Multiple Concepts per Category, but One Concept per Document

Note when collecting a set of documents for categorization analytics in Relativity while the group of documents may represent different aspects of a category, you should try to find individual documents that focus on a single concept. So while there can be one category for fraudulent communications and fraudulent accounting, each document should only discuss one concept. The individual document should fully explain the concept. Using a document that only addresses the concept in a few sentences is not sufficient.

The example documents in a categorization set should also not have sections of bad OCR or irrelevant footers.

A Typical Scenario For the Relativity Categorization Analytics Tool

The Relativity guide for analytics provides a typical scenario in which an admin would use a categorization set when faced with reviewing a raw document set for a legal dispute.

Categorization groups conceptually similar documents together and can assign the same document to more than one category.

A reasonable approach to categorization would involve the following steps:

1. If you have around 10M documents, a keyword search could narrow it down about 70% to 3 million.

2. Create a saved search with the narrowed down data set.

2. Ascertain the key concepts involved in the dispute.

3. Assemble a set of example documents that can be used to identify other documents that are representative of the same concepts.

4. Create a categorization set, with the saved search selected in the 'Documents to Be Categorized' field.

5. Create an analytics index for the saved search.

6. Select the index for the categorization set.

7. The other fields on the layout for the categorization set can be left at their default settings.

8. Create analytics categories for each of the key concepts involved in the dispute. These are added to the Analytics Category object in the section entitled, 'Analytics Category'.

9. Select documents which reference the concepts for each of the categories. These are added to the Analytics Example object in the section at the end of the layout.

10. Click 'Categorize All Documents' on the console for the layout.

Tags will be created for the categorized documents in the field tree.

words with numbers omitted from Relativity analytics index

When Relativity populates an analytics index it will not include any words which begin with a number. So for example, 7zip or 4ever, will omitted from an analytics index.

On the other hand, words which end with a number (such as gr8) or words which include a number in the middle (rome2rio) will be indexed without being modified.