Text-Independent Cross-Language Voice Conversion for Speech-to-Speech Translation
David Sündermann,
Universitat Politècnica de Catalunya –
Abstract:
For applications like multi-user speech-to-speech translation, it is helpful to individualize the output voice to make voices distinguishable. Ideally, this should be done by applying the input speaker’s voice characteristics to the output speech.
In general, a speech-to-speech translation system consists of three main modules: speech recognition, text translation, and speech synthesis.
Since the latter, the speech synthesis module, normally is based on a large speech corpus of a professional speaker manually corrected and carefully tuned, the output voice characteristics are static. This is overcome by a fourth module, the voice conversion unit, which processes the synthesizer’s speech according to the input voice characteristics.
Due to the nature of speech-to-speech translation, input and output voices use different languages leading to the following two challenges:
(i) As opposed to state-of-the-art voice conversion, whose statistical parameter training is based on parallel utterances of both involved speakers (text-dependent approach), here we have to rely on text-independent parameter training: There is no way to produce parallel utterances in different languages;
(ii) Most voice conversion techniques estimate conversion functions that depend on the phonetic class, either explicitly (e.g. using CART) or implicitly (e.g. using GMM). However, considering different languages, we face different phoneme sets that make it hard to find conversion functions for phonetic units, which are not covered by the other phoneme set.
In this talk, I present text-independent voice conversion techniques that are cross-language portable and aim at solving these challenges. In this context, I will
(i) introduce a speech alignment technique based on unit selection dealing with non-parallel speech
and (ii) show that vocal tract length normalization, which is applied to convert the source voice towards the target, can be directly applied to the time frames without the detour through frequency domain.
The techniques’ performance is assessed on several multi-lingual corpora in the framework of subjective evaluations. In addition to the evaluation results, speech samples will be used to illustrate the discussed techniques’ effectiveness.
Date: 2006-Nov-17 Time: 15:30:00 Room: 336
For more information:
Upcoming Events
Seminar: Combining Reasoning and Learning for Discovery
07 June, 1.30pm, at Sala José Tribolet in Pavilhão Informática II at IST.
Artificial Intelligence (AI) is a rapidly advancing field inspired by human intelligence. AI systems are now performing at human and even superhuman levels on various tasks, such as image identification, face and speech recognition, and chatbots such as chatGPT. The tremendous AI progress that we have witnessed in the last decade has been largely driven by deep learning advances and heavily hinges on the availability of large, annotated datasets to supervise model training. However, often we only have access to small datasets and incomplete data. We amplify a few data examples with human intuitions and detailed reasoning from first principles and prior knowledge for discovery. I will talk about our work on AI for accelerating the discovery for new solar fuels materials, which has been featured in Nature Machine Intelligence, in a cover article entitled, Automating crystal-structure phase mapping by combining deep learning with constraint reasoning [1]. In this work, we propose an approach called Deep Reasoning Networks (DRNets), which seamlessly integrates deep learning and reasoning via an interpretable latent space for incorporating prior knowledge. and tackling challenging problems. DRNets requires only modest amounts of (unlabeled) data, in sharp contrast to standard deep learning approaches. DRNets reach super-human performance for crystal-structure phase mapping, a core, long-standing challenge in materials science, enabling the discovery of solar-fuels materials. DRNets provide a general framework for integrating deep learning and reasoning for tackling challenging problems. For an intuitive demonstration of our approach, using a simpler domain, we also solve variants of the Sudoku problem. The article DRNets can solve Sudoku, speed scientific discovery [2] provides a perspective for a general audience about DRNets. DRNets is part of SARA, the Scientific Reasoning Agent for materials discovery [3]. Finally, I will also talk about the effectiveness of a novel curriculum learning with restarts strategy to boost a reinforcement learning framework [4]. We show how such a strategy is characterized by left heavy-tails and can outperform specialized solvers for Sokoban, a prototypical AI planning problem.
Professor Carla P. Gomes: Department of Computer Science, Cornell University
Carla Gomes is the Ronald C. and Antonia V. Nielsen Professor of Computing and Information Science, the director of the Institute for Computational Sustainability at Cornell University, and co-director of the Cornell University AI for Science Institute. Gomes received a Ph.D. in computer science in artificial intelligence from the University of Edinburgh. Her research area is Artificial Intelligence with a focus on large-scale constraint reasoning, optimization, and machine learning. Recently, Gomes has become deeply immersed in research on scientific discovery for a sustainable future and, more generally, in research in the new field of Computational Sustainability. Computational Sustainability aims to develop computational methods to help solve some of the key environmental, economic, and societal challenges to help put us on a path toward a sustainable future. Gomes was the lead PI of two NSF Expeditions in Computing awards. Gomes has (co-)authored over 200 publications, which have appeared in venues spanning Nature, Science, and a variety of conferences and journals in AI and Computer Science, including five best paper awards. Gomes was named the “most influential Cornell professor” by a Merrill Presidential Scholar (2020). Gomes was also the recipient of the Association for the Advancement of Artificial Intelligence (AAAI) Feigenbaum Prize (2021) for “high-impact contributions to the field of artificial intelligence, through innovations in constraint reasoning, optimization, the integration of reasoning and learning, and through founding the field of Computational Sustainability, with impactful applications in ecology, species conservation, environmental sustainability, and materials discovery for energy” and of the 2022 ACM/AAAI Allen Newell Award, for contributions bridging computer science and other disciplines. Gomes is a Fellow of the Association for the Advancement of Artificial Intelligence (AAAI), a Fellow of the Association for Computing Machinery (ACM), and a Fellow of the American Association for the Advancement of Science (AAAS).
INESC-ID ESR Talks – June 2023
If you are a masters/PhD student or a postdoctoral fellow, come and present your work in an informal and friendly environment – and savour some tasty snacks!
Individual talks will be 10-15 minutes plus time for feedback. Enroll on your selected date by emailing pedro.ferreira[at]inesc-id.pt.
Happening on the second Wednesday of every month (4pm-5pm):
- 14 June (Alves Redol, Room 9)
- 12 July (Alves Redol, Room 9)
We hope to see you there!
OLISSIPO Summer School in Lisbon | Computational phylogenetics to analyse the evolution of cells and communities
We are happy to announce the OLISSIPO Summer School on Computational phylogenetics to analyse the evolution of cells and communities, which will be held in Lisbon, Portugal, at INESC-ID, between July 2-7, 2023.
Keynote speakers:
David Posada, University of Vigo (class)
João Alves, University of Vigo (hands-on)
Nadia El-Mabrouk, Université de Montréal (class)
Mattéo Delabre, Université de Montréal (hands-on)
Ran Libeskind-Hadas, Claremont McKenna College (class and hands-on)
Russell Schwartz, Carnegie Mellon University (class and hands-on)
See the preliminary agenda at: https://olissipo.inesc-id.pt/tree-tango-school
Registration is mandatory. You can register at: https://forms.gle/VsASFHW5E7MJvaCc9
The registration fee is 250€ for students and OLISSIPO members and 350€ for postdocs or other researchers (meals indicated at the schedule of the school are included, accommodation and flights are not). All details will be made available upon registration.
We will have slots for flash talks (3-10 min depending on the number of submissions) to present yourself and the work you have been developing in your research.
The 13th Lisbon Machine Learning School | LxMLS 2023
The Lisbon Machine Learning Summer School (LxMLS) takes place yearly at Instituto Superior Técnico (IST). LxMLS 2023 will be a 6-day event (14-20 July, 2023), scheduled to take place as an in-person event.
The school covers a range of machine learning topics, from theory to practice, that are important in solving natural language processing problems arising in different application areas. It is organized jointly by Instituto Superior Técnico (IST), a leading Engineering and Science school in Portugal, the Instituto de Telecomunicações, the Instituto de Engenharia de Sistemas e Computadores, Investigação e Desenvolvimento em Lisboa (INESC-ID), the Lisbon ELLIS Unit for Learning and Intelligent Systems (LUMLIS), Unbabel, Zendesk, and IBM Research.
Check online for information about past editions: LxMLS 2011, LxMLS 2012, LxMLS 2013, LxMLS 2014, LxMLS 2015, LxMLS 2016, LxMLS 2017, LxMLS 2018, LxMLS 2019, LxMLS 2020, LxMLS 2021, LxMLS 2022 (you can also watch the videos of the lectures for 2016, 2017, 2018, and 2020).
31st International Conference on Information Systems Development (ISD 2023)
The 31st International Conference on Information Systems Development (ISD 2023) conference provides a forum for research and developments in the field of information systems. The theme of ISD 2023 is “Information systems development, organizational aspects and societal trends”. New trends in developing information systems emphasize the continuous collaboration between developers and operators in order to optimize the software delivery time. The conference promotes research on methodological and technological issues and how IS developers and operators are transforming organizations and society through information systems.
The ISD 2023 conference held this year also provides an opportunity for researchers and practitioners to promote their research, practical experience, and to discuss issues related to Information Systems through papers, posters, and journal-first paper presentations.
ISD 2023 will be hosted by Instituto Superior Técnico, in Lisbon, Portugal, on August 30–September 1, 2023.