Abstract: A method for providing automated support services by utilizing artificial intelligence is disclosed. The method includes receiving, via a graphical user interface, a request from a user; parsing, by using syntax analysis, the request; identifying, from the parsed request, a factor by using a model such as a machine learning model, the factor including a request context and a request sentiment; associating the request with a category corresponding to the factor; determining, by using the model, whether the request can be automatically resolved based on the factor and the category; and initiating at least one action based on a result of the determining.

Abstract: A training-stage technique trains a language model for use in an ASR system. The technique includes: obtaining a training corpus that includes a sequence of terms; determining that an original term in the training corpus is not present in a dictionary resource; segmenting the original term into two or more sub-terms using a segmentation resource; determining that the segmentation of the original term into the two or more sub-terms is a valid segmentation, based on two or more validity tests; and training the language model based on the terms that have been identified. A computer-implemented inference-stage technique applies the language model to produce ASR output results. The inference-stage technique merges a sub-term with a preceding term if these two terms are separated by no more than a prescribed interval of time.

Abstract: Techniques are described for performing complaint prioritization using one or more machine learning models for customer communications. For example, a computing system includes a memory and one or more processors in communication with the memory. The one or more processors are configured to receive communication data indicative of a service inquiry from a user device, generate a set of emotion factor values that indicate a measure of particular emotions in the service inquiry, determine, using a machine learning model and based on the set of emotion factor values, an emotional priority score for the service inquiry, and determine a response priority order for the service inquiry based on at least the emotional priority score.

Abstract: In embodiments of the present disclosure, there is provided an approach for fact checking based on semantic graphs. According to embodiments of the present disclosure, after obtaining a text to be fact checked, a plurality of evidence sentences related to the text are retrieved from an evidence database. Then, semantic graphs of the text and the evidence sentences are constructed based on the semantic analysis, and a veracity of a statement in the text can be determined based on the semantic graphs. Embodiments of the present disclosure propose a graph-based reasoning approach for fact checking, and use the constructed semantic graphs to facilitate verification of the truthfulness of the text, thereby improving the accuracy for fact checking.

Abstract: The present invention relates to methods and apparatus for dynamically detecting and/or mitigating threats in communications systems. Exemplary methods and apparatus of the present invention allow for a combination of automated and operator controlled responses to threats. While an operator is provided an opportunity to provide input on how to respond to a threat, after one or more threats of a given type are identified, the system will automatically take corrective action without waiting for operator input and/or in the absence of operator input following notification of a threat.

Abstract: A vehicle remote assistance system includes a plurality of vehicles and a remote assistance apparatus. Each of the plurality of vehicles includes a calling unit that transmits a call to the remote assistance apparatus. The remote assistance apparatus includes an assigning unit that schedules the calls that are transmitted from the plurality of vehicles and assigning the calls to any of a plurality of operators. The calling unit transmits a cancelation of the call to the remote assistance apparatus when a reason for the call is resolved after the call is made. The assigning unit cancels scheduling of the call that corresponds to the cancelation that is transmitted from the vehicle.

Abstract: A hearing aid may include a neural network chip having tiles arranged in an array, each tile including memory, 16-128 multiplier-accumulator circuits (MACs), and routing circuitry. The memory of each tile may be configured to store a portion of elements of a matrix A comprising weights of a recurrent neural network. Each tile may be configured to receive and store elements of an activation vector X, and all tiles in a column of the array may be configured to receive the same elements of X. The plurality of tiles may be configured to perform a matrix-vector multiplication A*X by performing multiply-and-accumulate sub-operations in parallel among the plurality of tiles. The routing circuitry from the tiles in each respective row of tiles may be configured to combine results of the multiply-and-accumulate sub-operations.

Abstract: Various embodiments of the present disclosure performing conversation sentiment monitoring for a conversation data object. In various embodiments, a text block that can be resized is identified within a conversation data object and successive regularized sentiment profile generation iterations are performed until a regularized sentiment score of the block exceeds a regularized sentiment score threshold. A current regularized sentiment profile generation iteration involves determining a regularized sentiment score for the block based on an initial sentiment score, a subjectivity probability value, and, optionally, a stage-wise penalty factor. A determination is then made as to whether the score exceeds the threshold. If so, then a regularized sentiment profile of the conversation data object is updated based on the regularized sentiment score. If not, then the text block is resized and a subsequent regularized sentiment profile generation iteration is performed based on the resized block.

Abstract: A method includes receiving, by a question answering system having a confidence threshold, plural questions from one or more user devices. The method includes processing each one of the questions by: generating an answer to the one of the questions; determining a confidence score of the answer; in response to determining the confidence score is greater than the confidence threshold, increasing the confidence threshold and returning the answer to the user device that generated the one of the questions; and in response to determining the confidence score is less than the confidence threshold, decreasing the confidence threshold and not returning the answer to the user device that generated the one of the questions. The increasing the confidence threshold and the decreasing the confidence threshold are performed such that the question answering system returns answers for the plural questions at a frequency that approximates a pre-defined target answering frequency.

Abstract: Techniques for providing a standardized voice user interface (VUI) that enables voice to voice natural language understanding (NLU) processing are described. The standardized VUI may be added to speechlets to enable a customer to interact with a business via server(s) using NLU processing. For example, a first user may initiate a first voice interaction using NLU processing with the server(s) and the system may initiate a second voice interaction using NLU processing between the server(s) and a second user. This enables a customer to initiate a transaction (e.g., request information, place an order, make a reservation, etc.) with the business using the speechlet. Thus, the business may use the speechlet to offer services without requiring additional infrastructure or complicated programming to implement.

Abstract: The embodiment of the present disclosure provides a method for smart gas call center feedback management and an Internet of things (IoT) system thereof. The method is implemented based on a smart gas management platform. The method includes: receiving a call message of a target customer through a call center, and a content of the call message being related to a gas business; determining a feedback mode by analyzing the call message through the call center; in response to the feedback mode being manual feedback, determining a target operator through the call center to feed back a call of the target customer; and in response to the feedback mode being automatic feedback, determining a feedback content through the call center and sending the feedback content to the target customer.

Abstract: There is a need for more effective and efficient natural language processing (NLP) solutions. This need can be addressed by, for example, solutions for performing NLP-based document prioritization by utilizing joint sentiment-topic (JST) modeling.

Abstract: Techniques that facilitate a dialogue system based on contextual information are provided. In one example, a system includes a contextual information component and a dialogue routing component. The contextual information component determines contextual information associated with a user identity based on a statement related to communication information received by a computing device associated with the user identity. The dialogue routing component generates a path traversal for a dialogue system based on the contextual information to facilitate generation of a response to the statement by the dialogue system.

Abstract: Verbal language analysis is provided to users. The user enrolls or subscribes for verbal language analysis or analytics. The user carries out or conducts a conversation with a third party. An intelligence device associated with the user records the conversation. The intelligence device performs verbal language analysis on the conversation. The verbal language analysis generates individual metrics for verbal factors of energy, word count, inflection, tone (e.g. pitch and sentiment), rate, and/or the like. A verbal intelligence index is determined from the individual metrics using aggregation, averaging, weighted averaging, and/or the like. An interface component generates views to display to the user for review of the conversation to facilitate better verbal performance during current and in future conversations.

Abstract: A computing system identifies an incoming voice call from a user device to an agent device associated with the computing system. The computing system generates a transcription of the incoming voice call using one or more natural language processing techniques. The computing system extracts a problem description from the transcription. The problem description indicates a topic for the incoming voice call. A first machine learning model estimates a situation vector from the problem description. A second machine learning model identifies a pre-existing situation vector that closely matches the estimated situation vector. The computing system retrieves a situation description that corresponds to the identified pre-existing situation vector.

Abstract: Methods for building and training a network of chatbots are provided. Methods may perform data analysis on a first chatbot in the network. The first chatbot may include a first domain of intents. The data analysis may identify and rank the intents in the first domain. The first domain may represent answers to a first domain of user queries. Methods may select a predetermined number of highest-ranking intents from the first domain based on the data analysis. Methods may input the selected intents into a second chatbot in the network. The second chatbot may include a second domain of intents. The second domain may represent answers to a second domain of user queries. Methods may input a hook into the second chatbot. The hook may include a trigger word and may correspond to the first domain. When the hook is triggered, the second chatbot may invoke the first chatbot.

Abstract: The present invention extends to methods, systems, and computer program products for detecting online contextual evolution of linguistic terms. Within messaging sources, some users may actively attempt to (relatively quickly) shift the meaning of a word or term. Some users may attempt to perjorate a word or term to have a more toxic meaning. Other users may attempt to reappropriate a word or term to have a less toxic or even a positive meaning. Aspects of the invention identify shifts in implied meanings of words and/or phrases over time. As such, emerging forms of harassment can be identified more quickly. Aspects of the invention can utilize users' behavioral histories as well as messaging structures to improve confidence when identifying term evolution. Machine learning algorithms can be configured to identify term evolution reducing workload on human moderators.

Abstract: A speech recognition device includes a sound input section, a sound output section, a communication control section that performs data transmission and reception with at least one of other recognition devices, a conversation-mode executing section that transmits sound data input to each of the other recognition devices and outputs sound data received from each of the other recognition devices, a speech recognition section that converts the sound input into text data, a hot word detecting section that detects a conversation activation hot word from the text data to activate the conversation-mode executing section, and a command transmitting section that transmits a control command to each of the other recognition devices. If the hot word detecting section detects the conversation activation hot word, the command transmitting section transmits the control command to activate a conversation-mode executing section provided in each of the other recognition devices.

Abstract: Conversation support systems and methods are operable to assist a user to more fully participate in an ongoing conversation. An exemplary embodiment receives a conversation map comprising a serially arranged plurality of conversation segments of audio information each associated with one of a plurality of conversation participants, and that includes dialogue of the ongoing conversation being spoken by the associated conversation participant; determining subject matter of a current portion of the ongoing conversation based on the dialogue; selects conversation support information that pertains to the current conversation subject; and generating a conversation queue that includes information that corresponds to the selected conversation support information, wherein the conversation queue is communicated from the conversational support system to at least one conversation queue output device that presents the conversation support information to the user.

Abstract: Systems and methods are disclosed for communication classification and escalation using a machine learning model. Metadata for a communication is obtained along with features of the content in the communication. Using a machine learning model, a likelihood that the communication comprises suspicious content or a topic of interest is determined from the metadata and or a communication transcript. If the likelihood that the communication comprises suspicious content or a topic of interest is at or above a predetermined threshold, then the communication is marked for review by a human agent; otherwise, the communication is ignored. A communication may have suspicious content or a topic of interest based upon the presence of keywords in the communication transcript either alone or in combined with metadata or other features of the content in the communication. The one or more keywords may be associated with illegal or unauthorized items, activities, or behaviors.