Tag Archives: datasets

Data-centric ML benchmarking: Announcing DataPerf’s 2023 challenges

Machine learning (ML) offers tremendous potential, from diagnosing cancer to engineering safe self-driving cars to amplifying human productivity. To realize this potential, however, organizations need ML solutions to be reliable with ML solution development that is predictable and tractable. The key to both is a deeper understanding of ML data — how to engineer training datasets that produce high quality models and test datasets that deliver accurate indicators of how close we are to solving the target problem.

The process of creating high quality datasets is complicated and error-prone, from the initial selection and cleaning of raw data, to labeling the data and splitting it into training and test sets. Some experts believe that the majority of the effort in designing an ML system is actually the sourcing and preparing of data. Each step can introduce issues and biases. Even many of the standard datasets we use today have been shown to have mislabeled data that can destabilize established ML benchmarks. Despite the fundamental importance of data to ML, it’s only now beginning to receive the same level of attention that models and learning algorithms have been enjoying for the past decade.

Towards this goal, we are introducing DataPerf, a set of new data-centric ML challenges to advance the state-of-the-art in data selection, preparation, and acquisition technologies, designed and built through a broad collaboration across industry and academia. The initial version of DataPerf consists of four challenges focused on three common data-centric tasks across three application domains; vision, speech and natural language processing (NLP). In this blogpost, we outline dataset development bottlenecks confronting researchers and discuss the role of benchmarks and leaderboards in incentivizing researchers to address these challenges. We invite innovators in academia and industry who seek to measure and validate breakthroughs in data-centric ML to demonstrate the power of their algorithms and techniques to create and improve datasets through these benchmarks.

Data is the new bottleneck for ML

Data is the new code: it is the training data that determines the maximum possible quality of an ML solution. The model only determines the degree to which that maximum quality is realized; in a sense the model is a lossy compiler for the data. Though high-quality training datasets are vital to continued advancement in the field of ML, much of the data on which the field relies today is nearly a decade old (e.g., ImageNet or LibriSpeech) or scraped from the web with very limited filtering of content (e.g., LAION or The Pile).

Despite the importance of data, ML research to date has been dominated by a focus on models. Before modern deep neural networks (DNNs), there were no ML models sufficient to match human behavior for many simple tasks. This starting condition led to a model-centric paradigm in which (1) the training dataset and test dataset were “frozen” artifacts and the goal was to develop a better model, and (2) the test dataset was selected randomly from the same pool of data as the training set for statistical reasons. Unfortunately, freezing the datasets ignored the ability to improve training accuracy and efficiency with better data, and using test sets drawn from the same pool as training data conflated fitting that data well with actually solving the underlying problem.

Because we are now developing and deploying ML solutions for increasingly sophisticated tasks, we need to engineer test sets that fully capture real world problems and training sets that, in combination with advanced models, deliver effective solutions. We need to shift from today’s model-centric paradigm to a data-centric paradigm in which we recognize that for the majority of ML developers, creating high quality training and test data will be a bottleneck.

Shifting from today’s model-centric paradigm to a data-centric paradigm enabled by quality datasets and data-centric algorithms like those measured in DataPerf.

Enabling ML developers to create better training and test datasets will require a deeper understanding of ML data quality and the development of algorithms, tools, and methodologies for optimizing it. We can begin by recognizing common challenges in dataset creation and developing performance metrics for algorithms that address those challenges. For instance:

  • Data selection: Often, we have a larger pool of available data than we can label or train on effectively. How do we choose the most important data for training our models?
  • Data cleaning: Human labelers sometimes make mistakes. ML developers can’t afford to have experts check and correct all labels. How can we select the most likely-to-be-mislabeled data for correction?

We can also create incentives that reward good dataset engineering. We anticipate that high quality training data, which has been carefully selected and labeled, will become a valuable product in many industries but presently lack a way to assess the relative value of different datasets without actually training on the datasets in question. How do we solve this problem and enable quality-driven “data acquisition”?

DataPerf: The first leaderboard for data

We believe good benchmarks and leaderboards can drive rapid progress in data-centric technology. ML benchmarks in academia have been essential to stimulating progress in the field. Consider the following graph which shows progress on popular ML benchmarks (MNIST, ImageNet, SQuAD, GLUE, Switchboard) over time:

Performance over time for popular benchmarks, normalized with initial performance at minus one and human performance at zero. (Source: Douwe, et al. 2021; used with permission.)

Online leaderboards provide official validation of benchmark results and catalyze communities intent on optimizing those benchmarks. For instance, Kaggle has over 10 million registered users. The MLPerf official benchmark results have helped drive an over 16x improvement in training performance on key benchmarks.

DataPerf is the first community and platform to build leaderboards for data benchmarks, and we hope to have an analogous impact on research and development for data-centric ML. The initial version of DataPerf consists of leaderboards for four challenges focused on three data-centric tasks (data selection, cleaning, and acquisition) across three application domains (vision, speech and NLP):

  • Training data selection (Vision): Design a data selection strategy that chooses the best training set from a large candidate pool of weakly labeled training images.
  • Training data selection (Speech): Design a data selection strategy that chooses the best training set from a large candidate pool of automatically extracted clips of spoken words.
  • Training data cleaning (Vision): Design a data cleaning strategy that chooses samples to relabel from a “noisy” training set where some of the labels are incorrect.
  • Training dataset evaluation (NLP): Quality datasets can be expensive to construct, and are becoming valuable commodities. Design a data acquisition strategy that chooses which training dataset to “buy” based on limited information about the data.

For each challenge, the DataPerf website provides design documents that define the problem, test model(s), quality target, rules and guidelines on how to run the code and submit. The live leaderboards are hosted on the Dynabench platform, which also provides an online evaluation framework and submission tracker. Dynabench is an open-source project, hosted by the MLCommons Association, focused on enabling data-centric leaderboards for both training and test data and data-centric algorithms.

How to get involved

We are part of a community of ML researchers, data scientists and engineers who strive to improve data quality. We invite innovators in academia and industry to measure and validate data-centric algorithms and techniques to create and improve datasets through the DataPerf benchmarks. The deadline for the first round of challenges is May 26th, 2023.


The DataPerf benchmarks were created over the last year by engineers and scientists from: Coactive.ai, Eidgenössische Technische Hochschule (ETH) Zurich, Google, Harvard University, Meta, ML Commons, Stanford University. In addition, this would not have been possible without the support of DataPerf working group members from Carnegie Mellon University, Digital Prism Advisors, Factored, Hugging Face, Institute for Human and Machine Cognition, Landing.ai, San Diego Supercomputing Center, Thomson Reuters Lab, and TU Eindhoven.

Source: Google AI Blog

PRESTO – A multilingual dataset for parsing realistic task-oriented dialogues

Virtual assistants are increasingly integrated into our daily routines. They can help with everything from setting alarms to giving map directions and can even assist people with disabilities to more easily manage their homes. As we use these assistants, we are also becoming more accustomed to using natural language to accomplish tasks that we once did by hand.

One of the biggest challenges in building a robust virtual assistant is identifying what a user wants and what information is needed to perform the task at hand. In the natural language processing (NLP) literature, this is mainly framed as a task-oriented dialogue parsing task, where a given dialogue needs to be parsed by a system to understand the user intent and carry out the operation to fulfill that intent. While the academic community has made progress in handling task-oriented dialogue thanks to custom purpose datasets, such as MultiWOZ, TOP, SMCalFlow, etc., progress is limited because these datasets lack typical speech phenomena necessary for model training to optimize language model performance. The resulting models often underperform, leading to dissatisfaction with assistant interactions. Relevant speech patterns might include revisions, disfluencies, code-mixing, and the use of structured context surrounding the user’s environment, which might include the user’s notes, smart home devices, contact lists, etc.

Consider the following dialogue that illustrates a common instance when a user needs to revise their utterance:

A dialogue conversation with a virtual assistant that includes a user revision.

The virtual assistant misunderstands the request and attempts to call the incorrect contact. Hence, the user has to revise their utterance to fix the assistant’s mistake. To parse the last utterance correctly, the assistant would also need to interpret the special context of the user — in this case, it would need to know that the user had a contact list saved in their phone that it should reference.

Another common category of utterance that is challenging for virtual assistants is code-mixing, which occurs when the user switches from one language to another while addressing the assistant. Consider the utterance below:

A dialogue denoting code-mixing between English and German.

In this example, the user switches from English to German, where “vier Uhr” means “four o’clock” in German.

In an effort to advance research in parsing such realistic and complex utterances, we are launching a new dataset called PRESTO, a multilingual dataset for parsing realistic task-oriented dialogues that includes roughly half a million realistic conversations between people and virtual assistants. The dataset spans six different languages and includes multiple conversational phenomena that users may encounter when using an assistant, including user-revisions, disfluencies, and code-mixing. The dataset also includes surrounding structured context, such as users’ contacts and lists associated with each example. The explicit tagging of various phenomena in PRESTO allows us to create different test sets to separately analyze model performance on these speech phenomena. We find that some of these phenomena are easier to model with few-shot examples, while others require much more training data.

Dataset characteristics

  1. Conversations by native speakers in six languages
    All conversations in our dataset are provided by native speakers of six languages — English, French, German, Hindi, Japanese, and Spanish. This is in contrast to other datasets, such as MTOP and MASSIVE, that translate utterances only from English to other languages, which does not necessarily reflect the speech patterns of native speakers in non-English languages.

  2. Structured context
    Users often rely on the information stored in their devices, such as notes, contacts, and lists, when interacting with virtual assistants. However, this context is often not accessible to the assistant, which can result in parsing errors when processing user utterances. To address this issue, PRESTO includes three types of structured context, notes, lists, and contacts, as well as user utterances and their parses. The lists, notes, and contacts are authored by native speakers of each language during data collection. Having such context allows us to examine how this information can be used to improve performance on parsing task-oriented dialog models.
    Each example in PRESTO consists of: Inputs — A user’s virtual state (context), one or more user utterances, and the corresponding virtual assistant responses (dialogue). Output — The semantic parsing of the last user utterance in the dialogue (parse).
  3. User revisions
    It is common for a user to revise or correct their own utterances while speaking to a virtual assistant. These revisions happen for a variety of reasons — the assistant could have made a mistake in understanding the utterance or the user might have changed their mind while making an utterance. One such example is in the figure above. Other examples of revisions include canceling one’s request (‘’Don’t add anything.”) or correcting oneself in the same utterance (“Add bread — no, no wait — add wheat bread to my shopping list.”). Roughly 27% of all examples in PRESTO have some type of user revision that is explicitly labeled in the dataset.

  4. Code-mixing
    As of 2022, roughly 43% of the world’s population is bilingual. As a result, many users switch languages while speaking to virtual assistants. In building PRESTO, we asked bilingual data contributors to annotate code-mixed utterances, which amounted to roughly 14% of all utterances in the dataset.
    Examples of Hindi-English, Spanish-English, and German-English code-switched utterances from PRESTO.
  5. Disfluencies
    Disfluencies, like repeated phrases or filler words, are ubiquitous in user utterances due to the spoken nature of the conversations that the virtual assistants receive. Datasets such as DISFL-QA note the lack of such phenomena in existing NLP literature and contribute towards the goal of alleviating that gap. In our work, we include conversations targeting this particular phenomenon across all six languages.
    Examples of utterances in English, Japanese, and French with filler words or repetitions.

Key findings

We performed targeted experiments to focus on each of the phenomena described above. We ran mT5-based models trained using the PRESTO dataset and evaluated them using an exact match between the predicted parse and the human annotated parse. Below we show the relative performance improvements as we scale the training data on each of the targeted phenomena — user revisions, disfluencies, and code-mixing.

K-shot results on various linguistic phenomena and the full test set across increasing training data size.

The k-shot results yield the following takeaways:

  1. Zero-shot performance on the marked phenomenon is poor, emphasizing the need for such utterances in the dataset to improve performance.
  2. Disfluencies and code-mixing have a much better zero-shot performance than user-revisions (over 40 points difference in exact-match accuracy).

We also investigate the difference between training monolingual and multilingual models on the train set and find that with fewer data multilingual models have an advantage over monolingual models, but the gap shrinks as the data size is increased.

Additional details on data quality, data collection methodology, and modeling experiments can be found in our paper.


We created PRESTO, a multilingual dataset for parsing task-oriented dialogues that includes realistic conversations representing a variety of pain points that users often face in their daily conversations with virtual assistants that are lacking in existing datasets in the NLP community. PRESTO includes roughly half a million utterances that are contributed by native speakers of six languages — English, French, German, Hindi, Japanese, and Spanish. We created dedicated test sets to focus on each targeted phenomenon — user revisions, disfluencies, code-mixing, and structured context. Our results indicate that the zero-shot performance is poor when the targeted phenomenon is not included in the training set, indicating a need for such utterances to improve performance. We notice that user revisions and disfluencies are easier to model with more data as opposed to code-mixed utterances, which are harder to model, even with a high number of examples. With the release of this dataset, we open more questions than we answer and we hope the research community makes progress on utterances that are more in line with what users are facing every day.


It was a privilege to collaborate on this work with Waleed Ammar, Siddharth Vashishtha, Motoki Sano, Faiz Surani, Max Chang, HyunJeong Choe, David Greene, Kyle He, Rattima Nitisaroj, Anna Trukhina, Shachi Paul, Pararth Shah, Rushin Shah, and Zhou Yu. We’d also like to thank Tom Small for the animations in this blog post. Finally, a huge thanks to all the expert linguists and data annotators for making this a reality.

Source: Google AI Blog

Datasets at your fingertips in Google Search

Access to datasets is critical to many of today's endeavors across verticals and industries, whether scientific research, business analysis, or public policy. In the scientific community and throughout various levels of the public sector, reproducibility and transparency are essential for progress, so sharing data is vital. For one example, in the United States a recent new policy requires free and equitable access to outcomes of all federally funded research, including data and statistical information along with publications.

To facilitate discovery of content with this level of statistical detail and better distill this information from across the web, Google now makes it easier to search for datasets. You can click on any of the top three results (see below) to get to the dataset page or you can explore further by clicking "More datasets." Here is an example:

When users search for datasets in Google search, they find a dedicated section highlighting pages with dataset descriptions. They can explore many more datasets by clicking on "More datasets" and going to Dataset Search.

Powered by Dataset Search

Dataset Search, a dedicated search engine for datasets, powers this feature and indexes more than 45 million datasets from more than 13,000 websites. Datasets cover many disciplines and topics, including government, scientific, and commercial datasets. Dataset Search shows users essential metadata about datasets and previews of the data where available. Users can then follow the links to the data repositories that host the datasets.

Dataset Search primarily indexes dataset pages on the Web that contain schema.org structured data. The schema.org metadata allows Web page authors to describe the semantics of the page: the entities on the pages and their properties. For dataset pages, schema.org metadata describes key elements of the datasets, such as their description, license, temporal and spatial coverage, and available download formats. In addition to aggregating this metadata and providing easy access to it, Dataset Search normalizes and reconciles the metadata that comes directly from the Web pages.

If you are a dataset author or provider and want others to find your datasets in Search, make sure that you publish your dataset in a way that makes it discoverable and specifies how others can reuse the data. Specifically, ensure that the Web page that describes the dataset has machine-readable metadata. The easiest way to ensure this is to publish your dataset in an established dataset repository. Some repositories cater to specific research communities, while others are "generalists" (figshare.com, zenodo.org, datadryad.org, kaggle.com, etc.). These repositories automatically include metadata in dataset pages for every dataset, which makes it easy for search engines to discover and include them in specialized result sections, as in the figure above.

As data sharing continues to grow and evolve, we will continue to make datasets as easy to find, access, and use as any other type of information on the web.


We are extremely grateful to the numerous Googlers who contributed to developing and launching this feature, including: Rachel Zax, Damian Biollo, Shiyu Chen, Jonathan Drake, Sunil Vemuri, Stephen Tseou, Amit Bapat, Will Leszczuk, Marc Najork, Sergei Vassilvitskii, Bruno Possas, and Corinna Cortes.

Source: Google AI Blog

FRMT: A Benchmark for Few-Shot Region-Aware Machine Translation

Many languages spoken worldwide cover numerous regional varieties (sometimes called dialects), such as Brazilian and European Portuguese or Mainland and Taiwan Mandarin Chinese. Although such varieties are often mutually intelligible to their speakers, there are still important differences. For example, the Brazilian Portuguese word for “bus” is ônibus, while the European Portuguese word is autocarro. Yet, today’s machine translation (MT) systems typically do not allow users to specify which variety of a language to translate into. This may lead to confusion if the system outputs the “wrong” variety or mixes varieties in an unnatural way. Also, region-unaware MT systems tend to favor whichever variety has more data available online, which disproportionately affects speakers of under-resourced language varieties.

In “FRMT: A Benchmark for Few-Shot Region-Aware Machine Translation”, accepted for publication in Transactions of the Association for Computational Linguistics, we present an evaluation dataset used to measure MT systems’ ability to support regional varieties through a case study on Brazilian vs. European Portuguese and Mainland vs. Taiwan Mandarin Chinese. With the release of the FRMT data and accompanying evaluation code, we hope to inspire and enable the research community to discover new ways of creating MT systems that are applicable to the large number of regional language varieties spoken worldwide.

Challenge: Few-Shot Generalization

Most modern MT systems are trained on millions or billions of example translations, such as an English input sentence and its corresponding Portuguese translation. However, the vast majority of available training data doesn’t specify what regional variety the translation is in. In light of this data scarcity, we position FRMT as a benchmark for few-shot translation, measuring an MT model’s ability to translate into regional varieties when given no more than 100 labeled examples of each language variety. MT models need to use the linguistic patterns showcased in the small number of labeled examples (called “exemplars”) to identify similar patterns in their unlabeled training examples. In this way, models can generalize, producing correct translations of phenomena not explicitly shown in the exemplars.

An illustration of a few-shot MT system translating the English sentence, “The bus arrived,” into two regional varieties of Portuguese: Brazilian (🇧🇷; left) and European (🇵🇹; right).

Few-shot approaches to MT are attractive because they make it much easier to add support for additional regional varieties to an existing system. While our work is specific to regional varieties of two languages, we anticipate that methods that perform well will be readily applicable to other languages and regional varieties. In principle, those methods should also work for other language distinctions, such as formality and style.

Data Collection

The FRMT dataset consists of partial English Wikipedia articles, sourced from the Wiki40b dataset, that have been translated by paid, professional translators into different regional varieties of Portuguese and Mandarin. In order to highlight key region-aware translation challenges, we designed the dataset using three content buckets: (1) Lexical, (2) Entity, and (3) Random.

  1. The Lexical bucket focuses on regional differences in word choice, such as the "ônibus" vs. "autocarro” distinction when translating a sentence with the word “bus” into Brazilian vs. European Portuguese, respectively. We manually collected 20-30 terms that have regionally distinctive translations according to blogs and educational websites, and filtered and vetted the translations with feedback from volunteer native speakers from each region. Given the resulting list of English terms, we extracted texts of up to 100 sentences each from the associated English Wikipedia articles (e.g., bus). The same process was carried out independently for Mandarin.
  2. The Entity bucket is populated in a similar way and concerns people, locations or other entities strongly associated with one of the two regions in question for a given language. Consider an illustrative sentence like, “In Lisbon, I often took the bus.” In order to translate this correctly into Brazilian Portuguese, a model must overcome two potential pitfalls:
    1. The strong geographical association between Lisbon and Portugal might influence a model to generate a European Portuguese translation instead, e.g., by selecting "autocarro” rather than "ônibus".
    2. Replacing “Lisbon” with “Brasília” might be a naive way for a model to localize its output toward Brazilian Portuguese, but would be semantically inaccurate, even in an otherwise fluent translation.
  3. The Random bucket is used to check that a model correctly handles other diverse phenomena, and consists of text from 100 randomly sampled articles from Wikipedia’s “featured” and “good” collections.

Evaluation Methodology

To verify that the translations collected for the FRMT dataset capture region-specific phenomena, we conducted a human evaluation of their quality. Expert annotators from each region used the Multi-dimensional Quality Metrics (MQM) framework to identify and categorize errors in the translations. The framework includes a category-wise weighting scheme to convert the identified errors into a single score that roughly represents the number of major errors per sentence; so a lower number indicates a better translation. For each region, we asked MQM raters to score both translations from their region and translations from their language’s other region. For example, Brazilian Portuguese raters scored both the Brazilian and European Portuguese translations. The difference between these two scores indicates the prevalence of linguistic phenomena that are acceptable in one variety but not the other. We found that in both Portuguese and Chinese, raters identified, on average, approximately two more major errors per sentence in the mismatched translations than in the matched ones. This indicates that our dataset truly does capture region-specific phenomena.

While human evaluation is the best way to be sure of model quality, it is often slow and expensive. We therefore wanted to find an existing automatic metric that researchers can use to evaluate their models on our benchmark, and considered chrF, BLEU, and BLEURT. Using the translations from a few baseline models that were also evaluated by our MQM raters, we discovered that BLEURT has the best correlation with human judgments, and that the strength of that correlation (0.65 Pearson correlation coefficient, ρ) is comparable to the inter-annotator consistency (0.70 intraclass correlation).

Metric    Pearson's ρ
chrF    0.48
BLEU    0.58
BLEURT    0.65

Correlation between different automatic metrics and human judgements of translation quality on a subset of FRMT. Values are between -1 and 1; higher is better.

System Performance

Our evaluation covered a handful of recent models capable of few-shot control. Based on human evaluation with MQM, the baseline methods all showed some ability to localize their output for Portuguese, but for Mandarin, they mostly failed to use knowledge of the targeted region to produce superior Mainland or Taiwan translations.

Google’s recent language model, PaLM, was rated best overall among the baselines we evaluated. In order to produce region-targeted translations with PaLM, we feed an instructive prompt into the model and then generate text from it to fill in the blank (see the example shown below).

    Translate the following texts from English to European Portuguese.
English: [English example 1].
European Portuguese: [correct translation 1].
English: [input].
European Portuguese: _____"

PaLM obtained strong results using a single example, and had marginal quality gains on Portuguese when increasing to ten examples. This performance is impressive when taking into consideration that PaLM was trained in an unsupervised way. Our results also suggest language models like PaLM may be particularly adept at memorizing region-specific word choices required for fluent translation. However, there is still a significant performance gap between PaLM and human performance. See our paper for more details.

MQM performance across dataset buckets using human and PaLM translations. Thick bars represent the region-matched case, where raters from each region evaluate translations targeted at their own region. Thin, inset bars represent the region-mismatched case, where raters from each region evaluate translations targeted at the other region. Human translations exhibit regional phenomena in all cases. PaLM translations do so for all Portuguese buckets and the Mandarin lexical bucket only.


In the near future, we hope to see a world where language generation systems, especially machine translation, can support all speaker communities. We want to meet users where they are, generating language fluent and appropriate for their locale or region. To that end, we have released the FRMT dataset and benchmark, enabling researchers to easily compare performance for region-aware MT models. Validated via our thorough human-evaluation studies, the language varieties in FRMT have significant differences that outputs from region-aware MT models should reflect. We are excited to see how researchers utilize this benchmark in development of new MT models that better support under-represented language varieties and all speaker communities, leading to improved equitability in natural-language technologies.


We gratefully acknowledge our paper co-authors for all their contributions to this project: Timothy Dozat, Xavier Garcia, Dan Garrette, Jason Riesa, Orhan Firat, and Noah Constant. For helpful discussion and comments on the paper, we thank Jacob Eisenstein, Noah Fiedel, Macduff Hughes and Mingfei Lau. For essential feedback around specific regional language differences, we thank Andre Araujo, Chung-Ching Chang, Andreia Cunha, Filipe Gonçalves, Nuno Guerreiro, Mandy Guo, Luis Miranda, Vitor Rodrigues and Linting Xue. For logistical support in collecting human translations and ratings, we thank the Google Translate team. We thank the professional translators and MQM raters for their role in producing the dataset. We also thank Tom Small for providing the animation in this post.

Source: Google AI Blog

The Data Cards Playbook: A Toolkit for Transparency in Dataset Documentation

As machine learning (ML) research moves toward large-scale models capable of numerous downstream tasks, a shared understanding of a dataset’s origin, development, intent, and evolution becomes increasingly important for the responsible and informed development of ML models. However, knowledge about datasets, including use and implementations, is often distributed across teams, individuals, and even time. Earlier this year at the ACM Conference on Fairness, Accountability, and Transparency (ACM FAccT), we published Data Cards, a dataset documentation framework aimed at increasing transparency across dataset lifecycles. Data Cards are transparency artifacts that provide structured summaries of ML datasets with explanations of processes and rationale that shape the data and describe how the data may be used to train or evaluate models. At minimum, Data Cards include the following: (1) upstream sources, (2) data collection and annotation methods, (3) training and evaluation methods, (4) intended use, and (5) decisions affecting model performance.

In practice, two critical factors determine the success of a transparency artifact, the ability to identify the information decision-makers use and the establishment of processes and guidance needed to acquire that information. We started to explore this idea in our paper with three “scaffolding” frameworks designed to adapt Data Cards to a variety of datasets and organizational contexts. These frameworks helped us create boundary infrastructures, which are the processes and engagement models that complement technical and functional infrastructure necessary to communicate information between communities of practice. Boundary infrastructures enable dataset stakeholders to find common ground used to provide diverse input into decisions for the creation, documentation, and use of datasets.

Today, we introduce the Data Cards Playbook, a self-guided toolkit for a variety of teams to navigate transparency challenges with their ML datasets. The Playbook applies a human-centered design approach to documentation — from planning a transparency strategy and defining the audience to writing reader-centric summaries of complex datasets — to ensure that the usability and utility of the documented datasets are well understood. We’ve created participatory activities to navigate typical obstacles in setting up a dataset transparency effort, frameworks that can scale data transparency to new data types, and guidance that researchers, product teams and companies can use to produce Data Cards that reflect their organizational principles.

The Data Cards Playbook incorporates the latest in fairness, accountability, and transparency research.

The Data Cards Playbook

We created the Playbook using a multi-pronged approach that included surveys, artifact analysis, interviews, and workshops. We studied what Googlers wanted to know about datasets and models, and how they used that information in their day-to-day work. Over the past two years, we deployed templates for transparency artifacts used by fifteen teams at Google, and when bottlenecks arose, we partnered with these teams to determine appropriate workarounds. We then created over twenty Data Cards that describe image, language, tabular, video, audio, and relational datasets in production settings, some of which are now available on GitHub. This multi-faceted approach provided insights into the documentation workflows, collaborative information-gathering practices, information requests from downstream stakeholders, and review and assessment practices for each Google team.

Moreover, we spoke with design, policy, and technology experts across the industry and academia to get their unique feedback on the Data Cards we created. We also incorporated our learnings from a series of workshops at ACM FAccT in 2021. Within Google, we evaluated the effectiveness and scalability of our solutions with ML researchers, data scientists, engineers, AI ethics reviewers, product managers, and leadership. In the Data Cards Playbook, we’ve translated successful approaches into repeatable practices that can easily be adapted to unique team needs.

Activities, Foundations, and Transparency Patterns

The Data Cards Playbook is modeled after sprints and co-design practices, so cross-functional teams and their stakeholders can work together to define transparency with an eye for real-world problems they experience when creating dataset documentation and governance solutions. The thirty-three available Activities invite broad, critical perspectives from a wide variety of stakeholders, so Data Cards can be useful for decisions across the dataset lifecycle. We partnered with researchers from the Responsible AI team at Google to create activities that can reflect considerations of fairness and accountability. For example, we've adapted Evaluation Gaps in ML practices into a worksheet for more complete dataset documentation.

Download readily-available activity templates to use the Data Cards Playbook in your organization.

We’ve formed Transparency Patterns with evidence-based guidance to help anticipate challenges faced when producing transparent documentation, offer best practices that improve transparency, and make Data Cards useful for readers from different backgrounds. The challenges and their workarounds are based on data and insights from Googlers, industry experts, and academic research.

Patterns help unblock teams with recommended practices, caution against common pitfalls, and suggested alternatives to roadblocks.

The Playbook also includes Foundations, which are scalable concepts and frameworks that explore fundamental aspects of transparency as new contexts of data modalities and ML arise. Each Foundation supports different product development stages and includes key takeaways, actions for teams, and handy resources.

Playbook Modules

The Playbook is organized into four modules: (1) Ask, (2) Inspect, (3) Answer, and (3) Audit. Each module contains a growing compendium of materials teams can use within their workflows to tackle transparency challenges that frequently co-occur. Since Data Cards were created with scalability and extensibility in mind, modules leverage divergence-converge thinking that teams may already use, so documentation isn’t an afterthought. The Ask and Inspect modules help create and evaluate Data Card templates for organizational needs and principles. The Answer and Audit modules help data teams complete the templates and evaluate the resulting Data Cards.

In Ask, teams define transparency and optimize their dataset documentation for cross-functional decision-making. Participatory activities create opportunities for Data Card readers to have a say in what constitutes transparency in the dataset’s documentation. These address specific challenges and are rated for different intensities and durations so teams can mix-and-match activities around their needs.

The Inspect module contains activities to identify gaps and opportunities in dataset transparency and processes from user-centric and dataset-centric perspectives. It supports teams in refining, validating, and operationalizing Data Card templates across an organization so readers can arrive at reasonable conclusions about the datasets described.

The Answer module contains transparency patterns and dataset-exploration activities to answer challenging and ambiguous questions. Topics covered include preparing for transparency, writing reader-centric summaries in documentation, unpacking the usability and utility of datasets, and maintaining a Data Card over time.

The Audit module helps data teams and organizations set up processes to evaluate completed Data Cards before they are published. It also contains guidance to measure and track how a transparency effort for multiple datasets scales within organizations.

In Practice

A data operations team at Google used an early version of the Lenses and Scopes Activities from the Ask modules to create a customized Data Card template. Interestingly, we saw them use this template across their workflow till datasets were handed off. They used Data Cards to take dataset requests from research teams, tracked the various processes to create the datasets, collected metadata from vendors responsible for annotations, and managed approvals. Their experiences of iterating with experts and managing updates are reflected in our Transparency Patterns.

Another data governance group used a more advanced version of the activities to interview stakeholders for their ML health-related initiative. Using these descriptions, they identified stakeholders to co-create their Data Card schema. Voting on Lenses was used to rule out typical documentation questions, and identify atypical documentation needs specific to their data type, and important for decisions frequently made by ML leadership and tactical roles within their team. These questions were then used to customize existing metadata schemas in their data repositories.


We present the Data Cards Playbook, a continuous and contextual approach to dataset transparency that deliberately considers all relevant materials and contexts. With this, we hope to establish and promote practice-oriented foundations for transparency to pave the path for researchers to develop ML systems and datasets that are responsible and benefit society.

In addition to the four Playbook modules described, we’re also open-sourcing a card builder, which generates interactive Data Cards from a Markdown file. You can see the builder in action in the GEM Benchmark project’s Data Cards. The Data Cards created were a result of activities from this Playbook, in which the GEM team identified improvements across all dimensions, and created an interactive collection tool designed around scopes.

We acknowledge that this is not a comprehensive solution for fairness, accountability, or transparency in itself. We’ll continue to improve the Playbook using lessons learned. We hope the Data Cards Playbook can become a robust platform for collaboratively advancing transparency research, and invite you to make this your own.


This work was done in collaboration with Reena Jana, Vivian Tsai, and Oddur Kjartansson. We want to thank Donald Gonzalez, Dan Nanas, Parker Barnes, Laura Rosenstein, Diana Akrong, Monica Caraway, Ding Wang, Danielle Smalls, Aybuke Turker, Emily Brouillet, Andrew Fuchs, Sebastian Gehrmann, Cassie Kozyrkov, Alex Siegman, and Anthony Keene for their immense contributions; and Meg Mitchell and Timnit Gebru for championing this work.

We also want to thank Adam Boulanger, Lauren Wilcox, Roxanne Pinto, Parker Barnes, and Ayça Çakmakli for their feedback; Tulsee Doshi, Dan Liebling, Meredith Morris, Lucas Dixon, Fernanda Viegas, Jen Gennai, and Marian Croak for their support. This work would not have been possible without our workshop and study participants, and numerous partners, whose insights and experiences have shaped this Playbook.

Source: Google AI Blog

Open Images V7 — Now Featuring Point Labels

Open Images is a computer vision dataset covering ~9 million images with labels spanning thousands of object categories. Researchers around the world use Open Images to train and evaluate computer vision models. Since the initial release of Open Images in 2016, which included image-level labels covering 6k categories, we have provided multiple updates to enrich annotations and expand the potential use cases of the dataset. Through several releases, we have added image-level labels for over 20k categories on all images and bounding box annotations, visual relations, instance segmentations, and localized narratives (synchronized voice, mouse trace, and text caption) on a subset of 1.9M images.

Today, we are happy to announce the release of Open Images V7, which expands the Open Images dataset even further with a new annotation type called point-level labels and includes a new all-in-one visualization tool that allows a better exploration of the rich data available.

Point Labels

The main strategy used to collect the new point-level label annotations leveraged suggestions from a machine learning (ML) model and human verification. First, the ML model selected points of interest and asked a yes or no question, e.g., “is this point on a pumpkin?”. Then, human annotators spent an average of 1.1 seconds answering the yes or no questions. We aggregated the answers from different annotators over the same question and assigned a final “yes”, “no”, or “unsure” label to each annotated point.

Illustration of the annotations interface.
(Image by Lenore Edman, under CC BY 2.0 license)

For each annotated image, we provide a collection of points, each with a “yes” or “no” label for a given class. These points provide sparse information that can be used for the semantic segmentation task. We collected a total of 38.6M new point annotations (12.4M with “yes” labels) that cover 5.8 thousand classes and 1.4M images.

By focusing on point labels, we expanded the number of images annotated and categories covered. We also concentrated the efforts of our annotators on efficiently collecting useful information. Compared to our instance segmentation, the new points include 16x more classes and cover more images. The new points also cover 9x more classes than our box annotations. Compared to existing segmentation datasets, like PASCAL VOC, COCO, Cityscapes, LVIS, or ADE20K, our annotations cover more classes and more images than previous work. The new point label annotations are the first type of annotation in Open Images that provides localization information for both things (countable objects, like cars, cats, and catamarans), and stuff categories (uncountable objects like grass, granite, and gravel). Overall, the newly collected data is roughly equivalent to two years of human annotation effort.

Our initial experiments show that this type of sparse data is suitable for both training and evaluating segmentation models. Training a model directly on sparse data allows us to reach comparable quality to training on dense annotations. Similarly, we show that one can directly compute the traditional semantic segmentation intersection-over-union (IoU) metric over sparse data. The ranking across different methods is preserved, and the sparse IoU values are an accurate estimate of its dense version. See our paper for more details.

Below, we show four example images with their point-level labels, illustrating the rich and diverse information these annotations provide. Circles ⭘ are “yes” labels, and squares are “no” labels.

Four example images with point-level labels.
Images by Richie Diesterheft, John AM Nueva, Sarah Ackerman, and C Thomas, all under CC BY 2.0 license.

New Visualizers

In addition to the new data release, we also expanded the available visualizations of the Open Images annotations. The Open Images website now includes dedicated visualizers to explore the localized narratives annotations, the new point-level annotations, and a new all-in-one view. This new all-in-one view is available for the subset of 1.9M densely annotated images and allows one to explore the rich annotations that Open Images has accumulated over seven releases. On average these images have annotations for 6.7 image-labels (classes), 8.3 boxes, 1.7 relations, 1.5 masks, 0.4 localized narratives and 34.8 point-labels per image.

Below, we show two example images with various annotations in the all-in-one visualizer. The figures show the image-level labels, bounding boxes, box relations, instance masks, localized narrative mouse trace and caption, and point-level labels. The + classes have positive annotations (of any kind), while classes have only negative annotations (image-level or point-level).

Two example images with various annotations in the all-in-one visualizer.
Images by Jason Paris, and Rubén Vique, all under CC BY 2.0 license.


We hope that this new data release will enable computer vision research to cover ever more diverse and challenging scenarios. As the quality of automated semantic segmentation models improves over common classes, we want to move towards the long tail of visual concepts, and sparse point annotations are a step in that direction. More and more works are exploring how to use such sparse annotations (e.g., as supervision for instance segmentation or semantic segmentation), and Open Images V7 contributes to this research direction. We are looking forward to seeing what you will build next.


Thanks to Vittorio Ferrari, Jordi Pont-Tuset, Alina Kuznetsova, Ashlesha Sadras, and the annotators team for their support creating this new data release.

Source: Google AI Blog

Crossmodal-3600 — Multilingual Reference Captions for Geographically Diverse Images

Image captioning is the machine learning task of automatically generating a fluent natural language description for a given image. This task is important for improving accessibility for visually impaired users and is a core task in multimodal research encompassing both vision and language modeling.

However, datasets for image captioning are primarily available in English. Beyond that, there are only a few datasets covering a limited number of languages that represent just a small fraction of the world’s population. Further, these datasets feature images that severely under-represent the richness and diversity of cultures from across the globe. These aspects have hindered research on image captioning for a wide variety of languages, and directly hamper the deployment of accessibility solutions for a large potential audience around the world.

Today we present and make publicly available the Crossmodal 3600 (XM3600) image captioning evaluation dataset as a robust benchmark for multilingual image captioning that enables researchers to reliably compare research contributions in this emerging field. XM3600 provides 261,375 human-generated reference captions in 36 languages for a geographically diverse set of 3600 images. We show that the captions are of high quality and the style is consistent across languages.

The Crossmodal 3600 dataset includes reference captions in 36 languages for each of a geographically diverse set of 3600 images. All images used with permission under the CC-BY 2.0 license.

Overview of the Crossmodal 3600 Dataset
Creating large training and evaluation datasets in multiple languages is a resource-intensive endeavor. Recent work has shown that it is feasible to build multilingual image captioning models trained on machine-translated data with English captions as the starting point. However, some of the most reliable automatic metrics for image captioning are much less effective when applied to evaluation sets with translated image captions, resulting in poorer agreement with human evaluations compared to the English case. As such, trustworthy model evaluation at present can only be based on extensive human evaluation. Unfortunately, such evaluations usually cannot be replicated across different research efforts, and therefore do not offer a fast and reliable mechanism to automatically evaluate multiple model parameters and configurations (e.g., model hill climbing) or to compare multiple lines of research.

XM3600 provides 261,375 human-generated reference captions in 36 languages for a geographically diverse set of 3600 images from the Open Images dataset. We measure the quality of generated captions by comparing them to the manually provided captions using the CIDEr metric, which ranges from 0 (unrelated to the reference captions) to 10 (perfectly matching the reference captions). When comparing pairs of models, we observed strong correlations between the differences in the CIDEr scores of the model outputs, and side-by-side human evaluations comparing the model outputs. , making XM3600 is a reliable tool for high-quality automatic comparisons between image captioning models on a wide variety of languages beyond English.

Language Selection
We chose 30 languages beyond English, roughly based on their percentage of web content. In addition, we chose an additional five languages that include under-resourced languages that have many native speakers or major native languages from continents that would not be covered otherwise. Finally, we also included English as a baseline, thus resulting in a total of 36 languages, as listed in the table below.

Arabic     Bengali*     Chinese     Croatian     Cusco
Danish     Dutch     English     Filipino     Finnish     French
German     Greek     Hebrew     Hindi     Hungarian     Indonesian
Italian     Japanese     Korean     Maori*     Norwegian     Persian
Polish     Portuguese     Romanian     Russian     Spanish     Swahili*
Swedish     Telugu*     Thai     Turkish     Ukrainian     Vietnamese
List of languages used in XM3600.   *Low-resource languages with many native speakers, or major native languages from continents that would not be covered otherwise.

Image Selection
The images were selected from among those in the Open Images dataset that have location metadata. Since there are many regions where more than one language is spoken, and some areas are not well covered by these images, we designed an algorithm to maximize the correspondence between selected images and the regions where the targeted languages are spoken. The algorithm starts with the selection of images with geo-data corresponding to the languages for which we have the smallest pool (e.g., Persian) and processes them in increasing order of their candidate image pool size. If there aren't enough images in an area where a language is spoken, then we gradually expand the geographic selection radius to: (i) a country where the language is spoken; (ii) a continent where the language is spoken; and, as last resort, (iii) from anywhere in the world. This strategy succeeded in providing our target number of 100 images from an appropriate region for most of the 36 languages, except for Persian (where 14 continent-level images are used) and Hindi (where all 100 images are at the global level, because the in-region images were assigned to Bengali and Telugu).


Photo by Chris Sampson

Photo by Henrik Palm

Photo by rojypala

Cusco Quechua

Photo by McKay Savage

Photo by Simon Schoeters

Photo by Stefan Krasowski
Sample images showcasing the geographical diversity of the annotated images. Images used under CC BY 2.0 license.

Caption Generation
In total, all 3600 images (100 images per language) are annotated in all 36 languages, each with an average of two annotations per language, yielding a total of 261,375 captions.

Annotators work in batches of 15 images. The first screen shows all 15 images with their captions in English as generated by a captioning model trained to output a consistent style of the form "<main salient objects> doing <activities> in the <environment>", often with object attributes, such as a "smiling" person, "red" car, etc. The annotators are asked to rate the caption quality given guidelines for a 4-point scale from "excellent" to "bad", plus an option for "not_enough_information". This step forces the annotators to carefully assess caption quality and it primes them to internalize the style of the captions. The following screens show the images again but individually and without the English captions, and the annotators are asked to produce descriptive captions in the target language for each image.

The image batch size of 15 was chosen so that the annotators would internalize the style without remembering the exact captions. Thus, we expect the raters to generate captions based on the image content only and lacking translation artifacts. For example in the example shown below, the Spanish caption mentions “number 42” and the Thai caption mentions “convertibles”, none of which are mentioned in the English captions. The annotators were also provided with a protocol to use when creating the captions, thus achieving style consistency across languages.

Photo by Brian Solis
    English     A vintage sports car in a showroom with many other vintage sports cars
The branded classic cars in a row at display
Spanish     Automóvil clásico deportivo en exhibición de automóviles de galería — (Classic sports car in gallery car show)
Coche pequeño de carreras color plateado con el número 42 en una exhibición de coches — (Small silver racing car with the number 42 at a car show)
Thai     รถเปิดประทุนหลายสีจอดเรียงกันในที่จัดแสดง — (Multicolored convertibles line up in the exhibit)
รถแข่งวินเทจจอดเรียงกันหลายคันในงานจัดแสดง — (Several vintage racing cars line up at the show.)
Sample captions in three different languages (out of 36 — see full list of captions in Appendix A of the Crossmodal-3600 paper), showcasing the creation of annotations that are consistent in style across languages, while being free of direct-translation artifacts (e.g., the Spanish “number 42” or the Thai “convertibles” would not be possible when directly translating from the English versions). Image used under CC BY 2.0 license.

Caption Quality and Statistics
We ran two to five pilot studies per language to troubleshoot the caption generation process and to ensure high quality captions. We then manually evaluated a random subset of captions. First we randomly selected a sample of 600 images. Then, to measure the quality of captions in a particular language, for each image, we selected for evaluation one of the manually generated captions. We found that:

  • For 25 out of 36 languages, the percentage of captions rated as “Good” or “Excellent” is above 90%, and the rest are all above 70%.
  • For 26 out of 36 languages, the percentage of captions rated as “Bad” is below 2%, and the rest are all below 5%.

For languages that use spaces to separate words, the number of words per caption can be as low as 5 or 6 for some agglutinative languages like Cusco Quechua and Czech, and as high as 18 for an analytic language like Vietnamese. The number of characters per caption also varies drastically — from mid-20s for Korean to mid-90s for Indonesian — depending on the alphabet and the script of the language.

Empirical Evaluation and Results
We empirically measured the ability of the XM3600 annotations to rank image captioning model variations by training four variations of a multilingual image captioning model and comparing the CIDEr differences of the models’ outputs over the XM3600 dataset for 30+ languages, to side-by-side human evaluations. We observed strong correlations between the CIDEr differences and the human evaluations. These results support the use of the XM3600 references as a means to achieve high-quality automatic comparisons between image captioning models on a wide variety of languages beyond English.

Recent Uses
Recently PaLI used XM3600 to evaluate model performance beyond English for image captioning, image-to-text retrieval and text-to-image retrieval. The key takeaways they found when evaluating on XM3600 were that multilingual captioning greatly benefits from scaling the PaLI models, especially for low-resource languages.

We would like to acknowledge the coauthors of this work: Xi Chen and Radu Soricut.

Source: Google AI Blog

TensorStore for High-Performance, Scalable Array Storage

Many exciting contemporary applications of computer science and machine learning (ML) manipulate multidimensional datasets that span a single large coordinate system, for example, weather modeling from atmospheric measurements over a spatial grid or medical imaging predictions from multi-channel image intensity values in a 2d or 3d scan. In these settings, even a single dataset may require terabytes or petabytes of data storage. Such datasets are also challenging to work with as users may read and write data at irregular intervals and varying scales, and are often interested in performing analyses using numerous machines working in parallel.

Today we are introducing TensorStore, an open-source C++ and Python software library designed for storage and manipulation of n-dimensional data that:

TensorStore has already been used to solve key engineering challenges in scientific computing (e.g., management and processing of large datasets in neuroscience, such as peta-scale 3d electron microscopy data and “4d” videos of neuronal activity). TensorStore has also been used in the creation of large-scale machine learning models such as PaLM by addressing the problem of managing model parameters (checkpoints) during distributed training.

Familiar API for Data Access and Manipulation
TensorStore provides a simple Python API for loading and manipulating large array data. In the following example, we create a TensorStore object that represents a 56 trillion voxel 3d image of a fly brain and access a small 100x100 patch of the data as a NumPy array:

>>> import tensorstore as ts
>>> import numpy as np

# Create a TensorStore object to work with fly brain data.
>>> dataset = ts.open({
... 'driver':
... 'neuroglancer_precomputed',
... 'kvstore':
... 'gs://neuroglancer-janelia-flyem-hemibrain/v1.1/segmentation/',
... }).result()

# Create a 3-d view (remove singleton 'channel' dimension):
>>> dataset_3d = dataset[ts.d['channel'][0]]
>>> dataset_3d.domain
{ "x": [0, 34432), "y": [0, 39552), "z": [0, 41408) }

# Convert a 100x100x1 slice of the data to a numpy ndarray
>>> slice = np.array(dataset_3d[15000:15100, 15000:15100, 20000])

Crucially, no actual data is accessed or stored in memory until the specific 100x100 slice is requested; hence arbitrarily large underlying datasets can be loaded and manipulated without having to store the entire dataset in memory, using indexing and manipulation syntax largely identical to standard NumPy operations. TensorStore also provides extensive support for advanced indexing features, including transforms, alignment, broadcasting, and virtual views (data type conversion, downsampling, lazily on-the-fly generated arrays).

The following example demonstrates how TensorStore can be used to create a zarr array, and how its asynchronous API enables higher throughput:

>>> import tensorstore as ts
>>> import numpy as np

>>> # Create a zarr array on the local filesystem
>>> dataset = ts.open({
... 'driver': 'zarr',
... 'kvstore': 'file:///tmp/my_dataset/',
... },
... dtype=ts.uint32,
... chunk_layout=ts.ChunkLayout(chunk_shape=[256, 256, 1]),
... create=True,
... shape=[5000, 6000, 7000]).result()

>>> # Create two numpy arrays with example data to write.
>>> a = np.arange(100*200*300, dtype=np.uint32).reshape((100, 200, 300))
>>> b = np.arange(200*300*400, dtype=np.uint32).reshape((200, 300, 400))

>>> # Initiate two asynchronous writes, to be performed concurrently.
>>> future_a = dataset[1000:1100, 2000:2200, 3000:3300].write(a)
>>> future_b = dataset[3000:3200, 4000:4300, 5000:5400].write(b)

>>> # Wait for the asynchronous writes to complete
>>> future_a.result()
>>> future_b.result()

Safe and Performant Scaling
Processing and analyzing large numerical datasets requires significant computational resources. This is typically achieved through parallelization across numerous CPU or accelerator cores spread across many machines. Therefore a fundamental goal of TensorStore has been to enable parallel processing of individual datasets that is both safe (i.e., avoids corruption or inconsistencies arising from parallel access patterns) and high performance (i.e., reading and writing to TensorStore is not a bottleneck during computation). In fact, in a test within Google’s datacenters, we found nearly linear scaling of read and write performance as the number of CPUs was increased:

Read and write performance for a TensorStore dataset in zarr format residing on Google Cloud Storage (GCS) accessed concurrently using a variable number of single-core compute tasks in Google data centers. Both read and write performance scales nearly linearly with the number of compute tasks.

Performance is achieved by implementing core operations in C++, extensive use of multithreading for operations such as encoding/decoding and network I/O, and partitioning large datasets into much smaller units through chunking to enable efficiently reading and writing subsets of the entire dataset. TensorStore also provides configurable in-memory caching (which reduces slower storage system interactions for frequently accessed data) and an asynchronous API that enables a read or write operation to continue in the background while a program completes other work.

Safety of parallel operations when many machines are accessing the same dataset is achieved through the use of optimistic concurrency, which maintains compatibility with diverse underlying storage layers (including Cloud storage platforms, such as GCS, as well as local filesystems) without significantly impacting performance. TensorStore also provides strong ACID guarantees for all individual operations executing within a single runtime.

To make distributed computing with TensorStore compatible with many existing data processing workflows, we have also integrated TensorStore with parallel computing libraries such as Apache Beam (example code) and Dask (example code).

Use Case: Language Models
An exciting recent development in ML is the emergence of more advanced language models such as PaLM. These neural networks contain hundreds of billions of parameters and exhibit some surprising capabilities in natural language understanding and generation. These models also push the limits of computational infrastructure; in particular, training a language model such as PaLM requires thousands of TPUs working in parallel.

One challenge that arises during this training process is efficiently reading and writing the model parameters. Training is distributed across many separate machines, but parameters must be regularly saved to a single object (“checkpoint”) on a permanent storage system without slowing down the overall training process. Individual training jobs must also be able to read just the specific set of parameters they are concerned with in order to avoid the overhead that would be required to load the entire set of model parameters (which could be hundreds of gigabytes).

TensorStore has already been used to address these challenges. It has been applied to manage checkpoints associated with large-scale (“multipod”) models trained with JAX (code example) and has been integrated with frameworks such as T5X (code example) and Pathways. Model parallelism is used to partition the full set of parameters, which can occupy more than a terabyte of memory, over hundreds of TPUs. Checkpoints are stored in zarr format using TensorStore, with a chunk structure chosen to allow the partition for each TPU to be read and written independently in parallel.

When saving a checkpoint, each model parameter is written using TensorStore in zarr format using a chunk grid that further subdivides the grid used to partition the parameter over TPUs. The host machines write in parallel the zarr chunks for each of the partitions assigned to TPUs attached to that host. Using TensorStore's asynchronous API, training proceeds even while the data is still being written to persistent storage. When resuming from a checkpoint, each host reads only the chunks that make up the partitions assigned to that host.

Use Case: 3D Brain Mapping
The field of synapse-resolution connectomics aims to map the wiring of animal and human brains at the detailed level of individual synaptic connections. This requires imaging the brain at extremely high resolution (nanometers) over fields of view of up to millimeters or more, which yields datasets that can span petabytes in size. In the future these datasets may extend to exabytes as scientists contemplate mapping entire mouse or primate brains. However, even current datasets pose significant challenges related to storage, manipulation, and processing; in particular, even a single brain sample may require millions of gigabytes with a coordinate system (pixel space) of hundreds of thousands pixels in each dimension.

We have used TensorStore to solve computational challenges associated with large-scale connectomic datasets. Specifically, TensorStore has managed some of the largest and most widely accessed connectomic datasets, with Google Cloud Storage as the underlying object storage system. For example, it has been applied to the human cortex “h01” dataset, which is a 3d nanometer-resolution image of human brain tissue. The raw imaging data is 1.4 petabytes (roughly 500,000 * 350,000 * 5,000 pixels large, and is further associated with additional content such as 3d segmentations and annotations that reside in the same coordinate system. The raw data is subdivided into individual chunks 128x128x16 pixels large and stored in the “Neuroglancer precomputed” format, which is optimized for web-based interactive viewing and can be easily manipulated from TensorStore.

A fly brain reconstruction for which the underlying data can be easily accessed and manipulated using TensorStore.

Getting Started
To get started using the TensorStore Python API, you can install the tensorstore PyPI package using:

pip install tensorstore

Refer to the tutorials and API documentation for usage details. For other installation options and for using the C++ API, refer to installation instructions.

Thanks to Tim Blakely, Viren Jain, Yash Katariya, Jan-Matthis Luckmann, Michał Januszewski, Peter Li, Adam Roberts, Brain Williams, and Hector Yee from Google Research, and Davis Bennet, Stuart Berg, Eric Perlman, Stephen Plaza, and Juan Nunez-Iglesias from the broader scientific community for valuable feedback on the design, early testing and debugging.

Source: Google AI Blog

Announcing the Patent Phrase Similarity Dataset

Patent documents typically use legal and highly technical language, with context-dependent terms that may have meanings quite different from colloquial usage and even between different documents. The process of using traditional patent search methods (e.g., keyword searching) to search through the corpus of over one hundred million patent documents can be tedious and result in many missed results due to the broad and non-standard language used. For example, a "soccer ball" may be described as a "spherical recreation device", "inflatable sportsball" or "ball for ball game". Additionally, the language used in some patent documents may obfuscate terms to their advantage, so more powerful natural language processing (NLP) and semantic similarity understanding can give everyone access to do a thorough search.

The patent domain (and more general technical literature like scientific publications) poses unique challenges for NLP modeling due to its use of legal and technical terms. While there are multiple commonly used general-purpose semantic textual similarity (STS) benchmark datasets (e.g., STS-B, SICK, MRPC, PIT), to the best of our knowledge, there are currently no datasets focused on technical concepts found in patents and scientific publications (the somewhat related BioASQ challenge contains a biomedical question answering task). Moreover, with the continuing growth in size of the patent corpus (millions of new patents are issued worldwide every year), there is a need to develop more useful NLP models for this domain.

Today, we announce the release of the Patent Phrase Similarity dataset, a new human-rated contextual phrase-to-phrase semantic matching dataset, and the accompanying paper, presented at the SIGIR PatentSemTech Workshop, which focuses on technical terms from patents. The Patent Phrase Similarity dataset contains ~50,000 rated phrase pairs, each with a Cooperative Patent Classification (CPC) class as context. In addition to similarity scores that are typically included in other benchmark datasets, we include granular rating classes similar to WordNet, such as synonym, antonym, hypernym, hyponym, holonym, meronym, and domain related. This dataset (distributed under the Creative Commons Attribution 4.0 International license) was used by Kaggle and USPTO as the benchmark dataset in the U.S. Patent Phrase to Phrase Matching competition to draw more attention to the performance of machine learning models on technical text. Initial results show that models fine-tuned on this new dataset perform substantially better than general pre-trained models without fine-tuning.

The Patent Phrase Similarity Dataset
To better train the next generation of state-of-the-art models, we created the Patent Phrase Similarity dataset, which includes many examples to address the following problems: (1) phrase disambiguation, (2) adversarial keyword matching, and (3) hard negative keywords (i.e., keywords that are unrelated but received a high score for similarity from other models ). Some keywords and phrases can have multiple meanings (e.g., the phrase "mouse" may refer to an animal or a computer input device), so we disambiguate the phrases by including CPC classes with each pair of phrases. Also, many NLP models (e.g., bag of words models) will not do well on data with phrases that have matching keywords but are otherwise unrelated (adversarial keywords, e.g., “container section” → “kitchen container”, “offset table” → “table fan”). The Patent Phrase Similarity dataset is designed to include many examples of matching keywords that are unrelated through adversarial keyword match, enabling NLP models to improve their performance.

Each entry in the Patent Phrase Similarity dataset contains two phrases, an anchor and target, a context CPC class, a rating class, and a similarity score. The dataset contains 48,548 entries with 973 unique anchors, split into training (75%), validation (5%), and test (20%) sets. When splitting the data, all of the entries with the same anchor are kept together in the same set. There are 106 different context CPC classes and all of them are represented in the training set.

Anchor Target Context Rating Score
acid absorption absorption of acid B08 exact 1.0
acid absorption acid immersion B08 synonym 0.75
acid absorption chemically soaked B08 domain related 0.25
acid absorption acid reflux B08 not related 0.0
gasoline blend petrol blend C10 synonym 0.75
gasoline blend fuel blend C10 hypernym 0.5
gasoline blend fruit blend C10 not related 0.0
faucet assembly water tap A22 hyponym 0.5
faucet assembly water supply A22 holonym 0.25
faucet assembly school assembly A22 not related 0.0
A small sample of the dataset with anchor and target phrases, context CPC class (B08: Cleaning, C10: Petroleum, gas, fuel, lubricants, A22: Butchering, processing meat/poultry/fish), a rating class, and a similarity score.

Generating the Dataset
To generate the Patent Phrase Similarity data, we first process the ~140 million patent documents in the Google Patent's corpus and automatically extract important English phrases, which are typically noun phrases (e.g., “fastener”, “lifting assembly”) and functional phrases (e.g., “food processing”, “ink printing”). Next, we filter and keep phrases that appear in at least 100 patents and randomly sample around 1,000 of these filtered phrases, which we call anchor phrases. For each anchor phrase, we find all of the matching patents and all of the CPC classes for those patents. We then randomly sample up to four matching CPC classes, which become the context CPC classes for the specific anchor phrase.

We use two different methods for pre-generating target phrases: (1) partial matching and (2) a masked language model (MLM). For partial matching, we randomly select phrases from the entire corpus that partially match with the anchor phrase (e.g., “abatement” → “noise abatement”, “material formation” → “formation material”). For MLM, we select sentences from the patents that contain a given anchor phrase, mask them out, and use the Patent-BERT model to predict candidates for the masked portion of the text. Then, all of the phrases are cleaned up, which includes lowercasing and the removal of punctuation and certain stopwords (e.g., "and", "or", "said"), and sent to expert raters for review. Each phrase pair is rated independently by two raters skilled in the technology area. Each rater also generates new target phrases with different ratings. Specifically, they are asked to generate some low-similarity and unrelated targets that partially match with the original anchor and/or some high-similarity targets. Finally, the raters meet to discuss their ratings and come up with final ratings.

Dataset Evaluation
To evaluate its performance, the Patent Phrase Similarity dataset was used in the U.S. Patent Phrase to Phrase Matching Kaggle competition. The competition was very popular, drawing about 2,000 competitors from around the world. A variety of approaches were successfully used by the top scoring teams, including ensemble models of BERT variants and prompting (see the full discussion for more details). The table below shows the best results from the competition, as well as several off-the-shelf baselines from our paper. The Pearson correlation metric was used to measure the linear correlation between the predicted and true scores, which is a helpful metric to target for downstream models so they can distinguish between different similarity ratings.

The baselines in the paper can be considered zero-shot in the sense that they use off-the-shelf models without any further fine-tuning on the new dataset (we use these models to embed the anchor and target phrases separately and compute the cosine similarity between them). The Kaggle competition results demonstrate that by using our training data, one can achieve significant improvements compared with existing NLP models. We have also estimated human performance on this task by comparing a single rater’s scores to the combined score of both raters. The results indicate that this is not a particularly easy task, even for human experts.

Model Training Pearson correlation
word2vec Zero-shot 0.44
Patent-BERT Zero-shot 0.53
Sentence-BERT Zero-shot 0.60
Kaggle 1st place single Fine-tuned 0.87
Kaggle 1st place ensemble Fine-tuned 0.88
Human 0.93
Performance of popular models with no fine-tuning (zero-shot), models fine-tuned on the Patent Phrase Similarity dataset as part of the Kaggle competition, and single human performance.

Conclusion and Future Work
We present the Patent Phrase Similarity dataset, which was used as the benchmark dataset in the U.S. Patent Phrase to Phrase Matching competition, and demonstrate that by using our training data, one can achieve significant improvements compared with existing NLP models.

Additional challenging machine learning benchmarks can be generated from the patent corpus, and patent data has made its way into many of today's most-studied models. For example, the C4 text dataset used to train T5 contains many patent documents. The BigBird and LongT5 models also use patents via the BIGPATENT dataset. The availability, breadth and open usage terms of full text data (see Google Patents Public Datasets) makes patents a unique resource for the research community. Possibilities for future tasks include massively multi-label classification, summarization, information retrieval, image-text similarity, citation graph prediction, and translation. See the paper for more details.

This work was possible through a collaboration with Kaggle, Satsyil Corp., USPTO, and MaxVal. Thanks to contributors Ian Wetherbee from Google, Will Cukierski and Maggie Demkin from Kaggle. Thanks to Jerry Ma, Scott Beliveau, and Jamie Holcombe from USPTO and Suja Chittamahalingam from MaxVal for their contributions.

Source: Google AI Blog

UVQ: Measuring YouTube’s Perceptual Video Quality

Online video sharing platforms, like YouTube, need to understand perceptual video quality (i.e., a user's subjective perception of video quality) in order to better optimize and improve user experience. Video quality assessment (VQA) attempts to build a bridge between video signals and perceptual quality by using objective mathematical models to approximate the subjective opinions of users. Traditional video quality metrics, like peak signal-to-noise ratio (PSNR) and Video Multi-Method Assessment Fusion (VMAF), are reference-based and focus on the relative difference between the target and reference videos. Such metrics, which work best on professionally generated content (e.g., movies), assume the reference video is of pristine quality and that one can induce the target video's absolute quality from the relative difference.

However, the majority of the videos that are uploaded on YouTube are user-generated content (UGC), which bring new challenges due to their remarkably high variability in video content and original quality. Most UGC uploads are non-pristine and the same amount of relative difference could imply very different perceptual quality impacts. For example, people tend to be less sensitive to the distortions of poor quality uploads than of high quality uploads. Thus, reference-based quality scores become inaccurate and inconsistent when used for UGC cases. Additionally, despite the high volume of UGC, there are currently limited UGC video quality assessment (UGC-VQA) datasets with quality labels. Existing UGC-VQA datasets are either small in size (e.g., LIVE-Qualcomm has 208 samples captured from 54 unique scenes), compared with datasets with millions of samples for classification and recognition (e.g., ImageNet and YouTube-8M), or don’t have enough content variability (sampling without considering content information, like LIVE-VQC and KoNViD-1k).

In "Rich Features for Perceptual Quality Assessment of UGC Videos", published at CVPR 2021, we describe how we attempt to solve the UGC quality assessment problem by building a Universal Video Quality (UVQ) model that resembles a subjective quality assessment. The UVQ model uses subnetworks to analyze UGC quality from high-level semantic information to low-level pixel distortions, and provides a reliable quality score with rationale (leveraging comprehensive and interpretable quality labels). Moreover, to advance UGC-VQA and compression research, we enhance the open-sourced YouTube-UGC dataset, which contains 1.5K representative UGC samples from millions of UGC videos (distributed under the Creative Commons license) on YouTube. The updated dataset contains ground-truth labels for both original videos and corresponding transcoded versions, enabling us to better understand the relationship between video content and its perceptual quality.

Subjective Video Quality Assessment
To understand perceptual video quality, we leverage an internal crowd-sourcing platform to collect mean opinion scores (MOS) with a scale of 1–5, where 1 is the lowest quality and 5 is the highest quality, for no-reference use cases. We collect ground-truth labels from the YouTube-UGC dataset and categorize UGC factors that affect quality perception into three high-level categories: (1) content, (2) distortions, and (3) compression. For example, a video with no meaningful content won't receive a high quality MOS. Also, distortions introduced during the video production phase and video compression artifacts introduced by third-party platforms, e.g., transcoding or transmission, will degrade the overall quality.

MOS= 2.052 MOS= 4.457
Left: A video with no meaningful content won't receive a high quality MOS. Right: A video displaying intense sports shows a higher MOS.
MOS= 1.242 MOS= 4.522
Left: A blurry gaming video gets a very low quality MOS. Right: A video with professional rendering (high contrast and sharp edges, usually introduced in the video production phase) shows a high quality MOS.
MOS= 2.372 MOS= 4.646
Left: A heavily compressed video receives a low quality MOS. Right: a video without compression artifacts shows a high quality MOS.

We demonstrate that the left gaming video in the second row of the figure above has the lowest MOS (1.2), even lower than the video with no meaningful content. A possible explanation is that viewers may have higher video quality expectations for videos that have a clear narrative structure, like gaming videos, and the blur artifacts significantly reduce the perceptual quality of the video.

UVQ Model Framework
A common method for evaluating video quality is to design sophisticated features, and then map these features to a MOS. However, designing useful handcrafted features is difficult and time-consuming, even for domain experts. Also, the most useful existing handcrafted features were summarized from limited samples, which may not perform well on broader UGC cases. In contrast, machine learning is becoming more prominent in UGC-VQA because it can automatically learn features from large-scale samples.

A straightforward approach is to train a model from scratch on existing UGC quality datasets. However, this may not be feasible as there are limited quality UGC datasets. To overcome this limitation, we apply a self-supervised learning step to the UVQ model during training. This self-supervised step enables us to learn comprehensive quality-related features, without ground-truth MOS, from millions of raw videos.

Following the quality-related categories summarized from the subjective VQA, we develop the UVQ model with four novel subnetworks. The first three subnetworks, which we call ContentNet, DistortionNet and CompressionNet, are used to extract quality features (i.e., content, distortion and compression), and the fourth subnetwork, called AggregationNet, maps the extracted features to generate a single quality score. ContentNet is trained in a supervised learning fashion with UGC-specific content labels that are generated by the YouTube-8M model. DistortionNet is trained to detect common distortions, e.g., Gaussian blur and white noise of the original frame. CompressionNet focuses on video compression artifacts, whose training data are videos compressed with different bitrates. CompressionNet is trained using two compressed variants of the same content that are fed into the model to predict corresponding compression levels (with a higher score for more noticeable compression artifacts), with the implicit assumption that the higher bitrate version has a lower compression level.

The ContentNet, DistortionNet and CompressionNet subnetworks are trained on large-scale samples without ground-truth quality scores. Since video resolution is also an important quality factor, the resolution-sensitive subnetworks (CompressionNet and DistortionNet) are patch-based (i.e., each input frame is divided into multiple disjointed patches that are processed separately), which makes it possible to capture all detail on native resolution without downscaling. The three subnetworks extract quality features that are then concatenated by the fourth subnetwork, AggregationNet, to predict quality scores with domain ground-truth MOS from YouTube-UGC.

The UVQ training framework.

Analyzing Video Quality with UVQ
After building the UVQ model, we use it to analyze the video quality of samples pulled from YouTube-UGC and demonstrate that its subnetworks can provide a single quality score along with high-level quality indicators that can help us understand quality issues. For example, DistortionNet detects multiple visual artifacts, e.g., jitter and lens blur, for the middle video below, and CompressionNet detects that the bottom video has been heavily compressed.

ContentNet assigns content labels with corresponding probabilities in parentheses, i.e., car (0.58), vehicle (0.42), sports car (0.32), motorsports (0.18), racing (0.11).
DistortionNet detects and categorizes multiple visual distortions with corresponding probabilities in parentheses, i.e., jitter (0.112), color quantization (0.111), lens blur (0.108), denoise (0.107).
CompressionNet detects a high compression level of 0.892 for the video above.

Additionally, UVQ can provide patch-based feedback to locate quality issues. Below, UVQ reports that the quality of the first patch (patch at time t = 1) is good with a low compression level. However, the model identifies heavy compression artifacts in the next patch (patch at time t = 2).

Patch at time t = 1 Patch at time t = 2
Compression level = 0.000 Compression level = 0.904
UVQ detects a sudden quality degradation (high compression level) for a local patch.

In practice, UVQ can generate a video diagnostic report that includes a content description (e.g., strategy video game), distortion analysis (e.g., the video is blurry or pixelated) and compression level (e.g., low or high compression). Below, UVQ reports that the content quality, looking at individual features, is good, but the compression and distortion quality is low. When combining all three features, the overall quality is medium-low. We see that these findings are close to the rationale summarized by internal user experts, demonstrating that UVQ can reason through quality assessments, while providing a single quality score.

UVQ diagnostic report. ContentNet (CT): Video game, strategy video game, World of Warcraft, etc. DistortionNet (DT): multiplicative noise, Gaussian blur, color saturation, pixelate, etc. CompressionNet (CP): 0.559 (medium-high compression). Predicted quality score in [1, 5]: (CT, DT, CP) = (3.901, 3.216, 3.151), (CT+DT+CP) = 3.149 (medium-low quality).

We present the UVQ model, which generates a report with quality scores and insights that can be used to interpret UGC video perceptual quality. UVQ learns comprehensive quality related features from millions of UGC videos and provides a consistent view of quality interpretation for both no-reference and reference cases. To learn more, read our paper or visit our website to see YT-UGC videos and their subjective quality data. We also hope that the enhanced YouTube-UGC dataset enables more research in this space.

This work was possible through a collaboration spanning several Google teams. Key contributors include: Balu Adsumilli, Neil Birkbeck, Joong Gon Yim from YouTube and Junjie Ke, Hossein Talebi, Peyman Milanfar from Google Research. Thanks to Ross Wolf, Jayaprasanna Jayaraman, Carena Church, and Jessie Lin for their contributions.

Source: Google AI Blog