Introduction
Key ML methods in use in investigations cases
Comment
The increasing regulatory governance of Chinese entities has resulted in a rise in eDiscovery projects as part of internal investigations (for further details on how eDiscovery projects can be used in this regard, please see "How to carry out internal investigations: eDiscovery considerations"). These cases usually involve large volumes of data and require pressing timelines, making the use of artificial intelligence (AI) critical.
When eDiscovery AI tools first emerged, most technology vendors focused on their English language capabilities. Now that investigations cases are increasingly taking place in China (and thus in Chinese), the focus of such tools is shifting, and their intelligence has greatly improved.
In the past decade, machine learning (ML)(1) in particular has proven to be very useful in investigations cases. Today, most eDiscovery cases in the context of investigations of Chinese entities make full use of the power of ML. The most common ML methods in use in China are email threading, conceptual searching and technology assisted reviews (TAR).
Key ML methods in use in investigations cases
Email threading
Email threading is used in most eDiscovery cases. It is a process that identifies email relationships and people involved in a conversation, groups them together and marks out the most representative emails, giving reviewers a full picture of the situation and reducing the number of documents to review.
Conceptual searching
In investigation cases, exploring data with conceptual searching may lead to finding hidden code names. Unlike traditional keyword searches, conceptual searching uses a "brain" to study the content of each document, understand their underlying meaning and return broader associated documents based on the input supplied.
TAR
TAR is a continuously learning model that helps improve early case assessment and expedite the pace of eDiscovery and review. It is often utilised for review quality control and privilege determinations. TAR studies document concepts combined with human input and predicts how documents should be classified. Based on these predictions, relevant documents can be prioritised for review teams, while the "brain" continuously studies human calls and feeds information back to the document pool for prioritisation until no more useful documents emerge.
The above methods have all been extensively applied in every aspect of Chinese investigations cases. These methods are powerful, and help clients to reduce time and cost spent on document reviews, potentially saving thousands or even millions of dollars, while ensuring quality and consistency of analysis and minimising the risks of error.
For further information on this topic please contact Jewel Zhu at AlixPartners by telephone (+8621 6171 7555) or email ([email protected]). The AlixPartners website can be accessed at www.alixpartners.com.
Endnotes
(1) ML is a type of AI that can learn from sample documents or human input, organise or classify them into different categories, and make educated guesses.