Our matching and recommendation engine combined a generic classification of content (articles, audio, video, etc.) according to a cross-language ontology and a fine-grained semantic analysis to detect relevant entities (people, places, organizations, concepts, products, brands, etc.). Documents were all classified and annotated so that their proximities could be determined in a complete, tailor-made “knowledge” graph. Each visitor profile could then be enriched based on its themes of interest, in order to provide them with customized content


After a setup phase, the recommendation engine went live a few weeks later. The immediate results let us directly measure ROI by the increase in visitor participation

Increase in visit length and number of page views
Decrease in bounce rate