Review validation

Review validation evaluates the accuracy of a Review Center queue. The goal of validation is to estimate the accuracy and completeness of your relevant document set if you were to stop the queue immediately and not produce any unreviewed documents. The primary statistic, elusion rate, estimates how many uncoded documents are actually relevant documents that you would leave behind if you stopped the queue. The other statistics give further information about the state of the queue.

For a detailed explanation of how the validation statistics are calculated, see Review validation statistics.

Note: Review validation does not check for human error. We recommend that you conduct your own quality checks to make sure reviewers are coding consistently.

Key definitions

The following definitions are useful for understanding review validation:

Discard pile—the set of documents that are uncoded, skipped, or coded as neutral. This also includes documents that are being reviewed when validation starts, but their coding has not been saved yet.
Already-coded documents—documents that have already been coded as either positive or negative. These are counted as part of the validation process, but they will not be served up to reviewers a second time. Neutral-coded documents are considered part of the discard pile instead, and those may be served up a second time.

Determining when to validate a Prioritized Review queue

When a Prioritized Review queue is nearing completion, it can become more difficult to find additional relevant documents. As you monitor your queue, the following dashboard charts can help you determine when the queue is ready for validation:

Rank Distribution—look for few or no unreviewed documents with a rank of 50 or higher.
Relevance Rate—you should see a decline in the relevance rate progress line indicating that very few responsive documents are being found.

When you believe you have found most of the relevant documents, run validation to estimate the accuracy and completeness of your relevant document set.

For more information on the dashboard charts, see Charts and tables.

Starting a validation queue

When you are ready to validate your progress in a Review Center queue, you can start a linked validation queue that samples documents from the discard pile and serves them to reviewers.

To set up the validation queue:

From the Review Center tab, click on the queue you want to validate.
Pause the queue.
- If auto-refresh is turned on, turn it off.
- If the queue is in the middle of refreshing, wait until the refresh has finished before starting validation.
On the right side of the Queue Summary section, click on the three-dot menu and select Set up Validation.

An options modal appears.
In the options modal, set the following:
1. Validation Reviewer Groups—the user groups you want reviewing the queue.
2. Set the sample size using three interconnected fields:
  1. Sample Size—this sets a fixed number of documents for the sample size. By default, this field is set to 1000 documents. The sample size must be larger than 5 and smaller than the size of the discard pile.
  2. Margin of Error Estimate (Elusion)—this calculates a size for the sample based on how accurate the Elusion statistic will be.
  3. Margin of Error Estimate (Recall)—this calculates a size for the sample based on how accurate the Recall statistic will be.
    
    Note: Each of these fields affects the others. For an explanation of how they work, see Choosing the validation settings.
Click Save.

Choosing the validation settings

Validation always samples a specific number of documents, but there are three ways to choose the sample size:

You can specify exactly how many documents you want to sample. Review Center automatically calculates the estimated margins of error for both Elusion and Recall based on the sample size you select.

Note: This is equivalent to choosing the “fixed” option when configuring an Elusion with Recall test in Active Learning. In contrast with an Active Learning Elusion with Recall test, Review Center only samples the discard pile. This means that the sample size is also the number of documents that will need to be coded.
You can specify the desired margin of error for the elusion estimate and let Review Center calculate an appropriate sample size. Review Center also automatically calculates the corresponding recall margin of error.

Note: This is equivalent to the “statistical” option when configuring an Elusion with Recall test in Active Learning.
You can specify the desired margin of error for the recall estimate and let Review Center calculate an appropriate sample size. Review Center also automatically calculates the corresponding elusion margin of error.

The final margin of error estimates may be slightly different from the ones chosen at setup, depending on the documents found during validation. All validation statistics aim for a 95% confidence interval alongside the margin of error.

The estimated elusion margin of error depends only on the sample size, and vice versa. Their relationship to the estimated recall margin of error depends on the number of relevant documents that have already been coded and the current size of the discard pile. It may vary among different validation samples, even within the same review.

For more information on how validation statistics are calculated, see Review validation statistics.

Inherited settings

Each validation queue inherits these settings from the main queue:

Queue Display Options
Reviewer Document View
Email Notification Recipients

To change them, edit the validation queue after creating it. For more information, see Editing a validation queue.

Monitoring a validation queue

Validation statistics are reported on the Review Center dashboard like any other queue. You can cancel validation from the three-dot menu, and you can also pause validation by clicking the Pause button. The validation statistics are reported in the Validation Progress section, and the charts and tables reflect the validation queue.

For reviewers, they access the validation queue from the Review Queues tab like all other queues. Reviewers code documents from the sample until all documents have been served up.

Note: For best results, we strongly recommend coding every document in the validation queue as positive or negative. Avoid skipping documents or coding them as neutral. For more information, see How validation handles skipped and neutral documents.

Editing a validation queue

You can change some of the queue settings at any time during validation.

To edit the validation queue:

On the right side of the Queue Summary section, click on the three-dot menu and select Edit.
Edit any of the following settings:
- Reviewer Groups
- Queue Display Options
- Reviewer Document View
- Email Notification Recipients
Click Save.

For descriptions of the queue settings, see Creating a Review Center queue.

Releasing unreviewed documents

If a reviewer falls inactive and does not review the last few documents in a validation queue, you can release those documents through the Queue Summary section of the dashboard. For more information, see Editing queues and other actions.

To see which documents are checked out to a reviewer, filter the Reviewed Documents table by the reviewer's name. Any documents that are still checked out will show the Coded Time as blank. For more information, see Reviewed Documents table.

Tracking sampled documents

If you want to run your own calculations or view documents in the validation sample, you can track the sampled documents from the Document list page. This process is optional.

To view sampled documents:

From the Documents tab, click on the Field Tree () icon.
Expand the Review Center folder.
Expand the folder for the queue you're validating.

Several subfolders appear.
Expand the folder titled [Queue Name] - Validation [Current Round Number]. If you have only run validation one time, the round number will be 1.

Each validation folder contains the documents selected for the sample. It also holds two sub-choices: one for documents coded positive or negative, and one for skipped or neutral documents. As documents are coded, they populate under these choices.

Viewing sampled documents through the Field Tree

Accepting or rejecting validation results

After all documents in the validation queue have been reviewed, a ribbon appears underneath the Queue Summary section. This ribbon has two buttons: one to accept the validation results, and one to reject them.

If you click Accept:

The queue status changes to Validation Complete.
The model remains frozen. Any future coding decisions will no longer be used to train the model, and the Review Progress statistics will not reflect any new coding.
The Validation Progress strip on the dashboard displays the final validation statistics.

If you click Reject:

The validation queue status changes to Rejected, and the main review queue changes to Paused.
The main review queue re-opens for normal coding, and you can build the model again at any time. Any documents coded since validation began, including those from the validation queue itself, will be included in the model build.
The Coding Progress strip on the dashboard displays the main queue's statistics.

You can run validation on the queue again at any later time, and you can reject validation rounds as many times as needed. Even if you reject the results, Review Center keeps a record of them. For more information, see Viewing results for previous validation queues.

Manually rejecting validation results

If you change your mind after accepting the validation results, you can still reject them manually.

To reject the results after accepting them:

On the right side of the Queue Summary section, click on the three-dot menu and select Reject Validation.
Click Reject.

After you have rejected the validation results, you can resume normal reviews in the main queue.

Reviewing validation results

After reviewers code all documents in the sample, the queue status changes to Complete. All validation results appear in the Validation Progress strip on the Review Center dashboard.

The results include:

Relevance Rate—percentage of sampled documents that were coded relevant by reviewers, out of all coded documents in the sample. If any documents were coded as neutral, this statistic also counts them as relevant.
Elusion Rate—the percentage of not-yet-coded documents that are estimated to be relevant. The range listed below it applies the margin of error to the sample elusion rate, which is an estimate of the discard pile elusion rate.
Note: Documents that are skipped or coded neutral in the validation queue are treated as relevant documents when calculating Elusion Rate. Therefore, coding all documents in the elusion sample as positive or negative guarantees the statistical validity of the calculated elusion rate as an estimate of the entire discard-pile elusion rate.
Eluded Documents—the estimated number of relevant documents that have not been found. This is calculated by multiplying the sample elusion rate by the number of documents in the discard pile. The range listed below it applies the margin of error to the document count.
Recall—percentage of documents that were coded relevant out of the total number of relevant documents, both coded and uncoded. The range listed below it applies the margin of error to the percentage.
Richness—the percentage of relevant documents across the entire review queue. The range listed below it applies the margin of error to the percentage.

For more information about how these statistics are calculated, see Review validation statistics.

Recalculating validation results

If you have re-coded any documents from the validation sample, you can recalculate the results without having to re-run validation. For example, if reviewers had initially skipped documents in the sample or coded them as neutral, you can re-code those documents outside the queue, then recalculate the validation results to include the new coding decisions.

To recalculate validation results:

On the right side of the Queue Summary section, click on the three-dot menu and select Recalculate Validation.
Click Recalculate.

Viewing results for previous validation queues

After you have run validation on a queue, you can switch back and forth between viewing the statistics for the main queue and any linked validation queues that were completed or rejected. Viewing the statistics for linked queues does not affect which queue is active or interrupt reviewers.

To view linked queues:

Click the triangle symbol near the right side of the Queue Summary section.

A drop-down menu listing all linked queues appears.
Select the queue whose stats you want to view.

When you're done viewing the linked queue's stats, you can use the same drop-down menu to select the main queue or other linked queues.

How adding or changing documents affects validation

Typically, review validation is linear: The administrator sets up the validation sample, the reviewers code the sample, and the results are calculated from those documents. However, if documents are added or removed, coded documents are re-coded, or other things happen to change the queue being validated, this can affect the validity of the results.

Scenarios that require recalculation

The following scenarios can be fixed by recalculating statistics:

Changing coding decisions on documents within the validation sample
Changing already-coded documents outside the sample from positive to negative or negative to positive
Adding already-coded documents to the queue after validation starts

In these cases, the sample itself is still valid, but the numbers have changed. For these situations, recalculate the validation results to see accurate statistics.

For instructions on how to recalculate results, see Recalculating validation results.

Scenarios that require a new validation queue

The following scenarios cannot be fixed by recalculation:

Adding uncoded or neutral documents to the queue after validation starts
Changing positive- or negative-coded documents outside the sample to skipped or neutral

In both of these cases, this means that the validation sample is no longer a random sample of all uncoded or neutral documents. For these situations, we recommend starting a new validation queue.