CommerceNet Smart Catalogs

                       Arthur M. Keller

                 Center for Information Technology

                      Stanford University

        URL: http://cit.stanford.edu/cit/commercenet.html

1. Current Approaches in Deployment
2. Online Catalogs
3. Online Catalogs Requirements
4. Smart Catalog Architecture
5. Smart Catalog Approach
6. Integrating Smart Catalogs with Distributors and Brokers
7. Benefits of Virtual Catalogs
8. How does all this work?
9. Agent Technology
10. Agent Roles
11. Facilitator
12. Ontologies
13. Conclusion

Online Catalogs

Current Approaches in Deployment:

• Use ordinary WWW protocols and clients such as Mosaic

• Navigation through documents that behave like menus

• Each vendor maintains its own (collection of) catalogs

Limitations of Current Approaches in Deployment:

• Hard to find what you want

• Navigation through many menus

• Each catalog is structured differently

• Limited or no capability for reverse-search (i.e., search by content)

• No capability for cross-catalog search

Online Catalogs

Examples:

• Give me the specs for Hewlett-Packard's DeskJet 1200C/PS printer.

• Give me a list of Hewlett-Packard's color PostScript printers for the Macintosh.

• Give me a list of dealers for the HP DeskJet 1200C/PS printer.

• Give me a list of color PostScript printers for the Macintosh made by any manufacturer and with list price under $3000.

• Notify me whenever anyone announces a new color PostScript printer for the Macintosh under $3000.

Technical Needs:

• Representation (specification, price, availability, etc.)

• Smart search (knowledge bases to set up search, efficiency techniques)

• Human computer interaction (natural language, forms, etc.)

Online Catalogs

Requirements (from CommerceNet Catalogs Working Group):

• Scalable, distributed search

• Heterogeneous information sources

• Up-to-date information

• Open architecture

• Variety of search techniques

• Cross-search of multiple catalogs

Online Catalog Architecture

Smart Catalog Approach

Supports reverse search: search by content

Supports translation among heterogeneous catalogs

Product information maintained up-to-date in database

Can support a variety of search techniques and interfaces

Supports cross-search of multiple catalogs

Integrating Smart Catalogs with Distributors and Brokers

Benefits of Virtual Catalogs

Allows extraction of product data by distributor from manufacturer catalogs

• Dynamically

• Upon user request

• Based on user's search criteria

• Using terminology of distributor or any connected vendor

• Retrieving data from any relevant connected vendor

Distributor maintains control over interaction with user

• User never interacts directly with manufacturer catalog

• Virtual smart catalog can have distributor's look and feel instead of that of each manufacturer

Distributor's catalog always kept up-to-date and in sync with each manufacturer's catalog

Distributor can choose to display all of a manufacturer's products or only a subset

Once product is chosen, can link to EDI process for negotiation of terms and conditions, order processing, and order fulfillment

• Between purchaser and distributor

• Can trigger resupply EDI transactions between distributor and manufacturer

• For large orders, purchaser's request can be given to manufacturer for drop shipment direct to purchaser

• Can handle case where distributor stocks item, and case where item is obtained only upon demand

How does all this work?

Agent Technology!

Agent Technology

Use of ACL (Agent Communication Language) = KQML + KIF + Ontologies

• KQML (Knowledge Query and Manipulation Language) gives the "performatives" or "verbs"

• KIF (Knowledge Interchange Framework) gives the "grammar"

• Ontologies (domains of discourse) give the "terminology" or "nouns" and "adjectives"

Catalog agents translate between ACL and the language of each catalog, e.g., SQL

User agents translate between ACL and the language of each user, e.g., HTTP and HTML

Agent Roles

Information source (catalog) agents' roles

• Advertise capabilities to facilitators

• Support protocols of information source (e.g., catalog)

• Interpret inquiries and implement on catalog

• Package responses into ACL

User agents' roles

• Support protocols of client software

• For WWW (HTTP and HTML) protocol viewers such as Mosaic
  • Support user interface model by creating appropriate query forms

  • Translate queries to ACL

  • Translate responses to dynamically generated HTML documents

Facilitator

Facilitators' roles

• Broker to determine which (catalog) agents are have relevant information for a given inquiry

• Translate queries into the terminology understood by each responding agent

• Translate responses into common terminology for comparison, or the terminology used by the requestor

• Decompose queries to be handled by multiple agents, if necessary

Scalability

• Multiple facilitators, each with its own domain of interest

• Facilitators communicate with each other so inquiries, etc., are directed to the appropriate facilitators and their agents

Ontologies

Base ontologies

• Terms and definitions of general utility
  • Examples: engineering math, standard business terminology, dates, standard legal terminology

  • Shared among all uses of this approach

  • Created by universities and research laboratories

Domain ontologies

• Terms and definitions relevant to a particular class of products
  • Examples: CPU speed, RAM, disk storage capacity

  • Shared among multiple vendors in that class

  • Created under the auspices of trade associations and standards bodies

Product ontologies

• Terms and definitions relevant to a particular vendor*s products and related products from other vendors
  • Examples: NuBus card, TSR, EISA

  • Created by one vendor

  • May be referred to by other vendors

Translation ontologies

• Translations between terms in multiple ontologies
  • Examples: How SVGA monitors can be used on Macintoshes

  • Created by a vendor who wants to market products in other areas

Conclusion

We have developed a smart catalog and brokering architecture.

We are implementing this architecture generically.

We are testing this architecture and implementation in at least 3 smart catalog pilots.

We are soliciting participants in pilots demonstrating distributors or brokers integrated with smart catalogs.

Our technology for the smart catalog and brokering architecture is based on its work on the ARPA Knowledge Sharing Initiative.