Opinion: How AI Helps Educators Reach Multilingual Populations

Two recent trends in education are diversifying how students learn, but converging to create a major challenge for educators:

• Multilingual education is on the rise. Since the 1990s, the number of U.S. residents who don’t speak English at home has grown by nearly 300 percent.
  
• E-learning has become more common for students of every language. By 2021 — the latest year for which the U.S. National Center for Education Statistics has released data — more than 60 percent of U.S. higher education students took at least one distance learning course. More than 20 percent exclusively learned from a distance.

Anyone who understands that it’s easiest to learn in your native language will immediately see the challenge: Educators must find ways to deliver online course content in multiple languages at once.

According to a February 2024 blog post by United Nations Educational, Scientific and Cultural Organization, by the end of primary school, students who learn in their own languages are up to 60 percent more likely to reach reading comprehension goals. That’s why it’s a problem that 40 percent of the world’s students lack access to courses conducted in the same language they speak at home. These aren’t just primary school students, either; language barriers affect education at every stage, including postsecondary and corporate learning.

A new set of AI tools can help. Below are a few AI solutions that make it possible to teach every student, whether you speak the same language or not.

AI LANGUAGE TOOLS FOR MULTILINGUAL EDUCATION

The computing architecture behind lots of today’s AI tools — referred to as deep neural networks — can be trained to perform a staggering variety of human-like tasks.

A large language model (LLM) like ChatGPT produces lifelike writing. A neural text-to-speech (TTS) voice can read that writing out loud, sounding a lot like a human speaker. And neural machine translation (NMT) models — which are trained on the correlations between languages — produce pretty good language-to-language translations.

Based on these capabilities, here are three classes of AI tools that can help educators more easily meet the challenges posed by diverse classrooms.

1. AI Translators

Computer translation has been around for decades, but the rise of AI has helped the technology get a lot faster and more accurate.

A NMT model can interpret words in context, successfully translating homonyms like left (is it the past tense of leave or the opposite of right?) and novel (the book or the adjective meaning new or original?)

Some NMTs (like Google Neural Machine Translation, or GNMT) can even apply its translation capability to languages that aren’t in the training data. That’s called zero-shot translation, and it can help expand the number of languages AI translation tools can handle.

These tools are fast enough to provide live translations for online lessons, automated captions for training video content and real-time translation between classmates who speak different languages.

2. AI Chatbots

The most famous AI chatbot is also the force that brought generative AI to the mainstream: ChatGPT.

Of course ChatGPT can write your students’ papers for them — but did you know it can probably do so in each student’s native language? The LLM behind ChatGPT was trained on many, many languages, and the chatbot can try zero-shot translation for the rest.

That makes ChatGPT and similar generative AI ideal for helping teachers create content outside of their own native language. (They just need to be sure to check the accuracy with another translation tool before publishing!)

It also opens the door for AI tutoring — which 85 percent of students preferred to in-person tutors in a 2023 survey. A ChatGPT study session in a student’s home language might help that student prepare for classroom work in a second language.

3. Multilingual Text to Speech

Taken on their own, AI translators and chatbots have a notable limitation: They’re both text-based.

Where does that leave the second-language learner with blindness or dyslexia or an auditory learning preference?

Another AI tool can help make multilingual content accessible to all. Text to speech reads digital text out loud — and neural TTS, which is built by feeding human voice recordings to deep neural networks, sounds almost like the real thing.

Neural TTS is also being used to build multilingual voices, which sound the same while speaking any number of languages. That’s a big deal, as earlier TTS voices had to be trained on each language independently, slowing down global reach.

With a multilingual neural TTS voice, students can listen to any written content in the language that helps them learn best. But beyond neural TTS, what sort of TTS tool should educators offer?

They should start with a TTS tool that integrates seamlessly with their learning management system. That way learners don’t have to open new apps or tabs to get the benefits of TTS.

Next, they should look for TTS with additional learning tools, such as integrated translation and speed control. That last feature is important for second-language learners, who may prefer to slow down tricky passages for greater comprehension.

In combination with text translators and generative AI, multilingual TTS allows e-learning content to speak virtually any language. That helps students learn any subject while also strengthening language skills. It can also help parents understand their children’s assignments and other communications from the school.

In a country where nearly 70 millionpeople don’t speak the national language at home, multilingual education is more than convenient — it’s essential for equitable student success.

Amy Foxwell is a marketing professional with experience working for companies such as Microsoft and ReadSpeaker, which makes text-to-speech tools.