewed as an effective technology

to integrate different data from multiple

heterogeneous sources (Das, M., et al, 2015), and

have been adopted to solve the problems of data

heterogeneity. The results of the integration are

determined by the quality of the ontology largely,

while the quality of the ontology is limited by the

experience of the ontology experts. The process of

constructing ontology is a task that requires a lot of

time and effort, hence it is practically impossible for

ontology experts to construct various domain

ontology manually, and implement the integration of

various legacy ERP systems on time. Ontology

learning framework (OLF) was proposed by Maedche

(Maedche, A., et al, 2001) to provide a (semi-)

Ma, C. and Molnár, B.

A Legacy ERP System Integration Framework based on Ontology Learning.

DOI: 10.5220/0007740602310237

In Proceedings of the 21st International Conference on Enterprise Information Systems (ICEIS 2019), pages 231-237

ISBN: 978-989-758-372-8

231

automated tools and method for ontology modelling,

and to achieve the integration of various data for

semantic Web.

In this paper, a legacy ERP system integration

framework based on ontology learning is proposed to

achieve the integration of the various legacy ERP

systems effectively and efficiently. The structure of

this paper is organized as follows. Section 2 gives a

brief description and discussion of the related work,

and illustrate the motivation and goals of this paper.

Section 3 proposes the general framework for legacy

ERP system integration based on ontology learning,

and the detailed process of legacy ERP system

integration based on ontology learning is illustrated.

The conclusion and future direction related to this

topic are reported in Section 4.

2 RELATED WORK

2.1 Legacy ERP System Integration

Legacy ERP systems play a significant role in

enterprise’s daily management, e.g., decision making,

production planning, plan executing, cost accounting,

and so forth. As we all know, the most prominent

elements of ERP systems are supply chain

management (SCM) systems and customer

relationship management (CRM) systems. However,

the large amounts of data are relevant for both

enterprise resource planning and supply chain

management, data is stored redundantly (Weske, M.,

2012). Worse, due to the traits of the legacy ERP

systems, including fragile, obsolete, and their

interface are unfriendly, the problems of the legacy

ERP system integration are becoming more and more

serious. Especially, it is crucial for enterprises to

develop and integrate a customized ERP system to

response the current dynamic business environment

(Tommi M.K., et al, 2014). For legacy ERP system

integration, there are three levels include, business

process integration, data integration, and system

integration.

2.1.1 Data Integration

Data integration is the fundamental level in the

hierarchical framework for legacy ERP system

integration. In generally, ERP system integration

could be achieved by the integration framework based

on data integration (Huang X.X, et al, 2013). XML is

viewed as a useful technology and tool are being

adopted to achieve data integration in business

information systems (Lampathaki, F., et al, 2009).

Meanwhile, technology and tools of data integration

based on extract-transform-load (ETL) are emerging,

so that a data integration framework was constructed

based on ETL tools (Dayal, U., et al, 2009) in order

to achieve the data integration by extracting data from

distributed database systems. By the development of

information technology, a data integration framework

based on ontology was proposed (Lv, Y., et al, 2016)

to achieve the data integration in wide range. To

improve the user friendliness of system interfaces and

to integrate the heterogeneous data at semantic level,

a lot of linked data were created. However, because

of the ontologies are used to standardize the linked-

data, and the ontologies underlying linked data are

often non-interoperable (He, Y., et al, 2018), which

bring a new challenge to integrate the linked-data

system.

2.1.2 Business Process Integration

Business process integration is a critical part in

integration framework for legacy ERP systems since

that is a precondition for ERP system integration. In

essence, the business process integration is strongly

and intimately connected to workflow integration,

hence, a business process framework was proposed

based on analysis of the workflow (Kobayashi, T., et

al, 2003).

There is a proposal for an integration framework

that is workflow oriented, which aims at business

integration and automation. The framework is

grounded in event process chain approach that

applied for ERP systems. This framework was

designed to encompass a model that is structure

oriented and dedicated to business process

applications and workflows (Samaranayake, P., et al,

2006).

In order to clarify the business process integration

within the ERP systems, a systematic architecture for

business process integration was built to give an

overview of the business process integration (Magal,

S.R., 2011). Recently, an ontology augmenting

XBRL (eXtensible Business Reporting Language)

extension model and a matching framework based on

semantic web technology and augmented ontology

were designed to achieve the process integration of

financial analysis at semantic level within ERP

systems (Bai, L., et al, 2018).

2.1.3 System Integration

System integration is the top-level integration

architecture of legacy ERP systems, its objective is to

realize centralized management and decision making.

System integration provides an approach for effective

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

232

integration among different heterogeneous legacy

ERP systems. In early research, a prototype of multi-

agent enterprise resource planning system that

employs software agent theory to achieve ERP

system integration (Lea, B.R, et al, 2005). The

progress of the cloud computing has created

opportunity, a collaborative manufacturing networks

was devised for cloud computing platform to achieve

the collaboration between of supply chain system and

MES (Manufacturing Excitation System, MES)

efficiently (Govindarajan, N., et al, 2017). In practice,

an increasing number of tools and platforms were

developed to accomplish integration along with

modernization of legacy ERP systems. In industrial

application fields, Mule ESB (Enterprise Service Bus,

ESB) is a useful system integration component that is

integrated by Anypoint platform (Riives, J., et al,

2012). This platform could integrate both the web-

based modern ERP systems and legacy systems

seamlessly by APIs (Application Program Interface,

API) and graphical data mapping and transformation

technology for legacy data. Nevertheless, the

efficiency and quality of the integration are

determined by the experience of system

administrator.

2.2 Ontology Learning

Ontology learning was proposed and considered as a

(semi-) automated tool to acquire and to extract

knowledge, to identify ontologies, to construct

ontologies, to recognize relationships among

ontologies, and so forth. This approach made it

possible to introduce the methods and algorithms of

machine learning into the discipline of ontology

engineering (Maedche, A., et al, 2001). The previous

researches on ontology learning were focused on

knowledge extraction automatically from text

database (Lehmann, J., et al, 2014). Initially, ontology

learning was applied on areas where Natural

Language Processing (NLP) plays an important role.

The fields of ontology learning are expanding

gradually from knowledge extraction to semantic

matching and ontology mapping accompanied the

progress of machine learning.

Ontology learning frameworks were built to

provide a useful tool for the purpose of ontology

learning. A flexible ontology learning framework was

proposed to generate and evaluate an ontology (semi-

)automatically through an integrated tool-suite that

exploiting ontology learning process (Gacitua, R., et

al, 2008). Through advancements of ontology

learning technology, frameworks have been enriched

greatly so that a framework can include functions for

learning, of elements – that could be either lexical or

pieces of ontological knowledge –, of methods, and

of evaluation of results (Shamsfard, M., et al, 2003).

Nowadays, in big data era, there are various

representations of knowledge and dynamic

relationships within different domain ontologies,

which increase the complexity of domain ontology

and bring a new challenge for ontology learning.

Furthermore, the research topics of ontology learning

are focused on the improvement of learning methods

and algorithms. For instance, an approach for

optimizing ontology learning frameworks is founded

on the method of seeking for near-optimal input

weights was proposed to integrate multiple and

heterogeneous evidence sources (Wohlgenannt, G., et

al, 2015). In addition, a domain ontology learning

method based on LDA (Latent Dirichlet Allocation,

LDA) model was proposed to improve the capability

of knowledge representation for ontology content

(Hong, W., et al, 2017). Similarly, a partial multi-

dividing ontology algorithm was proposed to

optimize the ontology learning model for improving

the efficiency of ontology learning (Gao, W., et al,

2018). However, limited number of studies regards to

ontology learning are available on NLP, moreover

rigorous studies for system integration based on

ontology learning have not been attempted to. It can

be seen from the aforementioned studies that research

about legacy ERP system integration and ontology

learning frameworks mainly focuses on data

integration, business process integration, and system

integration respectively. However, the results of

integration are restricted to descriptions of common

properties at simple syntactic level, thereby it can’t

meet the requirement of heterogeneous business

information system integration within the dynamic

business environment. While the research about

ontology learning framework mainly focuses on the

improvement of learning algorithms and methods, the

fields of application for ontology learning are

constrained on NLP, e.g., semantic analysis,

knowledge acquisition, content recommendation, and

so forth. To the best of our knowledge, there was no

attempt to explore the potential of Information

System Integration based on ontology learning.

The motivation of this paper is to solve the

problems of various legacy ERP system integration in

the dynamic business environment from the

perspective of ontology learning, and the objective is

to build a general integration framework for the

legacy ERP systems based on ontology learning.

A Legacy ERP System Integration Framework based on Ontology Learning

233

3 INTEGRATION

FRAMEWORKS BASED ON

ONTOLOGY LEARNING

3.1 Architecture for Legacy ERP

System Integration

For legacy ERP systems, data heterogeneity issues

will occur if a logical data item address is stored

multiple times in different sub-systems. Worse, the

majority heterogeneity of legacy ERP systems

usually caused by the various semantics of the

attributes, thereby these semantic differences need to

be eliminated in the process of ERP system

integration. Information systems integration based on

ontology learning does not only realize the system

integration at syntax level to meet the integrated

requirement for complex heterogeneous system in

dynamic business environment, but it also improve

automatically the quality and efficiency of the

information system integration at semantic level by

ontology learning and mapping.

In order to construct integration frameworks based

on ontology learning, the architecture for legacy ERP

system integration should be analysed and

considered. Therefore, an architecture for legacy ERP

system integration was illustrated in Figure 1.

As shown in Figure 1, the legacy ERP system

integration within different sub-organizations was

analysed. In this integration architecture for legacy

ERP systems, data integration was emphasized,

because it bridges business process integration and

information system integration. Meanwhile, there are

many-to-many mapping of relationships between

business processes and database entities. Thereby, a

conclusion can be drawn that data integration play a

vital role in legacy ERP system integration. Thus,

data integration was selected to illustrate the

integration framework based on ontology learning.

3.2 Integration Framework based on

Ontology Learning

For legacy ERP systems, there are various data

collections in heterogeneous databases that need to be

integrated. These data collections, namely, database

tables in relational database management systems,

were created under different kinds of database

management systems that exist various data

organization philosophies, so this fact causes a huge

obstacle for integrating. To solve this problem,

Extensible Mark-up Language (XML) was developed

to provide an opportunity for a unified description of

heterogenous data in a semi-structured way, and then

Figure 1: Architecture for legacy ERP system integration.

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

234

data can be transformed and loaded into heterogeneous

database structures. Similarly, there are several

heterogeneous data collections are represented of the

same entities, while the syntactic description differs

widely. The typical example of this phenomenon is that

there are various database field names and types in

different legacy ERP systems for the similar data items

of the same entities.

Figure 2: Legacy ERP system integration framework based

on ontology learning.

In the above described complex situation,

extensible mark-up languages are incapable of

accomplishing database integration effectively. To

tackle this problem, a legacy ERP system integration

framework based on ontology learning is proposed,

and then the schematic diagram of legacy ERP system

integration framework is given in Figure 2.

In Figure 2, the basic workflows of the legacy ERP

system integration based on ontology learning are

illustrated. The workflows of the legacy ERP system

integration based on ontology learning could be

summarized as follows: (ⅰ) Identify the entities from

legacy ERP systems database tables; (ⅱ) Extract the

ontology terms from database scripts documents; (ⅲ)

Match the different ontology terms of different entities;

(ⅳ) Map onto an integrated entity based on the

matching results of the ontology terms; (ⅴ) Create and

develop the integrated database schema and its

implementation.

3.3 Ontology Learning for Legacy ERP

System Integration

To depict the roles of ontology learning in integration

framework for legacy ERP systems proposed in

Section 3.2, the detailed process of ontology learning

for legacy ERP system integration is analysed and

illustrated in Figure 3.

It is visible from Figure 3 that the input of

ontology learning is text document sets, e.g., database

scripts, which contain system entities, and the output

of the ontology learning is an integrated data table.

The key steps of ontology learning for legacy ERP

system integration is described as follows.

Step 1 Pre-process of Entities Text Document:

There are various formats and naming conventions

among text documents for different entities, e.g. date

formats, fields name, and so forth, which produces a

huge obstacle for ontology learning. Hence, the text

document sets and text corpora should be pre-

processed (semi-) automatically at initial phase of

ontology learning for achieving the ontology con-

Figure 3: Ontology learning process for integration of the legacy ERP systems.

A Legacy ERP System Integration Framework based on Ontology Learning

235

struction and mapping based on linguistic methods.

The methods for text document pre-processing

include clustering, dimension reduction, and

linguistic processing. Due to the clustering method is

mainly used to extract the relations among data, thus

text notation, parsing, and lemmatization could be

adopted to pre-process the text document.

Step 2 Extract Ontology Terms from the Text

Document: It is a critical step to obtain ontology

terms for ontology learning, thereby, ontology terms

for representing knowledge, concepts within

ontology should be extracted after pre-processing text

documents. In this step, the natural language

processing techniques, such as semantic analysis, co-

occurrence analysis, and hyponymy detection, could

be applied to carry out term extraction from the text

documents and database scripts. Due to the results of

the ontology learning are largely determined by the

quality of the ontology terms. Thus, it is necessary for

us to consider how to improve the term quality and

the degree of automation in the process of ontology

term extraction.

Step 3 Recognize and Extract the Relationships

Among Ontology Terms: Relationships among

ontology terms are essential for ontology mapping

and matching, in this stage, all ontology terms will be

analysed with the help of association rule-base and

the domain ontology to recognize and extract the

relationships among ontology terms. The method for

recognition and extraction of relationships among

ontology terms could be classified in natural language

processing approaches and statistical approaches. The

method based on NLP techniques include dependency

analysis, lexicon for syntactic pattern, while method

based on statistical techniques include hierarchical

clustering, and association rule mining (Asim, M.N.,

et al, 2018). The typical relationship between

different ontology terms includes: X is similar with

Y, X is equal with Y, X is irrelevant considering Y,

and so forth.

Step 4 Matching Ontology Terms based on its

Relationships: Matching the different ontology terms

according to its relationships is conducive to integrate

the different legacy ERP systems at data integration

level. On performing the ontology terms matching

based on their relationships, the axioms for

relationship among the ontology terms should be

firstly extracted by applying inductive logic

programming. Then sets of relationships for ontology

terms will be obtained, these relationships will

provide a constructive advices and basis to support

the integrate process of the different database entities

within the different legacy ERP systems.

Step 5 Evaluate the Results of Matching

Ontology Terms: It is necessary to check the

consistency of ontology terms and then to evaluate

the results of matching ontology terms before the

process of integrating the different database entities

within heterogeneous legacy ERP systems. The

method for evaluating the results of matching

ontology terms could be classified into four types:

gold standard based, application based, data-driven

based and manual evaluation method. Meanwhile,

ontology learning is a multi-level process, which

increases the difficulty in the process of results

evaluation. Therefore, the domain ontology and

knowledge of ontology experts should be introduced

into the process of results evaluation.

4 CONCLUSIONS

In this paper, ontology learning technology was

employed to solve the integration problems of legacy

ERP systems. And then, a general integration

framework based on ontology learning is presented to

integrate various legacy ERP systems effectively and

efficiently. Data integration was selected to

demonstrate the integration process of the legacy ERP

systems based on ontology learning, and the key steps

of the legacy ERP system integration based on

ontology learning were given.

The presented study proposed a general system

integration framework based on ontology learning.

However, due to current study at its preliminary

stage, there are several issues that should be answered

in further investigation and study. According to the

current study, the process of ontology construction is

a precondition of the system integration based on

ontology learning. Therefore, the method of extract

and construct ontology terms from legacy ERP

systems database scripts should be investigated and

studied further. Moreover, the problem of how to

extract and learn the relationships between different

ontology terms from legacy ERP systems should be

considered and studied. The last but not least, the

problem of how to evaluate the efficiency and check

the consistency of the ontology terms in the process

of legacy ERP system integration based on ontology

learning also should be studied, because the

consistency between different ontology terms will

influence the quality of the mapping relationships,

eventually, it will determine the accuracy and

effectiveness of integration results of legacy ERP

systems.

ICEIS 2019 - 21st International Conference on Enterprise Information Systems

236

ACKNOWLEDGEMENTS

This work was supported by grants of the European

Union co-financed by the European Social Fund

(EFOP-3.6.3-VEKOP-16-2017-00002) and the China

Scholarship Council (201808610145).

REFERENCES

Brodie, M.L., 1993. The promise of distributed computing

and the challenges of legacy information systems.

Interoperable database systems, pp. 1-31.

Das, M., Cheng, J.C. and Law, K.H., 2015. An ontology-

based web service framework for construction supply

chain collaboration and management. Engineering,

Construction and Architectural Management, 22(5),

pp.551-572.

Maedche, A. and Staab, S., 2001. Ontology learning for the

semantic web. IEEE Intelligent systems, 16(2), pp.72-

79.

Weske, M., 2012. Business process management

architectures. In Business Process Management, pp. 25-

39. Springer, Berlin.

Tommi M.K., Maglyas, A., & Smolander, K., 2014. ERP

System Integration: An Inter-organizational Challenge

in the Dynamic Business Environment. International

Conference on Enterprise Information Systems, pp.39-

56. Springer: Lisbon.

Huang, X. X., Zhu, W., 2013. An enterprise data integration

ERP system conversion system design and

implementation. Applied Mechanics and Materials,

433-435 (5), pp. 1765-1769.

Lampathaki, F., Mouzakitis, S., Gionis, G., Charalabidis, Y.

and Askounis, D., 2009. Business to business

interoperability: A current review of XML data

integration standards. Computer Standards &

Interfaces, 31(6), pp.1045-1055.

Dayal, U., Castellanos, M., Simitsis, A. and Wilkinson, K.,

2009. Data integration flows for business intelligence.

12th International Conference on Extending Database

Technology: Advances in Database Technology, pp. 1-

11. ACM: Saint Petersburg.

Lv, Y., Ni, Y., Zhou, H. and Chen, L., 2016. Multi-level

ontology integration model for business collaboration.

The International Journal of Advanced Manufacturing

Technology, 84(1-4), pp.445-451.

He, Y., Xiang, Z., Zheng, J., Lin, Y., Overton, J.A. and Ong,

E., 2018. The eXtensible ontology development (XOD)

principles and tool implementation to support ontology

interoperability. Journal of biomedical semantics, 9(3),

p.1-10.

Kobayashi, T., Tamaki, M. and Komoda, N., 2003.

Business process integration as a solution to the

implementation of supply chain management systems.

Information & Management, 40(8), pp.769-780.

Samaranayake, P., & Chan, F. T. S., 2006. Business process

integration and automation in ERP system

environment: integration of applications and

workflows. The 2nd International Conference on

Information Management and Business, pp.155-162.

University of Western Sydney: Sydney.

Magal, S.R. and Word, J., 2011. Integrated business

processes with ERP systems. pp.315-348. Wiley

Publishing, Hoboken.

Bai, L., Koveos, P. and Liu, M., 2018. Applying an

ontology-augmenting XBRL model to accounting

information system for business integration. Asia-

Pacific Journal of Accounting & Economics, 25(1-2),

pp.75-97.

Lea, B.R., Gupta, M.C. and Yu, W.B., 2005. A prototype

multi-agent ERP system: an integrated architecture and

a conceptual framework. Technovation, 25(4), pp.433-

441.

Govindarajan, N., Ferrer, B. R., Xu, X., Nieto, A., & Lastra,

J. L. M. ,2017. An approach for integrating legacy

systems in the manufacturing industry. The 14th

International Conference on Industrial Informatics,

pp.683–688. IEEE: Poitiers.

Riives, J., Karjust, K., Küttner, R., Lemmik, R., Koov, K.

and Lavin, J., 2012. Software development platform for

integrated manufacturing engineering system. The 8th

International DAAAM Baltic Conference “Industrial

Engineering, pp. 555-560. Tallinn University of

Technology: Tallinn.

Maedche, A. and Staab, S., 2001. Ontology learning for the

semantic web. IEEE Intelligent systems, 16(2), pp.72-

79.

Lehmann, J. and Voelker, J., 2014. An introduction to

ontology learning. Perspectives on Ontology Learning.

pp.7-14. IOS Press: Amsterdam.

Gacitua, R., Sawyer, P. and Rayson, P., 2008. A flexible

framework to experiment with ontology learning

techniques. International Conference on Innovative

Techniques and Applications of Artificial Intelligence,

pp. 153-166.Springer: London.

Shamsfard, M. and Barforoush, A.A., 2003. The state of the

art in ontology learning: a framework for comparison.

The Knowledge Engineering Review, 18(4), pp.293-

316.

Wohlgenannt, G., Belk, S. and Rohrer, K., 2015.

Optimizing ontology learning systems that use

heterogeneous sources of evidence. International

Workshop on Multi-disciplinary Trends in Artificial

Intelligence, pp.137-148. Springer: Hanoi.

Hong, W., Hao, Z. and Shi, J., 2017. Research and

Application on Domain Ontology Learning Method

Based on LDA. Journal of Software, 12(4), pp.265-273.

Gao, W., Guirao, J.L., Basavanagoud, B. and Wu, J., 2018.

Partial multi-dividing ontology learning algorithm.

Information Sciences, 467, pp.35-58.

Asim,M.N., Wasim,M. and Khan,M.U.G., 2018. A survey

of ontology learning techniques and applications.

Database, 2018, pp.1-24.

A Legacy ERP System Integration Framework based on Ontology Learning

237