Dialogue & Speech
These technologies facilitate a natural interaction between a person and a computer and the methodologies provide the concepts, techniques and tools for speech processing via digital processing of the signal. Thanks to recent advances in Artificial Intelligence, the practical application of technologies –such as dialogue systems or speech recognition and synthesis in multiple sectors– is increasingly feasible, improving Human Computer Interaction or the processing and use of digital content in multiple languages, including Basque.
Dialogue Systems, Chatbots and Digital Assistants
Intelligent digital assistants are one of the most disruptive and enabling technologies in the new generation of solutions based on Artificial Intelligence. Thanks to Deep Learning and algorithms based on stochastic processes, they are able, among other things, to understand the user’s needs, extract their profile and generate recommendations taking the context into account. Given the transversal nature of the conversational voice assistants, they can be adapted to multiple domains (medical, administrative, commercial, business, industrial, etc.), creating smart interfaces allowing a more natural, direct and intuitive interaction with technology.
Automatic Transcription and Subtitling
Our team is highly specialised scientifically and has multiple cases of real transference and international experience at the highest level in technologies for enriched transcription and automatic subtitling of video and audio in multiple languages and operational modes (offline and online), technology based on constantly evolving proprietary Transkit library. These technological assets based on Deep Learning techniques have been applied in several scenarios with high technological challenges, such as telephone conversations, television content, public transparency portals, parliamentary sessions, meeting transcriptions, security environments, etc.
Voice Synthesis, Vocal Biometrics, Emotions, etc.
In Dialogue and Speech there are other technological assets allowing the development of relevant applications for the identified sectors. As in our speech recognition systems, the End-to-End architectures of our speech synthesisers allow us to generate natural and expressive synthetic voices in multiple languages or recognise emotions through speech. Also, out BioVoice library incorporates the functionalities to train biometric voice systems to recognise or verify the identity of a speaker.
- Noteworthy Projects
The Vicomtech-PRHLT Speech Transcription Systems for the IberSPEECH-RTVE 2018 Speech to Text Transcription Challenge
Online Learning of Attributed Bi-Automata for Dialogue Management in Spoken Dialogue Systems
Improving the Automatic Segmentation of Subtitles through Conditional Random Field