Sentiment Analysis: How Word Choice Can Leave a Trail to Fraud

Misty Carter, CFE, CIA, shares how sentiment analysis, an analytical tool that involves creating a system to analyze documents and other data to examine the authors' emotions and opinions, can be used to detect fraud.

By Misty Carter, CFE, CIA

Trying to figure out what people think has always been interesting to most people. This is true in social settings and when getting to know someone personally. And this information can be especially useful when investigating fraud. While some people are not willing to open up and show how they really feel through conversation, they might express their feelings through social media sites, forums, blogs and even emails. Fraud examiners are beginning to gather and analyze this type of data in investigations. This technique is called sentiment analysis, and it has proven to be an effective method in detecting red flags of fraud.

What is sentiment analysis?

Sentiment relates to a person's feelings, attitudes, and opinions. Sentiment analysis (also known as opinion mining) is an analytical tool that involves creating a system to analyze documents and other data to examine the authors' emotions and opinions. It is focused on identifying what people think or how they feel about something. This type of analysis can assist fraud examiners in uncovering existing fraud schemes. It can also serve as a trending tool to look at areas of concern that might indicate individuals at risk for suspicious or fraudulent behavior.

How can it be used to detect fraud?

Sentiment analysis includes a process called keyword spotting that includes developing a list of keywords that relate to a certain sentiment. These words, known as affect words, are usually positive or negative adjectives because such words can be strong indicators of sentiment. Fraud examiners search for affect words in employee emails or other communications as part of their fraud detection procedures. For example, a fraud examiner might receive a tip that a manager is manipulating his division's reported sales to meet company quotas. In evaluating the merit of the tip, the examiner might run a list of keywords — flexible, unreasonable, temporary and worried — against the alleged party's emails and, if there are positive results investigate the issues further.

Sentiment analysis using negative keywords can also identify potentially disgruntled employees. Fraud examiners can identify warning signs of fraud by searching for negative keywords that indicate the three elements of the Fraud Triangle (pressure, opportunity and rationalization). Examples might include: ignored, exhausted, inconsiderate, harassed, passed over or biased.

When preparing a keyword list, fraud examiners should include swear and slang words because these can carry strong sentiment. Fraudsters might also use acronyms in attempts to hide their fraud schemes; examples to consider searching for include:

• 411 (information)
• 8TB (ate the bait)
• UOME (you owe me)
• TYOP (tell you on phone)
• TOL (talk offline)
• LDL (let's discuss live)

Other considerations

While sentiment analysis is a great analytical tool, some caveats must be taken with its use. As with all investigative techniques, fraud examiners should gather complete evidence before coming to conclusions. After emails or documents are flagged based on keywords, the fraud examiner must still investigate further to determine if a fraud did occur. 

Additionally, employee privacy rights in certain jurisdictions might limit the ability to conduct sentiment analysis techniques on some types of employee communications, so fraud examiners should always consult legal counsel regarding any legal restrictions before undertaking any analysis of this type. Where such analysis is legally allowed, fraud examiners should ensure that the company has a policy stating that emails and other correspondence are company property and can be reviewed by the company at any point in time.