Methods and equipment for detecting conversational content with fraudulent intent, and computer-readable recordings.

Abstract

The invention relates to a device and method for detecting fraudulent conversational content in telephone calls. The method is implemented entirely on-device using machine learning techniques to analyze call content, assign intent and entity labels to each conversational phrase by comparing it to a database stored on the device, then calculate a risk score and stratify user alerts. This database can be updated by adding new sample phrases without retraining the model.

Images