Elsevier

Information Sciences

Volumes 346–347, 10 June 2016, Pages 286-301
Information Sciences

Modeling ontology evolution via Pi-Calculus

https://doi.org/10.1016/j.ins.2016.01.059Get rights and content

Highlights

  • We provide a new angle to model the operational semantics of the changes in ontology evolution.

  • The entities in the ontology is model as initiative agents that interact with each other with state transitions during the change.

  • We use Pi-Calculus to clarify the operation semantics of the changes.

  • The Mobile Work Bench is used as a tool to verify the applicability of our method.

Abstract

Extending and updating real world ontologies is unavoidable challenging with our ever-evolving understanding of the world and the evolution of the world by itself. In the current researches, changes are usually modeled as passive instant results of operations such as to add a child class or to delete a property, executed by the administrators. But this view neglects the real world facts that (1) an ontology evolves continuously over time, not just hopping instantly between static versions; (2) a change is a coherent procedure although it may be separated from the administrative point of view into different phases, such as to request, detect, represent, evaluate, implement and propagate the change. This paper provides an inside perspective of the ontology itself to model its evolution. The ontology entities are regarded as autonomous agents with find-grained state specifications. The operational semantics of the changes are formalized as series of information exchange actions of the agents. Pi-Calculus serves to describe the operational semantics in our modeling. Firstly, an ontology is encoded into a hyper graph. Then the nodes and edges in the hyper graph are formalized as Pi-Calculus processes. The replication operator is used to add new entities to the ontology and the communication rule works for the resolution of the information exchanges. Thus, a change is modeled as a coherent procedure as clarified in its operational semantics. A case study shows the feasibility of our method on the Mobility Workbench (MWB). We believe that the operational semantics in our scenario disclose the autonomous evolving nature of ontology evolution.

Introduction

Ontologies have been flourishing in recent years as the predominant pattern of knowledge bases in diverse domains such as agriculture, bio-medicine, defense and geology. Representative examples of ontology applications are the Gene Ontology [1], the Foundational Model of Anatomy [2], the National Cancer Institute Thesaurus [3] and GALEN [4]. Such ontologies are huge in their scales and constantly evolving [5]. Most of the real world ontologies are developed and managed by multiple groups collaboratively. The continual and asynchronous modifications present rigorous challenges to the administration in the detection, evaluation and implementation of the changes. For users and ontologists, the management work are laborious, time-intense and error-prone processes, even with the aid of some semi-automatic tools such as PROMPTdiff [6], COnto-diff [7], CoSWEM [8], DynarOnto [9], etc. Consequently, it is critical to address the issue to manage ontology evolution (OE) in the timely adaptation of an ontology to the arisen changes and preserve the consistency in the propagation of these changes to the relevant artifacts.

Ontology versioning [10] is one of the most fruitful research topics in OE studies. To handle changes, the management system has to detect the differences between versions of the ontologies first. Such a versioning system of ontologies compares and presents the structural changes [11]. Although it is a big step forward to view ontology changes in structure rather than in character as in software management, the main idea of ontology versioning and many of the following researches are based on the perspective of a human/user, i.e. the changes in OE are detected and handled passively as the administration tasks. Taking OE as the object to manage, the interrelated change procedure is separated into different phases such as change representation, semantic specification, implementation and propagation [12]. This perspective focus on the outcomes of the evolution, but not the procedure of the evolution itself. For example, there is a change to add a class A into the ontology as the sub-class of an existing class B. The change has to be represented in certain format in the representation phase, as some text or formula. Then it is evaluated for its semantics, as an intent to add a subsumption axiom AB into the ontology. Then it is implemented, i.e. the insertion of the axiom into the ontology. At last, the influence of this change is propagated, such as to inform the individuals of class B about this change. If an inconsistency arose, extra efforts should be spare to resolute the negative influence or even roll back the change. In our opinion, the procedure of evolution are driven autonomously, rather than requested, detected, separated, analyzed and implemented by outer forces. The entities that evolve should initiate the process and fulfill all the actions, such as inconsistency checking and side-effect propagation. To the best of our knowledge, there are few research results that have addressed OE in this way.

This paper takes the view of autonomous OE. It clarifies the operational semantics of the changes in OE, which is formalized as a series of information exchange actions during change. This is not the classic semantic definition in the ontology domain with set theory, but defined as a series of actions referred to as operational. We represent the elementary changes with the mathematical calculus Pi-Calculus [13] (shortened to Pi if no ambiguity exists). Firstly, an ontology is encoded as a hyper-graph in our method. Such a hyper-graphs consists of ordinary nodes and edges, plus hyper nodes as the encapsulation of multiple related nodes. An ordinary node in the graph represents an entity in the evolving ontology, such as a class, a property or an individual of a class. A hyper node is used to denote a compound class or a named pair as the individual of a property. From a structural aspect, elementary changes are differentiated into three categories, unary, binary and ternary, according to the number of nodes related. Secondly, the compositions of the ontology are modeled as Pi processes. Such processes may transit between different states, which are new, ready, busy and deserted as defined in the evolution automaton. Thirdly, the operational semantics are modeled with series of information exchanges among such Pi processes through Pi channels. To conquer the disability of Pi in creating new process names, the replication operator is exploited to create processes to be bound with new channel names. With the resolution of the communication through channels, the whole evolution procedure can be completed. Finally, a case study is provided to show the feasibility of our method on the Mobility Workbench (MWB) [14] in a controlled automated simulation test. The operational semantics of the changes in our scenario capture the continuous autonomous evolving nature of OE and provide the fundamental basis to the representation and evaluation of the more complex changes in the future.

The paper is organised as follows: Section 2 describes the related work; Section 3 gives the preliminaries of Pi-Calculus; Section 4 shows the modeling procedure; Section 5 evaluates the method with a case study on MWB; the discussions are in Section 6 and the conclusions are in the last section.

Section snippets

Related work

There have been many studies to explore the semantics of changes. Some focus on the analysis of different stages in handling the changes [12], [15], [16], [17], some emphasize on inconsistency management [18], [19], [20], [21], [22], [23], some simulate the version control of software engineering [10], [22], [24], [25], [26], some analyze the patterns of change operators [17], [27], and some provide calculus based analysis of the changes [28], [29], [30].

Stojanovic et el. proposed a user driven

Preliminaries

Pi-Calculus is designed to represent and evaluate communicating and mobile systems, concerning concurrent information synchronization [38]. It is a useful formalism for interacting systems with dynamically evolving communication topologies. It allows channels to be passed as data along other channels, with simple semantics and tractable algebraic theory. The classic Pi consists of processes (also known as agents) constructed as follows, with channels (also known as names or ports). P,Q:=mn.P|m

Modeling changes with Pi-Calculus

Ontology evolution will be presented in three steps: first to define the ontology itself with respect to its possible structure, then to clarify the operational semantics of the evolution and thirdly to model the implementation of the changes. We will describe these steps in detail in the following subsections.

Evaluation

In this section, a case study is explored to evaluate the feasibility of our method by the Mobile Work Bench.

Differences and characteristics?

The most notable research result in OE is the work done on ontology versioning. The idea is to borrow the mature version control method in software engineering into the field of OE. However, a classical versioning system depends on the textual differences between the versions, therefore there exist a great number of efforts to process structural or logical differences between ontologies such as [10], [16], [24], etc. in contrast to character-level differences between codes. Our method is

Conclusion

Ontology is no longer a static theoretical knowledge base but a practical evolving application of multiple modern information technologies in various domains. It becomes almost impossible to manage the evolution of huge intricate ontologies manually, therefore researchers tend to develop intelligent methods to manage the evolution. However, the runtime semantics during a change has not been well studied so far. This paper presents a new way to model the changes of ontology with the Pi-Calculus.

Acknowledgments

Thank Prof. David Weiss for proof reading and English refinement. The work of this paper has been funded by Jilin Province High-Performance Computing Platform Based Research in Related Fields: Cloud-based Distributed Ontology Reasoning Method 20140101206JC-10, by the Key Program for Science and Technology Development of Jilin Province of China under grant no. 20130206052GX, and by the Natural Science Research Foundation of Jilin Province of China under grant no.20150101054JC.

References (40)

  • C. Rosse et al.

    The Foundational Model of Anatomy Ontology Anatomy Ontologies for Bioinformatics

  • A. Rector et al.

    Ontological and practical issues in using a description logic to represent medical concept systems: Experience from GALEN

    Reasoning Web, Second International Summer School, Tutorial Lectures

    (2006)
  • M. Hartung et al.

    Analyzing the evolution of life science ontologies and mappings

    Proceedings of the 5th International Workshop on Data Integration in the Life Sciences (DILS)

    (2008)
  • M. Klein et al.

    Change Management for Distributed Ontologies

    PHD Thesis, Technical Report

    (2004)
  • P.-H. Luong et al.

    A rule-based approach for semantic annotation evolution in the CoSWEM system

  • R. Touhami et al.

    Ontology evolution for experimental data in food

    Metadata and Semantics Research. Proceedings of the 9th Research Conference, MTSR 2015

    (September 9-11, 2015)
  • M. Klein et al.

    Ontology versioning and change detection on the web

    Proceedings of the 13th International Conference on Knowledge Engineering and Knowledge Management (EKAW02)

    (2002)
  • N.F. Noy et al.

    Tracking changes during ontology evolution

    Proceeding of the 3rd International Semantic Web Conference (ISWC2004

    (2004)
  • L. Stojanovic et al.

    User-driven ontology evolution management

    Proceedings of the 13th International Conference on Knowledge Engineering and Knowledge Management. Ontologies and the Semantic Web, EKAW ’02

    (2002)
  • M. Abadi et al.

    A calculus for cryptographic protocols: The spi calculus

    Inf. Comput.

    (1999)
  • Cited by (7)

    View all citing articles on Scopus
    View full text