Events & Conferences

Pronunciation detection for Alexa’s new English-learning experience

Published

2 years ago

July 12, 2023

In January 2023, Alexa launched a language-learning experience in Spain that helps Spanish speakers learn beginner-level English. The experience was developed in collaboration with Vaughan, the leading English-language-learning provider in Spain, and it aimed to provide an immersive English-learning program, with particular focus on pronunciation evaluation.

We are now expanding this offering to Mexico and the Spanish-speaking population in the US and will be adding more languages in the future. The language-learning experience includes structured lessons on vocabulary, grammar, expression, and pronunciation, with practice exercises and quizzes. To try it, set your device language to Spanish and tell Alexa “Quiero aprender Inglés.”

Mini-lesson content page: lessons covering vocabulary, grammar, expression, and pronunciation.

The highlight of this Alexa skill is its pronunciation feature, which provides accurate feedback whenever a customer mispronounces a word or sentence. At this year’s International Conference on Acoustics, Speech, and Signal Processing (ICASSP), we presented a paper describing our state-of-the-art approach to mispronunciation detection.

Pronunciation correction: Blue highlighting indicates correct pronunciation. Red highlighting indicates incorrect pronunciation. For incorrectly pronounced phrases/words, Alexa will provide instructions on how to pronounce them.

Our method uses a novel phonetic recurrent-neural-network-transducer (RNN-T) model that predicts phonemes, the smallest units of speech, from the learner’s pronunciation. The model can therefore provide fine-grained pronunciation evaluation, at the word, syllable, or phoneme level. For example, if a learner mispronounces the word “rabbit” as “rabid”, the model will output the five-phoneme sequence R AE B IH D. It can then detect the mispronounced phonemes (IH D) and syllable (-bid) by using Levenshtein alignment to compare the phoneme sequence with the reference sequence “R AE B AH T”.

Related content

In a top-3% paper at ICASSP, Amazon researchers adapt graph-based label propagation to improve speech recognition on underrepresented pronunciations.

The paper highlights two knowledge gaps that have not been addressed in previous pronunciation-modeling work. The first is the ability to disambiguate similar-sounding phonemes from different languages (e.g., the rolled “r” sounds in Spanish vs. the “r” sound in English). We tackled this challenge by designing a multilingual pronunciation lexicon and building a massive code-mixed phonetic dataset for training.

The other knowledge gap is the ability to learn unique mispronunciation patterns from language learners. We achieve this by leveraging the autoregressiveness of the RNN-T model, meaning the dependence of its outputs on the inputs and outputs that preceded them. This context awareness means that the model can capture frequent mispronunciation patterns from training data. Our pronunciation model has achieved state-of-the-art performance in both phoneme prediction accuracy and mispronunciation detection accuracy.

L2 data augmentation

One of the key technical challenges in building a phonetic-recognition model for non-native (L2) speakers is that there are very limited datasets for mispronunciation diagnosis. In our Interspeech 2022 paper “L2-GEN: A neural phoneme paraphrasing approach to L2 speech synthesis for mispronunciation diagnosis”, we proposed bridging this gap by using data augmentation. Specifically, we built a phoneme paraphraser that can generate realistic L2 phonemes for speakers from a specific locale — e.g., phonemes representing a native Spanish speaker talking in English.

Related content

Parallel speech recognizers, language ID, and translation models geared to conversational speech are among the modifications that make Live Translation possible.

As is common with grammatical-error correction tasks, we use a sequence-to-sequence model but flip the task direction, training the model to mispronounce words rather than correct mispronunciations. Additionally, to further enrich and diversify the generated L2 phoneme sequences, we propose a diversified and preference-aware decoding component that combines a diversified beam search with a preference loss that is biased toward human-like mispronunciations.

For each input phone, or speech fragment, the model produces several candidate phonemes as outputs, and sequences of phonemes are modeled as a tree, with possibilities proliferating with each new phone. Typically, the top-ranked phoneme sequences are extracted from the tree through beam search, which pursues only those branches of the tree with the highest probabilities. In our paper, however, we propose a beam search method that prioritizes unusual phonemes, or phoneme candidates that differ from most of the others at the same depth in the tree.

From established sources in the language-learning literature, we also construct lists of common mispronunciations at the phoneme level, represented as pairs of phonemes, one the standard phoneme in the language and one its nonstandard variant. We construct a loss function that, during model training, prioritizes outputs that use the nonstandard variants on our list.

In experiments, we saw accuracy improvements of up to 5% in mispronunciation detection over a baseline model trained without augmented data.

Balancing false rejection and false acceptance

A key consideration in designing a pronunciation model for a language-learning experience is to balance the false-rejection and false-acceptance ratio. A false rejection occurs when the pronunciation model detects a mispronunciation, but the customer was actually correct or used a consistent but lightly accented pronunciation. A false acceptance occurs when a customer mispronounces a word, and the model fails to detect it.

Related content

Methods for learning from noisy data, using phonetic embeddings to improve entity resolution, and quantization-aware training are a few of the highlights.

Our system has two design features intended to balance these two metrics. To reduce false acceptances, we first combine our standard pronunciation lexicons for English and Spanish into a single lexicon, with multiple phonemes corresponding to each word. Then, we use that lexicon to automatically unannotated speech samples that fall into three categories: native Spanish, native English, and code-switched Spanish and English. Training the model on this dataset enables it to distinguish very subtle differences between phonemes.

To reduce false rejections, we use a multireference pronunciation lexicon where each word is associated with multiple reference pronunciations. For example, the word “data” can be pronounced as either “day-tah” or “dah-tah”, and the system will accept both variations as correct.

In ongoing work, we’re exploring several approaches to further improving our pronunciation evaluation feature. One of these is building a multilingual model that can be used for pronunciation evaluation for many languages. We are also expanding the model to diagnose more characteristics of mispronunciation, such as tone and lexical stress.

Source link

Events & Conferences

A New Ranking Framework for Better Notification Quality on Instagram

Published

1 day ago

September 2, 2025

Xian Sun

We’re sharing how Meta is applying machine learning (ML) and diversity algorithms to improve notification quality and user experience.
We’ve introduced a diversity-aware notification ranking framework to reduce uniformity and deliver a more varied and engaging mix of notifications.
This new framework reduces the volume of notifications and drives higher engagement rates through more diverse outreach.

Notifications are one of the most powerful tools for bringing people back to Instagram and enhancing engagement. Whether it’s a friend liking your photo, another close friend posting a story, or a suggestion for a reel you might enjoy, notifications help surface moments that matter in real time.

Instagram leverages machine learning (ML) models to decide who should get a notification, when to send it, and what content to include. These models are trained to optimize for user positive engagement such as click-through-rate (CTR) – the probability of a user clicking a notification – as well as other metrics like time spent.

However, while engagement-optimized models are effective at driving interactions, there’s a risk that they might overprioritize the product types and authors someone has previously engaged with. This can lead to overexposure to the same creators or the same product types while overlooking other valuable and diverse experiences.

This means people could miss out on content that would give them a more balanced, satisfying, and enriched experience. Over time, this can make notifications feel spammy and increase the likelihood that people will disable them altogether.

The real challenge lies in finding the right balance: How can we introduce meaningful diversity into the notification experience without sacrificing the personalization and relevance people on Instagram have come to expect?

To tackle this, we’ve introduced a diversity-aware notification ranking framework that helps deliver more diverse, better curated, and less repetitive notifications. This framework has significantly reduced daily notification volume while improving CTR. It also introduces several benefits:

The extensibility of incorporating customized soft penalty (demotion) logic for each dimension, enabling more adaptive and sophisticated diversity strategies.
The flexibility of tuning demotion strength across dimensions like content, author, and product type via adjustable weights.
The integration of balancing personalization and diversity, ensuring notifications remain both relevant and varied.

The Risks of Notifications without Diversity

The issue of overexposure in notifications often shows up in two major ways:

Overexposure to the same author: People might receive notifications that are mostly about the same friend. For example, if someone often interacts with content from a particular friend, the system may continue surfacing notifications from that person alone – ignoring other friends they also engage with. This can feel repetitive and one-dimensional, reducing the overall value of notifications.

Overexposure to the same product surface: People might mostly receive notifications from the same product surface such as Stories, even when Feed or Reels could provide value. For example, someone may be interested in both reel and story notifications but has recently interacted more often with stories. Because the system heavily prioritizes past engagement, it sends only story notifications, overlooking the person’s broader interests.

Introducing Instagram’s Diversity-Aware Notification Ranking Framework

Instagram’s diversity-aware notification ranking framework is designed to enhance the notification experience by balancing the predicted potential for user engagement with the need for content diversity. This framework introduces a diversity layer on top of the existing engagement ML models, applying multiplicative penalties to the candidate scores generated by these models, as figure1, below, shows.

The diversity layer evaluates each notification candidate’s similarity to recently sent notifications across multiple dimensions such as content, author, notification type, and product surface. It then applies carefully calibrated penalties—expressed as multiplicative demotion factors—to downrank candidates that are too similar or repetitive. The adjusted scores are used to re-rank the candidates, enabling the system to select notifications that maintain high engagement potential while introducing meaningful diversity. In the end, the quality bar selects the top-ranked candidate that passes both the ranking and diversity criteria.

Figure.1: Instagram’s diversity-aware ranking framework where the diversity layer sits on top of the existing modeling layer and penalizes notifications that are too similar to recently sent ones.

Mathematical Formulation

Within the diversity layer, we apply a multiplicative demotion factor to the base relevance score of each candidate. Given a notification candidate 𝑐, we compute its final score as the product of its base ranking score and a diversity demotion multiplier:

$\text{Score}(c) = R(c) \times D(c)$

where R(c) represents the candidate’s base relevance score, and D(c) ∈ [0,1] is a penalty factor that reduces the score based on similarity to recently sent notifications. We define a set of semantic dimensions (e.g., author, product type) along which we want to promote diversity. For each dimension i, we compute a similarity signal p_i(c) between candidate c and the set of historical notifications H, using a maximal marginal relevance (MMR) approach:

$p_i(c) = \mathrm{max}_{h \in H}\mathrm{sim}_i(c, h)$

where sim_i(·,·) is a predefined similarity function for dimension i. In our baseline implementation, p_i(c) is binary: it equals 1 if the similarity exceeds a threshold 𝜏_i and 0 otherwise.

The final demotion multiplier is defined as:

$D(c) = \prod_{i=1}^{m} \left( 1 - w_i \cdot p_i(c) \right)$

where each w_i∈ [0,1] controls the strength of demotion for its respective dimension. This formulation ensures that candidates similar to previously delivered notifications along one or more dimensions are proportionally down-weighted, reducing redundancy and promoting content variation. The use of a multiplicative penalty allows for flexible control across multiple dimensions, while still preserving high-relevance candidates.

The Future of Diversity-Aware Ranking

As we continue evolving our notification diversity-aware ranking system, a next step is to introduce more adaptive, dynamic demotion strategies. Instead of relying on static rules, we plan to make demotion strength responsive to notification volume and delivery timing. For example, as a user receives more notifications—especially of similar type or in rapid succession—the system progressively applies stronger penalties to new notification candidates, effectively mitigating overwhelming experiences caused by high notification volume or tightly spaced deliveries.

Longer term, we see an opportunity to bring large language models (LLMs) into the diversity pipeline. LLMs can help us go beyond surface-level rules by understanding semantic similarity between messages and rephrasing content in more varied, user-friendly ways. This would allow us to personalize notification experiences with richer language and improved relevance while maintaining diversity across topics, tone, and timing.

Source link

Events & Conferences

Simplifying book discovery with ML-powered visual autocomplete suggestions

Published

1 day ago

September 2, 2025

Mao Sheng Liu

Every day, millions of customers search for books in various formats (audiobooks, e-books, and physical books) across Amazon and Audible. Traditional keyword autocomplete suggestions, while helpful, usually require several steps before customers find their desired content. Audible took on the challenge of making book discovery more intuitive and personalized while reducing the number of steps to purchase.

We developed an instant visual autocomplete system that enhances the search experience across Amazon and Audible. As the user begins typing a query, our solution provides visual previews with book covers, enabling direct navigation to relevant landing pages instead of the search result page. It also delivers real-time personalized format recommendations and incorporates multiple searchable entities, such as book pages, author pages, and series pages.

1 of 2

Audible’s visual-autocomplete experience.

2 of 2

Amazon’s visual-autocomplete experience.

Our system needed to understand user intent from just a few keystrokes and determine the most relevant books to display, all while maintaining low latency for millions of queries. Using historical search data, we match keystrokes to products, transforming partial inputs into meaningful search suggestions. To ensure quality, we implemented confidence-based filtering mechanisms, which are particularly important for distinguishing between general queries like “mystery” and specific title searches. To reflect customers’ most recent interests, the system applies time-decay functions to long historical user interaction data.

Related content

Assessing the absolute utility of query results, rather than just their relative utility, improves learning-to-rank models.

To meet the unique requirements of each use case, we developed two distinct technical approaches. On Audible, we deployed a deep pairwise-learning-to-rank (DeepPLTR) model. The DeepPLTR model considers pairs of books and learns to assign a higher score to the one that better matches the customer query.

The DeepPLTR model’s architecture consists of three specialized towers. The left tower factors in contextual features and recent search patterns using a long-short-term-memory model, which processes data sequentially and considers its prior decisions when issuing a new term in the sequence. The middle tower handles keyword and item engagement history. The right tower factors in customer taste preferences and product descriptions to enable personalization. The model learns from paired examples, but at runtime, it relies on books’ absolute scores to assemble a ranked list.

Training architecture of the DeepPLTR model, which takes in paired examples (green and pink blocks). At runtime, the model scores only a single candidate at a time.

For Amazon, we implemented a two-stage modeling approach involving a probabilistic information-retrieval model to determine the book title that best matches each keyword and a second model that personalizes the book format (audiobooks, e-books, and physical books). This dual-strategy approach maintains low latency while still enabling personalization.

In practice, a customer who types “dungeon craw” in the search bar now sees a visual recommendation for the book Dungeon Crawler Carl, complete with book cover, reducing friction by bypassing a search results page and sending the customer directly to the product detail page. On Audible, the system also personalizes autocomplete results and enriches the discovery experience with relevant connections. These include links to the author’s complete works (Matt Dinniman’s author page) and, for titles that belong to a series, links to the full collection (such as the Dungeon Crawler Carl series).

Related content

Using reinforcement learning improves candidate selection and ranking for search, ad platforms, and recommender systems.

On Amazon, when the customer clicks on the title, the model personalizes the right book-format (audiobooks, e-books, physical books) recommendation and directs the customer to the right product detail page.

In both cases, after the customer has entered a certain number of keystrokes, the system employs a model to detect customer intent (e.g., book title intent for Amazon or author intent for Audible) and determine which visual widget should be displayed.

Audible and Amazon books’ visual autocomplete provides customers with more relevant content more rapidly than traditional autocomplete, and its direct navigation reduces the number of steps to find and access desired books — all while handling millions of queries at low latency.

This technology is not just about making book discovery easier; it is laying the foundation for future improvements in search personalization and visual discovery across Amazon’s ecosystem.

Acknowledgements: Jiun Kim, Sumit Khetan, Armen Stepanyan, Jack Xuan, Nathan Brothers, Eddie Chen, Vincent Lee, Soumy Ladha, Justine Luo, Yuchen Zeng, David Torres, Gali Deutsch, Chaitra Ramdas, Christopher Gomez, Sharmila Tamby, Melissa Ma, Cheng Luo, Jeffrey Jiang, Pavel Fedorov, Ronald Denaux, Aishwarya Vasanth, Azad Bajaj, Mary Heer, Adam Lowe, Jenny Wang, Cameron Cramer, Emmanuel Ankrah, Lydia Diaz, Suzette Islam, Fei Gu, Phil Weaver, Huan Xue, Kimmy Dai, Evangeline Yang, Chao Zhu, Anvy Tran, Jessica Wu, Xiaoxiong Huang, Jiushan Yang

Source link

Events & Conferences

Revolutionizing warehouse automation with scientific simulation

Published

1 week ago

August 26, 2025

Deniz Akyildiz

Modern warehouses rely on complex networks of sensors to enable safe and efficient operations. These sensors must detect everything from packages and containers to robots and vehicles, often in changing environments with varying lighting conditions. More important for Amazon, we need to be able to detect barcodes in an efficient way.

Related content

Generative AI supports the creation, at scale, of complex, realistic driving scenarios that can be directed to specific locations and environments.

The Amazon Robotics ID (ARID) team focuses on solving this problem. When we first started working on it, we faced a significant bottleneck: optimizing sensor placement required weeks or months of physical prototyping and real-world testing, severely limiting our ability to explore innovative solutions.

To transform this process, we developed Sensor Workbench (SWB), a sensor simulation platform built on NVIDIA’s Isaac Sim that combines parallel processing, physics-based sensor modeling, and high-fidelity 3-D environments. By providing virtual testing environments that mirror real-world conditions with unprecedented accuracy, SWB allows our teams to explore hundreds of configurations in the same amount of time it previously took to test just a few physical setups.

Camera and target selection/positioning

Sensor Workbench users can select different cameras and targets and position them in 3-D space to receive real-time feedback on barcode decodability.

Three key innovations enabled SWB: a specialized parallel-computing architecture that performs simulation tasks across the GPU; a custom CAD-to-OpenUSD (Universal Scene Description) pipeline; and the use of OpenUSD as the ground truth throughout the simulation process.