Abstract de la publi numéro 11774

It is common for humans to identify some content by listing examples of similar content. Querying by examples is an alternative way of querying which allows to identify more content as well as to expand knowledge. We experiment this approach over a noisy collection of extracted lists from the Web. We focus on lists of named entities of the same type. We estimate to have about 892 million lists of this type in the actual indexed Web. This is a new interesting collection to test such an approach and results are meaningful. The size of the collection of lists is shown to be more important than the quality as querying by examples avoids wrong matches.