Home/Events/Hugging Face and Voice Arena Launch Open ASR Evaluation for Hindi and Indian English

Hugging Face and Voice Arena Launch Open ASR Evaluation for Hindi and Indian English

Confirmed

Confidence

90%

Impact: 80%

Updated 1h ago

Consensus Brief

Hugging Face and Voice Arena have introduced the Open ASR Leaderboard's first evaluation sets for Hindi and Indian English, aimed at improving automatic speech recognition (ASR) for these languages. The new datasets, Monsoon en-IN and Monsoon hi-IN, include diverse speaker attributes and are designed to address biases in ASR performance across different demographics.

Sourced from

Hugging FaceT1 Hugging FaceT1+1 more

Primary: Hugging Face

What Changed Since Last Update

1h ago

New official source added: Hugging Face published an update on Fri, 28 Au ("The Open ASR Leaderboard Adds Its First Global South Language").

Claim Ledger

4 claims tracked across sources

Confirmed Fact

Hindi is spoken by more than half a billion people.

Confirmed Fact

The Monsoon dataset was built to vary along nine axes including geography, age, gender, and vocabulary.

Confirmed Fact

The dataset comprises 4,888 speakers with 12 speaker attributes recorded for each.

Independent Finding

The ASR error rates are not evenly distributed across different demographics, with commercial systems performing worse for Black speakers compared to white speakers.

Role-Based Impact Analysis

Source Timeline

2 sources corroborating

The Open ASR Leaderboard Adds Its First Global South Language

Hugging Face·22h ago

The Open ASR Leaderboard Adds Its First Global South Language

Hugging Face·22h ago