Epiq Transforms the Legal Industry Using Artificial Intelligence from AWS
Epiq provides proprietary solutions and managed legal services to law firms and government agencies worldwide, streamlining the administration of court reporting, business operations, mass torts, and other legal services. In the latter half of 2020, Epiq wanted to explore how modern technologies and artificial intelligence (AI) could transform its legal transcription services. Previously, Epiq transcribed courtroom proceedings using a combination of heuristic-based software and human transcribers. Manual transcription was time consuming and expensive, but software-produced transcriptions were often less accurate and challenging to adapt to regional accents.
Epiq also offers e-discovery services, or the process of finding and presenting electronic information as evidence for court cases. Multimedia assets, such as audio and video from police cameras or smartphones, are increasing in court cases, compared to text-based assets, such as articles and emails. Manually transcribing multimedia assets wouldn’t make financial sense. “Many legal professionals have been hesitant to fully adopt AI technology to process multimedia assets at scale,” says Don Young, vice president of engineering at Epiq. “Required accuracy levels are not being met by the available tools, and transcription and e-discovery solutions are often not well integrated. Epiq wants to treat audio and video as additional asset types in the ingestion phase so that we can apply the same search and identification capabilities we use for text-based documents.”
In a few weeks, Epiq delivered an AWS-powered transcription solution that is 5% more accurate than other third-party transcription engines.”
Vice President of Engineering, Epiq
Building an AI Transcription Solution
Having already used Amazon Web Services (AWS), Epiq chose Amazon Transcribe to automate and accelerate transcription for courtroom proceedings and multimedia assets. Powered by deep learning, Amazon Transcribe is an automatic speech recognition service that converts speech to text quickly and accurately, thus enabling developers to add voice AI to applications.
Epiq chose AWS for its transcription solution because other vendors would have required Epiq either to import subpetabytes of digital assets into the vendors’ cloud environments or to install the vendors’ proprietary transcription engines into Epiq’s environment on Amazon Elastic Compute Cloud (Amazon EC2), a web service that provides secure, resizable compute capacity. By choosing an AWS transcription service that scales on demand and makes raw data accessible, Epiq enjoys secure infrastructure and a quick time to market. “Using AWS for this solution has been fantastic,” says Shah Karim, chief technical officer at Epiq. “With these tools, we can innovate in the legal field with the power of the cloud and software as a service to deliver accurate, inexpensive automatic speech recognition to the industry.”
“Working Backwards” to Increase Accuracy and Reduce Costs by a Factor of 5–15 on AWS
Epiq consulted AWS Envision Engineering to create a proof of concept for its transcription solution. AWS introduced Epiq to the “Working Backwards” methodology, in which customers and their needs dictate the innovation process. Epiq president Roger Pilc invited Colin Bryar and Bill Carr—the authors of the book Working Backwards: Insights, Stories, and Secrets from Inside Amazon—to speak about the methodology to 140 leaders at Epiq. In just 6 weeks, Epiq developed and launched a solution on AWS which exceeded the level of accuracy it required.
To meet its standards, Epiq needed the transcription output to be at least 80 percent accurate; anything less would be too time consuming to correct compared to producing a transcription by hand. Using Amazon Transcribe, Epiq exceeded its goal, building a solution that produced 85 percent accuracy. Considering the fact that court recordings often involve multiple people speaking from a distance from recording devices, this was a major accomplishment. “In a few weeks, Epiq delivered an AWS-powered transcription solution that is 5 percent more accurate than other third-party transcription engines,” says Young. “With virtually unlimited scalability and superior multilingual support, it is immediately available in all geographical regions where Epiq does business—and at a much lower cost. Using other vendors would cost 5–15 times more.” Once Amazon Transcribe converts speech to text, Epiq uses Amazon Augmented AI (Amazon A2I), a machine learning service that makes it simple to build the workflows required for human review. Humans then correct the output and produce a final transcription that is 99 percent accurate. “Having an integrated correction loop is essential to a practical transcription toolset,” Young says.
Epiq also used AWS to augment its e-discovery solution by enabling efficient multimedia processing in only 3 months. “We can transcribe these media assets into text, and suddenly the entire system becomes uniform,” explains Young. “We now have a technical representation of audio and video that can be indexed and searched.” Epiq automated the workflow for the e-discovery solution using AWS Step Functions, a serverless function orchestrator that makes it simple to sequence functions into business-critical applications. “There are many things on AWS that are the building blocks for this solution,” Young continues. “Transcription is an engine, and we use AWS Step Functions as an adaptation tool to effectively automate the many flows within our system.”
Epiq can use AWS to transform a slow-to-modernize industry and accelerate its innovation timelines. “Any company using the cloud can increase the pace of innovation and become a cost leader in the legal industry,” says Young. “Proprietary noncloud stacks from established vendors are becoming liabilities.” Epiq has been able to reduce transcription costs while processing increasingly large amounts of multimedia asset data for its clients. For example, one Epiq client had hundreds of terabytes of data for transcription, and 40 percent of this data was comprised of media assets. “Transcribing audio-visual files is a resource-intensive process,” Young says. “Any cost advantage gives us a competitive edge in this industry and frees up resources to drive innovation in new areas.”
Continuing to Transform the Legal Industry
To further use its vast court reporting archives, Epiq plans to differentiate its delivery capabilities by continuing to innovate with new custom language models on Amazon Transcribe. “I’ve been in the industry for 30-plus years, and this is the best collaboration that I’ve ever experienced,” says Young. “AWS has provided so much technical and strategic help. We’re looking forward to a very fruitful 2021 using AWS.”
On AWS, Epiq not only built a cost-effective solution to automate, accelerate, and increase the accuracy of legal transcription, but it also innovated the way transcription is completed in the industry. Young continues, “Cloud computing introduced a different yet innovative formula into how we compete, not just for us, but for the entire industry.”
Founded in 1988, global legal services company Epiq offers legal firms a suite of e-discovery services to help them search and secure electronic data for use in legal cases.
Benefits of AWS
- Created a proof of concept for transcription in 6 weeks
- Developed an e-discovery solution in 3 months
- Provides court transcriptions that are 85% accurate
- Achieves 5% more accuracy than traditional legal tools
- Saves by using AWS rather than third-party vendors, which would cost 5–15x more
- Gained multilingual support
AWS Services Used
Amazon Transcribe is a fully managed service that makes it easy for developers to add accurate speech to text capabilities to their applications.
Amazon Elastic Compute Cloud (Amazon EC2) is a web service that provides secure, resizable compute capacity in the cloud.
Amazon Augmented AI (Amazon A2I) is a machine learning service which makes it easy to build the workflows required for human review.
AWS Step Functions
AWS Step Functions is a serverless function orchestrator that makes it easy to sequence AWS Lambda functions and multiple AWS services into business-critical applications.
To learn more, visit aws.amazon.com/transcribe/.