Since the introduction of Google Translate 13-years ago, methodologies such as neural machine translation, rewriting-based paradigms, and on-device processing have resulted in verifiable improvements in translation accuracy. The company’s developments have considerably improved the user experience in the 108 languages supported by Google Translate, with an average translation of 150 billion words daily. Google’s translation breakthroughs didn’t happen because of a single technological advancement, but rather because of the use of various technical approaches aimed at providing translations for low-resource, high-resource, general quality, and overall processing speed. Translate increased an average of five or more points across all languages, seven or more across the 50 lowest-resource languages from May 2019 to May 2020, as evaluated by human assessments and BLEU, a score based on the similarity between a system’s translation and human reference translations. In addition, Google has found that the machine translation hallucination problem has been lessened, making it easier for users to trust machine translations more.
At the company’s developer conference last month, Google I/O 2021, the company unveiled LaMDA, a new Language Model for Dialogue Applications. The application demonstrated the groundbreaking power of Artificial Intelligence (AI) in language learning. At the conference Google CEO, Sundar Pichai explained how natural language processing and transformer models have achieved numerous achievements and how Google has been unrelentingly striving to improve it.
Further working on this idea, Google has recently announced an AI language model that will enable people to have open-ended discussions with technology. However, people who have worked with the language model say making fluid, consistent and accurate human-to-technology interaction possible is still a considerable challenge.
For decades, AI researchers have attempted to build fluid communication between machines and humans that seems genuine, captures the nuances of human communication and streamlines chores. According to a report by Meticulous Research, the online language learning market is expected to reach $21.2 billion by 2027, growing at a CAGR of 18.7 percent.
Until now, Google has been fantastic in language translation, allowing users to web search in multiple languages, translating web pages, and proving accurate translation of phrases. Now, the company intends to use this ability to help its users learn languages. People familiar with the project say that Google will start using language learning within Google Search to address foreign language acquisition. According to The Information, the project, named Tivoli internally, emerged from the Google Research unit and is expected to launch later this year.
Tivoli will initially work over text and later be integrated into the company’s voice assistant and YouTube product lines. For example, it may offer language quizzes on YouTube in which viewers can videotape themselves after watching a video, and the AI will rate their performances.
Teaching foreign languages enables Google to move beyond superficial conversations and get into a viable but low-stakes argument for more fluid and conversational AI. Using the incorrect tense or phrase is unlikely to cause users serious harm.
Google is not new to language learning. In 2019, the tech giant had launched The Read-Along app, called Bolo, for Indian audiences, which aimed to teach children languages– English, Hindi, Spanish, and Portuguese. The app provided visual and audio cues for youngsters to help them learn a foreign language. It accelerated learning by featuring stories that focused on different topics. Bolo’s virtual assistant Diya analysed the text to discern whether a pupil has trouble pronouncing a word or phrase and provided them with appropriate suggestions. On the other hand, the Tivoli project looks to launch a language learning model at a broader and more integrated scale.
However, Google is not alone in its effort to launch an AI language model. Popular language learning platform Duolingo stands as a strong competitor against Google with its massive 300 million user base. Besides Duolingo, Google might face intense competition from key players like Berlitz Corporation, Rosetta Stone Inc, Memrise Inc, Inlingua International Ltd., Babbel, Busuu Ltd, iTutor Group, Open Education LLC, Linguistica 360, Inc., Mondly Languages, Sanako Corporation, and Mango Languages.
There has been no official release date for Tivoli. Still, Google hopes that the AI-based language learning project reaches a more extensive user base than any of the foreign language learning solutions available at the moment, through its existing Google services.