Adding Benchmaxxer Repellant to the Open ASR Leaderboard

Source: https://huggingface.co/blog/open-asr-leaderboard-private-data Back to Articles

• New high-quality, private datasets
• How can I evaluate my model on this data?
• Do models trained on the data providers have an advantage?
• What’s next?

“When a measure becomes a target, it ceases to be a good measure.” (Goodhart’s Law)

TLDR:Appen Inc.andDataoceanAIhave provided high-quality English ASR datasets covering scripted and conversational speech over multiple accents. To prevent potential risks of benchmaxxing or test-set contamination, we will keep these datasets private for a high-quality measure of performance on multiple tasks.

We’re not updating the average WER at this time: by default, the leaderboard’s Average WER remains computed on public datasets only. You can optionally include the private datasets using the toggle to see their impact 👀

Since its launch in September 2023, theOpen ASR Leaderboardhas been visited over 710K times. We’re blown away by the community’s interest and motivation to keep pushing speech recognition 🗣️

Two words sum up the objectives (but also challenges) in maintaining a benchmark like the Open ASR Leaderboard:

1. Standardization: models can have different conventions for their usage and outputs, e.g. with/without punctuation and casing. Datasets have the same challenges and can be structured differently. To this end, all test sets have been gathered into asingle dataseton the Hub for easy access and previewing. Moreover, to standardize model outputs and dataset transcripts, we use anormalizerthat (among other things) removes punctuation and casing, and maps to American spelling. It is based on the normalizer ofWhisper.
2. Openness: theUI codeandevaluation scriptsare open-sourced. This has helped not only to incorporate new models, but also to improve the quality of the evaluation procedure through community feedback and contributions.

Standardization and openness are essential for meaningful benchmarking, but they also make benchmarks more susceptible to benchmark-specific optimization (“benchmaxxing”), where models improve leaderboard performance without corresponding gains in real-world robustness. As models and use cases evolve, the Open ASR Leaderboard will continue incorporating high-quality datasets and new evaluation settings to better reflect real-world performance and improve robustness against benchmark-specific optimization.

As discussed in ourreport, there is no single “catch-all” ASR model: some perform better on American English, others on diverse accents and multilingual settings, while others are optimized for speed or conversational audio. Different applications also prioritize different capabilities, so a model that performs less well on one dimension is not necessarily a worse model overall. The goal of the Open ASR Leaderboard is to capture these nuances and provide a more holistic view of ASR performance.

https://huggingface.co/blog/open-asr-leaderboard-private-data#new-high-quality-private-datasetsNew high-quality, private datasets

To this end, we have worked with Appen Inc. and DataoceanAI to curate high-quality datasets for ASR benchmarking. Below is some information on the various splits.

DatasetAccentDuration [h]Male (%) / Female (%)StyleTranscriptionAppen Scripted AUAustralian1.4249 / 51ReadPunctuated, cased.Appen Scripted CACanadian1.5352 / 48ReadPunctuated, cased.Appen Scripted INIndian1.0249 / 51ReadPunctuated, cased.Appen Scripted USAmerican1.4549 / 51ReadPunctuated, cased.Appen Conversational INIndian1.3751 / 49Conversational, spontaneousPunctuated, disfluencies.Appen Conversational US003American1.6449 / 51Conversational, spontaneousPunctuated, cased, disfluencies.Appen Conversational US004American1.6549 / 51Conversational, spontaneousPunctuated, disfluencies.DataoceanAI Scripted USAmerican2.4354 / 46ReadPunctuated, cased (proper nouns), disfluencies.DataoceanAI Scripted GBBritish2.4347 / 53ReadPunctuated, disfluencies.DataoceanAI Conversational USAmerican8.82NAConversational, spontaneousPunctuated, disfluencies.DataoceanAI Conversational GBBritish5.96NAConversational, spontaneousPunctuated, disfluencies. Below are sample audio showing the variety of content (scripted, conversational, acronyms, disfluencies, proper nouns).

While private datasets may sound contrary to the spirit of openness, we believe that incorporating such datasets willincrease the trustworthinessof the Open ASR Leaderboard, as they are less likely to be exploited for benchmaxxing, whether by model developers who explicitly use the public test sets or who try to find training data that closely resembles a particular dataset to boost their score in the macroaverage.

With these datasets, we can also provide targeted metrics to highlight gaps and biases between controlled and often saturated settings (scripted, American accent) and more nuanced conditions (conversational and non-American accents). Below is a screenshot of the new “Private data” tab.

Below is how each column is computed.

• “Average WER” computes a macroaverage of the data provider averages, so that they are weighted equally.
• “Avg Scripted” performs a macroaverage of all scripted datasets.
• “Avg Conversational” performs a macroaverage of all conversational datasets.
• “Avg US” performs a macroaverage of all datasets with American accents.
• “Avg non-US” performs a macroaverage of all datasets with non-American accents.

We intentionally do not provide a score on each split, to avoid model developers from boosting their score with a specific data provider or accent.

https://huggingface.co/blog/open-asr-leaderboard-private-data#how-can-i-evaluate-my-model-on-this-dataHow can I evaluate my model on this data?

Get your model on the Open ASR Leaderboard, and we’ll run the evaluation! As before, the process for adding a model to the leaderboard takes place on the Open ASR LeaderboardGitHub:

1. Open a pull request, and amodel checklistwill appear. As before, you should report your results on the public datasets.
2. We will verify the results on the public sets and compute the metrics on the private ones.
3. Confirm the results we’ve obtained.

While you wait for your model to be added to the Open ASR Leaderboard, you can self-report your metrics on the public sets by adding a YAML file likethisto your model card. Your model will then appear on an (unverified) leaderboard that appears on thedataset page(see screenshot below). More on this approach to decentralized evaluation can be readhere.

https://huggingface.co/blog/open-asr-leaderboard-private-data#do-models-trained-on-the-data-providers-have-an-advantageDo models trained on the data providers have an advantage?

They could. We’ve asked Appen and DataoceanAI to not provide this data to their clients. But even if they do not provide this exact data, data from a similar distribution could still help the model on the corresponding evaluation set (similar to benchmaxxing by optimizing for a challenging task from the public sets). To this end, having multiple data providers balances out the advantage a model may get from having used data from one of the providers. And we are open to more data providers and eval sets for the “Private data” tab!

Moreover, to ensure that the private sets do not affect the model ranking, we’ve defaulted the Average WERto not include the Private sets in its macroaverage.

In the screenshot below, you can see that “Private data” is toggled off. This means that the macroaverage across datasets does not include it.

Simply toggle on “Private data” splits to include them in the macroaverage.

The “Rank Δ” column shows how the ordering changes relative to the default macroaverage configuration. Including or excluding public datasets also changes the macroaverage, allowing users to tailor the evaluation to the use cases and data distributions most relevant to their application.

https://huggingface.co/blog/open-asr-leaderboard-private-data#whats-nextWhat’s next?

We’re excited to hear the community’s feedback on how the new track and dataset toggling features help users identify the model(s) that best fit their application(s). We’re also looking into evaluations that better reflect real-world noisy conditions, and you can expect some news on that 😉

While preparing the private evaluation sets, we took extra care to ensure consistent audio and transcript quality across datasets, including developing tooling to identify challenging cases such as low signal-to-noise conditions or transcript mismatches, since these factors can meaningfully affect WER. More on that in a future post!