By Penelope Sibun
People are difficult enough to read when talking face-to-face. Trying to decode their emails for the purposes of electronic discovery is a whole other story. Meaning and nuance can be lost in a hastily written note and decoding someone’s true intentions can be a challenge.
Now, imagine that problem multiplied throughout an enterprise-level organization. Hundreds of employees sending thousands of emails filled with millions of words denoting different ideas, questions, and agendas.
With Emotive Tone Ontologies (sometimes referred to as Sentiment Analysis), organizations can identify the emotional tone of documents, especially email. Emotive language is a particularly good indication of the writer’s feelings and in turn gives insight into the attitudes of those inside an organization and their reactions to significant information or events.
Typically, the ontologies that forensic investigators build for an e-discovery matter focus on determining whether a document is responsive, non-responsive, or privileged, and this determination can usually be made directly, by examining the content of documents. (An ontology is a structured representation of linguistic and other features that is used to search data or locate evidence.)
Emotive Tone Ontologies, however, pick up indicators of emotion, independent of the subject matter of the documents: they cut across issues and detect underlying emotions. What is the emotional context behind this thread of e-mails? What is sitting below the surface of the document’s raw text? We know verbatim what our subjects said, but what emotions are bubbling beneath their words?
Emotions—particularly anger, surprise, or confusion—are incredibly valuable pieces of intelligence in litigation and investigation. Today, thanks to modern technology, these emotions are trapped in digital amber and accessible to those with the tools to extract them.
Since emotive language is independent of content, the text of emotional emails can vary widely. An expression of anger can appear anywhere: a berating email from a superior to a subordinate who has done something potentially illegal, or a personal note complaining to a spouse about a forgotten dinner reservation.
Looking for documents that contain both responsive language and emotive language can yield “smoking guns” on which a legal matter may turn. In the case of an irate spouse, the email is likely to be non-responsive, but can still be used in an investigative matter to build an accurate character profile of a key custodian: is this person volatile? foul-mouthed? abusive?
The process of exploring emotive tones—in English, at least—focuses on ten core feelings, including: Angry, Confused, Cursing, Derogatory, Frustrated, Problem, Secretive, Suspicious, Surprised, and Worried.
It’s easy to see how each of these emotions, preserved in time and recovered from an email server, could be useful in an e-discovery matter and in litigation readiness.
Context, as they say, is everything.
Penelope Sibun is an Associate Director in Ernst & Young LLP’s Fraud Investigation & Dispute Services focusing on advanced text analytics. The views expressed herein are her own and do not necessarily represent the views of Ernst & Young.