Extending sensitive data tagging without reannotating training data

A computer-implemented method comprising:

• hosting a plurality of models at a model endpoint in a machine learning service, each model of the plurality of models corresponding to a different type of sensitive data that is to be redacted from a transcript of content;
• adding a new model to the model endpoint, the new model trained to identify a new type of sensitive data in the transcript of content;
• identifying, by each of the plurality of models and the new model, sensitive data entities in the transcript of content;
• merging inference responses generated by each of the plurality of models and the new model using at least one inference policy, the inference responses from each of the plurality of models corresponding to respective ones of the different types of sensitive data, the inference responses from the new model corresponding to the new type of sensitive data;
• and
• returning a merged inference response identifying the sensitive data entities in the transcript of content.

A system comprising:

• a transcription service implemented by a first one or more electronic devices;
• and
• a machine learning service implemented by a second one or more electronic devices, the machine learning service including instructions that upon execution cause the machine learning service to:
• host a plurality of models at a model endpoint in a machine learning service, each model of the plurality of models corresponding to a different type of sensitive data that is to be redacted from a transcript of content;
• add a new model to the model endpoint, the new model trained to identify a new type of sensitive data in the transcript of content;
• identify, by each of the plurality of models and the new model, sensitive data entities in the transcript of content;
• merge inference responses generated by each of the plurality of models and the new model using at least one inference policy, the inference responses from each of the plurality of models corresponding to respective ones of the different types of sensitive data, the inference responses from the new model corresponding to the new type of sensitive data;
• and
• return a merged inference response identifying the sensitive data entities in the transcript of content.