Machine Translation Expands CRM Capabilities

More than 1.5 billion people speak English. And while English is still widely considered the dominant global language, some experts argue that it is losing that status due to China’s rise in geopolitical and economic prominence, increased multilingualism, rising internet and smartphone penetration around the world, and the growing accessibility of machine translation technology.

This last factor, the growth of the machine translation market, is significant. Market Research Future valued the technology at $1.9 billion in 2023 and expects it to reach $3.9 billion this year and $31.9 billion by 2032, growing at a compound annual rate of 26.3 percent.

While these numbers cover both consumer and business applications, companies are increasingly focusing on providing multilingual customer support to enhance user experience and expand their global reach. Machine translation enables them to communicate with customers in their preferred languages, driving customer satisfaction and loyalty. This is key, given that 75 percent of consumers worldwide responding to a Common Sense Advisory survey said they prefer products that are available in their native languages. Additionally, 60 percent said that they never purchase from English-only websites.

In response, companies are increasingly taking advantage of machine translation integration into many of the content management systems they use to create marketing and sales materials. The benefits of machine translation technology for marketers and sales representatives, web developers, technical writers, and content creators are expansive.

But the customer experience is comprised of activities not just before and during the purchase. Companies that use translation and localization technology to optimize these first two stages of the customer journey need to consider using it to optimize the third stage, which occurs after the purchase. Their customer service teams can also profit from machine translation technology.

In a customer service context, the chatbot is a perfect channel to use with machine translation. Translation technology today can be easily connected to most chatbot tools with just a few lines of code. Once connected, it enables multilingual communication within the chat window, enabling a seamless form of communication, even if the customer service representative natively speaks one language and the customer speaks another.

Another customer service component that is ripe for machine translation technology is the company knowledge base, which customers can use to find answers to common questions on their own and agents can use to push information out to customers.

Using machine translation technologies, customer service teams can upload knowledge base documents and have them translated into the languages they need to serve customers. These translations can even be reproduced with the same design and layout as the original documents, preserving brand-identifying elements like logos, color palettes, and fonts.

Beyond Text

And while text-based translations are common, using machine translation technology with voice- and video-based customer communications is a growing area of interest for many companies. It’s also the area that has seen some of the most significant technological advances in just the past year or two.

“Advances in text-to-speech, voice cloning, and visual dubbing are rapidly changing the customer relationship management paradigm,” says Jean-Louis Quéguiner, CEO and cofounder of Gladia, a French startup that offers a speech recognition application programming interface (API) for uses that include translations. “Combined, these technologies create hyper-realistic voiceovers that mimic a speaker’s natural voice and speech patterns, even in new languages. These applications enable companies to personalize and enhance customer experiences, from phone assistance for seniors to bridging language barriers in virtual meetings or e-learning environments.”

Modern machine translation technology enables companies to provide real-time support to customers in their native languages or create custom voices for their virtual agents, Quéguiner adds.

Among the many technological innovations impacting the translation technology ecosystem, few have been as transformative as generative artificial intelligence.

“[Large language models] have evolved considerably in language understanding and multilingual capabilities,” Quéguiner says. “[Speech recognition] systems have advanced from more acoustic/phonetic-based mechanisms to [large language model]-like complex architectures. This transition has increased the quality of transcription and translation, largely due to the enhanced contextual awareness of LLMs and gap-filling capabilities of these new models.”

Additional improvements are paving the way for more natural text-to-speech output that can be widely deployed across industries and use cases, Quéguiner adds. Enhanced voice cloning preserves the original voice, while lip-synching technology marks another significant breakthrough, enabling human-level quality of translation and dubbing.

But to achieve true human-level comparability, translation technologies need to be able to operate in real time, and modern artificial intelligence is bringing that closer to reality.

Real-time translation technology has advanced significantly in just the past year or two through improvements in AI that are making it faster, more seamless to the customer, more context-aware, and more accurate.

AI has also made it possible to create customized vocabularies for industry-specific terms.

Advanced genAI models can even help agents be more empathetic, regardless of the languages spoken, and take into account the cultural and social norms of different cultures.

In essence, generative AI has elevated real-time translation into a sophisticated, adaptive tool that allows for more natural, contextually aware, and personalized multilingual communication in the contact center.

But not everyone is convinced that real-time translation is viable just yet.

Rob McDougall, CEO of Upstream Works Software, a provider of omnichannel contact center desktop solutions, says real time only works in limited uses.

“If you’re using [translation technology] for chat, you can do it pretty much seamlessly. Someone on a web page can type in Spanish, the agent can read it in English and send English back, and the person on the web page just sees Spanish.”

The problem, though, is that in most customer interactions, translations require a third-party intermediary between the customer and the agent.

“You can have a third voice and they’re doing the translation in sort of real-time mode, but it’s a third voice. It’s not like [customers] can just talk to the agent who speaks seamlessly back to them in Spanish because it doesn’t work that way,” McDougall explains.

“When I want to do translations now, I gotta have a point-to-point telephony connection to an agent. And then I have to branch off that audio stream somewhere else to have it translated and then feed the new audio stream back into the audio stream that I have. That’s just not that easy to do,” he continues. “Although the capability is there, the required resources are expensive and complicated (to capture the audio stream, convert it to text, translate it, convert it back to audio, and re-inject it into the stream). As such, we still see most audio translations being done with actual people.”

So “near real time” is a more accurate description, according to McDougall.

Ripe for Innovation

But regardless of how quickly and dynamically the technology acts, AI-powered automatic translation has emerged as a game changer for companies. And the advances are coming from many vendors, both large and small.

In November, Microsoft launched the gated public preview of Microsoft Translator Pro, which offers a stand-alone, native experience, enabling speech-to-speech translated conversations among coworkers, users, or clients. In a related announcement, Microsoft said it was enhancing Azure AI Translator’s Document Translation (DT) feature. Previously, Document Translation supported fully digital documents and scanned PDFs, but now the service can also process mixed-content documents, translating both digital text and text embedded within images.

Similarly, Meta announced in the fall that it was testing an AI translation tool that will automatically translate audio content on Reels to enable people who speak different languages to enjoy the same content.

“With automatic dubbing and lip syncing, Meta AI will simulate the speaker’s voice in another language and sync their lips to match,” the company said in a statement. “We’re starting with small tests on Instagram and Facebook, translating some creators’ videos from Latin America and the U.S. in English and Spanish, and we plan to expand this to more creators and languages.”

Google’s YouTube video streaming platform also expanded its audio dubbing feature in late 2024 to include information-focused content. The feature uses AI to automatically transcribe YouTube videos between English and other languages.

And even a smaller company like D-ID, which provides a platform for AI-generated video creation, recently launched an AI model that will translate videos into 30 languages.

“As video content becomes increasingly central to digital communication, the importance of engaging with a multilingual audience has never been more significant,” said Gil Perry, cofounder and CEO of D-ID, in a statement. “D-ID Video Translate is a game changer for anyone who wants to create engaging and accessible video content for a global audience without incurring significant costs, redefining how we communicate around the world.”

Best Practices

Such communications, though, must consider several other factors, according to Quéguiner. “Traditional speech recognition systems struggle with accents, background noise, and overlapping conversations. In multilingual settings, they must handle language recognition, translation, and code switching.”

McDougall recommends focusing on languages that you already support: “If I can call a business and they will already serve me in Mandarin and Spanish, then the first thing to do would be to automate some of the translations for Spanish and Mandarin because that’s the market you’re going after and they’re the larger languages.

“Once you’ve got that solved, then you can start to focus on other languages,” McDougall adds.

Machine translation technology vendors are also helping corporate clients address another pressing concern: the increasing need for vertical-specific solutions that can handle the unique terminologies, jargon, and language requirements that generic machine translation models might not handle effectively. To address this, the market is witnessing the development of specialized machine translation solutions tailored to specific domains, such as legal, medical, technical, financial, and e-commerce.

Another ongoing concern that must be addressed quickly is the hallucinations that can affect the accuracy and reliability of applications. Quéguiner recommends that companies implement robust guardrails to prevent hallucinations and recognize that LLMs still tend to favor English and might require additional fine-tuning on non-English language data.

But the issues that must be considered don’t end there.

“Lip synchronization needs precise alignment of timestamps in transcripts,” Quéguiner says. “Additionally, enhancing user experience is critical by providing clear feedback to the audience/caller during automated translation processes, using appropriate visual/audio cues for different mediums and implementing callback mechanisms for phone-based translation, such as filler words for a natural conversational flow.”

V. Frank Sondors, founder of Salesforge, providers of a sales outreach platform that combines human work with AI, cautions companies against solely relying on AI for translations without human oversight. This can result in awkward wording or misunderstandings. The businesses getting the best results use AI as a helpful tool but still incorporate human reviews during the final output, he says.

Other mistakes some companies make in adding AI-based translation are underestimating the cost and complexity of the project. McDougall notes that the cost of supporting each new language grows exponentially, so supporting several languages can become extremely costly.

Experts, though, still contend that the machine translation market is poised for remarkable growth and holds significant potential for further advancements. Among the areas that hold the most promise for businesses are the following:

• Real-time translation in augmented and virtual reality, enabling cross-cultural collaboration, gaming, education, tourism, and remote assistance.
• Contextual and intent-based translation incorporating advanced natural language processing, sentiment analysis, and machine learning algorithms, across various domains.
• Customization and personalization that will allow users to train and fine-tune translation models according to their specific requirements, preferences, language styles, domain-specific terminologies, and brand voice.
• Enhanced translation that combines multiple modalities, such as text, speech, images, and videos, simultaneously, for multimedia content, such as social media posts, advertisements, and video subtitles.
• Continuous learning and adaptation from user feedback, post-editing suggestions, and professional translators’ inputs to make systems more intelligent, accurate, and adaptable to evolving language patterns, new terminologies, and changing linguistic contexts.
• Ethical and responsible machine translation that involves addressing biases in training data, ensuring inclusivity and representation, adhering to privacy and data protection regulations, and developing transparent and explainable models that can be audited and verified to promote trust and accountability.

Overall, experts believe the machine translation market is vast and exciting. Advancements in artificial intelligence, deep learning, NLP, and multimodal processing will drive innovation, resulting in more accurate, context-aware, and user-centric translation solutions that cater to diverse industries, domains, and communication channels. 

Phillip Britt is a freelance writer based in the Chicago area. He can be reached at spenterprises1@comcast.net.