The SCR Model - IEEE Xplore

9 downloads 0 Views 116KB Size Report
scale, multiform information resources. In this paper, we outline a notion of semantic-based context-role model for Web information integration and sharing.
Semantic-based Context-Role Model for Web Information Sharing Yuxin Mao1, Zhaohui Wu1, Huajun Chen1 , Zhao Xu1 1 Grid Computing Lab, College of Computer Science, Zhejiang University, Hangzhou 310027, China {maoyx,wzh, huajunsir, xuzhao}@zju.edu.cn

Abstract Various autonomous organizations have exposed heterogeneous information resources on the Web in a state out-of-order, which makes information sharing rather difficult. Semantics, organized as Web ontologies, can be used to integrate distributed, largescale, multiform information resources. In this paper, we outline a notion of semantic-based context-role model for Web information integration and sharing. The implementation of the model and its use case on Traditional Chinese Medicine are also described.

as a configurable graph. However, the functionalities of OntoRama are restricted to browsing only. It seems that utilizing ontologies to integrate information is no longer a novel technique; however, most of the existent methods just construct global models from the top down and haven’t take locality and compatibility of information into account. In this paper, we outline a notion of semantic-based contextrole model for domain-oriented Web information integration and sharing. An implementation and its use case on Traditional Chinese Medicine (TCM) are also described in detail. Throughout the paper, we will refer to an ontology about TCM (see figure 1).

1. Introduction Currently, the Web is expanding at a striking speed and a great volume of information has been generated across the Web. Various autonomous organizations expose heterogeneous information resources like databases and knowledge bases in a state out-of-order. In such a situation, integrating and sharing vast amount of information resources appears to be rather urgent issues on which we must focus our attention. Semantics, organized as Web ontologies, been broadly used to integrate distributed, independent, multiform information resources. A lot of works has been done in this field. Haystack [1] is an RDF based individual information management tool, which aims to enable users to define whichever arrangements of, connections between, and views of Web information they find most effective. Protégé [2] is an ontology development and knowledge acquisition environment with a graphical user interface. The server mode of Protégé supports multiple users working simultaneously and users can remotely browse and edit ontologies in Protégé, but it’s difficult to integrate information resources into server dynamically. OntoRama [3] is a prototype ontology browser, which takes RDF/XML as input format and models ontology

Figure 1. The TCM ontology

2. Basic Concepts To begin with our model, we must explain several key concepts at first.

Context The concept of context has been used abroad and presents multiple meanings within different fields like pervasive computing. We extended the theory of local model and local model semantics proposed by

Giunchiglia and Ghidini in [6] to the area of Web ontologies. An ontology O can be defined as the set of Semantic Pieces ( SP ). Let I be a countable set of indexes and then the context can be treated as a compatible set of SP s in an ontology. Context == {SPi SPi ∈ O}i∈I

(1)

The elements in a context set can be assembled on demand, which means that we can construct a context by extracting pieces of semantics from an ontology dynamically. Reasoning is always local to a subset of the known facts and the small subset is what determines the context of reasoning [7], so when we perform contextual reasoning on an ontology, a context can be also defined as a subset of the complete state described the domain ontology used for a given goal.

Role Contexts are relatively static elements in the model, so we add another ingredient role into our model, to enhance interactivities with users. A role itself doesn’t embody much knowledge other than a set of basic rules and an operative interface. Unlike traditional agent, role is a much more lightweight object and can be combined with other objects easily and freely, or several roles can be combined into an organization according to situation. Role Theory has been widely used for enterprise modeling and postulates that individuals occupy positions in an organization [8]. In our model, users can see role as a visual symbol (metaphor) with an operational interface, from which high-level processes can interact with it. Role projects an abstract view on a series of packed actions to users and it may partially simulate a social role or object. However, here the concept of role is mainly towards the Web with heterogeneous information resources, so we don’t project each social role to a role one by one and roles are formed due to the dynamic, multiform and distributed characteristics of Web information. On all accounts, a role is an atomic unit and when we use roles to construct high-level processes like inquiry or inference, we will not care about its inner structures.

3. Semantic-based Context-Role Model In this section, we present a Semantic-based Context-Role (SCR) Model for integrating and sharing

heterogeneous information resources in Web environment. A non-empty set of contexts CS is the context space of the model. RS is a set of roles available and active in the model. KS is the scope of knowledge to be used and generated in the model. Let MS be a set of mappings from 2 CS × 2 RS to KS . Let O be the Web ontology of a specific domain. Then, we can define the SCR Model as the following quadruple. SCRModel == O, CS , RS , MS

(2)

There’re two basic processes in the model, one constructing contexts and the other generating roles, and we describe them with two schemata. Assume that there is a global Web ontology o , and a context description d given by users. isNorm(d , o ) ∧ SPs(sps, spFilter (d , cc )) ∧

schema(cc, ConstructContext ) → context (c, o )

(3)

newContextSet (cs, c )

(4)

∧ elements(c, sps ) context (c, o ) ∧ contextSet (cs, o ) →

We need to construct different contexts dynamically due to the requirements of practical problems. If users submit a normative description, the model is able to construct a corresponding context by extracting SP s and returns a URI [9] as the handle of the context. isNorm(d , o ) ∧ roleSet (rs, rsFilter (d , gr )) ∧ schema(gr , GenerateRole ) → role(r , o )

(5)

New constructed contexts will extend CS and in contrast roles are not constructed dynamically but generated from RS . So rather than construction, the process of role generation is much more like selection. The choice of role is always relative to some contexts and focuses on specific tasks. Several roles can share a context and the same role can act under several contexts. Although role is not composed of semantics, we can still refer to it with a URI, as everything on the Web can be treated as resources.

3.1 Knowledge Extension Protocol If we want to deal with practical and complex problems, we should refer to large-scale information and organized it as knowledge. In a distributed environment, heterogeneous information resources

may have their own schema and structure, but we can abstract generalities and match them with a common Web ontology to extend KS , as long as they belong to the same domain. We can describe the Knowledge Extension Protocol in the SCR Model with the following schema. abstractDescription(d , ir ) ∧ constructContext (c, contextFilter (d , o )) → extendKS (d , c ) context (c, o ) ∧ contextSet (cs, o ) → updateContextSet (cs, c )

(6) (7)

4.1. Semantic Registration

3.2 Information Refinement Protocol As the goal of the SCR Model is to cover largescale Web information, the automatic refinement of information for end-users is very important. constructContexts(cl , contextFilter (d , o )) ∧ generateRoles(rl , roleFilter (d , o )) ∧ refineSet (rs, o ) → evolve(rs, refine(rl , cl ))

visual interactive environment for large-scale information sharing in the Web. Most of the complex functionalities and processes in the SCR Model are implemented as Grid Services. For various Grid Services, we accordingly develop semantic plug-ins, which are independent and optional functional modules in Semantic Browser to access services remotely. The most related, most recent and most specific ones are preferred on choosing contexts and roles in practice. Semantic Browser visualizes contexts as Semantic Graphs ( SG ) and assist users to browse semantic information graphically.

(8)

Roles act under given contexts and interact with original information to generate knowledge. The refined set of valuable information can be utilized to resolve practical problems or used as important references. And during the process, KS of the model has evolved to a higher level, though there may be no extra information resources being generated.

The Knowledge Extension Protocol is implemented as Semantic Registration [4], which is the process to extend the scope of the browser. Semantic Registration Service (SemRS) establishes mappings from the schemata of external information resources to the shared ontology and provides the service of registering and inquiring about the mapping information. Semantic Browser provides users with a visual interface to facilitate the registration. During the process of Semantic Browsing (SemBS), contexts are constructed on demand and users can dynamically register local information resources at a context through the semantic mapping protocol. Semantic Browser will get resource schemata from SemRS and dynamically list them below the semantic graph. Simply by dragging and dropping a concept or relation from the context onto a list item, we establish a semantic mapping.

4.2. Semantic Query

Figure 2. A screen shot of Semantic Browser

4. Implementation According to the model proposed above, we have implemented a system dedicated to integrating and sharing large-scale information, which is called Semantic Browser [10](see figure 2). We built our system based on several Semantic Web [11] standards and Grid [12] technologies and provide users with a

The Information Refinement Protocol is going to be implemented as several parts and Semantic Query [13] is an important one for querying information at semantic layer. Semantic Query Service (SemQS) accepts semantic queries, inquires of SemRS to determine which resources are capable of providing the answer, and convert semantic queries into local queries. The results will be wrapped as semantic information and returned to Semantic Browser. We offer a Semantic Query Language (Q3) to describe querying statements with semantics. Querying statements can be generated dynamically during the process of SemBS [5]. Users can dynamically construct contexts and the graphical components of contexts offer several query operations in the context menu. When users carry out one of operations at a graph component, a corresponding

query segment will be automatically generated in a dynamic query panel. Meanwhile, users must attach one or more roles to the context (e.g. TCM Pharmacist to Chinese Patent Drugs) to complete the query request.

Medicine; Intel / University Sponsored Research Program: DartGrid: Building an Information Grid for Traditional Chinese Medicine; and China 211 core project: Network-based Intelligence and Graphics.

References 5. Use Case We can show the usage of Semantic Browser based on our model for China Academy of Traditional Chinese Medicine by a use case. If a patient finds his knees and loin debilitated and thus wants to look for some drugs to treat his pains, then he can turn to Semantic Browser for consultation. As Debility of the loins and knees is an instance of the class Pain of the sacrum and coccyx (a subclass of Symptoms of internal diseases), he can construct two contexts Disease and TCM Prescription, and select the role TCM Physician to perform a semantic query visually in Semantic Browser. According to user’s description, the role TCM Physician may find following two rules in its rule set: 1) Debility of the loins and knees is caused by ColdDamp. 2) Cold-Damp medicinal instant granule treats ColdDamp. Cold-Damp is an instance of the class Etiological factors. As the ontology used above has been extended and related with several distributed databases about TCM prescriptions and diseases, so Semantic Browser will get useful information for the user through SemQS. As a result, two kinds of Cold-Damp medicinal instant granule will be displayed as intuitive SG s, and the patient may use either of them as an important recommendation for his disease.

6. Conclusion In this paper, we mainly describe a semantic-based context-role model for sharing and integrating largescale information under the Web environment. We have implemented the SCR Model as Semantic Browser and built an information-sharing platform for TCM based upon Semantic Browser, which involves several large databases distributed across the Web.

Acknowledgement This work is supported in part by China 973 fundamental research and development project: The research on application of semantic grid on the knowledge sharing and service of Traditional Chinese

[1] D. F. Huynh, D. Quan and D. R. Karger, User Interaction Experience for Semantic Web Information, The Twelfth International World Wide Web Conference, May 2003. [2] N. F. Noy, M. Sintek, S. Decker, M. Crubezy, R. W. Fergerson and M. A. Musen, Creating Semantic Web Contents with Protege2000, IEEE Intelligent Systems 16(2): 60-71, 2001. [3] P. Eklund, N. Roberts and S. Green, OntoRama: Browsing RDF Ontologies Using a Hyperbolic-style Browser, The First International Symposium on CyberWorlds (CW2002), pp.405-411, Theory and Practices, IEEE press, 2002. [4] Y. Mao, Z. Wu and H. Chen, Semantic View for Grid Services, Proceedings of the International Conference on Services Computing, p 329-335, 2004. [5] Y. Mao, Z. Wu and H. Chen, Visual Semantic Query Construction in Dart Database Grid, GRID AND COOPERATIVE COMPUTING, LECTURE NOTES IN COMPUTER SCIENCE 3252: 768-744, 2004. [6] C. Ghidini and F. Giunchiglia, Local Models Semantics, or contextual reasoning= locality+ compatibility, Artificial Intelligence, 127(2001): 221-259 2001. [7] F. Giunchiglia, Contextual Reasoning. Epistemologia, Special Issue on "I Linguaggi e le macchine". Vol. XVI, pages 45-364, Tilgher-Genova, Italy, 1993. [8] B. J. Biddle and E. J. Thomas, Role Theory: Concepts and Research, New York: Robert E. Krieger Publishing Company, pp. 450, 1979. [9] F. Manola and E. Miller, RDF Primer, W3C Recommendation, 2004. [10] Y. Mao, Z. Wu and H. Chen, Semantic Browser: an Intelligent Client for Dart-Grid, COMPUTATIONAL SCIENCE - ICCS 2004, PT 1, PROCEEDINGS LECTURE NOTES IN COMPUTER SCIENCE 3036: 470-473, 2004. [11] T. Berners-Lee, J. Hendler and O. Lassila, The Semantic Web, Scientific American, May 2001. [12 I. Foster, C. Kesselman and S. Tuecke, The Anatomy of the Grid: Enabling Scalable Virtual Organizations, Lecture Notes in Computer Science, Vol. 2150: 1-26, 2001. [13] H. Chen, Z. Wu and Y. Mao, Q3: a Semantic Query Language for Dart Database Grid, GRID AND COOPERATIVE COMPUTING, LECTURE NOTES IN COMPUTER SCIENCE 3251: 372-380, 2004.