Discovery is an enhanced search engine for internal databases. Its main features include natural language query parsing, web browser client interface, and high-integration with existing Microsoft products. This last feature may be seen as good or bad depending on what Microsoft products are already in place. See System Requirements.
The server side of Discovery presumably polls the parts of a network and document storage areas specified. It then creates an index of the searchable documents which is presumably updated at specifiable intervals. The details were not disclosed on the website but inferences can be made. Discovery obviously works with existing collections of documents. The effort and authorization required to add documents to the collection is no more or less than it was before the system was installed.
Discovery is an enhanced search engine. A little categorization is applied to the results, but no major structural changes are applied to the organization of the knowledge before it is shown to the user. Once the knowledge is retrieved, the user can do nothing to change it or its organization. The user never has to know about the original structure of the knowledge (ie: the directory structure that holds the documents) so this simplifies the user's life by obviating the need to search through the documents by hand.
Discovery shows the knowledge it finds in a slight modification of standard web search engine results. The top ten results are at the top of the results frame, then come categorized results, then the remaining results. The title of a found document is listed, then a short description of the document, usually consisting of the first few sentences, then a link to retrieve the document and one to send it to another person.
This method of visualization is used in favor of a more complex, graphically oriented, display to ease learning and speed acceptance by new users.
Discovery's big claim to fame is its ability to parse natural language queries and discover concepts relevant to the searcher. From some tests, I found it to pick out the content of my queries quite well. Unfortunately it took 1.5 to 2 minutes to complete a search and return results. Discovery is very easy to use with it's text area for queries of arbitrary length. The methods it employs to generate the results are hidden from the user (eg. the website didn't get that specific).
The only collaborative use of Discovery is its little "Send this item to a Friend" link after each document found. This is only slightly better than nothing.
Discovery accesses currently existing collections of knowledge. It does allow a searcher to specify which collection(s) to search at query time, but it is not explicitly a tool for the administration of knowledge.
