Download presentation
Presentation is loading. Please wait.
Published byJody Gray Modified over 8 years ago
1
Database Technologies for E-Commerce Rakesh Agrawal IBM Almaden Research Center
2
Outline Overview of the Pangea B2B marketplace Technology dives: Interactive parametric search Catalog integration Storage & retrieval of e-commerce data Research opportunities
3
Pangea Architecture
4
Product Selection
5
Brand and Loyalty Creation
6
Features (1) Eureka: Interactive parametric search Searches based on example products Similarity search for product substitution Dynamic Content generation/presentation (on the fly from database): Catalog Hierarchy pages with product counts Side-by-Side Product Comparison Product Listings/Details Shopper groups through preference layer
7
Features (2) Server side Searching: Functional description search using classification Key word & Part Number Search Restriction of the search to sub-trees of the category hierarchy (pushed down to the database) Real time Price and Availability Engine: Ability to crawl, interrogate, extract price & availability data from various suppliers/distributors in real time and present them in side-by-side comparison format Easily add new distributors/suppliers Completely client side implementation to prevent blocking
8
Features (3) Design Worksheet (Generalized shopping cart): List of items: an item could be a specific part, alternative set of parts, specifications, other design worksheets (nesting) Completely integrated with all relevant components (search, content, price, etc.) Aggregate constraints (e.g. total price) Multiple projects saved on server Share projects with other (authorized) users Mining for suggestions
9
Features (4) Design data warehouse creation and maintenance Crawling technology for extracting part information from websites of various manufacturer/distributor/suppliers Data extraction from Manufacturer Spec Sheets (pdf files) Classification technology to build, merge, manage and populate category hierarchies.
10
Features (5) Soft content creation and maintenance Crawling to acquire articles/news/postings from various web news sources, usenet newsgroups, etc. Agents to cluster, classify, extract concepts, identify associations. Personalized news/data presentation (based on user defined profile, channels, etc.) Complete interleaving and integration with part content. Automatic generation of summary pages for manufacturers (e.g. related news articles, URLs, product categories).
11
Prototype Data for nearly 2 million parts, in 2000 categories Focused crawling of more than 175 manufacturers for datasheets/ application notes/manuals (160,000) 1/4 Terabyte database
Similar presentations
© 2024 SlidePlayer.com. Inc.
All rights reserved.