Abstract: A method for determining a class to which an input image belongs in an inference process, includes: storing a frequent feature for each class in a frequent feature database; inputting an image as the input image; extracting which class the input image belongs to; extracting a plurality of features that appear in the inference process; extracting one of the features that satisfies a predetermined condition as a representative feature; reading out the frequent feature corresponding to an extracted class from the frequent feature database; extracting one or a plurality of ground features based on the frequent feature and the representative feature; storing a concept data representing each feature in an annotation database; reading out the concept data corresponding to the one or the plurality of ground features from the annotation database; generating explanation information; and outputting the ground feature and the explanatory information together with the extracted class.

Abstract: Systems and methods are described herein for disambiguating a voice search query that contains a command keyword by determining whether the user spoke a quotation from a content item and whether the user mimicked or approximated the way the quotation is spoken in the content item. The voice search query is transcribed into a string, and an audio signature of the voice search query is identified. Metadata of a quotation matching the string is retrieved from a database that includes audio signature information for the string as spoken within the content item. The audio signature of the voice search query is compared with the audio signature information in the metadata to determine whether the audio signature matches the audio signature information in the quotation metadata. If a match is detected, then a search result comprising an identifier of the content item from which the quotation comes is generated.

Abstract: Efficient assignment of bit numbers is performed even under a low bit rate condition. A quantizer 12 obtains a quantized spectral sequence from a frequency spectral sequence. An integer transformer 13 obtains a unified quantized spectral sequence by obtaining, by a bijective transformation, a transformed integer for each of the sets, each being made up of integer values, obtained from the quantized spectral sequence. An integer encoder 15 obtains an integer code by encoding the unified quantized spectral sequence using a bit assignment sequence. An object-to-be-encoded estimator 18 obtains an estimated unified spectral sequence from the frequency spectral sequence by a transformation which is performed by the integer transformer 13 or a transformation that approximates the magnitude relationship between values before and after the above transformation. A bit assigner 14 obtains a bit assignment sequence and a bit assignment code from the estimated unified spectral sequence.

Abstract: A label assignment model generation device extracts a plurality of feature amounts for a word from a document as an extraction source, and generates, based on an appearance frequency of each of the extracted feature amounts, a label assignment model that is a machine learning model and assigns a label to a word included in a document as an assignment target. The label assignment model generation device adjusts a degree of influence of the feature amount on the label assignment model based on a deviation of the appearance frequency of each of the plurality of feature amounts. The label assignment model generation device extracts a plurality of feature amounts for a word from a remaining document excluding a predetermined document from a plurality of documents as extraction sources, and generates the label assignment model based on the appearance frequency of the extracted feature amounts.

Abstract: Techniques for joining a device of a third user to a communication between a device of a first user and a device of a second user are described herein. For instance, two or more users may utilize respective computing devices to engage in a telephone call, a video call, an instant-messaging session, or any other type of communication in which the users communicate with each other audibly and/or visually. In some instances, a first user of the two users may issue a voice command requesting to join a device of a third user to the communication. One or more computing devices may recognize this voice command and may attempt to join a device of a third user to the communication.

Abstract: A text search query including one or more words may be received. An ASR index created for an audio recording may be searched over using the query to produce ASR search results including words, each word associated with a confidence score. For each of the words in the ASR search results associated with a confidence score below a threshold (and in some cases having one or more preceding words in the ASR index and one or more subsequent words in the ASR index), a phonetic representation of the audio recording may be searched for the word having the confidence score below the threshold, where it occurs in the audio recording, possibly after the one or more preceding words and in the audio recording before the one or more subsequent words, to produce phonetic search results. Search results may be returned include ASR and phonetic results.

Abstract: A text search query including one or more words may be received. An ASR index created for an audio recording may be searched over using the query to produce ASR search results including words, each word associated with a confidence score. For each of the words in the ASR search results associated with a confidence score below a threshold (and in some cases having one or more preceding words in the ASR index and one or more subsequent words in the ASR index), a phonetic representation of the audio recording may be searched for the word having the confidence score below the threshold, where it occurs in the audio recording, possibly after the one or more preceding words and in the audio recording before the one or more subsequent words, to produce phonetic search results. Search results may be returned include ASR and phonetic results.

Abstract: A system, method, and wireless earpieces for implementing a virtual assistant. A request is received from a user to be implemented by wireless earpieces. A virtual assistant is executed on the wireless earpieces. An action is implemented to fulfill the request utilizing the virtual assistant. The wireless earpieces may be a set of wireless earpieces and the virtual assistant may be implemented independently by the wireless earpieces.

Abstract: Methods and systems are presented for generating a machine learning model using enhanced gradient boosting techniques. The machine learning model is configured to receive inputs corresponding to a set of features and to produce an output based on the inputs. The machine learning model includes multiple layers, wherein each layer includes multiple models. To generate the machine learning model, multiple models are built and trained in parallel for each layer of the machine learning model. The multiple models use different subsets of features to produce corresponding output values. After a layer in built and trained, a collective error may be determined for the layer based on the output values from the different models in the layer. An additional layer of models may be added to the machine learning model to reduce the collective error of a previous layer.

Abstract: There are provided methods and apparatus for performing modified cosine transformation (MDCT) with an analysis/synthesis windowing function, using an analysis windowing function having a meandering portion which passes a linear function in correspondence of at least four points.

Abstract: Techniques are described herein for enabling autonomous agents to generate conclusive answers. An example of a conclusive answer is text that addresses concerns of a user who is interacting with an autonomous agent. For example, an autonomous agent interacts with a user device, answering user utterances, for example questions or concerns. Based on the interactions, the autonomous agent determines that a conclusive answer is appropriate. The autonomous agent formulates the conclusive answer, which addresses multiple user utterances. The conclusive answer provided to the user device.

Abstract: Embodiments of the present disclosure provide an interactive method of a robot, an interactive device of a robot and a device. The method includes: obtaining voice information input by an interactive object, and performing semantic recognition on the voice information to obtain a conversation intention; obtaining feedback information corresponding to the conversation intention based on a conversation scenario knowledge base pre-configured by a simulated user; and converting the feedback information into a voice of the simulated user, and playing the voice to the interactive object.

Abstract: The present invention allows for the capture and sentiment analysis of text the customer inputs into a chat, but never actually sends to the customer service representative (ghost text). The system captures this ghost text with a ghost capture system (GCS) software module. The GCS module analyzes the ghost text to generate metadata. The ghost text and metadata are used by a sentiment analysis engine to apply appropriate sentiment to the ghost text. The sentiment and ghost text are routed to a customer service representative (CSR). This provides the customer service agent with additional detail and information about a customer's emotions during a text chat conversation, allowing the CSR to determine a court of interaction not only based on the customer's response, but also based on the ghost text and the sentiment from the ghost text.

Abstract: A method and apparatus for obtaining word vectors based on a language model, a device and a storage medium are disclosed, which relates to the field of natural language processing technologies in artificial intelligence. An implementation includes inputting each of at least two first sample text language materials into the language model, and outputting a context vector of a first word mask in each first sample text language material via the language model; determining the word vector corresponding to each first word mask based on a first word vector parameter matrix, a second word vector parameter matrix and a fully connected matrix respectively; and training the language model and the fully connected matrix based on the word vectors corresponding to the first word masks in the at least two first sample text language materials, so as to obtain the word vectors.

Abstract: One embodiment of the invention provides a method for terminology ranking for use in natural language processing. The method comprises receiving a list of terms extracted from a corpus, where the list comprises a ranking of the terms based on frequencies of the terms across the corpus. The method further comprises accessing a domain ontology associated with the corpus, and re-ranking the list based on the domain ontology. The resulting re-ranked list comprises a different ranking of the terms based on relevance of the terms using knowledge from the domain ontology. The method further comprises generating clusters of terms via a trained model adapted to the corpus, and boosting a rank of at least one term of the re-ranked list based on the clusters to increase a relevance of the at least one term using knowledge from the trained model.

Abstract: Systems and methods are disclosed for enabling verbal interaction with an NLUI application without relying on express wake terms. The NLUI application receives an audio input comprising a plurality of terms. In response to determining that none of the terms is an express wake term pre-programmed into the NLUI application, the NLUI application determines a topic for the plurality of terms. The NLUI application then determines whether the topic is within a plurality of topics for which a response should be generated. If the determined topic of the audio input is within the plurality of topics, the NLUI application generates a response to the audio input.

Abstract: A system and method to process audio data received over the ARC or eARC interface of HDMI from audio sources are provided. A media device may receive compressed audio data in a number of data formats. The media device may convert between the audio formats provided by the audio sources and the audio formats supported by audio playback devices. The media device may inspect frames of audio data to determine if the frames are to be decoded. The frame may be decoded and subsequently encoded into the data formats supported by the audio playback devices. To reduce latency, the media device may enable a pass-through mode to bypass the decoding of the frames to allow the frames to be decoded at the audio playback devices. A bi-directional loopback application may route audio data received over the ARC or eARC interface from the audio sources to the audio playback devices.

Abstract: The present disclosure discloses a method and system for generating mixed voice data, and belongs to the technical field of voice recognition. In the method for generating mixed voice data according to the present disclosure, a pure voice and noise are collected first, normalization processing is performed on the collected voice data, randomization processing is performed on processed data, then GAIN processing is performed on the data, and finally filter processing is performed to obtain mixed voice data. The system for generating mixed voice data according to the present disclosure includes a collecting unit, a calculating unit, and a storage unit, the collecting unit being electrically connected to the calculating unit, and the calculating unit being connected to the storage unit through a data transmitting unit. The present disclosure provides the method and the system to meet the data requirement of deep learning.

Abstract: Analyzing communications to determine appropriateness and provide recommendations based on prior communications. Prior communications may create a baseline tone based on the sentiment, words, or phrases used in previous communications between the user and the intended recipient and/or other user recipient pairs. Machine learning techniques may be used to continuously update a baseline tone.

Abstract: A method for fitness assistance to a user in a conversation session, comprising: receiving, in the conversation session, at least one message, wherein the conversation session is between the user and an electronic conversational agent (1210); updating, based on the conversation session and one or more factors associated with the at least one message, a fitness intention associated with the user (1220); extracting, based on one or more factors associated with the at least one message, fitness information associated with the user (1230); obtaining at least one response, through applying a set of rules on the fitness information and the fitness intention (1240); generating, according to the at least one response, a response message including a fitness suggestion to the user (1250); and providing the response message including the fitness suggestion to the user (1260).