Martin Sundberg and Mikael Ohlsson have, as part of their studies at Uppsala University, published a master’s thesis in collaboration with Scila. Their thesis “Automatic Voice Trading Surveillance: Achieving Speech and Named Entity Recognition in Voice Trade Calls Using Language Model Interpolation and Named Entity Abstraction” explores the integration of domain-specific models alongside a generic speech recognition model to achieve accurate transcription and extraction of named entities from voice trading recordings. It demonstrates the proof-of-concept model’s flexibility in tailoring configurations to meet specific business needs, ensuring compliance with market regulations.
