摘要 |
Structured and/or unstructured data about enterprises are acquired from one or more sources such as commercial data providers, enterprise web sites, and/or directory web sites. Strings are extracted from the unstructured data. The strings contain key, value pairs describing facts about the enterprises. The extracted strings are parsed to normalize the keys and values and place them in a machine-understandable structured representation. Some keys and/or values cannot be normalized. The facts are clustered with the enterprise to which they pertain. Normalized facts from different sources are compared and confidence levels and/or weights are assigned to the facts. These confidence levels and weights are used to select the facts that are displayed on a page for the enterprise in a directory.
|