Common Terminology

Term

Description

AI-Driven Batches

This Dashboard feature allows you to create batches of documents that need to be reviewed for the purpose of training a Classifier. Reveal uses AI techniques to intelligently select documents for training.

Optionally, you can use the AI-Driven Batch feature to create a Control Set for the purpose of measuring the Classifier training progress against a set of target metrics such as Recall, Precision, and F1 Score.

AI Tags

To create a Classifier you must first create an AI Tag. When creating a tag, selecting Prediction enabled connects the tag to Reveal's AI engine to treat the setting of a tag choice by a reviewer as training for evaluating further documents. When you create the AI Tag, Reveal will automatically create a new Classifier to train the machine on how to score documents (score between 0 and 100) based upon the application of the tag to documents by reviewers.

Assignments

Assigning documents is the practice of dividing up sets of documents to be reviewed by multiple users in a controlled manner. Commonly referred to as batching, it is managed in Reveal with the Assignments feature. Users may have direct assignments, or may check batches out from a pool. Assignment folders for a project will appear under Assignments in the user's sidebar.

Brain Explorer

Brain Explorer is an interactive graphical tool affording the user the ability to view and manipulate cluster information and send it to a search.

Children

Children are sub-documents grouped with a higher-level document or email. For example, attachments are children of emails.

Classifiers

Classifiers are the objects used to manage the machine learning process. To create a Classifier you must first create an AI Tag. When you create the AI Tag, Reveal will automatically create a new Classifier to train the machine on how to score documents (score between 0 and 100). After your new Classifier is created, you can begin training your new Classifier by creating AI-Driven Batches. As you train your new Classifier, Reveal will create an AI Model that's associated to your Classifier. This AI Model will generate predictive scores for each document in your project. AI Models can be reused on other projects to quickly prioritize documents without all the work involved with training a Classifier from scratch.

Cluster

Clustering analyzes textual documents and groups conceptually similar documents. Clusters are automatically generated and most readily seen in the Clusters data visualization.

Clustered documents may relate to a subject or a type of communication. For example, documents in an Earnings Call cluster would be given a higher cluster score for being closer to the center of the cluster (for example, a report or transcript of such a call as opposed to preparatory discussions).

Reveal utilizes Advanced AI to group conceptually close documents together.

When clustering at thread level, an attachment could go to a different cluster.

Communication

A communication is a pair of senders/receivers that engage in active conversation. Reveal collects and analyzes and presents visualizations of communications weighted for frequency, time, topic, sentiment and other factors.

Concepts

Instead of simply responding to a search keyword or phrase with hits on a single term, Reveal analyzes its index to suggest documents containing patterns that it infers share relevance with the term entered. These may be proximate terms found in documents containing the entered term, or terms that appear in documents containing similar content to those retrieved.

Dashboard

The Dashboard is the first screen that opens when a project is selected.

The Dashboard is the control center of Reveal: project data, visualizations, filters, search and analytical tools are all available on this screen. Users can organize, select, search and manage documents here.

Document

A document is an electronic document or email that was originally created and may contain one or more segments. In the case of an email, a reply has both the new segment and contains the previous segment. Reveal AI globally deduplicates all documents based on given hash value. Reveal AI maintains the link between segments and documents.

Duplicates

A Duplicate is a document which generates the same hashed value when one version is compared to another. These are identical to the last comma and space and are therefore essentially one document from different sources.

Email Threads

Email threads are conversations comprised of email replies and forwards.

Entity

A named entity is an extracted piece of data identified by proper name by Reveal.

Named Entities are an efficient method for searching (filtering) for specific people, places or things.

For example, imagine keyword searching for Washington. A keyword search cannot distinguish between a state, street, person or school. Using a named entity, the search results will be more precise.

Reveal analyzes every piece of data in the system and identifies the named entities.

Family Documents

Family Documents are a set comprised of a primary or parent document with attached or appended child documents. Examples include an email and attachments, or a memorandum with appendices.

Features

Features may be keywords (which can be a phrase or a string of several words), entities or metadata elements associated with a document. When documents are tagged with an AI-enabled choice, these features contribute to the training of a classifier’s model. The model can be subsequently used to score documents based on the features present in the model and in the document.

Heatmap

The Heatmap accessed under the Dashboard displays the co-occurrence of the top terms of one type of metadata with the top values of a second type of metadata or a set of search terms. With color emphasis indicating “heat” or preponderance of intersections within the grid, users can see at a glance the strength of relationships between one type of data and another.

Inclusive Document

An inclusive email is a document that contains the maximum amount of distinct segments for a thread. There may be multiple inclusive documents within a thread. For example, multiple conversation branches may create multiple inclusive documents.

As an example, an exchange of emails will create multiple documents. However, the final email will have all segments from the previous communications. An end-user only has to interrogate the final document to read all possible segments.

Labels

Reveal uses AI capabilities to provide labels for object detection and image classification. Image labeling provides a great time and cost saving in the review and analysis of non-textual image content by labeling objects identified in each image.

Locked

Retrieves documents having settings in effect for preventing parts of the document record from being changed, including Fields, Notes, Annotations, Redactions and the various Tag Sets associated with the document.

Model

A portable package of knowledge, trained by application and analysis of classier tags, used by Reveal to apply a set of categorization rules in gathering and analyzing classes of information; model examples set a mathematical start and end point for relevance.

Model Library

Models created from classifiers in projects may be published to the model library of the project or for all of a company’s projects to be reused, adapted and re-saved.

Near Duplicates

Near Duplicates may be thought of as different drafts or versions of a document, assessed within roughly 80% similarity. These versions may be compared in Reveal Review to highlight differences between any two versions at a time.

Notes

Notes are reviewer comments added to the space provided in the Review Screen.

Parents

Parents are documents to which other documents are attached or appended to form families. For example, an email is a parent to its attachments.

Privileged

This retrieves documents in the current review where privilege tags have been set.

Segment

A segment is an individual email; every segment has only one writer; without any replies or forwards. Reveal counts individual segments to tell the story of how many times something was written about, using a content-based algorithm to deduplicate segments in the project. This greatly reduces the amount of duplicative data in the project.

As an example, I write about Acme Corp one time and three people receive this email. This email is considered one segment. If someone replies to the email and includes the original email, there is only one new segment added to the Reveal AI project.

Sentiment

Sentiment is the tone of a communication. Reveal uses an algorithm to calculate a score of the sentiment for the overall segment.

Negative sentiment equals a writer using words and tone associated with negative connotation. A segment with a score lower than -1 will equal a negative segment.

Positive sentiment equals a writer using words and tone associated with positive connotation. A segment with a score higher than 1 will equal a positive segment.

For example, a writer may say:

“Everything is a mess and we need to shut this down right away. However, the staff is nice.”

The segment contains both negative and positive sentiment. Reveal calculates an overall score to take this scenario into account.

Shingles

Shingles are the n-grams found in the body text and can be used to compare the body content of messages. In this approach each document is represented as a set of unique shingles, which break the document into sequences of n (for example, 3) words, and then measures the percentage of one document’s shingles that appear in another.

Single Sign-On (SSO)

Reveal's single sign-on (SSO) provides a single Reveal ID across all modules within the Reveal platform even if your firm still requires application-based authentication. This capability is secured in part by use of two-factor authentication.

Supervised Learning

Supervised learning, also known as supervised machine learning, is a subcategory of machine learning and artificial intelligence. It is the process of using manually coded document examples to train Classifiers (algorithms) to predict whether a document is relevant, privileged, or otherwise categorically classifiable.

Tag Rules

Setting rules for tag validation allows the project administrator to make sure that proper coding protocols and dependencies are consistently observed. These validation rules can be tested for each Tag Profile within the Tags window under Project Admin.

Team

Users' scope of action (profiles, assignments, views) is defined by their Team membership. Users may be added to or removed from teams on the fly to enable or disable access to certain resources.

Term List

Search tool that may be used in building a complex query under Advanced Search. The text version may be copied and pasted from a text file; toggling to the table version allows totaling of each search value or the entire value, within the retrieved subset or for the entire project dataset. Can be added to a current search, or used to create a saved search or hit report.

Thread

A thread is a document or series of documents that shares at least one exact segment.

For example, an email may be sent to a recipient, who in turn forwards to another recipient and starts a parallel conversation. Reveal detects threads and creates a ThreadID.

Time chart

This interactive visualization graphically displays communications as distributed over a selected period of time.

Two-factor Authentication

Reveal requires that a second factor be used to verify access to the user's account. Reveal’s default option at present is to use an Authenticator application, usually installed on the user’s handheld device, to provide a code to be entered to verify the user's identity.

Visualization

The visualization graphically displays communications between people or about subjects. Reveal creates a link chart based on the data being interrogated. Such visualizations include Clusters and Dashboard graph widgets.

Wordlist

Wordlists are lists of keyword terms, including Boolean and proximity searches, set out separated by hard returns with specified text and highlight colors to readily show the presence of these terms within any document in the dataset viewed by a Team member assigned to the wordlist.