RelOne PERSONAL INFORMATION DETECT Certification Flashcards
1
Q
- What is Personal Information (PI) Detect?
a. PI Detect is Relativity’s integrated solution for contract
review.
b. PI Detect is an AI-powered solution that identifies and
redacts personal information.
c. PI Detect is an AI-powered solution used to reduce the
time, cost, and risk to produce an entity notification list.
A
?
2
Q
- What can PI Detect’s pre-trained detectors and machine
learning models help with?
a. Reducing false positives when identifying personal
information.
b. Deduplicating entities across documents.
c. Collecting data for personal information analysis
A
?
3
Q
- What unit is billed for PI Detect?
a. Number of PI detections predicted.
b. Number of documents ingested.
c. Gigabytes of data in PI Detect per month.
A
?
4
Q
- What is an end product of PI Detect?
a. Automatically-redacted personal information.
b. An automatically-generated entity notification list.
c. Automated amendment or notice generation across all
your agreements
A
?
5
Q
- What information is found in the Document Report?
a. An entity notification list.
b. A saved search.
c. A list of all ingested documents with identified personal
information.
A
?
6
Q
- What other Relativity application does PI Detect leverage for
its end product?
a. Contracts
b. Redact
c. Legal Hold
A
?
7
Q
- What format is PI Detect available in?
a. A Relativity application (RAP) available for Server.
b. A Relativity application (RAP) available for RelativityOne.
c. A Relativity application (RAP) available for RelativityOne
and Server.
A
?
8
Q
- How can you try out this product with a RelativityOne Flex
Commit subscription?
a. Use some of your free tier of documents according to
your Flex Commit level.
b. Use one of your free Proofs of Concept (POCs) included
with Flex Commit plans.
c. Use less features to reduce billing amount
A
?
9
Q
- Which method improves the precision of PI predictions?
a. Reviewing all documents that do not have predicted
personal information.
b. Using façade redactions.
c. Re-running PI detectors after reviewers make changes.
A
?
10
Q
- What method does PI Detect use to understand the context
and nuance in unstructured data to provide more accurate
results and remove false positives?
a. TAR 2.0
b. A combination of machine learning and natural language
processing
c. Topic modeling
A
?
11
Q
- What documents can PI Detect support?
a. Unprocessed PST files
b. Microsoft documents only
c. PDFs and spreadsheets
A
?
12
Q
- When reviewing spreadsheets, in what order should
reviewers validate the AI predicted table boundaries and
personal information?
a. Table boundaries and table columns with predicted
personal information can be reviewed in any order.
b. First review table columns with predicted personal
information, then review table boundaries.
c. First review table boundaries, then review table columns
with predicted personal information.
A
?
13
Q
- What is the goal of reviewers during non-spreadsheet
review?
a. To validate PI predictions created by the machine
learning pipeline.
b. To QC annotations made by other reviewers.
c. Non-spreadsheets do not need to be manually reviewed.
A
?
14
Q
- Where would you review PI results in the Document Viewer?
a. The PI columns in the Document Reports tab.
b. The cards available in the PI Detection panel or highlights
on the document itself in the Native, Extracted Text, or
PDF view.
c. The cards available in the Persistent Highlight Sets panel
or highlights on the Extracted Text view of the document.
A
?
15
Q
- If a document contains poor OCR or handwriting, how would
you capture personal information in the Viewer?
a. Flag the document as a technical issue, add the personal
information in comments, re-run OCR, and re-ingest the
document.
b. Highlight the extracted text to capture it as displayed,
manually enter the value, and then select the PI type.
c. Click Draw Annotation in the PI Detection panel, draw
a box over the personal information, manually enter the
value, and then select the PI type.
A
?
16
Q
- In a PI and Entity Search, what is a shortcut to search for
all documents that contain at least one piece of personal
information?
a. CONTAINS PI
b. ‘PI Count’ > 0
c. CONTAINS ENTITY
A
?
17
Q
- What syntax should you use if you want to search for all
documents that contain Social Security Number and Phone
Number PI types?
a. ‘PI Types’ CONTAINS ‘Social Security Number’ AND
‘Phone Number’
b. ‘PI Types’ CONTAINS ‘Social Security Number’ AND ‘PI
Types’ CONTAINS ‘Phone Number’
c. ‘PI Types’ CONTAINS [‘Social Security Number’, ‘Phone
Number’]
A
?
18
Q
- What is a PI detector?
a. A combination of AI models, regular expressions, and
keywords that detect a string of text and classify it as a
form of personal information.
b. A combination of keywords that detect a string of text and
tag the document as containing the text. It is also referred
to as a document category.
c. A label reviewers can apply manually to a document for
classification purposes
A
?
18
Q
- Your document set contains documents that have low OCR
quality. How would you search for these documents?
a. ‘Unable to map PI to viewer’
b. ‘Document Flags’ CONTAINS ‘Unable to map PI to viewer’
c. ‘Unable to map PI to viewer’ = TRUE
A
?
19
Q
- During a project, when is the best time to configure PI
detectors?
a. At the beginning of a project and iteratively throughout
review.
b. Before the first run of Incorporate Feedback.
c. Before manual review begins.
A
?
20
Q
- A company-specific employee ID appears in your data set
and is not being captured. What is the best way to ensure
this information is being captured?
a. Edit an out of the box detector to capture employee IDs.
b. Create a custom employee ID detector.
c. Create a tag and manually capture each instance of an
employee ID when reviewing documents.
A
?
21
Q
- What RelativityOne feature does PI Detect use to mark
redactions on documents?
a. Façade redactions
b. Full page redactions
c. Blackout
A
?
22
Q
- Where can redactions be applied within PI Detect?
a. Façade redactions tab
b. Document report tab
c. Document list page
A
?
23
Q
- What mass operation do you use to redact documents with PI
Detect?
a. PI Detect Redact
b. Convert Spreadsheet Markups
c. Prepare to Redact
A
?
24
26. How long does it take to run a report for PI Detect?
a. One minute per document.
b. It is dependent on how large the dataset is.
c. Reports will take at least six hours to generate
?
25
25. What must be configured before redacting with PI Detect?
a. A Markup Set and a Processing Set.
b. A Saved Search and a Processing Set.
c. A Markup Set and a Markup Type
?
26
27. What is the purpose of the Document Report?
a. It provides detailed insights on the document review status
and PI types detected within the data.
b. It lists documents in the dataset predicted to have personal
information and whether the document is responsive.
c. It links entities found within the dataset to their
corresponding personal information.
?
27
29. What is the most efficient way to track reviewers’ progress
during document review?
a. Filter the Document Report to all reviewed documents,
paying special attention to who is reviewing the
documents.
b. Use the Reviewer Progress Report to obtain statistics on all
documents reviewed to date in the dataset.
c. Ask your reviewers to self-report a list of documents they
reviewed that day, so you can spot check the documents
for accuracy.
?
27
28. How do you ensure that the Document Report has the most
up-to-date information?
a. Run a Personal Information - Report job.
b. Create a saved search.
c. Run the Incorporate Feedback pipeline before running a
report job.
?
28
30. Why is it important to manually annotate unstructured
documents where text cannot be highlighted in the native
viewer?
a. To ensure the documents successfully load in PI Detect.
b. To add a technical issue flag to the document and discard
it from the project.
c. To ensure redactions are properly mapped onto the native.
?
29
31. Which statement is true about first-level and quality controllevel review?
a. First-level reviewers confirm machine predictions and their
work is checked by quality control-level reviewers.
b. Documents can only be reviewed once: either by a firstlevel reviewer or a quality control-level reviewer.
c. First-level and quality control-level review are both
mandatory and cannot be skipped.
?
30
32. How is the RelativityOne coding layout configured for PI
Detect?
a. It is embedded within the RelativityOne coding layout.
b. It is the same as the RelativityOne coding layout.
c. It is non-existent. You cannot capture document codings
when using PI Detect.
?
31
33. When should you use a document category?
a. When you are trying to isolate a specific document through
its unique control number.
b. When you are trying to locate a certain type of document,
such as a resume or a specific tax form, rather than a PI
type located on the document.
c. When you are trying to identify documents where PI
detectors attempted to fetch results, but an error occurred.
?
32
34. What is the difference between a global and local keyword
when building detectors?
a. Global keywords look for a phrase and translate it across
Relativity’s database of 100+ non-English languages. Local
keywords only look for a phrase in the English language.
b. Global keywords look for a regular expression on a
document. Local keywords look for a phrase located next
to a regular expression.
c. Global keywords look for a phrase in the entire document,
regardless of its proximity to a regular expression. Local
keywords look for a phrase within a designated character
distance of a regular expression.
?
33
35. How many rounds of detector QC does Relativity typically
recommend a user perform?
a. 0-1 rounds
b. 2-3 rounds
c. 4-5 rounds
?
34
36. Assume you are only interested in capturing personal phone
numbers in your project. During the QC of detectors, what is
the best method for eliminating repeating business phone
numbers from your PI results?
a. Locate the phone numbers on the blocklist and block it as
personal information in the dataset, which will remove the
prediction from every unreviewed document in the dataset.
b. Update the phone number detector to eliminate all phone
numbers beginning with a specific area code from being
predicted as personal information.
c. Instruct human reviewers to delete a bad phone number
annotation on each document it appears in as there is no
way to remove the annotation in bulk.
?
35
38. What can be captured using the pre-trained AI detectors in
PI Detect and not by most manual methods of identifying
personal information?
a. An incorrectly formatted credit card number.
b. A Social Security Number.
c. An Excel file with large amounts of personal information.
?
35
37. What is not an intended use case for PI Detect?
a. Expedite PI identification as part of investigations for
reporting purposes.
b. Automatically customize PI identification settings per the
latest privacy regulation.
c. Create a streamlined process for PI identification and
redaction before production
?
36
39. What is an intended use case for PI Detect?
a. A set of documents are deemed to contain privileged
information and need to be logged for litigation purposes.
b. A cyber incident has occurred and impacted individuals
need to be notified.
c. A set of documents with personal information need to be
redacted for upcoming litigation.
?
37
40. Which statement is true about workflows using PI Detect?
a. You can only remove duplicate entities at the final Entity
Centric Report stage.
b. You can work within the RelativityOne environment and
integrate your current processes.
c. Photos or image files with poor quality OCR should be
reviewed the same as spreadsheet files.
?
