A "precedent" search is a search for a model document.
Generally, the key to finding a good precedent is knowing the context in which a document was previously used, rather than text in the document itself.
An example is: " a purchase and sale agreement for a retail shopping center in Florida". "Purchase and Sale Agreement" will be in the text of the document and the name of the document. But "Florida" and "retail shopping center" may not appear in the text of the document. If they do appear, they would appear infrequently.
They key to making a precedent search working is leveraging the document metadata against other systems. For instance, we require users to assign a document to a particular matter. We plan to use that matter identification to pull information from other sources and impute that information on the document.
The other key to a precedent search is using a faceted search to narrow the search results using the additional metadata.