Send an email
Artificial Intelligence is changing the way businesses operate across industries, sectors, and geographies. Automation software is reducing manual labor in fields from finance to legal, and the ever-increasing amount of data is driving a new wave of business analytics.
One area where these technologies are driving massive change and efficiencies is in legal document review. During litigation, legal document review is part of the eDiscovery process where attorneys wade through hundreds or thousands of documents ranging from emails and texts to images, spreadsheet files, all manner of electronic documents, Slack or Teams messages, social media posts, and even CAD drawings and other digital files.
“Document review is part of the discovery phase of a legal matter where attorneys read, analyze, and categorize potentially discoverable material for responsiveness or relevance, confidentiality, privilege, and issues involved in an investigation or dispute,” said Wes Kiplinger, Director of Review Services at Sandline Global.
With the explosion of data over the last 20 years, the task of document review has grown much more challenging, which has required significant change in the balance between humans and machines in the work of reviewing these documents. Leading legal software companies have built entire new platforms to assist review attorneys with this task. In recent years, machine learning has become an increasingly important part of this process.
Machine Learning algorithms can greatly reduce the number of documents attorneys must review. “Some matters see a reduction of 80% or more in the amount of data needing eyes-on review. Typically, however, one can expect a reduction of 40-60%,” Kiplinger said.
Kiplinger recalls a time before ML-driven technology was the industry standard. “I started working in this space when electronic review was just getting off the ground and even did a few paper reviews where I sat in a room with banker boxes full of printed emails and memos. Technology has brought us a long way. Concept clusters and Active Learning (machine learning algorithms that evolve based on human inputs – constantly refining models via human inputs) have dramatically reduced the amount of material needing to be reviewed.”
“Long gone are the days,” Kiplinger said, “when I saw emails to your family or just what you were shopping for online through your work email.”
Legal Software maker Reveal uses AI to help organizations uncover more useful information faster by incorporating machine learning into every aspect of the eDiscovery process. Reveal’s review platform includes features like image recognition and labeling, active learning, and audio and video transcription. The tool leverages multiple models to identify specific types of human behavior, automatically identifying documents that include language that suggests threatening behavior, comments on appearance, hate and discrimination, and other categories potentially relevant to litigation.
George Socha, Senior Vice President of Brand Awareness at Reveal, said, “Machine learning has brought unprecedented speed to actionable insight in document review. Reveal’s unsupervised ML puts a wide array of previously difficult-to-locate information about people and events at your fingertips.”
The availability of this information is key to document reviewers. Kiplinger said, “Analytics puts the most interesting and relevant material in front of attorneys first while setting aside irrelevant information and junk (such as spam emails and news alerts).”
The analytics process has become a significant part of document review, Kiplinger said. “There are many different tools in the analytics toolbox, and they all assist in reducing the amount of material that needs to be reviewed by attorneys. In the past few years there have been stunning advancements in analytical tools. Before even looking at the documents themselves, we can see visual maps (concept clusters) of who sent emails to each other, how often, the topics of those emails and even the tone (e.g. these emails were emotionally negative). Once a review begins, Active Learning applies coding choices to unreviewed documents to further drill down into material that is likely relevant and also what is not relevant which drastically reduces the amount of data that needs eyes-on review,” he said.
Reveal’s document review software offers an array of visual analytics that can inform attorneys before the documents have even been presented to human eyes. Tools like the Cluster Wheel that allows reviewers to quickly explore topics of interest, the Meta Dashboard, which gives reviewers an at-a-glance view of key search terms, and Transparent Concept Search, which lets users start with a phrase, a paragraph, or even entire document, then expand their search automatically to reveal related concepts, which could potentially uncover key concepts that weren’t previously being tracked.
Socha said Reveal continues to work on making their product more powerful and easier to use. “Reveal’s platform will continue to get easier to use, offer richer sets of features, and work with wider arrays of data. Analytics, including ML, will become nearly pervasive, more actively guide investigators and lawyers, and yet more effective at feeding up insights quickly,” he said.
Kiplinger points out, however, that document review will require human involvement for the foreseeable future. “Years ago, the consensus seemed to be that analytics would do away with human document review altogether. That isn’t going to happen. The amount of data we create has grown exponentially and this is a continuing trend. Analytics will continue to reduce the noise in data and will become even better at identifying potentially relevant material and suppressing the ever-increasing amount of irrelevant data,” he said.
As in other industries and sectors, we are seeing increased power from the continually expanding capabilities of AI, but for the near future we see more and more industries using this powerful new technology to augment and enhance human capabilities. As with other professions, we are seeing the need for review attorneys to expand their skillset beyond knowledge of the law, and to include a deeper understanding of the technology that powers what we do.