Our Analysts' Insights

Most people frown upon the idea of talking to machines, yet they are already encountering them in everyday conversations. For example, when you call a bank or utility company, you probably start the conversation with a bot that triages the request and escalates the conversation to a human call agent when it reaches the limits of its capabilities. There is no longer a need for a human to tell you your credit card balance. Likewise, there is no need for a human interpreter if a nurse checks on the temperature of a patient or if an airline attendant is helping you process a basic flight rebooking. This is where machine interpreting comes into play.

What Is Machine Interpreting?

Machine interpreting (MI), often referred to as spoken translation, is the stepchild of machine translation applied to spoken content. While we include it as an interpreting modality, the intent is not to pit it against professional human interpreters but rather to simply explain its usage in contrast to traditional interpreting modes. 

MI is the output resulting from an amalgam of four technologies: 1) speech recognition to convert the speech to text; 2) transcript cleanup and normalization to make the text machine-ready; 3) machine translation to translate the text into the desired language; and 4) speech synthesis to voice the translation. 

Although MI is a voice-to-voice communication method, the process produces by-products that end up more useful in the eyes of many clients. Voice-to-script enables speech to last longer than its utterance and be searchable for future use. Voice-to-caption provides increased accessibility for the deaf and hard of hearing. And voice-to-subtitle is the key to language access – whether for recorded or live events. The full speech-to-speech roundtrip tends to be more limited to blind or visually impaired users or those otherwise occupied and unable to read (for example, people driving cars or operating machinery). 

State of the Market for Machine Interpreting

The market for machine interpreting remains very small. However, we’re seeing an uptick in terms of user comfort with and acceptance of MI technology, thanks in great part to the advances made with speech recognition and neural machine translation. In a survey of 40 enterprises in the second half of April, we found that 15% of them already use MI and 5% plan to do so. In addition, technology providers are reporting an uptick in interest from a broad range of client types.

• Media coverage of MI wearables is common. Techies like to talk about ear buds, watches, and other wearables enabled with the ability to translate conversations in real time. However, because these devices target mostly mass consumers, the interest is more akin to gadgets that come and go. MI technology vendors have not yet managed to create evangelists that can convince the market to replicate their use cases and adopt the new functionality for everyday use. In other cases, developers in large enterprises have struggled to convey the value of their work to their corporate leaders, leading their work to sit on the sidelines.
   
• App stores offer plenty of “pocket translators.” These solutions appeal to tourists or people who deal with language access issues for basic conversations on a daily basis. Examples include Google Translate, iTranslate, SayHi, Speak & Translate, TranslateMe, TripLingo, and Voice Translator. While some of these apps have great download numbers, their usage and market share remain modest.
   
• Other MI offerings still struggle to go mainstream. Standalone applications focus on niche markets and traction is slow despite some great successes. Integrated solutions in conference and video chats – such as with Skype – are more common, but many of these are difficult to set up and use. Mainstream adoption of such tools once again remains in its infancy. However, all of these options help build awareness that will boost MI adoption in the long run.

Impact of COVID-19 on Machine Interpreting

Machine interpreting remains a sci-fi concept for most, but COVID-19 presents a noteworthy opportunity for the technology. While it may not gain as much traction as remote interpreting solutions that rely on professional human interpreters, it is catching the eye of organizations with limited budgets or that are thinking outside the box to solve the complexities of dealing with the pandemic.

• Some firms are just discovering the solution. Due to the adoption of virtual meeting platforms and videoconferencing products, some organizations stumble upon built-in features to automatically translate speech. Others are more proactively seeking solutions to address a need not easily met by other interpreting modalities or that are too costly for them given the basic quality they need. For companies in this group, we recommend that you test tools such as Wordly.ai, which can offer a great way to add multilingual capability to meetings to boost comprehension when participants possess only an imperfect knowledge of the presentation language.
   
• Some vendors offer their technology for free during the pandemic. Developers such as iTranslate and TranslateLive are making a version of their product available to help on the frontlines of fighting the virus. They allow you to take your own device into quarantine and use it for communication without wait times. This is a smart move from companies trying to bring awareness to this type of solution. However, it’s also a great opportunity for enterprises to experiment with the technology and identify potential use cases. Organizations just have to remember that MI shouldn’t be about replacing human interpreters but adding language support when otherwise there is none.

What’s the Future?

COVID-19 may turn out to be a game changer for machine interpreting, but more so in the long-term as users may start to build some tolerance for the shortcomings of MI interpreting, especially in cases where the alternative is no access to language support.

• The pandemic will bring awareness. At the same time, technology vendors with MI business solutions are few and far between and overall are not prepared to leverage the scenario for deeper market penetration.
   
• Proliferation will depend on the pace of progress. Giant technological strides have been made, but many more are required for users to believe in the tools outside of times of crisis. MI developers have yet to produce a “killer app” that is able to take the technology to the next step that can be used in business, education, healthcare, or travel.
   
• Mainstream success will come from integrations. High-usage conferencing and messaging tools are the most likely candidates as they enable real-time communication. This is particularly useful when organizations don’t want to invest in professional interpreting, but access could improve for deaf, heard-of hearing, or individuals with imperfect knowledge of the language being used.

MI has long been in the public eye, thanks to Star Trek and other science fiction franchises, but it is only now that it is becoming a mature and usable technology that responds to real-world needs. Its usability is rapidly improving and it will enable organizations and individuals to extend language support into environments where they would previously have left much of their audience unable to interact and communicate.