Over the last two weeks K-Praxis has been focusing on the using unstructured information management to derive intelligent actionable decision support out of commercial information (please see:Information Classification Standards: An Overview and Supply and Procurement Intelligence: Analytics from Unstructured and Semi-structured Information). The idea that there is a huge of amount of textual data around various e-business commerce that remains untapped and needs to be exploilted to squeeze out more commercial gains, also applies to product information and electronic catalog management.
An Anatomy of a Typical Product Information and Catalog Management Solution
So what are the key ingredients of a product information management or an electronic catalog management solution? These solutions are trying to solve a well-defined problem (which became a classic catch-phrase in the pre-Ecommerce-boom-&-bust days (remember those over-quoted 3 C's of e-commerce: content, commerce and collaboration!). A complete solution in this field should be able to help companies capture, create, assemble, manage, leverage and publish (to multiple media: Web, print, PDF, CD-ROM and wireless devices) rich and personalized product/catalog information for use through their various supply, procurement and e-business channels. Information on a company's products are often stored on many disparate locations: word processing files or publishing files, websites and databases. This various formats of information could send different and confusing signals to various e-business channel patterns - both inward-looking as well as outward-looking channels and customers. Product information management solutions fuse all the product information related to a specific product is kept consistent across all information channels.
Vendors Offering Product Information and Cataloging Solutions
Arbortext;Ascential;ElCom; Enigma; Enterworks; Eprise; Intigma; Mercada; Object Publisher; Percussion; Pindar;Poet;Requisite;Saqqara;Stibo Catalog.
Out of these, solutions provided by Ascential, Poet, Requisite, and Stibo Catalog seem to be much more complete and advanced.
Product Information Management and Unstructured Data Management/Exploitation
Given the anatomy of a classic product and cataloging solution, linguistics and machine learning based unstructured information management technologies can provide solutions which can radically increase the revenue generating potential and re-usability of product information and electronic catalogs. The following are a few key elements of an advanced automated solution:
Automated Product Information Extraction: this involves extracting both structured and unstructured product content from various e-business channels (e.g., short newswires, form based data, and textual data) by using language processing (both statistical and grammar-driven) and machine learning techniques and automatic translation of unstructured information into structured product information
Auto-classification of product information: This involves classifying the product information by using variety of machine learning and AI-based methods.
Reclassification, personalization and discovery : Reclassifying helps personalize product information and also adds the layers of retrievalbility and discoverability to product information
Intelligently Integrating product information into E-Business channels: This involves intelligent and automated syndication of product information and electronic catalogs through all the ebusiness channels and ebusiness applications, making it available in real-time to any body who needs it.
Business Analytics through Product Information: This involves deriving decision support and actionable information items from the classified and organized product information.
Vendors providing next generation product information and catalog management solutions
A2i ; Cardonet; Softface; Zycus.
Out of these A2i uses Inxight's language processing technology to help in retrieving, classifying and leveraging textual data that exists inside catalogs. Cardonet uses a unique rule-based technology to integrate product from demand and supply chains focusing on reducing costs and improving through automation extraction and retrieval. Zycus and Softface use Bayesian classification, rule-based methods to classify and organize product information, Softface also uses syntactic natural language processing to classify and tag sentences in product data.
Our analysis shows that most of the next generation product information management solutions deploy various automated methods to aggregate, cleans, normalize, classify and syndicate product information. Although companies like Zycus and Softface are using some more advanced AI-based technologies to leverage unstructured information in product catalogs, there is a room to achieve much more by using various other technologies like latent semantic processing or Hidden Markov Models or more advanced ontology based language processing, etc. Besides offering cutting edge technology vendors will need to evangelize, explain and proselytize the uses and benefits of such technologies to customers and help them move to a grater degree of automation in product information management.