unthinkable
Case study · A Nigerian legal technology company (unnamed in the source) · software for courts and lawyers

Nigerian courts get transcripts of hearings and depositions as they are spoken, across the country's accents.

The company wanted judges and lawyers in Nigerian courts to have accurate transcripts of every proceeding, so judgments are delivered on a correct record.

What changed

100,000+words and speech samples in the training set (as reported)
80%accuracy of speech to text (as reported)
35%shorter documentation turnaround (as reported)
What stood in the way

What stood in the way.

What stood in the way
  • Manual transcription of depositions was slow, and missed details led to errors in the record.
  • Accents change from region to region in Nigeria, and transcribers sometimes missed whole sentences.
What changed
  • 100,000+words and speech samples in the training set (as reported)
  • 80%accuracy of speech to text (as reported)
  • 35%shorter documentation turnaround (as reported)
What we built

What we built.

  • A speech recognition engine built on a transfer-learning machine learning model, transcribing streaming audio and video.
  • Automated preparation of thousands of hours of deposition audio and past transcripts into 16 kHz mono audio for training.
  • Pre-training and data augmentation on 100,000+ words and speech samples from speakers with different accents.
  • Inverse text normalisation to make transcripts readable, noise handling, and retraining on new proceedings over time.
  • Transfer learning
  • Automatic speech recognition (ASR)
  • Inverse text normalisation
What it shows

What this work shows about us.

We train on the real setting

The model learned from Nigerian court audio and accents, not a generic voice set.

Readable, not only correct

Post-processing turns raw recognition into transcripts lawyers and judges can read.

Want to build a similar speech recognition engine for transcripts? Tell us what you want to change. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with the right team.

Start your plan ↗
If we built it today

If we built it today, with AI.

What we would add now, moment by moment.

During the hearing

An agent would mark each speaker and flag passages the engine heard with low confidence for the clerk to check during the session. The clerk certifies every transcript.

After the hearing

An agent would turn the certified transcript into a summary of issues and exhibits for the judge and counsel. The judge still decides the case.

Related

Where this work fits.

    Facts and figures from the project records; nothing added.

    Ask Unthinkable

    Want this for your business?

    Tell us what you run or what you're building. Our AI assistant, Ask Unthinkable, helps you define your requirements, then connects you with a senior AI engineer.

    Send us a note instead

    Required

    Thanks. A senior AI engineer will reply to you.
    u
    Ask UnthinkableClarity first, then the right team
    Your conversation

    An AI agent. This prototype uses prepared answers and sends nothing.