Lefebvre
Legal
Classification, labeling and automated document search
The legal world is undergoing a profound process of adapting to the new technologies offered by machine learning. This transformation affects all actors but has a particular effect on legal publishers, who must serve large firms by managing millions of documents quickly and flexibly.
In reality, these companies have to devote increasing efforts to be able to meet the expectations of their customers. As expectations rise in the digital world, those efforts become impractical at the company level.
Foqum offers those legal actors a full range of document processing functionalities, from automatic classification to free text document search.
Those tools significantly reduce operational effort and facilitate business decisions (such as a review of document classification criteria) that were previously impractical and are now possible in a matter of days.
Challenge
A leading legal publisher must manage an ever-growing body of judgments and legal documents, process all the judgments as they are published daily, and be capable of managing the document search needs of its customers by linking both new and old documents.
Those search and organizational tasks must tackle a large number of variables and contexts, such as semantic and legal issues, the legal roles involved in the case, the meaning of the judgment, legal evidence, etc. In practice, this means dealing with databases of millions of documents.
Processing the entire corpus for new business decisions (like adding a new classification criterion) is thus impractical.
Solution
The complex needs of this customer require the problem to be tackled using numerous Foqum algorithms for document processing. Our classification system achieves accuracy of more than 97% in the main categories of work and documentation.
Our multi-label classifier is able to classify judgments and documents by more than 10,000 legal categories organized by numerous hierarchical classification levels. Despite covering such an extreme thematic range, the classifier is flexible, enabling it to be used to process new documents live and to be applied to the customer’s entire corpus. It is able to process millions of documents in less than 6 hours, which is impractical by manual means.
