Laurent Besacier

Title: Innovations of Laurent Besacier

Introduction

Laurent Besacier is a notable inventor based in Seyssinet-Pariset, France. He has made significant contributions to the field of neural machine translation and text-to-speech technologies. With a total of 3 patents, his work is at the forefront of innovation in multilingual processing and speech generation.

Latest Patents

One of his latest patents is titled "Multilingual unsupervised neural machine translation with denoising adapters." This invention involves methods and systems for unsupervised training of a neural multilingual sequence-to-sequence (seq2seq) model. Denoising adapters for each language are integrated into the encoder and/or decoder of the seq2seq model. The parameters of these denoising adapters are trained on language-specific tasks using monolingual text. The model is then fine-tuned on translation tasks using parallel data.

Another significant patent is the "End-to-end neural text-to-speech model with prosody control." This invention focuses on generating an end-to-end neural text-to-speech (TTS) model that processes input text to produce speech representations. The TTS model is trained using an annotated dataset that includes prosodic features, allowing it to learn the association between prosody and the annotations.

Career Highlights

Laurent Besacier is currently employed at Naver Corporation, where he continues to push the boundaries of technology in his field. His work has garnered attention for its innovative approach to machine learning and artificial intelligence applications.

Collaborations

He collaborates with talented individuals such as Ioan Calapodescu and Inyoung Kim, contributing to a dynamic and innovative work environment.

Conclusion

Laurent Besacier's contributions to neural machine translation and text-to-speech technologies highlight his role as a leading inventor in the field. His patents reflect a commitment to advancing technology and improving communication across languages.