Abstract: Systems and methods for training an object detection network are described. Embodiments train an object detection network using a labeled training set, wherein each element of the labeled training set includes an image and ground truth labels for object instances in the image, predict annotation data for a candidate set of unlabeled data using the object detection network, select a sample image from the candidate set using a policy network, generate a labeled sample based on the selected sample image and the annotation data, wherein the labeled sample includes labels for a plurality of object instances in the sample image, and perform additional training on the object detection network based at least in part on the labeled sample.

Abstract: Embodiments of the technology described herein are directed to a persona-specific navigation interface for a document. Initially, a user may select a persona associated with a document through a document navigation interface. A machine-learning model may identify an interest within a portion of the document. The interest may be mapped to the persona. A navigation interface that includes a navigable link to the portion of the document is generated and output for display. A user interaction with the navigable link is received. In response to the interaction, the portion of the document corresponding to the navigable link is output for display.

Abstract: Embodiments provide systems, methods, and computer storage media for extracting semantic labels for field widgets of form fields in unfilled forms. In some embodiments, a processing device accesses a representation of a fillable widget of a form field of an unfilled form. The processing device generates an encoded input representing text and layout of a sequence of tokens in a neighborhood of the fillable widget. The processing device uses a machine learning model to extract a semantic label representing a field type of the fillable widget in view of the encoded input. The processing device causes execution of an action using the semantic label.

Abstract: Methods, computer systems, computer-storage media, and graphical user interfaces are provided for facilitating data visualization generation. In one implementation, dataset intent data, visual design intent data, and insight intent data determined from a user input natural language query are obtained. A set of candidate intent recommendations is generated using various combinations of the dataset intent data, visual design intent data, and insight intent data. Each of the candidate intent recommendations is incorporated into a set of visualization templates to determine eligibility of the candidate intent recommendations. For eligible candidate intent recommendations, a score associated with a corresponding visualization template is determined. Based on the scores, a candidate intent recommendation and corresponding visualizations template is selected to use as a visual recommendation for presenting a data visualization.

Abstract: Embodiments of the technology described herein are directed to a persona-specific navigation interface for a document. Initially, a user may select a persona associated with a document through a document navigation interface. A machine-learning model may identify an interest within a portion of the document. The interest may be mapped to the persona. A navigation interface that includes a navigable link to the portion of the document is generated and output for display. A user interaction with the navigable link is received. In response to the interaction, the portion of the document corresponding to the navigable link is output for display.

Abstract: Various embodiments are directed to assigning offers to marketing deliveries utilizing new features to describe offers in the marketing deliveries. Marketing deliveries can be described at a finer level to thus enhance the effectiveness of building and conducting marketing campaigns. The approaches facilitate matching content to recipients, predicting content performance, and measuring content performance after dispatching a marketing delivery.

Abstract: In some embodiments, a data visualization system detects insights from a dataset and computes insight scores for respective insights. The data visualization system further computes insight type scores, from the insight scores, for insight types in the detected insights. The data visualization system determines a selected insight type for the dataset having a higher insight type score than unselected insight types and determines, for the selected insight type, a set of selected insights that have higher insight scores than unselected insights. The data visualization system determines insight visualizations for the set of selected insights and generates, for inclusion in a user interface of the data visualization system, selectable interface elements configured for invoking an editing tool for updating the determined insight visualizations from the dataset. The selectable interface elements are arranged in the user interface according to the insight scores of the set of selected insights.

Abstract: This disclosure describes methods, systems, and non-transitory computer readable media for automatically parsing infographics into segments corresponding to structured groups or lists and displaying the identified segments or reflowing the segments into various computing tasks. For example, the disclosed systems may utilize a novel infographic grouping taxonomy and annotation system to group elements within infographics. The disclosed systems can train and apply a machine-learning-detection model to generate infographic segments according to the infographic grouping taxonomy. By generating infographic segments, the disclosed systems can facilitate computing tasks, such as converting infographics into digital presentation graphics (e.g., slide carousels), reflow the infographic into query-and-response models, perform search functions, or other computational tasks.

Abstract: Systems and methods for machine learning are described. The systems and methods include receiving target training data including a training image and ground truth label data for the training image, generating source network features for the training image using a source network trained on source training data, generating target network features for the training image using a target network, generating at least one attention map for training the target network based on the source network features and the target network features using a guided attention transfer network, and updating parameters of the target network based on the attention map and the ground truth label data.

Abstract: Certain embodiments involve a template-based redesign of documents based on the contents of documents. For instance, a computing system selects a template for modifying an input document. To do so, the computing system uses a generative adversarial network to generate an interpolated layout image from an input layout image, which represents the input document, and a template layout image, which represents the selected template. The computing system matches the input element to an interpolated element from the interpolated layout image. The computing system generates an output document by, for example, modifying a layout of the input document to match the interpolated layout image, such as by fitting the input element into a shape of the interpolated element.

Abstract: Methods, computer systems, computer-storage media, and graphical user interfaces are provided for facilitating data visualization generation. In one implementation, dataset intent data, visual design intent data, and insight intent data determined from a user input natural language query are obtained. A set of candidate intent recommendations is generated using various combinations of the dataset intent data, visual design intent data, and insight intent data. Each of the candidate intent recommendations is incorporated into a set of visualization templates to determine eligibility of the candidate intent recommendations. For eligible candidate intent recommendations, a score associated with a corresponding visualization template is determined. Based on the scores, a candidate intent recommendation and corresponding visualizations template is selected to use as a visual recommendation for presenting a data visualization.

Abstract: Techniques and systems for determining paywall metrics are described. In an implementation, a candidate paywall metric is created that corresponds to an increased propensity of users to engage in a paid transaction when exposed to a paywall. In this way, providers of digital content may increase the proportion of users that perform a transaction when exposed to the paywall.

Abstract: Certain embodiments involve a template-based redesign of documents based on the contents of documents. For instance, a computing system selects a template for modifying an input document. To do so, the computing system uses a generative adversarial network to generate an interpolated layout image from an input layout image, which represents the input document, and a template layout image, which represents the selected template. The computing system matches the input element to an interpolated element from the interpolated layout image. The computing system generates an output document by, for example, modifying a layout of the input document to match the interpolated layout image, such as by fitting the input element into a shape of the interpolated element.

Abstract: Systems and methods for training an object detection network are described. Embodiments train an object detection network using a labeled training set, wherein each element of the labeled training set includes an image and ground truth labels for object instances in the image, predict annotation data for a candidate set of unlabeled data using the object detection network, select a sample image from the candidate set using a policy network, generate a labeled sample based on the selected sample image and the annotation data, wherein the labeled sample includes labels for a plurality of object instances in the sample image, and perform additional training on the object detection network based at least in part on the labeled sample.

Abstract: In some embodiments, a data visualization system detects insights from a dataset and computes insight scores for respective insights. The data visualization system further computes insight type scores, from the insight scores, for insight types in the detected insights. The data visualization system determines a selected insight type for the dataset having a higher insight type score than unselected insight types and determines, for the selected insight type, a set of selected insights that have higher insight scores than unselected insights. The data visualization system determines insight visualizations for the set of selected insights and generates, for inclusion in a user interface of the data visualization system, selectable interface elements configured for invoking an editing tool for updating the determined insight visualizations from the dataset. The selectable interface elements are arranged in the user interface according to the insight scores of the set of selected insights.

Abstract: Embodiments are disclosed for determining an answer to a query associated with a graphical representation of data. In particular, in one or more embodiments, the disclosed systems and methods comprise obtaining a visual embedding for a graphical representation of data, the visual embedding representing a plurality of graphical elements. The one or more embodiment further include obtaining a query embedding for a query associated with the graphical representation of data, the query embedding representing a plurality of textual elements of the query with at least one textual element substituted with an identifier for at least one graphical element of the set of graphical elements. The one or more embodiment further include generating a chart sequence from the visual embedding and a query sequence from the query embedding, generating an output sequence based on the graph and the query sequences, and determining an answer to the query from the output sequence.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media that generate expressive audio for input texts based on a word-level analysis of the input text. For example, the disclosed systems can utilize a multi-channel neural network to generate a character-level feature vector and a word-level feature vector based on a plurality of characters of an input text and a plurality of words of the input text, respectively. In some embodiments, the disclosed systems utilize the neural network to generate the word-level feature vector based on contextual word-level style tokens that correspond to style features associated with the input text. Based on the character-level and word-level feature vectors, the disclosed systems can generate a context-based speech map. The disclosed systems can utilize the context-based speech map to generate expressive audio for the input text.

Abstract: Embodiments are disclosed for determining an answer to a query associated with a graphical representation of data. In particular, in one or more embodiments, the disclosed systems and methods comprise obtaining a visual embedding for a graphical representation of data, the visual embedding representing a plurality of graphical elements. The one or more embodiment further include obtaining a query embedding for a query associated with the graphical representation of data, the query embedding representing a plurality of textual elements of the query with at least one textual element substituted with an identifier for at least one graphical element of the set of graphical elements. The one or more embodiment further include generating a chart sequence from the visual embedding and a query sequence from the query embedding, generating an output sequence based on the graph and the query sequences, and determining an answer to the query from the output sequence.

Abstract: A method, apparatus, and non-transitory computer readable medium for chart question answering are described. The method, apparatus, and non-transitory computer readable medium may receive a text query about a chart, identify a plurality of chart elements in the chart, associate a text string from the text query with corresponding chart elements from the plurality of chart elements, replace the text string in the text query with arbitrary rare words based on the association to produce an encoded query, generate an embedded query based on the encoded query, generate an image feature vector based on the chart, combine the embedded query and the image feature vector to produce a combined feature vector, compute an answer probability vector based on the combined feature vector, and provide an answer to the text query based on the answer probability vector.

Abstract: The present disclosure relates to systems, methods, and non-transitory computer-readable media that generate expressive audio for input texts based on a word-level analysis of the input text. For example, the disclosed systems can utilize a multi-channel neural network to generate a character-level feature vector and a word-level feature vector based on a plurality of characters of an input text and a plurality of words of the input text, respectively. In some embodiments, the disclosed systems utilize the neural network to generate the word-level feature vector based on contextual word-level style tokens that correspond to style features associated with the input text. Based on the character-level and word-level feature vectors, the disclosed systems can generate a context-based speech map. The disclosed systems can utilize the context-based speech map to generate expressive audio for the input text.