Overview
Search engines in the field of e-commerce have been managed for years in two main ways ; using a linguistic or statistical approach. In spite of the undeniable contribution of these two technologies, they are beginning to outlive their usefulness.
A technological breakthrough was necessary. Sight’Up’s scientific team started looking for a new approach ; a third way, which would combine precision of results, low labour costs and flexibility of use. Such was the task ahead of us.
One of the longest-standing challenges in the field of artificial intelligence and the approach to "machine-learning" has been : how to obtain results close to 100% reliability while basing machine-learning on a small number of samples.
We found the way to do it …
Sight’Up uses the genetic behaviour of the documents. Each text is first transformed into a chromosomal chain which is then processed by a set of specific algorithms.
This approach makes it possible to deal with extremely complex problems, such as the extraction of “product” characteristics for e-commerce, with very little human intervention.
Sight’Up’s scientific breakthrough has the following characteristics :
- Average precision often close to 99% with a Recall of more than 95%.
- Independent of language ; can be used in Romance, Asian or other language groups.
- Independent of product families or documents, we can manage clothes as well as data-processing products, e-mails, SMS etc…
- Human intervention limited in volume and in levels of skills required ; very small learning corpus to be labelled ; no knowledge of data-processing required.
- Processing capacity : one to two million documents per hour.
These years of research and effort were rewarded when Kelkoo, European leader in e-shopping engines and a subsidiary of Yahoo, chose our patented technology. After a period of tender and benchmarking lasting more than nine months and from a worldwide selection of candidat text-mining engines, Kelkoo finally chose Taggis : Sight’Up’s characteristic extraction engine.
