Abstract: It is possible to achieve high analysis accuracy on the whole irrespective of a coverage ratio of a morpheme dictionary. A morpheme prediction section 38 outputs a morpheme label string as a morphological analysis result of input text by inputting a resultant vector generated for each character by a vector combination section 36 into a series predictor and sequentially performing calculation. At this point, there is used the series predictor which is subjected to learning in advance such that a morpheme label string obtained by inputting a resultant vector in which a character vector and a morpheme information vector are combined and which is generated without being combined with the morpheme information vector with a specific probability into the series predictor and sequentially performing the calculation coincides with a morpheme label string of learning data.

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: Method and system for determining a conversation system from a multi-conversation system using Artificial Intelligence (AI) is provided. The method includes receiving a user query associated with a domain and creating a hierarchical tree comprising a root node and a child node using a first pre-trained machine learning model. The method further includes traversing the hierarchical tree for a path between root node and one leaf child node to identify a topic hierarchy. The path is associated with a confidence score corresponding to mapping between user query and match data of nodes in the path. The method further includes determining a conversation system from the multi-conversation system for outputting data to answer the user query corresponding to one leaf child node of one path with a highest confidence score.

Abstract: Various implementations disclosed herein include devices, systems, and methods for providing domain-specific responses. In various implementations, a device includes a non-transitory memory and a processor coupled with the non-transitory memory. In some implementations, a method includes determining a user interest value indicative of a pharmaceutical article that the user is interested in learning about. In some implementations, the pharmaceutical article is associated with a set of media content items that provide information regarding the pharmaceutical article. In some implementations, the method includes setting a configuration parameter of a virtual agent to a configuration value that is based on the pharmaceutical article indicated by the user interest value.

Abstract: In an embodiment, communication data including an interaction between at least two individuals in an organization is received via at least two communication platforms. Content of the interaction is extracted to produce extracted content. A frequency of interaction and a type of interaction between the at least two individuals is determined based on the extracted content. A map representative of social interaction within the organization is generated. The map includes an indication of the frequency of interaction and the type of interaction between the at least two individuals. A new frequency of the interaction and a new type of interaction for at least a first individual of the at least two individuals with another individual is determined based on the map. A signal representing the new frequency of interaction and the new type of interaction with the other individual in the organization is provided to a compute device.

Abstract: Systems and methods are directed to improving speech signal recognition. A method includes obtaining a spatial audio signal; separating a continuous speech signal and a corresponding directivity flag signal for a sound source direction from the spatial audio signal; and combining the continuous speech signal with the corresponding directivity flag signal for the sound source direction to generate a speech activation detection signal for the sound source direction. Because the speech activation detection signal of the sound source direction is obtained by combining the continuous speech signal with the directivity flag signal of the sound source direction, the speech activation detection signal has directivity, reducing the interference from continuous speech signal in other sound source directions.

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: Described herein is a system for sentiment detection in audio data. The system processes audio frame level features of input audio data using a machine learning algorithm to classify the input audio data into a particular sentiment category. The machine learning algorithm may be a neural network trained using an encoder-decoder method. The training of the machine learning algorithm may include normalization techniques to avoid potential bias in the training data that may occur when the training data is annotated for a perceived sentiment of the speaker.

Abstract: Systems, devices, and methods discussed herein provide improved autonomous agent applications that are configured to provide explanations in response to user-submitted questions. A user query may be received and a classification (e.g., a general question, a specific question) for the user query may be determined based at least in part on a predefined rule set (or a classification model). A set of decision features (e.g., particular user data) associated with a decision generated by a machine-learning model may be identified. An explanation chain may be identified from a plurality of explanation chains based at least in part on the user query. The explanation chain may describe a logical chain of explanations associated with a decision making process related to the machine-learning model. A response to the user query may be provided based at least in part on the explanation chain and the set of decision features.

Abstract: A sound processing method obtains note data representative of a note; obtains an audio signal to be processed; specifies, in accordance with the note, an expression sample representative of a sound expression to be imparted to the note and an expression period, of the audio signal, to which the sound expression is to be imparted to the note; and specifies, in accordance with the expression sample and the expression period, a processing parameter relating to an expression imparting processing for imparting the sound expression to a portion corresponding to the expression period in the audio signal. The method then generates a processed audio signal by performing the expression imparting processing in accordance with the expression sample, the expression period, and the processing parameter to the audio signal.

Abstract: Splitting a natural language sentence into primitive phrases retaining relations of terms includes receiving a natural language sentence, building a parse tree from the natural language sentence using a natural language parser, and recursively identifying discourse markers in subtrees of the parse tree, starting with the highest ranking discourse marker in the parse tree, thereby separating each of the respective subtrees at the respective discourse marker using a set of predefined rules until a set of basic subtrees remains. The recursive identification includes looking-ahead for identifying long ranging discourse markers before identifying local discourse markers.

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 method for authoring a conversational bot including: receiving conversation data; receiving seed intent data that comprises seed intents having a seed intent label and sample intent-bearing utterances; using an intent mining algorithm to mine the conversation data to determine new utterances to associate with the seed intent; augmenting the seed intent data to include the mined new utterances associated with the seed intents; and uploading the augmented seed intent data into the conversation bot. The intent mining algorithm may include: identifying intent-bearing utterances; identifying candidate intents; for each of the seed intents, identifying seed intent alternatives from the sample intent-bearing utterances; associating the intent-bearing utterances from the conversation data with the seed intents via determining a degree of semantic similarity between the candidate intents of the intent-bearing utterances and the seed intent alternatives.

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: An electronic device for natural language understanding includes at least one memory and at least one processor coupled to the at least one memory. The at least one processor is configured to process an utterance using a trained model. The at least one processor is also configured to replace a first portion of the utterance with a first token, where the first token represents a semantic role of the first portion of the utterance based on a slot vocabulary. The at least one processor is further configured to determine a slot value in the utterance based on the first token. In addition, the at least one processor is configured to perform a task corresponding to the utterance based on the determined slot value.

Abstract: A computer-implemented, machine learning-based method of converting an unstructured technical report into a structured technical report includes obtaining an unstructured technical report, tokenizing the unstructured technical report into an n-gram array, identifying and filtering non-interesting n-grams from the first n-gram array based on common language usage of the non-interesting n-grams and a determination that the non-interesting n-grams do not appear on a confirmed technical entity database, generating and displaying a technical entity candidate list from the filtered n-gram array, displaying, obtaining, from a pattern matching model and/or a graphical user interface, an indication that a technical entity candidate is a technical entity of interest, appending the technical entity of interest to the confirmed technical entity database, generating and displaying a structured technical report with the confirmed technical entities and corresponding technical entity value parameters, and iterating the proc

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 processing natural language sentences and a computer understanding state machine with computer understanding state machine states are disclosed. The computer understanding state machine receives as inputs the parsed sentence including abstractions of the parsed sentence including weights of the abstractions. These inputs are used to find the best fitting state to process the input sentence. Subsequent sentences also are parsed and these are input to the state machine to allow processing of sequences of sentences.

Abstract: Method and system for extracting valuable words and forming a valuable word net, wherein a server is employed to collect contents of articles from internet sources, EDM texts (email direct marketing), product descriptions and other texts, and extract valuable words by machine learning. Each valuable word performs connection weight based on the number of times the text has been read, the number of times the text has been clicked, the number of times the text has been cited, the correlation of external sites, the conversion of expert knowledge, the probability space, the Shannon entropy, the spatial distribution and other values and their conversion by machine learning; then the linked valuable words are integrated to form a valuable word net. When the valuable words are needed to be used, the valuable words and the valuable word net can be retrieved from the database for subsequent applications.

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.