Audio Transcription

Financial Market Speech Recognition

Artificial Intelligence Powering Voice Work-Flows

picture1
SPEECH RECOGNITION

Scribe’s models are trained on hundreds of hours of financial conversations to achieve greater accuracy for financial English other engines.

SPEAKER SEPARATION

Scribe automatically detects speaker transitions in audio files, improving accuracy and creating a logical, timestamped transcript.

FAST DECODING

Scribe uses proprietary decoding techniques to transcribe audio 10x faster, enabling transcription of hour long audio in minutes allowing fast turnaround.

METADATA EXTRACTION

Scribe uses proprietary natural language processing libraries called Asset Class Interpreters to extract quote & trade data from a conversation.

Example Use Cases

Audio File Transcription

Scribe’s Historical Transcription engine creates time-stamped, searchable, speaker separated transcripts from audio files. GreenKey’s engine can be used via our public API or in an on-premise environment through our Docker container image. You can even dial Scribe into your phone calls and automatically receive transcriptions to your email account.

transcript
Quote and Trade Extraction (NLP)

Scribe’s natural language processing listens to audio streams in real-time to extract quote or trade metadata to populate a ticket or capture for general compliance purposes or more specifically for MiFID II compliance.  Feed the engine audio files or real time streams from a computer or turret connected to our websocket API.

“Deck double you tea eye fifty bid”  becomes  Dec WTI 50b

Insight Engine 

GreenKey’s Insight engine returns sentiment, speakers, summarizations and keywords set by your organization and found in any audio file.

screen-shot-2017-11-07-at-8.21.16-pm

Why Us?

We leverage GPU clusters to train on thousands of hours of actual trader and broker conversations. We offer a very unique financial market language and acoustic model that will return unrivaled accuracy. We also offer natural language processing to run on the transcribed data to extract quotes, trades and other meaningful insights. Further customization, error corrections and continuous training enable customers to build highly accurate, domain specific work flows.

See It Work

Interested in learning more?