Translated, the founding company of Pi Campus, has been selected by the European Parliament to automatically transcribe and translate parliamentary multilingual debates in real-time, covering the 24 official languages used by the institution. The purpose of this endeavor is to facilitate communication between members, make debates transparent to all EU citizens, and provide deaf and hard of hearing people, who currently have no direct access to debates, with a tool to follow the discussions.
It will be the first human-in-the-loop speech translation system: during the debates, members’ audio will be transcribed and translated in real-time and distributed via fully-localized web and mobile applications and live streaming APIs for third-party developers. This will be the first human-in-the-loop speech translation system: a team of professional translators will provide feedback on the translation that machines will take into consideration in less than 1 second. The system is not limited to sentences but adapts to the context to improve the translation.
The product will be developed in collaboration with two companies that have already worked with Translated in building innovative products for professional translators: Fondazione Bruno Kessler (FBK), a world-leading research center in MT and automatic speech recognition (ASR); and PerVoice, an ASR world-leading provider. Within the next 12 months, the Italian consortium will release a prototype to be tested by the European Parliament together with up to two other solutions as per the rules of the call for tender. The best-performing tool will be confirmed as the official one for the next 2 years.
The new product is not a simple concatenation of ASR and MT, but a new, fully integrated system where the MT algorithms are tolerant of ASR errors, this way avoiding propagating imperfections. This approach will not only help deliver much more contextualized translations, but it will also present an opportunity to improve the quality of the output while the plenary session is happening. This is possible thanks to the human correction feedback that is facilitated by the tool, allowing end-users and the team of professional translators to share corrections.
«For this project, we are bringing together 10 years of research in machine translation and speech recognition that allows our system to achieve great results in terms of performance and translation quality,» says Translated VP of Product Management Simone Perone. Some of the new AI models that will be used have already been put at work successfully in products such as ModernMT (an MT that improves from corrections and adapts to the context, named a Gartner Cool Vendor in 2019), Matecat (a computer-assisted translation tool that makes post-editing easy), and Matesub (the first subtitling tool offering suggestions during transcription, now in beta and due to be released in September 2020).
Envisaged for decades, speech translation technology is now developed enough to be put to use. Not only is it real-time, robust, and language-independent, it also supports all EU languages, all speakers, all domains as well as both spontaneous and formal speech. «We are proud of being selected by the European Parliament and grateful for the opportunity to move forward with our mission: open up language to everyone – says Translated co-founder and CEO Marco Trombetti – All the tools developed by Translated over the years have been based on the belief that human creativity is precious and software must be designed to unlock it. We want to take a step further in this direction, allowing professional translators to add to the transcription and translation of a speech those nuances that only a human ear can catch.»