Searching for 32 character plus strings in dtSearch

Note that the dtSearch index in Relativity is configured so that it will cutoff any words which are longer than 32 characters. For example, this string:

aeccd639e83d0540ba407a252a23e53d1cfe71

. . .is 38 characters long. It appears in the EDRM Enron email data set.

But if you run a dtSearch search to find it in the searchable extracted text of a workspace, you'll get no results.

Modify the search by simply searching for the first 32 characters:

aeccd639e83d0540ba407a252a23e53d

. . . and the document in which it appears will come up in the search results:

It's not necessary to add an asterisk for a wildcard search after the 32nd character.

Resource Scaling in RelativityOne

One of the advantages of RelativityOne is that it performs automated resource scaling - a server is directed to prioritize the completion of certain functions. So if a mass operation is run involving a lot of data, server resources can be increased, and re-directed from other jobs so the mass operation completes sooner. Automated scaling will be performed for:

1. Branding

2. Indexing

3. OCRing

4. PDF mass exports

5. Processing

6. Productions

Relativity provides some very specific information here, on how fast scaling will be implemented for different functions. When a Processing job takes longer than 10 minutes, and there are other jobs in progress, scaling should be initiated within 45 minutes.

If it takes more than 2 hours to run a branding, production, OCR, or mass PDF job, scaling will be performed.

Reviewing dupes for Continuous Active Learning in Relativity

When conducting active learning (CAL) in Relativity, keep in mind that you have a choice between whether or not the reviewers will be given documents which are textually similar to those which a coder has already determined to be responsive or non-responsive.

The Suppress Duplicate Documents setting should be set to Yes you are performing 'coverage review' in order to train the system as quickly as possible. In this model, the documents which the system is most uncertain about will be added to the review queue first - those with a rank around 50 on the scale from 0 to 100.

Prioritized review will prompt reviewers to focus on the highest ranking documents. When this type of review is performed the reviewers will code the highest ranking documents first.

Once a project has been started, the suppress duplicates setting cannot be changed.

The documents which are suppressed for a classification will be tagged in the field tree, so they can be reviewed earlier.