Number of Examples to Use for Categorization

When preparing a categorization set in Relativity, don't be reluctant to use a lot of examples. Relativity suggests that a minimum of 5 to 20 documents be used for any one category, and that as many as 15,000 documents can be effectively used as examples for a single category.

There's no actual limit to the number of documents which can be used as examples for a category, but greater numbers of examples will slow down the categorization process. Relativity provides a general rule of thumb by noting that a workspace of a few million documents may require 2,000 example documents.

Good examples for categorization

When adding examples to use for categorization in Relativity be sure to note that certain kinds of documents will NOT make good examples:

1. Each document should be at least as long as a few paragraphs.

2. An example document should focus on a particular concept.

3. The selection of a document as an example should not be based on the document's metadata.

4. Categorization will not consider email headers or repeated content.

5. Don't use documents with a lot of numbers such as spreadsheets.

6. Don't assign duplicate or similar documents as examples for a category.

7. Don't use the same document for more than one category.

Don't Rename Category Names

When working with categorization in a Relativity workspace, be sure to never change the name of a manually added category after the categorization process has been run.

If this is done, errors may result when the category is used in a search.