Grenoble, France - 02 September 2005 - Kelkoo, Europe's largest shopping search engine and a wholly owned subsidiary of Yahoo! Inc., has chosen technology from the French editor Sight'Up after extensive benchmarking and analysis of text mining products from an international selection of companies. Using Taggis® artificial intelligence from Sight'Up, Kelkoo will benefit from a product characteristic extraction system that is unique in Europe. Kelkoo handles over 100 million products in 9 languages in 10 countries every day.
As the leading shopping search engine in Europe, Kelkoo faces a similar challenge as all other search engines: how to achieve 100% precise results with a search system that is simple to implement and maintain. The challenge is all the more complex in the e-commerce sector where databases containing product information are not necessarily organised in a structured manner (unlike information on patents, for example, where the data is normally highly structured).
Jean-Marc Potdevin, Kelkoo V.P. Engineering Europe explains, "Precision and relevancy are of paramount importance to Kelkoo. We have successfully maintained our market leadership through continued investment in our technology. After testing several solutions, we chose Sight'Up technology because it meets our needs perfectly. Unlike other technologies, Taggis uses self-learning and does not rely on management rules or dictionaries. Sight'Up tools will helps us match the 100s of millions of products we provide prices and product details for cover with precision and keep our results up-to date as soon as new products become available."
Taggis is a unique patented technology, that offers Kelkoo the following advantages:
- Simple operation based on a unique self-learning system: No more need to create management rules or dictionaries in order to handle searches within product categories. This makes maintenance extremely simple: when a new product category comes along, the learning mechanism only needs a few samples to be updated. This allows Kelkoo to adjust product categories easily (for example adding bathing suits as summer gets closer), and above all to enhance its coverage of new categories, such as new lines of clothing and accessories.
- Language independence:Given the international coverage of the Kelkoo group, with presence in 10 countries using 9 different languages, a language independent solution is critical. Based on a self-learning system, Sight'Up technology can be used with any language and any alphabet. Taggis was tested successfully with some of the main Asian languages recently, including Chinese, Japanese, and Vietnamese.
- Maintenance autonomy for product category managers: The new system lets product category managers update and maintain the search system themselves, instead of having to get help from technical teams to create rules and dictionaries. This gives them better and faster control over their offers.
Above all, no computer knowledge is required to use Taggis. In designing its system, Sight'Up made sure that people do not need expertise in computers to handle it easily. All users need is some basic knowledge about the product families they have to manage, and some common sense.
- Search results with precision higher than 95%: Consider this product example: "a red silk dress with a black belt."
Traditional search engines will take the above text and display results on "black dress", whereas Taggis only provides results that actually correspond to the on-line shopper's request for a special type of red dress.
The goal is for machines to be able to reproduce the behaviour of in-store salespeople when they try to find goods that are as close as possible to a customer's request. With Taggis, artificial intelligence is really at work.
- 1 to 2 million documents per hour, while other technologies reach their limit at a few tens of thousands.
