Alexa’s text-to-speech research at Interspeech 2022

Highlighted papers focus on transference — of prosody, accent, and speaker identity.

Interspeech, the world’s largest and most comprehensive conference on the science and technology of spoken-language processing, took place last week in Incheon, Korea, with Amazon as a platinum sponsor. Amazon Science asked three of Alexa AI’s leading scientists — in the fields of speech, spoken-language-understanding, and text-to-speech — to highlight some of Amazon’s contributions to the conference.

In this installment, Antonio Bonafonte, a senior applied scientist in the Amazon Text-to-Speech group, highlights work on transference — of prosody, accent, and speaker identity — in text-to-speech.

This year, the Amazon Text-to-Speech organization presented more than a dozen papers at Interspeech 2022. Amazon TTS gives Alexa its voice, working every day to add more expressiveness and conversational awareness. Here we highlight some of papers that illustrate what we are doing in those directions.

Expressive and contextually appropriate prosody

Neural text-to-speech (TTS) techniques have made the speech produced by TTS systems much more natural. To make the prosody of the speech more expressive and context appropriate as well, researchers have done considerable work on learning prosody representations from ground-truth speech.

The paper “CopyCat2: A single model for multi-speaker TTS and many-to-many fine-grained prosody transfer”, by Sri Karlapati and coauthors, proposes a model that learns word-level speaker-independent prosody representations from multispeaker speech. These representations can be used for fine-grained prosody transfer from multiple source speakers to multiple target speakers. Furthermore, predicting the word-level prosody representations from text results in a TTS model with improved naturalness and appropriateness.

The CopyCat2 architecture.

The word-level prosodic representation is split into two components, one for timing and rhythm and a second for other prosodic characteristics. The figure above shows how the second component is learned using a conditional variational autoencoder. The input mel-spectrogram (X), which represents the speech signal as energies in certain frequency bands, is compressed into a sequence of vectors (Z), one per word. Those vectors are then used to reconstruct the mel-spectrogram.

Related content
New voice for Alexa’s Reading Sidekick feature avoids the instabilities common to models with variable prosody.

The decoder is conditioned on the phonemes and the speaker, so it captures speaker-independent prosody information, and a similar approach is used to learn speaker-independent word-level representations of timing aspects.

To use CopyCat2 as a text-to-speech model, the researchers train an additional model to predict the parameters of the prosodic-word-embedding distribution (Z) from BERT embeddings. In tests involving a multispeaker US English dataset of varied styles, including news, facts, and greetings, they compared their approach to a strong TTS baseline with contextually appropriate prosody and copy-synthesized speech. They found that their model reduced the gap in naturalness between synthetic and real speech by 22.79%.

Reducing the data required to build expressive voices

Training a state-of-the-art TTS model is usually a data-intensive process, and building a portfolio of voices in multiple styles and languages compounds the data requirement.

In the paper “Low-data? No problem: low-resource, language-agnostic conversational text-to-speech via F0-conditioned data augmentation”, Giulia Comini et al. propose a methodology to build expressive text-to-speech voices using only one hour of expressive speech from the target speaker. The method requires 8–10 hours of neutral speech — that is, speech with a limited range of expression — from another speaker, a significant reduction from previous methods.

Low data.png
A new approach to building expressive text-to-speech voices can make do with only an hour of expressive speech from the target speaker.

The authors propose to convert the neutral data from the supporting speaker to the target-speaker identity, while maintaining the target speaker’s expressive style. They use a modification of the original CopyCat prosody transfer model. As shown in the figure, the CopyCat parallel decoder regenerates the mel-spectrogram from the speaker embedding; the fundamental frequency (F0), or perceived pitch of individual phonemes; the phonetic representation; and the output of the CopyCat reference encoder. The reference encoder captures the information from the source mel-spectrogram that is not explicitly given to the decoder, — i.e., phonemes, with their duration and F0, and the speaker embedding.

Related content
Users find speech with transferred expression 9% more natural than standard synthesized speech.

The model is trained with the expressive speech of the target speaker and neutral speech from the supporting speaker. Once the model is trained, the mel-spectrogram of the supporting data is transformed into augmented expressive data for the target speaker. The CopyCat decoder is conditioned on the target speaker embedding and on an expressive F0 contour generated from the text and the speaker embedding by an independent model trained with the same data.

The paper shows that the F0 distribution of the augmented data resembles that of the target speaker. They also show that their data augmentation approach improves on one that does not use F0 conditioning.

Alexa multilingual models

Amazon has developed a shared neural TTS model for several speakers and languages that can extend a synthetic voice trained on data in only one language into other languages. For instance, the technology allows the English-language Alexa feminine-sounding voice to speak fluent Spanish in US multilingual homes. Similarly, Alexa’s English-language US masculine-sounding voice already has a British accent in the UK and speaks Spanish in the US, French in Canada, and German in Germany.

Related content
Neural text-to-speech enables new multilingual model to use the same voice for Spanish and English responses.

Alexa communicates on a wide variety of topics, and the style of speech should match the textual content. Transferring styles across languages while maintaining a fixed speaker identity, however, is challenging.

In the paper “Cross-lingual style transfer with conditional Prior VAE and style loss”, Dino Ratcliffe et al. propose an architecture for cross-lingual style transfer. Specifically, they improve the Spanish representation across four styles — newscaster, DJ, excited, and disappointed — while maintaining a single speaker identity for which only English samples are available.

Cross-lingual style transfer.png
A new approach to cross-lingual style transfer groups utterances of the same style together irrespective of language.

They achieve this by using a learned-conditional-prior variational autoencoder (LCPVAE), a hierarchical variational-autoencoder (VAE) approach.

The approach introduces a secondary VAE, which is conditioned on one-hot-encoded style information; that is, the style code has as many bits as there are styles, and a 1 at exactly one spot denotes a particular style. This results in a structured embedding space, which groups together utterances of the same style irrespective of language.

Related content
Papers focus on speech conversion and data augmentation — and sometimes both at once.

As can be seen in the figure, the TTS decoder generates the mel-spectrogram from the speaker embedding, language, phonemes, and the style embedding. During training, the style embeddings are generated by the LCPVAE using the one-hot code and the reference mel-spectrogram; at inference, the style embedding is the centroid of the embeddings for a particular style. The model’s loss function includes a style classification term that steers the generated mel-spectrogram toward the same style as the reference spectrogram.

Based on subjective evaluations (MUSHRA), this approach shows significant improvements on cross-lingual style representation in all four styles, DJ (2.8%), excited (5.3%), disappointed (3.5%) and newscaster (2.3%), without compromising speaker similarity and in-lingual style representation.

Creating new characters

Current TTS technology can produce realistic synthetic speech for sample voice identities seen during training. But speech synthesis with speakers unseen during training, without post-training adaptation, remains a big challenge. Synthesis with a new voice often means creating high-quality data to train a generative model.

Related content
Thanks to a set of simple abstractions, models with different architectures can be integrated and optimized for particular hardware accelerators.

Normalizing flows are generative models with tractable distributions, where sampling and density evaluation can be both exact and efficient. In “Creating new voices using normalizing flows”, Piotr Biliński and his colleagues investigate the ability of normalizing flows in TTS and voice conversion modes to extrapolate from speakers observed during training to unseen speaker identities — without any recordings of those speakers, and therefore without the possibility of target speaker adaptation.

Their approach is based on the Flow-TTS model, but instead of using it to generate synthetic speech of seen speakers, they adapted it to create new voices. Key contributions include adding the ability to sample new speakers, introducing voice conversion mode, and comparing it to TTS mode.

Normalizing flows.png
Instead of using normalizing flows to synthesize the speech of seen speakers, Amazon researchers adapted them to create new voices.

The architecture of the model consists of an invertible transformation based on normalizing flows. The design allows for lossless reconstruction of a mel-spectrogram from a representational space (z) given conditions (θ) such as speaker embedding. In text-to-speech mode, sampling z from the prior distribution and running the inverse transformation allows us to generate the mel-spectrogram given the conditions θ.

To apply the model in voice conversion mode, we map the source mel-spectrogram to a latent representation z using as condition the source-speaker embedding. Then, the latent representation z is converted back to a mel-spectrogram using the speaker embedding of the target speaker. To generate speaker embeddings of new voices, we train a separate neural network that generates plausible speaker embeddings for a given regional English variant.

Extensive evaluations demonstrate that the proposed approach systematically obtains state-of-the-art performance in zero-shot speech synthesis and allows us to create voices distinct from those in the training set. In addition, the authors find that as the level of conditioning to the model is increased, voice conversion and TTS modes can be used interchangeably.

Related content

US, WA, Seattle
Are you excited about building high-performance robotic systems that can perceive, learn, and act intelligently alongside humans? The Robotics AI team is creating new science products and technologies that make this possible, at Amazon scale. We work at the intersection of computer vision, machine learning, robotic manipulation, navigation, and human-robot interaction.The Amazon Robotics team is seeking broad, curious applied scientists and engineering interns to join our diverse, full-stack team. In addition to designing, building, and delivering end-to-end robotic systems, our team is responsible for core infrastructure and tools that serve as the backbone of our robotic applications, enabling roboticists, applied scientists, software and hardware engineers to collaborate and deploy systems in the lab and in the field. Come join us!
US, VA, Arlington
The Central Science Team within Amazon’s People Experience and Technology org (PXTCS) uses economics, behavioral science, statistics, and machine learning to proactively identify mechanisms and process improvements which simultaneously improve Amazon and the lives, well-being, and the value of work to Amazonians. We are an interdisciplinary team, which combines the talents of science and engineering to develop and deliver solutions that measurably achieve this goal. As Director for PXT Central Science Technology, you will be responsible for leading multiple teams through rapidly evolving complex demands and define, develop, deliver and execute on our science roadmap and vision. You will provide thought leadership to scientists and engineers to invent and implement scalable machine learning recommendations and data driven algorithms supporting flexible UI frameworks. You will manage and be responsible for delivering some of our most strategic technical initiatives. You will design, develop and operate new, highly scalable software systems that support Amazon’s efforts to be Earth’s Best Employer and have a significant impact on Amazon’s commitment to our employees and communities where we both serve and employ 1.3 million Amazonians. As Director of Applied Science, you will be part of the larger technical leadership community at Amazon. This community forms the backbone of the company, plays a critical role in the broad business planning, works closely with senior executives to develop business targets and resource requirements, influences our long-term technical and business strategy, helps hire and develop engineering leaders and developers, and ultimately enables us to deliver engineering innovations.This role is posted for Arlington, VA, but we are flexible on location at many of our offices in the US and Canada.
US, VA, Arlington
Employer: Services LLCPosition: Data Scientist IILocation: Arlington, VAMultiple Positions Available1. Manage and execute entire projects or components of large projects from start to finish including data gathering and manipulation, synthesis and modeling, problem solving, and communication of insights and recommendations.2. Oversee the development and implementation of data integration and analytic strategies to support population health initiatives.3. Leverage big data to explore and introduce areas of analytics and technologies.4. Analyze data to identify opportunities to impact populations.5. Perform advanced integrated comprehensive reporting, consultative, and analytical expertise to provide healthcare cost and utilization data and translate findings into actionable information for internal and external stakeholders.6. Oversee the collection of data, ensuring timelines are met, data is accurate and within established format.7. Act as a data and technical resource and escalation point for data issues, ensuring they are brought to resolution.8. Serve as the subject matter expert on health care benefits data modeling, system architecture, data governance, and business intelligence tools. #0000
US, TX, Dallas
Employer: Services LLCPosition: Data Scientist II (multiple positions available)Location: Dallas, TX Multiple Positions Available:1. Assist customers to deliver Machine Learning (ML) and Deep Learning (DL) projects from beginning to end, by aggregating data, exploring data, building and validating predictive models, and deploying completed models to deliver business impact to the organization;2. Apply understanding of the customer’s business need and guide them to a solution using AWS AI Services, AWS AI Platforms, AWS AI Frameworks, and AWS AI EC2 Instances;3. Use Deep Learning frameworks like MXNet, PyTorch, Caffe 2, Tensorflow, Theano, CNTK, and Keras to help our customers build DL models;4. Research, design, implement and evaluate novel computer vision algorithms and ML/DL algorithms;5. Work with data architects and engineers to analyze, extract, normalize, and label relevant data;6. Work with DevOps engineers to help customers operationalize models after they are built;7. Assist customers with identifying model drift and retraining models;8. Research and implement novel ML and DL approaches, including using FPGA;9. Develop computer vision and machine learning methods and algorithms to address real-world customer use-cases; and10. Design and run experiments, research new algorithms, and work closely with engineers to put algorithms and models into practice to help solve customers' most challenging problems.11. Approximately 15% domestic and international travel required.12. Telecommuting benefits are available.#0000
US, WA, Seattle
MULTIPLE POSITIONS AVAILABLECompany: AMAZON.COM SERVICES LLCPosition Title: Manager III, Data ScienceLocation: Bellevue, WashingtonPosition Responsibilities:Manage a team of data scientists working to build large-scale, technical solutions to increase effectiveness of Amazon Fulfillment systems. Define key business goals and map them to the success of technical solutions. Aggregate, analyze and model data from multiple sources to inform business decisions. Manage and quantify improvement in the customer experience resulting from research outcomes. Develop and manage a long-term research vision and portfolio of research initiatives, with algorithms and models that to be integrated in production systems. Hire and mentor junior is an Equal Opportunity-Affirmative Action Employer – Minority / Female / Disability / Veteran / Gender Identity / Sexual Orientation #0000
US, VA, Arlington
MULTIPLE POSITIONS AVAILABLECompany: AMAZON.COM SERVICES LLCPosition Title: Data Scientist IILocation: Arlington, VirginiaPosition Responsibilities:Design and implement scalable and reliable approaches to support or automate decision making throughout the business. Apply a range of data science techniques and tools combined with subject matter expertise to solve difficult business problems and cases in which the solution approach is unclear. Acquire data by building the necessary SQL / ETL queries. Import processes through various company specific interfaces for accessing Oracle, RedShift, and Spark storage systems. Build relationships with stakeholders and counterparts. Analyze data for trends and input validity by inspecting univariate distributions, exploring bivariate relationships, constructing appropriate transformations, and tracking down the source and meaning of anomalies. Build models using statistical modeling, mathematical modeling, econometric modeling, network modeling, social network modeling, natural language processing, machine learning algorithms, genetic algorithms, and neural networks. Validate models against alternative approaches, expected and observed outcome, and other business defined key performance indicators. Implement models that comply with evaluations of the computational demands, accuracy, and reliability of the relevant ETL processes at various stages of is an Equal Opportunity-Affirmative Action Employer – Minority / Female / Disability / Veteran / Gender Identity / Sexual Orientation #0000
US, IL, Chicago
MULTIPLE POSITIONS AVAILABLECompany: AMAZON.COM SERVICES LLCPosition Title: Data Scientist ILocation: Chicago, IllinoisPosition Responsibilities:Build the core intelligence, insights, and algorithms that support the real estate acquisition strategies for Amazon physical stores. Tackle cutting-edge, complex problems such as predicting the optimal location for new Amazon stores by bringing together numerous data assets, and using best-in-class modeling solutions to extract the most information out of them. Work with business stakeholders, software development engineers, and other data scientists across multiple teams to develop innovative solutions at massive is an Equal Opportunity-Affirmative Action Employer – Minority / Female / Disability / Veteran / Gender Identity / Sexual Orientation #0000
US, WA, Seattle
Are you motivated to explore research in ambiguous spaces? Are you interested in conducting research that will improve the employee and manager experience at Amazon? Do you want to work on an interdisciplinary team of scientists that collaborate rather than compete? Join us at PXT Central Science!The People eXperience and Technology Central Science Team (PXTCS) uses economics, behavioral science, statistics, and machine learning to proactively identify mechanisms and process improvements which simultaneously improve Amazon and the lives, wellbeing, and the value of work to Amazonians. We are an interdisciplinary team that combines the talents of science and engineering to develop and deliver solutions that measurably achieve this goal.We are seeking a senior Applied Scientist with expertise in more than one or more of the following areas: machine learning, natural language processing, computational linguistics, algorithmic fairness, statistical inference, causal modeling, reinforcement learning, Bayesian methods, predictive analytics, decision theory, recommender systems, deep learning, time series modeling. In this role, you will lead and support research efforts within all aspects of the employee lifecycle: from candidate identification to recruiting, to onboarding and talent management, to leadership and development, to finally retention and brand advocacy upon exit.The ideal candidate should have strong problem-solving skills, excellent business acumen, the ability to work independently and collaboratively, and have an expertise in both science and engineering. The ideal candidate is not methods-driven, but driven by the research question at hand; in other words, they will select the appropriate method for the problem, rather than searching for questions to answer with a preferred method. The candidate will need to navigate complex and ambiguous business challenges by asking the right questions, understanding what methodologies to employ, and communicating results to multiple audiences (e.g., technical peers, functional teams, business leaders).About the teamWe are a collegial and multidisciplinary team of researchers in People eXperience and Technology (PXT) that combines the talents of science and engineering to develop innovative solutions to make Amazon Earth's Best Employer. We leverage data and rigorous analysis to help Amazon attract, retain, and develop one of the world’s largest and most talented workforces.
US, WA, Bellevue
Job summaryThe Global Supply Chain-ACES organization aims to raise the bar on Amazon’s customer experience by delivering holistic solutions for Global Customer Fulfillment that facilitate the effective and efficient movement of product through our supply chain. We develop strategies, processes, material handling and technology solutions, reporting and other mechanisms, which are simple, technology enabled, globally scalable, and locally relevant. We achieve this through cross-functional partnerships, listening to the needs of our customers and prioritizing initiatives to deliver maximum impact across the value chain. Within the organization, our Quality team balances tactical operation with operations partners with global engagement on programs to deliver improved inventory accuracy in our network. The organization is looking for an experienced Principal Research Scientist to partner with senior leadership to develop long term strategic solutions. As a Principal Scientist, they will lead critical initiatives for Global Supply Chain, leveraging complex data analysis and visualization to:a. Collaborate with business teams to define data requirements and processes;b. Automate data pipelines;c. Design, develop, and maintain scalable (automated) reports and dashboards that track progress towards plans;d. Define, track and report program success metrics.e. Serve as a technical science lead on our most demanding, cross-functional projects.
US, MA, Cambridge
Job summaryMULTIPLE POSITIONS AVAILABLECompany: AMAZON.COM SERVICES LLCPosition Title: Data Scientist IILocation: Cambridge, MassachusettsPosition Responsibilities:Utilize code (Python, R, etc.) to build ML models to solve specific business problems. Build and measure novel online & offline metrics for personal digital assistants and customer scenarios, on diverse devices and endpoints. Research and implement novel machine learning algorithms and models. Collaborate with researchers, software developers, and business leaders to define product requirements and provide modeling solutions. Communicate verbally and in writing to business customers and leadership team with various levels of technical knowledge, educating them about our systems, as well as sharing insights and is an Equal Opportunity-Affirmative Action Employer – Minority / Female / Disability / Veteran / Gender Identity / Sexual Orientation #0000