Abstract: In one respect, there is provided a system for training a neural network adapted for classifying one or more scripts. The system may include at least one processor and at least one memory. The memory may include program code that provides operations when executed by the at least one memory. The operations may include: extracting, from an icon associated with a file, one or more features; assigning, based at least on the one or more features, the icon to one of a plurality of clusters; and generating, based at least on the cluster to which the icon is assigned, a classification for the file associated with the icon. Related methods and articles of manufacture, including computer program products, are also provided.

Abstract: Centroids are used for improving machine learning classification and information retrieval. A plurality of files are classified as malicious or not malicious based on a function dividing a coordinate space into at least a first portion and a second portion such that the first portion includes a first subset of the plurality of files classified as malicious. One or more first geometric regions are defined in the first portion that classify files from the first subset as not malicious. A file is determined to be malicious based on whether the file is located within the one or more first geometric regions.

Abstract: Systems, methods, and articles of manufacture, including computer program products, are provided for classification systems and methods using modeling. In some example embodiments, there is provided a system that includes at least one processor and at least one memory including program code which when executed by the at least one memory provides operations. The operations can include generating a representation of a sequence of sections of a file and/or determining, from a model including conditional probabilities, a probability for each transition between at least two sequential sections in the representation. The operations can further include classifying the file based on the probabilities for each transition.

Abstract: In one respect, there is provided a system for classifying an instruction sequence with a machine learning model. The system may include at least one processor and at least one memory. The memory may include program code that provides operations when executed by the at least one processor. The operations may include: processing an instruction sequence with a trained machine learning model configured to detect one or more interdependencies amongst a plurality of tokens in the instruction sequence and determine a classification for the instruction sequence based on the one or more interdependencies amongst the plurality of tokens; and providing, as an output, the classification of the instruction sequence. Related methods and articles of manufacture, including computer program products, are also provided.

Abstract: In one respect, there is provided a system for training a neural network adapted for classifying one or more instruction sequences. The system may include at least one processor and at least one memory. The memory may include program code which when executed by the at least one processor provides operations including: training, based at least on training data, a machine learning model to detect one or more predetermined interdependencies amongst a plurality of tokens in the training data; and providing the trained machine learning model to enable classification of one or more instruction sequences. Related methods and articles of manufacture, including computer program products, are also provided.

Abstract: In one respect, there is provided a system for training a machine learning model to detect malicious container files. The system may include at least one processor and at least one memory. The at least one memory may include program code that provides operations when executed by the at least one processor. The operations may include: training, based on a training data, a machine learning model to enable the machine learning model to determine whether at least one container file includes at least one file rendering the at least one container file malicious; and providing the trained machine learning model to enable the determination of whether the at least one container file includes at least one file rendering the at least one container file malicious. Related methods and articles of manufacture, including computer program products, are also disclosed.

Abstract: In one respect, there is provided a system for training a machine learning model to detect malicious container files. The system may include at least one processor and at least one memory. The memory may include program code which when executed by the at least one processor provides operations including: processing a container file with a trained machine learning model, wherein the trained machine learning is trained to determine a classification for the container file indicative of whether the container file includes at least one file rendering the container file malicious; and providing, as an output by the trained machine learning model, an indication of whether the container file includes the at least one file rendering the container file malicious. Related methods and articles of manufacture, including computer program products, are also disclosed.

Abstract: Techniques for converting text to speech having emotional content. In an aspect, an emotionally neutral acoustic trajectory is predicted for a script using a neutral model, and an emotion-specific acoustic trajectory adjustment is independently predicted using an emotion-specific model. The neutral trajectory and emotion-specific adjustments are combined to generate a transformed speech output having emotional content. In another aspect, state parameters of a statistical parametric model for neutral voice are transformed by emotion-specific factors that vary across contexts and states. The emotion-dependent adjustment factors may be clustered and stored using an emotion-specific decision tree or other clustering scheme distinct from a decision tree used for the neutral voice model.

Abstract: Multi-voice font interpolation is provided. A multi-voice font interpolation engine allows the production of computer generated speech with a wide variety of speaker characteristics and/or prosody by interpolating speaker characteristics and prosody from existing fonts. Using prediction models from multiple voice fonts, the multi-voice font interpolation engine predicts values for the parameters that influence speaker characteristics and/or prosody for the phoneme sequence obtained from the text to spoken. For each parameter, additional parameter values are generated by a weighted interpolation from the predicted values. Modifying an existing voice font with the interpolated parameters changes the style and/or emotion of the speech while retaining the base sound qualities of the original voice.

Abstract: Multi-voice font interpolation is provided. A multi-voice font interpolation engine allows the production of computer generated speech with a wide variety of speaker characteristics and/or prosody by interpolating speaker characteristics and prosody from existing fonts. Using prediction models from multiple voice fonts, the multi-voice font interpolation engine predicts values for the parameters that influence speaker characteristics and/or prosody for the phoneme sequence obtained from the text to spoken. For each parameter, additional parameter values are generated by a weighted interpolation from the predicted values. Modifying an existing voice font with the interpolated parameters changes the style and/or emotion of the speech while retaining the base sound qualities of the original voice.

Abstract: Techniques for converting text to speech having emotional content. In an aspect, an emotionally neutral acoustic trajectory is predicted for a script using a neutral model, and an emotion-specific acoustic trajectory adjustment is independently predicted using an emotion-specific model. The neutral trajectory and emotion-specific adjustments are combined to generate a transformed speech output having emotional content. In another aspect, state parameters of a statistical parametric model for neutral voice are transformed by emotion-specific factors that vary across contexts and states. The emotion-dependent adjustment factors may be clustered and stored using an emotion-specific decision tree or other clustering scheme distinct from a decision tree used for the neutral voice model.

Abstract: Multi-voice font interpolation is provided. A multi-voice font interpolation engine allows the production of computer generated speech with a wide variety of speaker characteristics and/or prosody by interpolating speaker characteristics and prosody from existing fonts. Using prediction models from multiple voice fonts, the multi-voice font interpolation engine predicts values for the parameters that influence speaker characteristics and/or prosody for the phoneme sequence obtained from the text to spoken. For each parameter, additional parameter values are generated by a weighted interpolation from the predicted values. Modifying an existing voice font with the interpolated parameters changes the style and/or emotion of the speech while retaining the base sound qualities of the original voice.

Abstract: Dynamic features are utilized with CRFs to handle long-distance dependencies of output labels. The dynamic features present a probability distribution involved in explicit distance from/to a special output label that is pre-defined according to each application scenario. Besides the number of units in the segment (from the previous special output label to the current unit), the dynamic features may also include the sum of any basic features of units in the segment. Since the added dynamic features are involved in the distance from the previous specific label, the searching lattice associated with Viterbi searching is expanded to distinguish the nodes with various distances. The dynamic features may be used in a variety of different applications, such as Natural Language Processing, Text-To-Speech and Automatic Speech Recognition. For example, the dynamic features may be used to assist in prosodic break and pause prediction.

Abstract: Dynamic features are utilized with CRFs to handle long-distance dependencies of output labels. The dynamic features present a probability distribution involved in explicit distance from/to a special output label that is pre-defined according to each application scenario. Besides the number of units in the segment (from the previous special output label to the current unit), the dynamic features may also include the sum of any basic features of units in the segment. Since the added dynamic features are involved in the distance from the previous specific label, the searching lattice associated with Viterbi searching is expanded to distinguish the nodes with various distances. The dynamic features may be used in a variety of different applications, such as Natural Language Processing, Text-To-Speech and Automatic Speech Recognition. For example, the dynamic features may be used to assist in prosodic break and pause prediction.

Abstract: According to one embodiment, an apparatus for fusing voiced phoneme units in Text-To-Speech, includes a reference unit selection module configured to select a reference unit from the plurality of units based on pitch cycle information of the each unit and the number of pitch cycles of the target segment. The apparatus includes a template creation module configured to create a template based on the reference unit selected by the reference unit selection module and the number of pitch cycles of the target segment, wherein the number of pitch cycles of the template is same with that of pitch cycles of the target segment. The apparatus includes a pitch cycle alignment module configured to align pitch cycles of each unit of the plurality of units except the reference unit with pitch cycles of the template by using a dynamic programming algorithm.

Abstract: According to one embodiment, an apparatus for synthesizing a speech, comprises an inputting unit configured to input a text sentence, a text analysis unit configured to analyze the text sentence so as to extract linguistic information, a parameter generation unit configured to generate a speech parameter by using the linguistic information and a pre-trained statistical parameter model, an embedding unit configured to embed information into the speech parameter, and a speech synthesis unit configured to synthesize the speech parameter with the information embedded by the embedding unit into a speech with the information.

Abstract: The present invention provides a method and apparatus for enrollment and evaluation of speaker authentication. The method for enrollment of speaker authentication, comprising: generating a plurality of acoustic feature vector sequences respectively based on a plurality of utterances of the same content spoken by a speaker; generating a reference template from said plurality of acoustic feature vector sequences; generating a corresponding pseudo-impostor feature vector sequence for each of said plurality of acoustic feature vector sequences based on a code book that includes a plurality of codes and their corresponding feature vectors; and selecting an optimal acoustic feature subset based on said plurality of acoustic feature vector sequences, said reference template and said plurality of pseudo-impostor feature vector sequences.

Abstract: The present invention provides a method and apparatus for enrollment and verification of speaker authentication. The method for enrollment of speaker authentication, comprising: extracting an acoustic feature vector sequence from an enrollment utterance of a speaker; and generating a speaker template using the acoustic feature vector sequence; wherein said step of extracting an acoustic feature vector sequence comprises: generating a filter-bank for the enrollment utterance of the speaker for filtering locations and energies of formants in the spectrum of the enrollment utterance based on the enrollment utterance; filtering the spectrum of the enrollment utterance by the generated filter-bank; and generating the acoustic feature vector sequence from the filtered enrollment utterance.

Abstract: The present invention provides a method and apparatus for verification of speaker authentication. A method for verification of speaker authentication, comprising: inputting an utterance containing a password that is spoken by a speaker; extracting an acoustic feature vector sequence from said inputted utterance; DTW-matching said extracted acoustic feature vector sequence and a speaker template enrolled by an enrolled speaker; calculating each of a plurality of local distances between said DTW-matched acoustic feature vector sequence and said speaker template; nonlinear-transforming said each local distance calculated to give more weights on small local distances; calculating a DTW-matching score based on said plurality of local distances nonlinear-transformed; and comparing said matching score with a predefined discriminating threshold to determine whether said inputted utterance is an utterance containing a password spoken by the enrolled speaker.

Abstract: The present invention provides a method and apparatus for enrollment and evaluation of speaker authentication. The method for enrollment of speaker authentication, comprising: generating a plurality of acoustic feature vector sequences respectively based on a plurality of utterances of the same content spoken by a speaker; generating a reference template from said plurality of acoustic feature vector sequences; generating a corresponding pseudo-impostor feature vector sequence for each of said plurality of acoustic feature vector sequences based on a code book that includes a plurality of codes and their corresponding feature vectors; and selecting an optimal acoustic feature subset based on said plurality of acoustic feature vector sequences, said reference template and said plurality of pseudo-impostor feature vector sequences.