Better Feedback Makes for Faster Reviews

Testing on the Toilet Presents...

Healthy Code on the Commode



Better Feedback Makes for Faster Reviews

by Felipe Sodré and Adam Bender

Have you ever received a code review with a comment like this?

server_is_alive = InitializeServer()

Comment:

This doesn’t make any sense. Why don't you use InitializeServerWithAFewExtraSteps() instead?


If so, it probably left you scratching your head. Why is the reviewer asking this question? Are they testing you? Did you make a big mistake? The problem is the comment is vague (and maybe even a bit rude). It also leaves out important context as to what the reviewer is thinking, making it difficult to respond to. Here are a few simple ways to ensure your review comments are effective, helpful, and clear. 

  • Be kind! People are more receptive to feedback if you assume competence and treat them with respect. 

  • Focus your comments on the code, not the author. Avoid statements with the word ‘you’ which can give the impression that you are judging the person and not the artifact.

  • Explain why you are making the comment. You may be aware of alternatives that are not obvious to the author, or they may be aware of additional constraints.

  • Express the tradeoffs your suggestion entails and take a pragmatic approach to working with the author to achieve the right balance.

  • Approach your role as a guide, not a test grader. Balance giving direct guidance with leaving some degrees of freedom for the author.

  • Consider marking low priority comments with severity like Nit, Optional, or FYI to help the author prioritize the important ones. 

Some comments can be difficult to express concisely. If you have a rewrite in mind, like a subtle API usage or structural change (e.g., split this big change into smaller changes), consider providing an example. If you find you are stuck after more than two rounds of feedback, consider moving to a more direct communication channel like chat or live call to discuss next steps.

When done well, comments can make code reviews more efficient and collaborative:

server_is_alive = InitializeServer()

Comment:

InitializeServerWithAFewExtraSteps() seems to achieve the same result but with built-in logging and auditing (see http://short_link_here). Would that be a better option here?


The goal of reviews of any kind is to get the best possible outcome for you and your team. Software Engineering is a team effort, and by helping each other more effectively you will deliver better, together!

References: Code Health: Respectful Reviews == Useful Reviews, Google's guide on Code Review Comments

More information, discussion, and archives: testing.googleblog.com

Copyright Google LLC. Licensed under a Creative Commons

Attribution–ShareAlike 4.0 License (http://creativecommons.org/licenses/by-sa/4.0/).


Keeping Austin Wired! 5 Gig now available in Austin as Google Fiber continues growth and investment across Central Texas

Greetings from Texas — where everything is bigger, including our Austin fiber network, which has grown in leaps and bounds since you last heard from us. It’s not just bigger, it’s also faster. As of today, customers across Austin can sign up for 5 Gig which comes with symmetrical upload and download speeds up to 5,000 mbps, an included Wi-Fi 6 router, up to two mesh extenders, a 10 Gig Fiber Jack, and professional installation for $125 month.


Thumbnail

And 5 Gig will be available in a lot more places as new construction areas are connected to our network and we open sign-ups for service in new neighborhoods. We’ve continued to grow our network within the City of Austin, and we’ve also begun serving customers in several Travis County neighborhoods located just outside of Austin’s city limits. And we’ll be kicking off construction in both Round Rock and West Lake Hills in the coming months.


We also recently announced expansion to Rollingwood, and our team is in conversations with many other surrounding communities about bringing GFiber to their residents and businesses. We’re working to bring better internet to more people across Central Texas, and hope to have more to share soon.


Our investment in the Austin community has continued to deepen as our network has grown. GFiber’s Gigabit Community program, in collaboration with the Housing Authority of the City of Austin (HACA), continues to expand — launching three new communities over the last two years serving hundreds of additional households in need with free gigabit internet, along with devices and digital skills training. This brings the total number of Austin families served by the program to over 1,300. 


Next up is Espero at Rutland, which will support our community's goal to transition our homeless neighbors into long term housing.


Our Community Connections program, which kicked off in 2013, also continues to grow — GFiber now provides gigabit internet to more than 50 local community organizations and institutions in Austin at no cost. Since our last city update, we’ve connected 18 new Community Connections, including: 



As our network reaches new neighborhoods and cities, more organizations become eligible for this program, and we’ll continue to help them meet their internet needs so they can serve their constituencies and their missions.


Along with these two primary programs, Google Fiber continues to work with local organizations to increase access to quality internet and to build digital literacy and STEM skills across Austin. 


Working with Foundation Communities, Google Fiber is providing gigabit internet to those transitioning to permanent housing. Additionally, we’re supporting Community Tech Network’s efforts to get more Austinites enrolled in the Affordable Connectivity Program, a federal program that provides $30 a month for internet service to those who are eligible.


However, internet access without the ability to navigate the online world won’t get you anywhere. Recent partnerships with amazing local organizations working to ensure everybody has the digital skills they need to thrive include: 



As you can see, we’ve got a lot going on in Austin — and it’s all speeding up (literally!) with 5 Gig coming online now. Additionally, we’ll start testing 20 Gig service in Austin later this year.



If you are looking for how to keep up on what’s happening in your neighborhood, you can sign up for updates here. And follow us on social media (Twitter, Facebook, Instagram, LinkedIn) to get the latest news from throughout the Central Texas region. 


Posted by John-Michael Cortez, Government & Community Affairs Manager


Long Term Support Channel Update for ChromeOS

LTS-108 is being updated in the LTS channel to 108.0.5359.240 (Platform Version: 15183.103.0) for most ChromeOS devices. Want to know more about Long Term Support? Click here.


This update contains multiple Security fixes, including:



1464113 High CVE-2023-4074 Use after free in Blink Task Scheduling
1450899 High CVE-2023-3732 Out of bounds memory access in Mojo
1459124 High CVE-2023-4076 Use after free in WebRTC




Giuliana Pritchard 

Google Chrome OS

Latest ARTwork on hundreds of millions of devices

Posted by Serban Constantinescu, Product Manager

Wouldn’t it be great if each update improved start-up times, execution speed, and memory usage of your apps? Google Play system updates for the Android Runtime (ART) do just that. These updates deliver performance improvements, the latest security fixes, and unify the core OpenJDK APIs across hundreds of millions of devices, including all Android 12+ devices and soon Android Go.

ART is the engine behind the Android operating system (OS). It provides the runtime and core APIs that all apps and most OS services rely on. Both Java and Kotlin are compiled down to bytecode executed by ART. Improvements in the runtime, compiler and core API benefit all developers making app execution faster and bytecode compilation more efficient.

While parts of Android are customizable by device manufacturers, ART is the same for all devices and Google Play system updates enable a path to modular updates.

Modularizing the OS

Android was originally designed for monolithic updates, which meant that OS components did not need to have clear API boundaries. This is because all dependent software would be built together. However, this made it difficult to update ART independently of the rest of the OS. Our first challenge was to untangle ART's dependencies and create clear, well-defined, and tested API boundaries. This allowed us to modularize ART and make it independently updatable.

Illustration of a racecar with an engine part hovering above the hood. A curved arrow points to where this part should go

As a core part of the OS, ART had to blaze new trails and engineer new OS boundaries. These new boundaries were so extensive that manually adding and updating them would be too time-consuming. Therefore, we implemented automatic generation of those through introspection in the build system.

Another example is stack unwinding, which reports the functions last executed when an issue is detected. Before modularizing the OS, all stack unwinding code was built together and could change across Android versions. This made the transition even more challenging, since there is only one version of ART that is delivered to many versions of Android, we had to create a new API boundary as well as design it to be forward-compatible with newer versions of the ART APEX module on devices that are no longer getting full OS updates.

Recently, for Android 14, we refactored the interface between the Package Manager, the service that determines how to install and update apps, and ART. This moves the OS boundary from the ART dex2oat command line to a well-defined interface that enables future optimizations, such as finer-grained control over the compilation mode.

ART updatability also introduced new challenges. For example, the collection of Java libraries, referred to as the Boot Classpath, had to be securely recompiled to ensure good performance. This required introducing a new secure state for compilation during boot as well as a fallback JIT compilation mode.

On older devices, the secure compilation happens on the first reboot after an ART update. On newer devices that support the Android Virtualization Framework, the compilation happens while the device is idle, in an enclave called Isolated Compilation – saving up to 20 seconds of boot-time.

Testing the ART APEX module

The ART APEX module is a complex piece of software with an order of magnitude more APIs than any other APEX module. It also backs a quarter of the developer APIs available in the Android SDK. In addition, ART has a compiler that aims to make the most of the underlying hardware by generating chipset-specific instructions, such as Arm SVE. This, together with the multiple OS versions on which the ART APEX module has to run, makes testing challenging.

We first modularized the testing framework from per-platform release (e.g. Android CTS) to per module. We did this by introducing an ART-specific Mainline Test Suite (MTS), which tests both compiler and runtime, as well as core OpenJDK APIs, while collecting code coverage statistics.

Our target is 100% API coverage and high line coverage, especially for new APIs. Together with HWASan and fuzzing, all of the tests described above contribute to a massive test load that needs to be sharded across multiple devices to ensure that it completes in a reasonable amount of time.

Illustration of modularized testing framework

We test the upcoming ART release every day by compiling over 18 million APKs and running app compatibility tests, and startup, performance, and memory benchmarks on a variety of Android devices that replicate the diversity of our ecosystem as closely as possible. Once tests pass with all possible compilation modes, all Garbage Collector algorithms, and supported OS versions, we begin gradually rolling out the next ART release.

Benefits of ART Google Play system updates

By updating ART independently of OS updates, users get the latest performance optimizations and security fixes as quickly as possible, while developers get OpenJDK improvements and compiler optimisations that benefit both Java and Kotlin.

As shown in the graph below, the runtime and compiler optimizations in the ART 13 update delivered real-world app start-up improvements of up to 30% on some devices.

Graph of average app startup time showing startup time in milliseconds with improvement up to 30% across 12 weeks on devices running the latest ART Google Play system update

ART updates allow us to frequently deploy fixes with little additional effort from our ecosystem partners. They include propagating upstream OpenJDK fixes to Android devices as quickly as possible, as well as runtime and compiler security fixes, such as CVE-2022-20502, which was detected by our automated fuzzing tests.

For developers, ART updates mean that you can now target the latest programming features. ART 13 delivered OpenJDK 11 core language features, which was the fastest-ever adoption of a new OpenJDK release on Android devices.

What’s next

In the coming months, we'll be releasing ART 14 to all compatible devices. ART 14 includes OpenJDK 17 support along with new compiler and runtime optimizations that improve performance while reducing code size. Stay tuned for more details on ART 14!

Java and OpenJDK are trademarks or registered trademarks of Oracle and/or its affiliates.

Dev Channel Update for ChromeOS / ChromeOS Flex

The Dev channel is being updated to OS version: 15572.10.0 Browser version: 117.0.5938.13 for most ChromeOS devices.

If you find new issues, please let us know one of the following ways

  1. File a bug
  2. Visit our ChromeOS communities
    1. General: Chromebook Help Community
    2. Beta Specific: ChromeOS Beta Help Community
  3. Report an issue or send feedback on Chrome

Interested in switching channels? Find out how.

Matt Nelson,
Google ChromeOS

Meet the student leaders building apps using Google technology

Posted by Kübra Zengin, North America GDSC Regional Lead

Serving as a Google Developer Student Clubs (GDSC) Lead at the university level builds technical skills and leadership skills that serve alumni well in their post-graduate careers. Four GDSC Alumni Leads from universities in Canada and the U.S. have gone on to meaningful careers in the tech industry, and share their experiences.

Image of Daniel Shirvani (right) with Ayman Bolad (left)at a Google Developer Students event

Daniel Shirvani: The Next Frontier in Patient Data

Daniel Shirvani graduated from the University of British Columbia (UBC) in Vancouver, Canada, in 2023, with a Bachelor’s of Science in Pharmacology, and will soon return to UBC for medical school. He served as Google Developer Student Clubs (GDSC) Lead and founding team member. He also launched his own software company, Leftindust Systems, in 2019, to experiment with creating small-scale electronic medical record software (EMR) for the open source community. This project is now closed.

“I built a startup to rethink the use of medical software,” he says.

As a summer student volunteer at a Vancouver-area heart clinic, Shirvani was tasked with indexing hundreds of medical records, who had specific blood glucose HBA1C levels and factors related to kidney disease, to see who would be eligible for the new cardiac drug. However, the clinic’s medical records software didn’t have the capability to flag patients in the system, so the only way to register the hundreds of files on Shirvani’s final list would be to do so manually–and that was impossible, given the size of the list and the time remaining in his work term. He believed that the software should have been able to not only flag these patients, but also to automatically filter which patients met the criteria.

“Two to three hundred patients will not receive this life-saving drug because of this software,” Shirvani says. “My father is a patient who would have been eligible for this type of drug. His heart attack put things into perspective. There are families just like mine who will have the same experience that my father did, only because the software couldn’t keep up.”

Shirvani decided to combine his medical knowledge and programming skills to develop an electronic medical software, or EMR, that could store patient data numerically, instead of within paragraphs. This allows doctors to instantly analyze the data of patients, both at the individual and group-level. Doctors across North America took notice, including those from UBC, Stanford, UCLA, and elsewhere.

“During the North America Connect conference, a 2-day in-person event bringing together organizers and members across North America from the Google for Developers community programs including Google Developer Group, Women Techmakers, Google Developer Experts, and Google Developer Student Clubs, I met with many GDEs and Googlers, such as Kevin A. McGrail, who is now a personal mentor,” says Shirvani, who continues to look for other ways to make change in the healthcare community.

"When systems disappoint, we see not an end, but a new beginning. It’s in that space that we shape the future.


Image of Alexandra Cusell presenting at Carnegie Mellon University Swartz Center for Entrepreneurship

Alex Cussell: Becoming a tech entrepreneur

Alex Cussell graduated from the University of Central Florida in 2020, where she was a GDSC Lead her senior year. She says the experience inspired her to pursue her passion of becoming a tech entrepreneur.

“Leading a group of students with such differing backgrounds, addressing the world’s most pervasive problems, and loving every second of it taught me that I was meant to be a tech entrepreneur,” she says. “We were on a mission to save the lives of those involved in traffic accidents, when the world as we knew it came to a screeching halt due to the COVID-19 pandemic.

After her virtual graduation, Cussell moved to Silicon Valley and earned a Master’s in Technology Ventures from Carnegie Mellon University. She studied product management, venture capital, and startup law, with a vision of building a meaningful company. After getting engaged and receiving multiple gift cards as bridal shower gifts, Cussell found herself confused about each card’s amount and challenged trying to keep them organized.

She created the Jisell app, which features a universal gift card e-wallet, allowing users to digitize their gift cards. The app has had over five thousand dollars in gift cards uploaded to date and a partnership with the largest gift card distributor in the U.S. Jisell product manager Emily Robertson was Cussell’s roommate at the GDSC summit.

“Without Google Developer Student Clubs, I might never have realized how much I love problem-solving or technical leadership or known so much about the great tools offered by Google,” Cussell says. "Thank you to everyone who contributes to the GDSC experience; you have truly changed the lives of so many.”


Headshot of Angela Busheska, smiling

Angela Busheska: Founding a nonprofit to fight climate change

Angela Busheska is double majoring in electrical engineering and computer science, with a minor in mathematics, at Lafayette College in Easton, Pennsylvania, and anticipates graduating in 2025. A Google intern this summer and last summer, Busheska participated in Google’s Computer Science Research Mentorship Program from September 2021-January 2022, which supports the pursuit of computing research for students from historically marginalized groups through career mentorship, peer-to-peer networking, and building awareness about pathways within the field. Busheska investigated the computing processes across four different projects in the field of AI for Social Good.

During the pandemic, in 2020, Busheska founded EnRoute, a nonprofit to harness the power of everyday actions to fight climate change and break down the stigma that living sustainably is an expensive and challenging commitment. She also built a mobile app using Android and Flutter that helps users make simple daily transportation and shopping choices to reduce their carbon footprints. Since 2020, the app has guided thousands of users to reduce more than 100,000kg of CO2 emissions.

EnRoute honors Busheska’s aunt, who passed away when Busheska was 17. Busheska grew up in Skopje, in North Macedonia, one of the world’s most polluted cities.

“When I was 17 years old, Skopje’s dense air pollution led my aunt, who suffered from cardiovascular difficulties, to complete blood vessel damage, resulting in her swift passing,” says Busheska. “Inspired by my personal loss, I started researching the causes of the pollution.”

EnRoute has been featured on the Forbes 30 Under 30 Social Impact List and has been publicly recognized by Shawn Mendes, Prince William, One Young World, and the United Nations.


Headshot of Sapphira Ching, smiling

Sapphira Ching: Advancing Environmental, Social, and Government standards (ESG)

Sapphira Ching, a senior at the University of Pennsylvania’s Wharton School, spent her junior year as UPenn’s GDSC Lead, after joining GDSC her first year, leading social media for the club that spring and heading marketing and strategy her sophomore year. As a GDSC Lead, Sapphira expanded GDSC's campus membership and partnerships to reach an audience of over 2,000 students. In line with her passion for Environmental, Social, and Government standards (ESG) and Diversity, Equity, and Inclusion (DEI), Sapphira built a leadership team from different areas of study, including engineering, business, law, medicine, and music.

Ching’s passions for ESG, technology, and business drive her choices, and she says, “I am eager to incorporate ESG into tech to bring people together using business acumen.”

The Wharton School appointed her as an inaugural undergraduate fellow at the Turner ESG Initiative, and she founded the Penn Innovation Network, an ESG innovation club. Her summer internships have focused on ESG; her 2021 summer internship at MSCI (formerly known as Morgan Stanley Capital International) centered on on ESG, and her 2022 summer internship was at Soros Fund Management, an ESG juggernaut in finance. She is also a NCAA Division I student-athlete and Olympic hopeful in sabre fencing.

“I attribute my growth in ESG, tech, and business to how GDSC has helped me since my first year of college,” Ching says.

Are you an Alumni or current GDSC Lead? You can join the Google Developer Student Clubs (GDSC) LinkedIn Group here. The group is a great place to share ideas and connect with current and former GDSC Leads.

Interested in joining a GDSC near you? Google Developer Student Clubs (GDSC) are university based community groups for students interested in Google developer technologies. Students from all undergraduate or graduate programs with an interest in growing as a developer are welcome. Learn more here.

Interested in becoming a GDSC Lead? GDSC Leads are responsible for starting and growing a Google Developer Student Club (GDSC) chapter at their university. GDSC Leads work with students to organize events, workshops, and projects. Learn more here.

Google at Interspeech 2023

This week, the 24th Annual Conference of the International Speech Communication Association (INTERSPEECH 2023) is being held in Dublin, Ireland, representing one of the world’s most extensive conferences on research and technology of spoken language understanding and processing. Experts in speech-related research fields gather to take part in oral presentations and poster sessions and to build collaborations across the globe.

We are excited to be a Platinum Sponsor of INTERSPEECH 2023, where we will be showcasing more than 20 research publications and supporting a number of workshops and special sessions. We welcome in-person attendees to drop by the Google Research booth to meet our researchers and participate in Q&As and demonstrations of some of our latest speech technologies, which help to improve accessibility and provide convenience in communication for billions of users. In addition, online attendees are encouraged to visit our virtual booth in Topia where you can get up-to-date information on research and opportunities at Google. Visit the @GoogleAI Twitter account to find out about Google booth activities (e.g., demos and Q&A sessions). You can also learn more about the Google research being presented at INTERSPEECH 2023 below (Google affiliations in bold).


Board and Organizing Committee

ISCA Board, Technical Committee Chair: Bhuvana Ramabhadran

Area Chairs include:
    Analysis of Speech and Audio Signals: Richard Rose
    Speech Synthesis and Spoken Language Generation: Rob Clark
    Special Areas: Tara Sainath


Satellite events

VoxCeleb Speaker Recognition Challenge 2023 (VoxSRC-23)
Organizers include: Arsha Nagrani

ISCA Speech Synthesis Workshop (SSW12)
Speakers include: Rob Clark


Keynote talk – ISCA Medalist


Survey Talk

Speech Compression in the AI Era
Speaker: Jan Skoglund


Special session papers

Cascaded Encoders for Fine-Tuning ASR Models on Overlapped Speech
Richard Rose, Oscar Chang, Olivier Siohan

TokenSplit: Using Discrete Speech Representations for Direct, Refined, and Transcript-Conditioned Speech Separation and Recognition
Hakan Erdogan, Scott Wisdom, Xuankai Chang*, Zalán Borsos, Marco Tagliasacchi, Neil Zeghidour, John R. Hershey


Papers

DeePMOS: Deep Posterior Mean-Opinion-Score of Speech
Xinyu Liang, Fredrik Cumlin, Christian Schüldt, Saikat Chatterjee

O-1: Self-Training with Oracle and 1-Best Hypothesis
Murali Karthick Baskar, Andrew Rosenberg, Bhuvana Ramabhadran, Kartik Audhkhasi

Re-investigating the Efficient Transfer Learning of Speech Foundation Model Using Feature Fusion Methods
Zhouyuan Huo, Khe Chai Sim, Dongseong Hwang, Tsendsuren Munkhdalai, Tara N. Sainath, Pedro Moreno

MOS vs. AB: Evaluating Text-to-Speech Systems Reliably Using Clustered Standard Errors
Joshua Camp, Tom Kenter, Lev Finkelstein, Rob Clark

LanSER: Language-Model Supported Speech Emotion Recognition
Taesik Gong, Josh Belanich, Krishna Somandepalli, Arsha Nagrani, Brian Eoff, Brendan Jou

Modular Domain Adaptation for Conformer-Based Streaming ASR
Qiujia Li, Bo Li, Dongseong Hwang, Tara N. Sainath, Pedro M. Mengibar

On Training a Neural Residual Acoustic Echo Suppressor for Improved ASR
Sankaran Panchapagesan, Turaj Zakizadeh Shabestary, Arun Narayanan

MD3: The Multi-dialect Dataset of Dialogues
Jacob Eisenstein, Vinodkumar Prabhakaran, Clara Rivera, Dorottya Demszky, Devyani Sharma

Dual-Mode NAM: Effective Top-K Context Injection for End-to-End ASR
Zelin Wu, Tsendsuren Munkhdalai, Pat Rondon, Golan Pundak, Khe Chai Sim, Christopher Li

Using Text Injection to Improve Recognition of Personal Identifiers in Speech
Yochai Blau, Rohan Agrawal, Lior Madmony, Gary Wang, Andrew Rosenberg, Zhehuai Chen, Zorik Gekhman, Genady Beryozkin, Parisa Haghani, Bhuvana Ramabhadran

How to Estimate Model Transferability of Pre-trained Speech Models?
Zih-Ching Chen, Chao-Han Huck Yang*, Bo Li, Yu Zhang, Nanxin Chen, Shuo-yiin Chang, Rohit Prabhavalkar, Hung-yi Lee, Tara N. Sainath

Improving Joint Speech-Text Representations Without Alignment
Cal Peyser, Zhong Meng, Ke Hu, Rohit Prabhavalkar, Andrew Rosenberg, Tara N. Sainath, Michael Picheny, Kyunghyun Cho

Text Injection for Capitalization and Turn-Taking Prediction in Speech Models
Shaan Bijwadia, Shuo-yiin Chang, Weiran Wang, Zhong Meng, Hao Zhang, Tara N. Sainath

Streaming Parrotron for On-Device Speech-to-Speech Conversion
Oleg Rybakov, Fadi Biadsy, Xia Zhang, Liyang Jiang, Phoenix Meadowlark, Shivani Agrawal

Semantic Segmentation with Bidirectional Language Models Improves Long-Form ASR
W. Ronny Huang, Hao Zhang, Shankar Kumar, Shuo-yiin Chang, Tara N. Sainath

Universal Automatic Phonetic Transcription into the International Phonetic Alphabet
Chihiro Taguchi, Yusuke Sakai, Parisa Haghani, David Chiang

Mixture-of-Expert Conformer for Streaming Multilingual ASR
Ke Hu, Bo Li, Tara N. Sainath, Yu Zhang, Francoise Beaufays

Real Time Spectrogram Inversion on Mobile Phone
Oleg Rybakov, Marco Tagliasacchi, Yunpeng Li, Liyang Jiang, Xia Zhang, Fadi Biadsy

2-Bit Conformer Quantization for Automatic Speech Recognition
Oleg Rybakov, Phoenix Meadowlark, Shaojin Ding, David Qiu, Jian Li, David Rim, Yanzhang He

LibriTTS-R: A Restored Multi-speaker Text-to-Speech Corpus
Yuma Koizumi, Heiga Zen, Shigeki Karita, Yifan Ding, Kohei Yatabe, Nobuyuki Morioka, Michiel Bacchiani, Yu Zhang, Wei Han, Ankur Bapna

PronScribe: Highly Accurate Multimodal Phonemic Transcription from Speech and Text
Yang Yu, Matthew Perez*, Ankur Bapna, Fadi Haik, Siamak Tazari, Yu Zhang

Label Aware Speech Representation Learning for Language Identification
Shikhar Vashishth, Shikhar Bharadwaj, Sriram Ganapathy, Ankur Bapna, Min Ma, Wei Han, Vera Axelrod, Partha Talukdar


* Work done while at Google

Source: Google AI Blog