Artificial Intelligence Technology -
Conformité24 janvier, 2024

Lenders leveraging AI and machine learning to overcome language barriers

For Americans with Limited English Proficiency (LEP), getting a mortgage can present significant challenges due to the language barrier, the complexity of the process, and the significant number of documents they need to comprehend. Currently, there is no federal requirement to provide mortgage documents and disclosures in any language other than English. Despite not having a requirement in place, lenders recognize the benefits of providing translated documents, particularly in Spanish. 

According to the Migration Policy Institute, there are more than 25 million people with LEP, representing 9% of the U.S. population, many of whom are ready and willing to be homeowners. The National Association of Hispanic Real Estate Professionals (NAHREP) has estimated that, as of 2022, there were approximately 8 million mortgage-ready Latinos aged 45 and under and 2.8 million who are near mortgage-ready. Further, Hispanic homeownership rates have steadily grown over the last several years, reaching 49% recently, and, according to the Urban Institute, Latinos will account for 70% of homeownership growth over the next 20 years. 

Given the market potential of prospective Spanish-speaking homeowners, as well as growing industry and government pressure to increase services for LEP populations, we sat down with Wolters Kluwer LEP expert, Crystal Coker, Senior Technology Product Manager, to shed some light on what lenders and servicers should know. 

Why should lenders invest in translation resources?

Coker: The data and numbers bear out the significant opportunity that exists with America’s Spanish-speaking population, as well as with other LEP consumers who may be prime candidates for mortgage lending. We’re seeing more and more lenders studying this issue and making plans to offer document and content translations. Not only is it the right thing to do from a customer service perspective, it makes good business sense for lenders looking to tap new customers, particularly in a tight lending environment. And if improving customer service and finding new customers aren’t reason enough, most industry observers expect laws and regulations to eventually force the requirement. 

The Consumer Financial Protection Bureau (CFPB), Federal Housing Finance Agency (FHFA), Fannie Mae and Freddie Mac all encourage financial institutions to provide more services in non-English languages. Specifically, the CFPB recommends that lenders provide a clear and timely disclosure to consumers that describes the extent and limitations of their translation services. The FHFA provides a model Language Translation Disclosure. And the GSEs introduced a document called the Supplemental Consumer Information Form that allows consumers to indicate their language preference. While initially optional, the Supplemental Consumer Information Form became required for loans with application dates on or after March 1, 2023, for any transactions sold to Fannie and Freddy on the secondary market. For many lenders, though, the question becomes what do they do with that information once a consumer has checked their language preference? 

Within a couple of years, the CFPB may identify as many as eight target languages based on U.S. census trends that should be supported with translated mortgage and lending documents. If we look at census data, of course, Spanish is number one, but then if you drill down to states, depending on where lenders do business and the footprint they have, they might also need translations in Chinese, Korean, Tagalog, Vietnamese and other languages. 

How is Wolters Kluwer helping its customers address the growing need for translation services? 

Coker: Our Expere® Language Translation Solution is a financial document translation system that uses artificial intelligence (AI) and machine learning (ML), combined with our linguistic and compliance experts, to deliver highly accurate and scalable translation capabilities. Because we’ve been at this for more than five years, we’ve been able to create proprietary, curated and customized document templates, dictionaries and even full customer-specific libraries with documents that are not only compliant, but also meet the unique specifications of individual lenders.     

What are the potential pitfalls lenders may face when tackling language translations? 

Coker: Translating nuanced legal financial documents for non-English speaking consumers is expensive, time consuming, and prone to human error. Most lenders don’t have in-house translation services, and even if they have Spanish-speaking employees, that’s not enough to provide the nuanced translation services required. Nor does it address the human-error that can occur, which is why leveraging AI and ML technology is so critical in the process. But relying solely on AI and ML isn’t the answer either, which is why we also rely on the expertise of our specialized linguistic and compliance professionals who are able to complement and correct what is generated from AI.   

How, specifically, does the Expere Language Translation process work? 

Coker: The process starts with our proprietary compliance dictionaries that have been curated over time – pulling content from the CFPB, Fannie and Freddie – and are continually updated and maintained. Our translation workbench uses AI to provide the initial translation which also incorporates custom ML algorithms that build off of prior translations, and corrections made to those translations, in order to produce more precise results moving forward. 

The documents are then reviewed by bilingual linguistic experts who work closely with bilingual compliance experts for contextual expertise to accept or reject the translation AI has provided. Once validated by our expert teams, the translations are stored in Wolters Kluwer’s translation hub to improve the ongoing accuracy and efficient delivery of future translations. The integrated, all-in-one translation process is part of an automated doc prep workflow that allows Wolters Kluwer to provide translations quickly and efficiently without sacrificing accuracy. 

Interestingly, when Wolters Kluwer first began our translation services in 2018, roughly 80% of the translations were manual and 20% were assisted by AI or ML. Today, it’s flipped and about 96% of translations are provided via AI/ML and only 4% are manual translations. That proves how much the machines have learned over time to improve the overall output.

Are all of the documents that lenders need to translate about the same? 

Coker: Not at all. It’s important to understand that every month the Expere platform publishes revisions to content, which means every lender may be on a different version based on when they started using that version. As a result, each lender has a unique library of translated content that includes standard, Uniform Instrument-based documents as well as documents that have been customized to the lender’s specifications. As new translations occur, the content in that lender’s library is updated.

Learn more about Expere Language Translation services.

Back To Top