(As published in The MReport, October 2019)
Regulators are increasingly requiring, and the market is expecting, the financial services industry to do a better job meeting the needs of Limited English Proficiency (LEP) borrowers by providing translated, non-executable copies of loan documents, also known as convenience documents.
One reason the industry is slow to embrace non-English language (NEL) convenience documents is because of the perceived compliance risk and the potential high costs of translating technical content. Some lenders fear that if they only translate some of their documents, they may be vulnerable to Fair Lending claims for not offering all product types to all customers. Others have raised UDAAP concerns. If the lender cannot guarantee that customers will be provided with translated documents throughout the life of the loan, then might the lender be accused of being deceptive by providing advertising and application materials in NEL?
Despite the regulatory uncertainties, many lenders want to enter the LEP consumer space. And, for good reason—it’s a big market. The percentage of LEP consumers is growing and shows no signs of slowing down anytime soon.
What is the Mortgage Translation Clearinghouse?
While many in the industry want to pursue LEP customers, they are dissuaded by the technical and financial challenges of translating their loan document collection. In response, the Federal Housing Finance Agency (FHFA), Freddie Mac, and Fannie Mae created the Mortgage Translations Clearinghouse, a collection of resources that includes a standardized glossary of mortgage terms and an archive of translated documents.
Although the Mortgage Translation Clearinghouse collection is a good starting point for an NEL document program, it’s not a complete solution. The documents are incomplete, not customized and possibly outdated. While many translated Fannie Mae model notes and security agreements are available, state-specific disclosures are often missing. With few exceptions, the documents that are present in the archive cannot be used out-of-the-box and need revisions to reflect the institution-specific content that is present on the lender’s English versions. Most model forms, including English language forms, lack the state-mandated content required to actually use them in commerce.
Maintenance is always an issue for loan documents, but it is particularly challenging for translated content. State legislatures, regulators and the courts routinely publish new requirements ensuring that document compliance is a moving target. Document vendors frequently publish updated content, so maintaining parity between the English and translated versions can be burdensome.
A better alternative
Machine-assisted translation software offers some promising solutions to address many of these issues. Machine learning and artificial intelligence (AI) are increasingly being leveraged to assist skilled bank compliance personnel. Machine-assisted translations follow a similar trajectory.
One of the most practical ways that technology can assist is through translation memory software, which is essentially a database of different groupings of text. The software recognizes when text has already been translated and suggests reusing the translated text. This is particularly useful for legal documents, given that particular phrasings have precedential value or their meaning is well-understood. Further, states often require that documents use specific phrasings or text. This can result in a significant amount of text being eligible for reuse from one mortgage document to another—a great scenario for leveraging machine learning. Translation memory software also assists with content updates and maintenance issues by isolating content that has changed and suggesting appropriate updates.
While machine translation software has made significant strides, it will not replace human translators—at least in the near term. In fact, by lowering production costs, it may help grow the market for translated content and lead to increased opportunities. Translating financial documents requires people with a deep understanding of the target language and subject matter to ensure that the translation maintains the original intent and context. Much of this work is done post-editing, where the machine produces a translation and then a human corrects the mistakes. Over time, the machine learns from the mistakes and creates more accurate translations.
This is the same work pattern that we see in other areas of financial services. Modern compliance departments rely heavily on software to sort and prioritize the firehose of legislative and regulatory change into a manageable stream of categorized information to be reviewed by the appropriate subject matter expert. Although much of this information sorting used to be performed by paralegals or entry-level associates, it is now automated. Leading businesses are taking the process even further, using AI to suggest the specific documents that should be reviewed in response to regulatory change.
As demographics, customer expectations, and new regulatory requirements evolve, the industry is obliged to provide a higher level of service to LEP communities. Lenders would be well served to consider how to leverage the many technological advances in machine-assisted translations—all of which are lowering barriers to entry by significantly reducing the cost of maintaining documents in multiple languages.