Case Study

Generative AI Enables Rapid Multilingual Document Review in U.K. High Court Litigation

In a high profile fraud and bribery case in the U.K., the defendant’s law firm needed to rapidly process and review documents in multiple languages. FTI Technology was engaged to manage the e-discovery and document review effort and to ensure generative artificial intelligence could be applied in a defensible manner.

Situation

The matter faced significant public attention and the underlying disputes spanned parties in multiple countries, which exacerbated the pressure and scrutiny that typically accompany disclosure exercises for high court litigation. Several data issues created additional challenges for the legal team. Many of the relevant documents were poor quality scans of hand-written materials in numerous languages. Without an efficient way to clean and translate these, the team would not be able to complete an effective review under the case’s tight deadlines.

Our Role

Working closely with outside counsel, FTI Technology’s team of e-discovery experts and data scientists established several workflows to use large language models and generative AI. These were applied to hone the dataset to only the relevant documents, conduct review, translate multilingual materials, extract legible text from low quality scans and validate results for defensibility checks.

FTI Technology’s expert team provided:

  • Initial first-pass review using generative AI features within the e-discovery platform being used for the matter to support case planning and prioritisation for secondary review.
  • Continuous active learning models to analyse the documents identified as potentially relevant for further review, helping to reduce the overall document population and provide quality control.
  • Flexible workflows that combined multiple technologies in parallel, including IQ.AI by FTI Technology — to perform objective coding and extract key metadata from PDFs — advanced optical character recognition to convert poor quality scans into a reviewable format, and translation capabilities for documents in foreign languages.
  • Examination of hallucinations to identify any flaws in the results and continuously improve the system for enhanced output.
Related topics: