How Artificial Intelligence Is Helping to Translate Animal Languages

Which could lead to improvements in human language translations

  • Researchers are making progress in using artificial intelligence to translate animal sounds into human speech. 
  • AI is taking on many human-to-human translation tasks. 
  • But some experts say the nuances in speech mean there’s a long way to go before AI can replace human translators.
Two people sitting on a couch with coffee cups and a dog looking at one of them.

svetikd / Getty Images

Your dog's bark might one day be as understandable as human speech, thanks to recent advances in artificial intelligence (AI). 

Researchers are applying machine learning techniques to translate the sounds animals make. It's part of the rapidly expanding field of AI translation.  

"AI helps the translation process in two main ways," Nataly Kelly, the author of the book "Found in Translation: How Language Shapes Our Lives and Transforms the World," told Lifewire in an email interview. "First, by improving the grammar and flow of the source text. Second, by automating the translation process fully. Products like Grammarly can improve the original writing, while online tools such as Google Translate automate the translation."

Will Everyone Be Doctor Doolittle?

The nonprofit Earth Species Project recently published a roadmap that it hopes will enable researchers to translate animal speech using AI machine learning techniques. 

"We hypothesize that ML models can provide new insights into non-human communication, and believe that discoveries fueled by these new techniques have the potential to transform our relationship with the rest of nature," the organization wrote in the blog post about the roadmap. "We also expect that many of the techniques we develop will be useful tools for processing data in applied conservation settings." 

AI models can learn from large amounts of human translations and "do pretty well" translating between many languages automatically today, Wei Xu, a natural language processing researcher at Georgia Tech, told Lifewire in an email. Xu said that Google Translate is an excellent example of such a model.

"For machine translation, the primary technique used is a neural network architecture called 'sequence-to-sequence,' also known as 'encoder-decoder,' which is often based on the Transformer model first developed at Google," Xu said.

Despite the advances, many human translators tend to sniff at their computerized counterparts. Alain J. Roy, the president of ASTA-USA Translation Services, said via email that within the translation industry, AI translation technologies "are still regarded as fairly infantile at their core. They are utilized as a tool to streamline and improve the efforts of human translators—particularly when translating materials into a multitude of languages simultaneously—but they cannot replace the human element when accuracy and localization are imperative."

We expect in the future AI interpretation to be more real-time, more accurate, and more capable at interpreting a lot more languages...

Human Language Translation Goes High-Tech

The future looks bright for AI translation. Steven Toy, the CEO of Memrise, a company that uses AI to teach foreign languages, said in an email interview that advances in AI would allow better and faster translation results. He pointed out that AI systems are only as good as their training data. 

"There is a lot of data for English, but less so for other languages, and that which is out there hasn't yet been fed to computers," Toy said. "As these systems consume that data, they can move between two languages faster and more precisely."

Another area expected to improve soon, thanks to AI, is 'sentiment analysis,' which will allow for more nuanced translations. 

"Words have different meanings based on their context and surrounding words, and as computers evolve to take this into account, they will be able to more precisely interpret the intended meaning of the source words and offer better and more precise translations," Toy said.

Business people sitting at a table with laptops, talking on headsets.

ljubaphoto / Getty Images

Xu said that while AI can impressively translate between 100 or so widely spoken languages (e.g., English, French, Chinese), it is not flawless and can not replace professional human translators at high stake scenarios. 

"It can not yet work as well as the simultaneous interpreter at United Nations meetings," Xu added. "However, we are getting there. We expect in the future AI interpretation to be more real-time, more accurate, and more capable at interpreting a lot more languages that are spoken by a smaller population."

But Roy said the technology still has a long way to go before it can be viable for conducting translations or interpretations about critical business, legal, certified, or governmental affairs. 

"This is because AI technology cannot take matters like cultural and colloquial subtleties or nuances into account when generating translations," Roy added. "Likewise, AI does not have the capacity to understand context, tone, or style—all of which are pertinent in business, legal, or other certified correspondences."

Was this page helpful?