XML
;
NSC 89-2213-E-009-019
88 8 1 89 7 31
!"#$%&'()*+,
-./0123456!"78
-9:;<0123456!"78
0=.>?@AB56!"CDE*FGH78
.> I3J./3J!"K78
LM.N)O9?PQRS3JT
89 10 30
XML !
!
!
!
Research on XML-based Electronic Catalog Systems
Incorpo-rated with Personalized Merchandise Suggestions
U NSC 89-2213-E-009-019
88 8 1 89 7 31
)O9? PQRS3JT
UVWXYZ[\]^G\_`a )9PRT
!" #$%&'()*+,-./0 &'123456789: ;<=>7 ?>7@<A XML BCD&EFGH=)IH =JKLMNO=1&EFGH=P IQRS>7TU&VWX=Y Z [\]^_FG`abcd^e f1!)IH=g)IhR=)I ijkelmFG)In&Vop Jqrstj1234uIMNO =vw @<&EFGH=W)IH= mxy89: z{1|}89: 234/5 z{ ~_;<= 1 )Iijke XMLMNO=Metadata AbstractSupporting electronic catalogs (e-catalogs) to provide product-search services on the Internet is emerging as an important function of electronic brokers. However, the explosive growth of information is creating difficulties for customers to search for goods informa-tion they desire. In this work, an integrated framework is proposed to develop personal-ized e-catalogs. The proposed framework integrates XML-based metadata models, user models and agents. The proposed metadata model has a multi-level architecture that can facilitate resource-discovery and format
translation, and can flexibly model the defi-nitions of diverse attributes. The proposed user model uses three-level user profiles to model a user's shopping interests to resources, categories and goods. This work integrates the agent technology with the proposed metadata models and user models to develop a personalized e-catalog system. With the support of personalized services, the novel e-catalog system allows an integrated product search to be performed effectively.
KeywordsE-catalog, User Profile, XML,
E-Broker, Agent, Metadata
}W }W}W }W
Electronic brokers (e-brokers) provide various important brokerage services, such as merchandise filtering, suggestions, search, negotiation and electronic catalogs. The pri-mary function of an electronic catalog (e-catalog) system is to support Inter-net-based product searches. However, the explosive growth of information is creating difficulties for customers in searching for goods information they require. The char-acteristics of goods may vary in distinct product categories. Resources (virtual stores) may also employ different methods and data formats to query and store goods information. Relevant issues include the integration of heterogeneous goods information as well as interoperation among the search system and diverse resources.
Metadata [10,11] describes the character-istics of data that has been collected for a specific purpose. It can be used to facilitate the search service and integrate
heterogene-ous information. The extensible markup lan-guage (XML) [9,13] is designed to facilitate the interchange of structured documents over the Internet. Software agent technology [4,5] is well suited to search and integrate data from heterogeneous environments. Agents communicate through agent communication language (ACL)[3].
Keller et al. have proposed smart catalogs and virtual catalogs to support cross-catalog searches [6]. Lincke et al.[7] also presents an architecture for mediating electronic product catalogs. Recently, Web-based e-catalog sys-tems in B2B procurement have also been proposed [1].
The filtering of goods information can be realized by incorporating user profiles into search systems. Most studies have focused on text-based document filtering. Barra et al. [2] have proposed a vector of configurations in-dicating customer preference to buy product items from specific categories. However, their model is inadequate to model user needs.
The objective of this research is the fol-lowing. (1) Develop an XML-based metadata model. (2) Design a personalized merchan-dise information filtering and suggestion system on the basis of user shopping profile. (3) Apply agent technology to develop a personalized e-catalog system enabling users to conduct structured query and personalized product search and browsing on the Internet.
h34J h34Jh34J h34J
The research results of this work have been published in conference [8]. The main research results are summarized as follows.
(1) In this work, an integrated framework is proposed to develop personalized e-catalogs. The proposed framework integrates XML- based metadata models, user models and agents. The proposed metadata model is a multi-level architecture to describe separate levels of goods-related information. In the multi-level metadata model, XML is em-ployed as the meta-language to describe do-main-level metadata, resource-level metadata, category-level metadata, and goods-level
metadata. The metadata model can facilitate resource-discovery and format translation, and can flexibly model the definitions of di-verse attributes to describe goods in various product categories and data resources.
(2) User models are designed to support personalized filtering within e-catalogs. The proposed user model uses three-level user profiles to model shopping interests to a certain level of goods information, including preferred resources, categories and goods. The user profile records the user's preference (weight) to the characteristics (described by attribute and value) of goods information. To provide personalized filtering, via referenc-ing user profiles, a match process is designed to rank users' preferred product items in the search of the metadata model.
(3) On the basis of the proposed metadata model and user model, this work integrates the agent technology to develop a personal-ized e-catalog system. The proposed meta-data is used as the basis of the ontology (vo-cabulary). The proposed e-catalog system employs agents to achieve search results that satisfy the needs of customers. Based on user profiles, the novel system provides structured query by the attributes of selected product category as well as supports personalized product search and browsing for each user. The multi-level design flexibly models goods-related information and users' shop-ping interests
The major works of this research are summarized in the following.
3.1 XML-based metadata model
XML is used herein to define the pro-posed multi-level metadata framework.
Users with the same interest may form an Internet community. The domain-level meta- data has been designed to describe each community (domain). The resource-level metadata describes the sorts of product a vir-tual store or virvir-tual mall sells. Products that have common attributes are grouped within the same category. Category-level metadata is used to describe searchable attributes of each category and class relationship within
categories. Goods-level metadata provides detailed product information. To facilitate a search, product information formats of di-verse virtual stores must be translated to a unified format that is described by the goods-level metadata.
3.2 Personalized e-catalog design
The setting options of the user profile are based upon the metadata architecture. The information stored in user profile includes the basic information of each user, password, and preferred locations, virtual stores, and categories. In addition, we propose a three- level user profile that has a corresponding level to the merchandise metadata, to de-scribe a user's interest to a certain level of merchandise. Under an attribute-based de-scription, the profile describes users' shop-ping interests in detail. This implies that the profile records every weight of an attribute and its value of a certain item.
To separate the interests of a user into resource, category and goods level, the user model is therefore, a three-level profile. The resource-level user profile describes the user shopping interest to virtual stores when buy-ing items of a certain product category. In addition, the goods-level user profile can fil-ter out certain product items that the user may not be interested in. Category-level user profile is used to achieve a symmetric mod-eling and provide a more specific filtering to category-level goods information.
The system establishes a Directory In-formation Tree (DIT), through processing the Resource-level metadata and Category-level metadata. The system constructs a tree struc-tured virtual catalog by consulting the DIT. Once a user profile is established, the system will use the personalized information stored therein as filtering constraints to search the DIT and dynamically construct a personal-ized virtual catalog.
3.3 Agent-based e-catalog framework
The proposed brokering architecture con-sists of the user agent, the facilitator, and the
resource agent. The facilitator comprises five components, the control agent, broker agent, category model agent, virtual catalog agent and domain agent. The user agent connects the system to the client. The resource agent is responsible for retrieving data from a par-ticular resource. The control agent controls the query flow, gathers the result from re-source agents and communicates with other agent communities. In addition, the control agent cooperates with the broker agent for resource discovery and communicates with the category model agent to verify the capa-bilities of the probable resources. The broker agent checks resource-level metadata and category-level metadata to find possible data resources that may sell goods, which satisfy the request. The category model agent man-ages the capability of each resource and spe-cific (searchable) attributes of diverse prod-uct categories. The domain agent uses a do-main table to do-maintain dodo-main-related infor-mation regarding other agent communities. The virtual catalog agent maintains user pro-files to construct personalized virtual cata-logs. Furthermore, several functions on the agents were enhanced to support the person-alized filtering service. Personperson-alized con-straints including the resource, category and goods-level user profiles, are used by agents to cooperate in the ranking and filtering of product items.
We used JAVA and Voyager, a JAVA-based agent class by ObjectSpace Inc [13]. to develop the proposed system. We also developed an Applet in a Web browser to communicate with the user agent. The demonstration of the prototype system is de-scribed in [8].
@W @W@W @W
The proposed XML-based metadata can facilitate resource-discovery and format translation, and flexibly model the definitions of diverse attributes to describe goods in various product categories and data resources. Furthermore, in supporting personalized fil-tering within e-catalogs, the designed user model, with three-level user profiles, can
model users' shopping interests. With the multi-level design of both a user and mer-chandise model, the match process can be carried out efficiently to search and rank us-ers' preferred product items.
Moreover, with the integration of XML-based metadata and user models, the designed agents can communicate and coop-erate with other agents in resource discovery, format translation, ranking and filtering to achieve search results that satisfy customer needs. The novel system provides structured query via the attributes of selected product category, as well as supports personalized product search and browsing for each user based on their profiles.
We have accomplished 90% of the work described in the proposal. The research achievements include (1) proposing novel approach: A personalized e-catalog frame-work is proposed to satisfy diverse custom-ers’ preference. XML is used to describe the metadata to facilitate the interoperability between heterogeneous systems. Moreover, an effective merchandise filtering and sug-gestion technique is developed based on user shopping profile; (2) applying new technol-ogy: We used Java and agent technology to develop the system; and (3) deploying a pro-totype system.
The e-catalog system is a crucial function for e-brokers in Internet shopping. Our work will be a basis for further research on Internet commerce. Our work not only contributes to further research on Internet commerce but also contributes to the application of elec-tronic commerce. In summary, we have pro-posed novel idea, investigated new technol-ogy and developed a prototype system. The work has been published in the conference [8]. We are highly satisfied with the research achievements of the project.
[1] J. P. Baron, M. J. Shaw, A. D. Bailey, Jr., “Web-based E-catalog Systems in B2B Procure-ment”, Communications of the ACM, Vol. 43, No
5, May 2000.
[2] M. Barra, G. Cattaneo, M. Izzo, A. Negro, V. Scarano , "Symmetric Adaptive Customer Mod-eling for Electronic Commerce in a Distributed Environment", Proc. of Intl. IFIP Working
Con-ference: Trends in Electronic Commerce,
Ham-burg , Germany, June 1998.
[3] Tim Finin, Richard Fritzson, Don McKay, and McEntire, "KQML as an Agent Communication Language", Proc. of the Third Intl. Conf. on
In-formation and Knowledge Management (CIKM'94), Nov. 1994.
[4] Michael R. Genesereth and Steven P. Ketchpel, "Software Agents", Communications of the ACM, Vol. 37, No. 7, July 1994.
[5] Michael R. Genesereth, Narinder P. Singh, and Mustafa A. Syed, "A Distributed and Anonymous Knowledge Sharing Approach to Software Inter-operation", Logic Group, Computer Science De-partment, Stanford University, 1993, http://Logic.Stanford.edu /sharing/papers/fgcs.ps
[6] Arthur M. Keller and Micheal R. Genesereth, "Multi-Vendor Catalogs: Smart Catalogs and Vir-tual Catalogs", EDI FORUM: Journal of
Elec-tronic Commerce, Vol.9, No.3, Sept. 1996.
[7] D. M. Lincke and B. Schmid, "Mediating Elec-tronic Product Catalogs", Communications of the
ACM, Vol. 41, No 7, July 1998.
[8] Duen-Ren Liu, Y.-J. Lin, Y.-W. Hwang, C.-M. Chen, “A Framework for Personalized E-Catalogs: an Integration of XML-based Metadata, User Models and Agents”, Proc. of Hawai'i Intl.
Con-ference on System Sciences, Maui, Hawaii. Jan.
2001.
[9] Bart Meltzer and Robert Glushko, "XML and Electronic Commerce: Enabling the Network Economy", ACM SIGMOD Record, Vol. 27, No. 4, December 1998
[10] "Component part of ISO CD10303 - Exchange of Product Model Data", ISO, TC184/SC4/WG5 EXPRESS Language Reference Manual, April 1991
[11] "Directory Interchange Format (DIF) Writer's Guide, Version 5.0a", Jan. 1997, http://gcmd.gsfc. nasa. gocv.difguide/difman.html
[12] Voyager 2.0B2, ObjectSpace Inc., http://www. objectspace.com/products/voyager1.htm
[13] World Wide Web Consortium, "Extensible Markup Language (XML)", http://www.w3.org/ XML
