Abstract: A method of generating an output movement layout for a traffic intersection is disclosed which includes receiving intersection geographical data, establishing a center point for the intersection, receiving vehicle global positioning system (GPS) data, establishing a radius of interest for the intersection based on the received vehicle GPS data, establishing entry and exit headings for each vehicle based on the GPS data, generating an angular cluster chart based on the entry and exit headings of each vehicle, and generating an output movement layout for the intersection based on the generated angular cluster chart. Further, a method for green time reallocation at traffic signals is disclosed that is based on the congestion experienced by vehicles identified as following specific movements.

Abstract: Certain aspects of the disclosure provide systems and methods for training an information extraction transformer model architecture directed to pre-training a first multimodal transformer model on an unlabeled dataset, training a second multimodal transformer model on a first labeled dataset to perform a key information extraction task processing the unlabeled dataset with the second multimodal transformer model to generate pseudo-labels for the unlabeled dataset, training the first multimodal transformer model based on a second labeled dataset comprising one or more labels, the pseudo-labels generated, or combinations thereof to generate a third multimodal transformer model, generating updated pseudo-labels based on label completion predictions from the third multimodal transformer model, and training the third multimodal transformer model using a noise-aware loss function and the updated pseudo-labels to generate an updated third multimodal transformer model.

Abstract: Aspects of the subject technology provide for suggested home screens and/or lock screens of an electronic device. The suggested home screens and/or lock screens can include suggested backgrounds, user interface elements, colors, styles, photos, and the like, and can be generated and/or suggested for a user of an electronic device based on activity history information for the user and/or the electronic device. In one or more implementations, the home screens and/or lock screens may be generated and/or suggested specifically for various operational modes of the electronic device.

Abstract: Systems and methods are described for generating dynamic web clippings. A web clipping service can obtain from a client computing device metadata for a dynamic web clipping, which includes an identifier of a web document and state information of a web browser that has rendered the web document on the client computing device, the state information enabling replication of a browsing session of the client computing device. The web clipping service can use the metadata to recreate the browsing session of the client device on the web clipping service, and capture an image of the web document as rendered within he recreated browsing session. The web clipping service can then provide the image of the web document as the dynamic web clipping.

Abstract: Generating granular clusters for real-time processing is provided. The systems can identify tokens based on aggregating input from computing devices over a time interval. The systems can identify, based on metrics, a subset of tokens for cluster generation. The systems can generate, via a clustering technique, token clusters from the subset of the tokens, each of the token clusters comprising two or more tokens from the subset of the tokens. The systems can apply a de-duplication technique to each of the token clusters. The systems can apply a filtering technique to the token clusters to remove tokens erroneously grouped in a token cluster. The systems can assign, based on a selection process, a label for each of the token clusters. The systems can activate, based on a number of remaining tokens in each of the token clusters, a subset of the token clusters for real-time content selection.

Abstract: Generating granular clusters for real-time processing is provided. The systems can identify tokens based on aggregating input from computing devices over a time interval. The systems can identify, based on metrics, a subset of tokens for cluster generation. The systems can generate, via a clustering technique, token clusters from the subset of the tokens, each of the token clusters comprising two or more tokens from the subset of the tokens. The systems can apply a de-duplication technique to each of the token clusters. The systems can apply a filtering technique to the token clusters to remove tokens erroneously grouped in a token cluster. The systems can assign, based on a selection process, a label for each of the token clusters. The systems can activate, based on a number of remaining tokens in each of the token clusters, a subset of the token clusters for real-time content selection.

Abstract: An indication from a live agent in a communication session with a user to invoke a specified automated agent interactive conversation flow is received. The communication session with the user is transitioned from the live agent to the specified automated agent interactive conversation flow. During the specified automated agent interactive conversation flow, the live agent retains control over the communication session. An indication of an interaction of the user with the specified automated agent interactive conversation flow is provided. The communication session with the user is returned back to the live agent.

Abstract: A method includes receiving a permission request from a user to perform an operation in a computer environment. Data elements of operations previously performed by a plurality of users are received. A set of operation types are generated by identifying data elements of operations previously performed by other users of the plurality of users having operation types performed by the user. The data elements in the set are mapped into a graph database. Permission messages are generated with a correct response identified from a data attribute from a specific operation type previously performed by the user, and wrong responses identified from the specific operation type previously performed by the other users. An accumulated probability based on the permission respond message is computed using the graph database. The user is tagged based on the accumulated probability as a permissioned user for performing the operation, or a non-permissioned violator user.

Abstract: Techniques for creating and calibrating natural-language understanding (NLU) machine learning models are described. In certain embodiments, a training service tunes parameters of a function, taking the output from an NLU machine learning model as an input of the function, to calibrate the NLU machine learning model's output to optimize the interpretability of the resulting output, e.g., confidence score(s). Embodiments herein include generating, by the NLU machine learning model, an output based at least in part on an input (e.g., utterance) from a user, and applying a tuned, output modifying function to the output from the NLU machine learning model to generate a modified output. An inference may be generated based at least in part on the modified output.

Abstract: Systems and methods for providing contextual information for computerized document understanding. The systems and methods can be used to assist users in filling out documents by providing contextual information based on anomalies identified in a provided document. The methods and systems may identify the deficiency in the document and automatically generate a query related to the anomaly. The query can be fed as an input to a question-answering (QA) model that can provide an answer as the contextual information.

Abstract: Certain aspects of the present disclosure provide techniques for providing sentiment analysis of content. In order to determine the overall sentiment of content, a request is received by a sentiment analyzer, which then identifies a content identification number and retrieves comments associated with the content identification number. The sentiment analyzer pre-processes the comments, which includes removing all personal identifying information from the comments. The sentiment analyzer sends the pre-processed comments to a natural language processing service, and in turn, receives sentiment indications corresponding to the comments provided. Based on the sentiment scores, the sentiment analyzer generates a sentiment analysis and displays the sentiment analysis in the graphical user interface generated by the sentiment analyzer.

Abstract: Certain aspects of the present disclosure provide techniques for providing sentiment analysis of content. In order to determine the overall sentiment of content, a request is received by a sentiment analyzer, which then identifies a content identification number and retrieves comments associated with the content identification number. The sentiment analyzer pre-processes the comments, which includes removing all personal identifying information from the comments. The sentiment analyzer sends the pre-processed comments to a natural language processing service, and in turn, receives sentiment indications corresponding to the comments provided. Based on the sentiment scores, the sentiment analyzer generates a sentiment analysis and displays the sentiment analysis in the graphical user interface generated by the sentiment analyzer.

Abstract: Utilizing quick actions includes receiving user input via a user input module of an agent workspace, wherein the user input comprises an action input comprising an action command, and in response to receiving the user input, determining a secondary menu for the action command, wherein the secondary menu comprises a plurality of selectable parameters for the action command, and presenting the secondary menu in association with the user input module. Utilizing quick actions also includes detecting a selection of a parameter from the selectable parameters, and triggering an action based on the action command and the selected parameter, wherein the action is associated with a process for the agent workspace.

Abstract: A method for performing question answer (QA) tasks that includes entering an input into an encoder portion of an adaptive memory network, wherein the encoder portion parses the input into entities of text for arrangement of memory banks. A bank controller of the adaptive memory network organizes the entities into progressively weighted banks within the arrangement of memory banks. The arrangement of memory banks may be arranged to have an initial memory bank having lowest relevance for lowest relevance entities being closest to the encoder, and a final memory bank having a highest relevance for highest relevance entities being closes to a decoder. The method may continue with inferring an answer for the question answer (QA) task with the decoder analyzing only the highest relevance entities in the final memory bank.

Abstract: Utilizing quick actions includes receiving user input via a user input module of an agent workspace, wherein the user input comprises an action input comprising an action command, and in response to receiving the user input, determining a secondary menu for the action command, wherein the secondary menu comprises a plurality of selectable parameters for the action command, and presenting the secondary menu in association with the user input module. Utilizing quick actions also includes detecting a selection of a parameter from the selectable parameters, and triggering an action based on the action command and the selected parameter, wherein the action is associated with a process for the agent workspace.

Abstract: A method of generating a supplemental user interface is disclosed. At a client system, it is detected that an alert identifying a suspicious activity that is to be reviewed by a user of the client system has been received. A template for a user interface is accessed. The template includes database queries and instructions for presenting data items derived from query results corresponding to the database queries in the user interface. The instructions include a mapping of the data items to data fields of the user interface. Data items derived from the query results are presented in the user interface based on the instructions included in the template. An audit log is updated based on an acceptance by the user of the data items presented in the user interface.

Abstract: A method for performing question answer (QA) tasks that includes entering an input into an encoder portion of an adaptive memory network, wherein the encoder portion parses the input into entities of text for arrangement of memory banks. A bank controller of the adaptive memory network organizes the entities into progressively weighted banks within the arrangement of memory banks. The arrangement of memory banks may be arranged to have an initial memory bank having lowest relevance for lowest relevance entities being closest to the encoder, and a final memory bank having a highest relevance for highest relevance entities being closes to a decoder. The method may continue with inferring an answer for the question answer (QA) task with the decoder analyzing only the highest relevance entities in the final memory bank.

Abstract: A solid elastomeric reinforcing material for use in the manufacture of compounded rubber goods that is made from an elastomeric material coagulated from a natural or synthetic latex, a natural or synthetic fibrous filler that has a linear density of at least 0.5 dtex, a tenacity of at least 1.0 gram per dtex, a fiber length between 0.1 and 6 mm and a specific surface area range between 0.1-25 square meters per gram and a modulus enhancing additive.

Abstract: The claimed utility is called back-hair-trim-guide with an adjustable mirror for man. It trims man's back and side hair safely and effectively with a clipper. It comprises the following five parts, 1. An utility holding bar to hold the back mirror and back hair guides 2. An adjustable back hair mirror, 3. Two adjustable back hair protection cover (guides) 4. A front belt with both ends connecting to either ends of the said utility holding bar. Its length is adjustable. 5. A top belt with both ends connecting to either ends of the said utility holding bar. Its length is adjustable.