aiOla Unveils Whisper-NER AI Model That Masks Sensitive Information

aiOla, a speech artificial intelligence technology provider, has launched the Whisper-NER AI model for automatic speech recognition with built-in named entity recognition and the automatic detection and masking of sensitive information, such as names, phone numbers, and addresses, during the transcription of audio.

With aiOla's Whisper-NER model, users input an audio file along with the names of entities they want to be identified. The model then transcribes the audio while simultaneously masking the entities so that sensitive personal information isn't stored, even temporarily.

Whisper-NER, built on OpenAI's Whisper, was trained using a synthetic dataset that combines large amounts of synthetic speech with open NER text datasets. This approach allowed the model to learn both transcription and entity recognition in parallel.;

"Whisper-NER is the first open-source AI model that not only detects and masks sensitive data but can ensure that sensitive information is never generated in the first place," said Gill Hetz, vice president of research at aiOla, in a statement. "Our approach allows us to structure unstructured transcriptions without relying on generic models like ChatGPT and without requiring separate ASR and NER processes, which can negatively impact privacy and security. Whisper-NER operates as a zero-shot solution, combining both tasks in one elegant step, significantly improving efficiency while maintaining supreme accuracy. This innovation not only boosts performance but also strengthens ethical AI practices, fostering trust in the secure and responsible collection of speech data."