accent sherlock NEW

accent sherlock is an AI accent guesser and detector that analyzes a short spoken passage (about 20 seconds) to estimate a speaker’s likely regional or country accent.Users tap the microphone, read a brief scripted passage, and receive a "sounds-like" region or country estimate with approximate confidence rather than definitive identification.

The tool stores anonymized voice clips to support open research and model improvement and is trained on public corpora (Speech Accent Archive, L2-Arctic, EDAcc) with a WavLM encoder.Useful for language learners and pronunciation practice, teachers and educators for classroom demonstrations, and linguists or speech researchers for accent comparison and dataset exploration.

Outputs include probable country/region labels and visual cues to compare accents across speakers; accuracy is approximate and not intended for identifying individuals.Keywords.accent detection, accent recognition, AI accent analysis, pronunciation practice, speech research.

Key Features

• Analyzes ~20-second spoken passages to estimate a speaker's regional or country accent
• Provides 'sounds-like' region/country estimates with approximate confidence scores
• Stores anonymized voice clips to support research and model improvement
• Trained on public corpora (Speech Accent Archive, L2-Arctic, EDAcc) using a WavLM encoder
• Outputs probable country/region labels and visual cues for comparing accents across speakers

Use Cases

• Help language learners refine pronunciation by uploading ~20-second speech samples to Accent Sherlock to receive regional/country accent estimates with confidence scores, compare visually against native-speaker models, and focus practice on specific phonetic targets without extra tools
• Enable teachers to create personalized pronunciation lesson plans by analyzing students' short recordings, tracking accent shifts over time with visual comparisons and confidence metrics, and storing anonymized clips for progress reporting and class-wide benchmarking
• Support speech researchers by building anonymized accent corpora from short recordings, exploring accent distributions with the classifier's confidence outputs, and generating visual comparisons across speaker groups for dialectology and pronunciation studies

Who is it for?

• Language learners
• Teachers
• Linguists
• Speech researchers
• Voice analysts