While the firm can never find evidence that a company's board of directors or C-suite executives intentionally decided not to inform the public about bad news, it can find shifts in sentiment around performance strategy following the news, Craig Macaulay, a data scientist at the Melbourne-based firm, told Law360 Pulse in a November interview.
"The sentiment analysis has allowed us to … zero in on documents and emails so that we can then paint a picture of the knowledge," Macaulay said.
More e-discovery companies are building AI tools to detect negative and positive sentiments as well as certain emotions in documents, which can help attorneys more quickly identify key communications when investigating potential wrongdoing.
Daniel Linna, senior lecturer and director of law and technology initiatives at Northwestern University Pritzker School of Law, said in a December interview that e-discovery tools like sentiment analysis have gained popularity in the legal industry because human document review is expensive, time-consuming and tedious work.
Sentiment analysis tools allow attorneys to more quickly identify documents where someone is expressing intent to do something or is trying to pressure someone else to do something, Linna said.
"It's hard to pick those needles out of a haystack when you're [manually] reviewing hundreds or thousands of emails," he said.
In October, Chicago-based Relativity announced at its annual users' conference that it was developing a sentiment analysis tool using natural language processing and machine learning that can detect positive and negative tones, anger and desire.
The company told Law360 Pulse in November that it started developing its sentiment analysis tool at the beginning of 2022 after customers requested it. The tool was launched last week to all of Relativity's clients, except for those in the government sector.
Relativity's sentiment analysis tool is different from those of other e-discovery companies in that it can detect anger and desire within sentences or paragraphs of documents. Some text analysis tools can only determine whether a whole document has positive or negative sentiments.
Macaulay, who got a chance to test out Relativity's sentiment analysis tool before it was released to the rest of the company's users, said he was surprised by how effective it is for document review, because it provides attorneys with analytics before they even have to look at a document themselves.
Macaulay noted that in addition to shareholder litigation, Phi Finney used Relativity's sentiment analysis to analyze court rulings in cases that it was not involved in. For instance, in one of Phi Finney's cases, a judge recused himself ahead of an important hearing, so the firm used sentiment analysis to analyze the new judge's sentiment in previous rulings and determine how to best present its case to the judge, Macaulay said.
Most e-discovery software developers offer data analytics on their platforms, but not all of them have sentiment analysis among their data analytics offerings, according to their websites.
Chicago-based Reveal-Brainspace was one of the first e-discovery companies to offer sentiment analysis, launching its tool about 10 years ago, according to Irina Matveeva, chief of data science and AI at Reveal and an adjunct professor at the Illinois Institute of Technology. Reveal merged with Brainspace in 2021.
Reveal's sentiment analysis uses machine learning and natural language processing to give documents scores for positive and negative sentiment, pressure, opportunity, rationalization, intent and roundabout writing, Matveeva said.
Matveeva noted that the company's sentiment analysis tool gives scores for different sentiments and emotions on a scale of zero to 10, because a document is rarely only negative or only positive. The tool also identifies which parts of a document are negative or positive, or signal other emotions.
"We should think about communications between human beings as a mixture of emotions," she said.
San Francisco-headquartered legal research software company Casetext released a conceptual search tool called AllSearch in June that can be used to search for certain types of sentiments, according to Pablo Arredondo, the company's co-founder and chief innovation officer.
Arredondo said attorneys can input a sentence like "I'm uneasy" or "I'm very angry," and AllSearch will single out documents that express similar sentiments in them.
"When you're going through millions of emails, it's sometimes useful to quickly isolate ones that might be more legally functional," he said, adding, "That's of only so much value because a lot of the most important evidence is said without any emotion."
Geoff Freedman, senior product manager at Austin, Texas-based legal technology company Disco, told Law360 in a December email that attorneys have been asking the company for "sentiment-driven AI classifications" and that it plans to release a sentiment analysis tool "in the first half of 2023."
"We anticipate seeing wider adoption of sentiment analysis in the near future, as these tools can help drive greater efficiencies for legal professionals, particularly when it comes to gaining stronger insights around positive, negative or mixed sentiments," Freedman said in the email.
While sentiment analysis tools are currently gaining popularity in e-discovery, one day they will be made obsolete by AI tools called large language models that can do multiple tasks, according to Arredondo.
"The way AllSearch works is it is one of these large language models, and that's why you can use it to search for sentiment, or you can use it to search for privilege, or you can use it to search for people saying, 'We're going to get in legal trouble,'" Arredondo said. "When you have tools that operate at the level of language itself, you don't have to have a bunch of specialized tools — you can just have one big one."
--Editing by Alanna Weissman.
For a reprint of this article, please contact reprints@law360.com.
