Abstract: An anti-communications fraud apparatus includes: an analysis unit that analyzes a voiceprint of a communication voice of a calling party; a determination unit that acquires, from a database in which the voiceprint and a degree of fraud risk are stored in association with each other, the degree of fraud risk corresponding to the voiceprint of the calling party, and determines whether the degree of fraud risk exceeds a predetermined threshold; and a notification unit that notifies that the calling party is dangerous when the degree of fraud risk exceeds the threshold.

Abstract: A system is disclosed for identifying a user based on the classification of user characteristic data. An identity verification system receives a request from a requesting target user for access to an operational context and characteristic data describing actions of the requesting target user. The identity verification system inputs the characteristic data to an identity confidence model to determine an identity confidence value describes a likelihood that an identity of the requesting target user matches an authenticating identity and determines a false match rate and false non-match rate, which represent a performance of the identity confidence model. The identity verification system determines a match probability for the requesting target user by adjusting the identity confidence value based on the determined false match rate and false non-match rate and grants the requesting target user access to the operational context if match probability is greater than the operational security threshold.

Abstract: Technology for determining an insurance fraud risk associated with a user comprises receiving, at an enterprise, a call or a chat from a user device associated with the user. Data associated with the call or the chat is analyzed to determine if at least one factor indicating fraud is present. A weighted level of possible fraud associated with the at least one factor is determined. The weighted level of possible fraud is compared to at least one weight threshold. The user is identified as an increased fraud risk based on the weighted level of possible fraud meeting or exceeding the at least one weight threshold.

Abstract: A computer-implemented method includes obtaining, using a hardware processor, training data including utterances of speakers and performing tasks to train a machine learning model that converts an utterance into a feature vector, each task using one subset of multiple subsets of training data. The subsets of training data include a first subset of training data including utterances of a first number of speakers and at least one second subset of training data. Each second subset of training data includes utterances of a number of speakers that is less than the first number of speakers.

Abstract: A text-independent speaker verification system utilizes mel frequency cepstral coefficients analysis in the feature extraction blocks, template modeling with vector quantization in the pattern matching blocks, an adaptive threshold and an adaptive decision verdict and is implemented in a stand-alone device using less powerful microprocessors and smaller data storage devices than used by comparable systems of the prior art.

Abstract: The invention relates to a method, a computer program product and a computer system for structuring an unstructured text by making use of statistical models trained on annotated training data. Each section of text in which the text is segmented is further assigned to a topic which is associated to a set of labels. The statistical models for the segmentation of the text and for the assignment of a topic and its associated labels to a section of text explicitly accounts for: correlations between a section of text and a topic, a topic transition between sections, a topic position within the document and a (topic-dependent) section length. Hence structural information of the training data is exploited in order to perform segmentation and annotation of unknown text.