2010 International Conforence on Computer Application and System Modeling (ICCASM 2010)

Semantic Grid for Chemical Domain Knowledge Service

Xuelin Shi YingZhao

School of Information Science and Technology School of Information Science and Technology
Beijing University of Chemical Technology Beijing University of Chemical Technology
Beijing, 100029, China Beijing, 100029, China
e-mail: shixl@mail. buct. edu. cn e-mail: zhaoy@mail. buct. edu. cn

Abstract-Chemical Grid is a collaborative grid platform flexible collaboration and computation on a global scale.
based on CGSP (China Grid Support Platform) and provides a Fig. l shows the progress of Semantic Web.
grid computing environment for various chemical applications.
On this platform we constructed a knowledge service
architecture using Semantic Web technology, in particular
RDF (Resource Description Framework), to integrate -
::l
heterogeneous chemical information, generate domain (trn Semantic Semantic
.... (1)
knowledge and provide knowledge service for researchers. o 3 Web Grid
o po
Furthermore, we designed a Chemical Knowledge � ::l

. g.
�
Representation Model consisting of a stack of languages, and �
0

� �
brought out a RDF/S-Graph model to represent semantic ::l
property and relationships of these data. All extracted
chemical domain knowledge, i.e. RDF data and service
mechanism compose a Semantic Grid, which is an experience
on grid computing platform and is a innovative effort of the Computing Performance
Semantic Web.

Figure 1. The progress of Semantic Grid.

Keywords-Grid Computing; Semantic Web;
Knowledge service; RDF
I. INTRODUCTION
The objective of Grid computing is to bring a variety of
computational and data resources together to create new
capabilities. Now it presents a new trend to distributed
computing and Internet applications [1]. In grid
environment how to integrate large-scale heterogeneous
resources and provide users knowledge service is becoming
more and more attractive. In the past few years there have
been some efforts to provide the knowledge and semantic
framework to enable the automated management and sharing
of complex resources using Semantic Web technology.
Semantic Web has been proposed with rapid growth of
Web, in order to provide enhanced information service by
applying machine-processable meta-information [2]. It is the
idea of having data on the Web defined and linked in a way
that it can be used for more effective discovery, automation,
integration and reuse across various applications. The Web
can reach its full potential if it becomes a place where data
can be processed by automated tools as well as people.
Based on grid computing and Semantic Web, Semantic Grid
[3 , 4, 5, 6] was brought up as an infrastructure where
complex applications and services can be deployed with
minimal manual intervention. It is characterized by an open
system, with a high degree of automation, which supports
Chemical;
This paper reports on our experiences in building a
Semantic Grid infrastructure: Chemical Knowledge Service
Semantic Grid (CKSSG) for chemical research as part of the
Chemical Grid Project of Beijing University of Chemical
Technology (BUCT) . Chemical Grid is a grid computing
platform based on China Grid Support Platform (CGSP)[7].
It integrates most of chemical research resources of BUCT
and provides high collaborative grid computing platform for
chemical applications. Our work is one application on
ChinaGrid. Based on Chemical Grid, the CKSSG especially
aims to provide semantic retrieval services for chemical
domain knowledge. It is concerned with the way that
knowledge is acquired, used, retrieved, published and
maintained to assist chemical researchers to achieve their
particular goals and objectives. To implement such function,
we designed knowledge service architecture, which is easy
to deploy on the grid platform. Then a chemical domain
knowledge representation model (CKRM) is brought up, the
key of CKRM is RDF/S-Graph.
This paper is organized as follows. Section 2 introduces
the background of Chemical Grid and the knowledge service
architecture of CKSSG. In section 3 , we describe
representation model for chemical domain knowledge based
on a stack ofianguages, especially introduced RDF/S-Graph.
The implementation of the CKSSG and its application
scenario is given in Section 4. Finally, concluding remarks
and future work of research are described in Section 5.

978-1-4244-7237-6/10/$26.00 ©2010 IEEE V7-549

 2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

II. KNOWLEDGE SERVICE ARCHITECTURE OF CKSSG CKSSG is designed to improve current information service
systems and help users get what he (she) wants by providing
In this section, we introduce the background of Chemical
semantic search and knowledge classification. Fig. 3 shows
Grid platform and present the knowledge service architecture
the architecture of CKSSG.
of CKSSG.
A. Chemical grid
Chemical Grid aims to provide a unified and
collaborative environment to access different chemical Knowledge Services
applications across different physical domains and security
firewalls. It is a collaborative computing framework based
on CGSP. CGSP is a grid middleware developed to build the
ChinaGrid, which integrates all kinds of resources in Meta-knowledge
education and research environments, makes the
heterogeneous and dynamic nature of resource transparent to
the users, and provides high performance, high reliable,
secure, convenient and transparent grid service for the RDFData Repository
scientific computing and engineering research. CGSP is
based on the core of Globus Toolkit.
RDFModeling
Fig. 2 gives the framework of Chemical Grid [8]:

Domain Specific Service Layer Extraction semi-structured

(Chemical Engineering Applications)

,
,
___________________ 1____________________
,
,
Data Source
(WWW)
,
, Collaboration Resource
Support Service Sharing Service

1
,
, Figure 3. CKSSG architecture
'------------------- -------------------

The architecture of consists of three layers: extraction

CGS P
Globus Toolkit
Figure 2. Framework of Chemical Grid
The framework consists of 3 tiers: domain specific
service layer, common service layer and grid middleware
layer. We used CGSP as the lowest layer, grid middleware.
Above the 3 layers, there is the portal, web interface for
users, which is the entrance for the end user to use grid
services.
The first layer is CGSP acting as grid middleware layer,
which implements the characteristic of grid for collaborative
resource sharing. The second layer is common service,
which includes collaboration support service and resource
sharing service. This layer is based on grid middleware layer
for collaborative chemical grid computing environment. We
used some professional software to construct the second
layer. The third is application layer, which oriented to special
application domain. Many chemical engineering applications
are deployed on this layer.
B. Knowledge service architecture
Based on Chemical Grid, we construct the knowledge
service architecture CKSSG for chemical researchers.
layer, knowledge layer and knowledge services layer, which
implement functions of all the key stages of the knowledge
management lifecycle, i. e. knowledge acquisition,
knowledge representation and knowledge use.
The bottom is extraction layer, which integrated digital
resources on chemical engineering. At present, the major
source is Web pages, which are immense, heterogeneous and
dynamic, and it is semi-structured, such as html, xml or other
semi-structured text. This layer extracts these semi­
structured data to implement knowledge acquisition.
After knowledge has been acquired form these semi­
structured data, it is then required to represent the knowledge
in a special model easy to implement semantic retrieval. This
is the function of the second layer, knowledge layer, which
consists of two sub-layers: RDF data repository and meta­
knowledge.
The bottom of the knowledge layer is the RDF data
repository, which stores knowledge from the extraction layer.
We used RDF as knowledge representation model, making
data machine-understandable and enhancing the retrieve
services.
The top level of the knowledge layer is meta-knowledge.
Meta-knowledge is a process of critical thinking, reasoning
and understanding of knowledge, through which it turns into
a kind of skills of problem-solving and decision making.
Meta-knowledge serves not only knowledge management,

V7-550
 2010 International Conference on Computer Application and System Modeling (ICCASM 2010)
but more importantly provIsIOn of knowledge services,
problem-solving methods, and decision making processes.
Meta-knowledge is the analysis, evaluation, comprehension,
inference, and interpretation of inherent knowledge and is
the exploitation, analysis and manifestation of users wisdom
and experience [9].
Shortly, just like metadata, meta-knowledge is
description of knowledge. In our CKSSG, the meta­
knowledge is created, maintained and populated in such way:
firstly the knowledge engineer constructed category tree of
chemical domain knowledge. Based on key words the RDF
data is classified into deferent category. Secondly, while
operation of CKSSG, based on feedbacks of users, the meta­
knowledge is improved.
The top is knowledge services layer, which provides
knowledge sharing, searching, browsing, summarizing,
visualizing and organizing services for users. Importantly,
the knowledge service interface component will record basic
operations when a user uses the system. These logs, which
will display the interests of the user in certain information
resources, will be the major source of meta-knowledge.
Furthermore, as the meta-knowledge accumulating, the RDF
data repository will be updated with new meta-knowledge
rules.
This architecture is concerned with the knowledge is
acquired, represented, used, retrieved and maintained to
assist chemical scientists to achieve their particular goals and
objectives. It extracted knowledge from WWW, and
represented them in RDF format. In order to provide users
knowledge navigation service, it construed meta-knowledge.
Then users can access the knowledge by knowledge service
interface on the top layer.
III. CHEMICAL KNOWLEDGE REPRESENTATION MODEL
CKSSG acquires chemical engineering domain
knowledge from Internet, such as Web sites, open chemical
resource library and so on. These data are characterized as
semi-structured, i. e. they are often semi-structured texts,
HTML documents and XML documents. Furthermore the
data often includes some description information, i. e.
metadata, such as subject, updating time, source, and so on.
Based on such features, it is effective to use RDF as
knowledge representation model. Using such model makes
data machine-understandable and enhances the retrieve
services.
Therefore in our CKSSG we designed a knowledge
model, Chemical Knowledge Representation Model
(CKRM) , based on RDF to represent knowledge and provide
users more flexible semantic retrieve services. In the
following, we introduced the layered structure of our
chemical domain knowledge model, then described such the
RDF/S-Graph model, at last the retrieval mechanism are
presented.
A. Layered structure of CKRM
According to the requirements of CKRM, it consists of a
stack of languages and data model, likes a pyramid. The
layer of CKRM is showed in Fig. 4.
V7-551
I
Retrieval Mechanism
I Domain Knowledge Category
I RDFS (RDFSchema)
1 RDF(Resource Description Framework)
I XML (eXtensible Markup Language)
Semi-Structured Data
Figure 4. Layer of CKRM
As shown in figA, the bottom is XML (Semi-structured
data is information source, not included in CKRM). XML is
already widely known and designed for make-up in
documents of arbitrary structure. In our CKRM, XML is
used as serialization syntax of RDF/S data.
The second layer is RDF, which is a W3C
recommendation for the formulation of meta-data on the
World Wide Web [10]. RDF is an XML application
customized for adding meta-information to Web documents.
When it comes to semantic interoperability, RDF has
significant advantages over XML The basic building block
.
in RDF is an object-attribute-value triple, commonly written
as A(O, '1. That is, an object ° has an attribute A with value
V. For example, the following is a relationship expressing in
A(O, '1 format:
hasName{'http://202.4. I 30. 20SlperiodList',
'periodic table of the elements ')
The above triple can be serialized in XML as followed:
<rdfDescription rdfabout=
''http://202. 4. 130. 20S/pieriodList'' >
<hasName rdfresource=
" periodic table of the elements">
<lrdfDescription>
Above RDF is RDFS. Just as XML Schema provides a
vocabulary definition facility for XML RDF Schema ,
provides a similar facility for RDF. RDF Schema extends
this standard with the means to specifY domain vocabulary
and object structures [11]. These techniques will enable the
enrichment of the Web with machine-understandable
semantics, thus giving rise to what has been dubbed the
Semantic Web.
The upper layer is chemical domain knowledge category,
which not only is used for knowledge classification and
knowledge navigation, but also facilitates users' feedbacks
processing. Currently the domain knowledge category is
constructed by knowledge engineer according to chemical
part of Chinese Book Category, i. e. TPI to TP2.
 2010 International Conference on Computer Application and System Modeling (ICCASM 2010)

The top layer is retrieval mechanism of knowledge of

CKRM format. In fact it just needs design a query
mechanism for RDF/S data, which should implement
&r3
querying at the semantic level. Now there are a lot of query Dark-red sediment
mechanisms for RDF, such as RQL [12], RDQL [13] and so
on. In our CKSSG, we used a simple retrieval mechanism for
RDF/S-Graph [14].

B. RDFIS-Graph
is-a
As described above, the basic building block in RDF is
ch:ao
an object-attribute-value triple, commonly written as A(O, V).
That is, an object ° has an attribute A with value V. The
triple also can be presented as graph: objects (resources) are
presented as nodes and attributes are denoted by labeled Pyrolysis oil
edges between nodes, Le. {O/ -A -{V/. Usually one object
can be a value of another object, so it is clear to presents
rl: catalyst for oil production
such relation using graph.
r2: Fe203
RDFS defines classes of RDF resources and the range of
r3: CUS04
attributes. In nature RDFS is also RDF documents, so RDFS
also can be presented by graph: nodes present classes and
Figure 6. RDF/S-Graph of RDF Document
edges denote types of attributes. Therefore, RDF/S-Graph
can be used as knowledge representation model. A RDF/S­
Fig.5 is a RDF document serialized in XML, which
Graph is a Directed Acyclic Graph (DAG) with labeled
describes molecule type, physical property and chemical
edges, which can be presented as RG (V, E, D, 1), and V
=

property of catalyst for oil production (Fe]03), especially

denotes the set of nodes, E denotes the set of edges, D is
including the information source (Le. URL address) of the
domain of node, 1 is mapping from edges to nodes.
digital object. Fig.6 gives the graph of this RDF document
Fig.5 and fig.6 give an example of RDF/S-Graph to
based on above RDF/S-Graph definition.
represent chemical knowledo-e:
It is clearly to note that using RDF/S-Graph is fit to
<rdf:RDF> implement querying at the semantic level, because it
<rdf:Description about="catalyst for pyrolysis oil represents the relationships among properties and objects.
production" Importantly, such data are easy to be stored in RDBMS
<ch:molecule>Fe203</ ch:molecule> (Relationship Database System). In CKSSG we used
< ch:physical�rop> MySQL.
dark-red sediment
</ ch:physical�rop>
IV. IMPLEMENTAnON AND APPLICATION
< ch:chemical�rop>
oil yield can be improved As an application deployed on Chemical Grid platform,
about 76.7% when using Fe203 as catalyst GKSSG provides chemical domain knowledge, knowledge
</ ch:chemical�rop> retrieval and navigation services, and it can be accessed by
<url_addr>http://......</url_addr> all users of our Chemical Grid. When the user need access
</rdf: Description>
these knowledge service, he/she just submit a request on
Chemical Grid Web portal, then the grid job schedule system
</rdf:RDF>
will manage the job execution and implement load balance
of widespread computational resources.

A. Implementation on Chemical Grid platform

Figure 5. An Example of RDF Document Based on Chemical Grid, we implement CKSSG. From
users' view, CKSSG is also a service provided by Chemical
Grid: users can submit request in portal, then Chemical Grid
system actives CKSSG. When CKSSG receives request, it
begins search RDF data repository and meta-knowledge.
Then the results are presents to users. Fig.7 shows the system
architecture ofCKSSG.

V7-552
 2010 International Conforence on Computer Application and System Modeling (ICCASM 2010)
Portal
ROBMS
� ROFOata
� Repository
Figure 7. System Overview of CKSSG
As shown in the above figure, ChemExtract module
extracts semi-structure data from WWW. Then these data are
processed and stored in database. Now ChemExtrace is a
executable application developed with Java.
ChemSearch module is the core component of the system.
It not only provides knowledge query service, but also
provides knowledge navigation. Furthermore it is designed
as a WSDL compatible and SOAP compatible web services
deployed on grid platform.
The web portal is the interface by which users can access
knowledge service of CKSSG. Unlike typical web subject
portals, a grid portal may also provide access to grid resource.
B. Application scenario
A scenario of CKSSG application is derived from
requirements of one chemical researcher, who wants get
some information about catalyst for oil production.
The researcher logs into the Grid platform by Web portal,
and he can choose service. By looking through the services
list, he decides to access CKSSG service.
Maybe he can firstly use knowledge navigation. From the
Domain Knowledge Category he finds the catalyst, and he
goes to the sub-category until he gets what he wants.
He can also use knowledge retrieval service by
submitting a query request. When ChemSearch module gets
the results, he can get what he wants.
Various chemical researchers maybe express different
interest in knowledge. By recording users' feedback, these
logs will be an important source to improve knowledge.
V. CONCLUSION AND FUTURE WORK
In this paper we report our CKSSG application on grid
platform. It is concerned with the way that knowledge is
acquired, used, retrieved, published and maintained to assist
chemical researchers to achieve their particular goals and
V7-553
objectives. To implement such function, we designed
knowledge service architecture, which is easy to deploy on
the grid platform. Then a chemical domain knowledge
representation model (CKRM) is brought up, the key of
CKRM is RDF/S-Graph. Based on the architecture and the
knowledge mode, we implement CKSSG system.
At present CKSSG just integrated a little of chemical
digital resources from WWW. However, the grid metaphor
intuitively gives rise to the view of the CKSSG as a set of
services that are provided by particular individuals or
institutions for consumption by others. On the other hand a
grid is useless without appliances to plug in [3]. Our CKSSG
is a useful knowledge service which is able to deploy on grid
computing platform and provide services using distributed
computing resources. All extracted chemical domain
knowledge, i.e. RDF data and service mechanism compose a
Semantic Grid, which is an experience on grid computing
platform and is an innovative effort of the Semantic Web.
REFERENCES
[I] 1. Foster, C. Kesselman. The Grid: Blueprint for a Future Computing
Infrasturcture. Mogan Kaufmann Publisher, 1999.
[2] John Davies, Dieter Fensel, Frank van Harmelen. Towards the
semantic web: ontology-driven knowledge management. John Wiley
& Sons, Ltd, 2003.
[3] David De Roure, Nicholas Jennings, Nigel Shadbolt. Research
Agenda for the Semantic Grid: A Future e-Science Infrastructure.
Technical Report, URL:
http://www.semanticgrid.org/vI. 9/semgrid. pdf
[4] K. R Taylor, J. W. Essex, etc. al. The Semantic Grid and Chemistry:
Experiences with CombeChem. Web Semantics: Sciences and Agents
on the World Wide Web 4 (2006), 84-10I.
[5] D. De Roure, N. Jennings, N.R Shadbolt. The Semantic Grid: Past,
Present, and Future. Proc. IEEE 93 (2005), 669-681.
[6] C.A. Goble, D. De Roure, etc. al.: Enhancing services and
applications with knowledge and semantics. The Grid2: Blueprint for
a New Computing Infrastructure. Morgan-Kaufinann, 2004, 431-458.
[7] Jin HaL: China Grid: making grid computing a reality. International
Collaboration and Cross-Fertilization (ICADL2004), 2005, 13-24.
[8] Ying Zhao, Xuelin ShL Collaborative Computational Chemical Grid
Based on CGSP. 2007 IFIP International Conference on Network and
Parallel Computing Workshops (NPC 2007), 199-202.
[9] Xiaoxing Zhang. Knowledge Service and Digital Library: A
Roadmap for the Future. 7th International Conference on Asian
Digital Libraries (ICADL 2004), Shanghai, China, December 13-17,
2004.
[10] O.Lassila, Ralph Swick. Resource Description Framework (RDF)
Model and Syntax Specification. URL: http://www.w3.orgITRlREC­
rdf-syntax!.
[II] D. Brickley, RV. Guha. RDF Vocabulary Description Language 1.0:
RDF Schema. URL http://www.w3.orglTR/2004/REC-rdf-schema-
200402101.
[12] G. Karvounarakis, A. Magganaraki, et. al. Query the Semantic Web
with RQL. Computer Networks 42(2003), pp.617-640.
[13] A. Seaborne. RDQL-A Query Language for RDF. W3C Member
Submission. http://www.w3c.org/ SubmissionlRDQL.
[14] Xuelin Shi, Ying Zhao. RDF based integrated information retrieval in
grid computing environment. 2008 International Joint Conference on
Neural Networks (IJCNN2008).