Abstract: A technique selects and serves local information items (e.g., news articles) to users. In a first stage, the technique uses a machine-trained localness-determining system to determine whether a candidate information item contains the kind of information that qualifies as locally-themed. In a second stage, a scope-determining system determines a particular geographic region associated with the information item. The technique then selectively serves the information item to a particular consumer upon determining that the particular consumer is located in the particular geographic region associated with the item. In some implementations, the scope-determining system describes the particular geographic region of the information item using a set of geohashes, and describes the location of the consumer using at least one geohash. The technique uses an ensemble approach to identify the particular geographic region of the item, and to generate training examples for use in training the localness-determining system.

Abstract: A technique selects and serves local information items (e.g., news articles) to users. In a first stage, the technique uses a machine-trained localness-determining system to determine whether a candidate information item contains the kind of information that qualifies as locally-themed. In a second stage, a scope-determining system determines a particular geographic region associated with the information item. The technique then selectively serves the information item to a particular consumer upon determining that the particular consumer is located in the particular geographic region associated with the item. In some implementations, the scope-determining system describes the particular geographic region of the information item using a set of geohashes, and describes the location of the consumer using at least one geohash. The technique uses an ensemble approach to identify the particular geographic region of the item, and to generate training examples for use in training the localness-determining system.

Abstract: The example embodiments are directed toward improvements in document classification. In an embodiment, a method is disclosed comprising generating a set of sentences based on a document; predicting a set of labels for each sentence using a multi-label classifier, the multi-label classifier including a self-attended contextual word embedding backbone layer, a bank of trainable unigram convolutions, a bank of trainable bigram convolutions, and a fully connected layer the multi-label classifier trained using a weakly labeled data set; and labeling the document based on the set of labels. The various embodiments can target multiple use cases such as identifying related entities, trending related entities, creating ephemeral timeline of entities, and others using a single solution. Further, the various embodiments provide a weakly supervised framework to train a model when a labeled golden set does not contain a sufficient number of examples.

Abstract: The example embodiments are directed toward improvements in document classification. In an embodiment, a method is disclosed comprising generating a set of sentences based on a document; predicting a set of labels for each sentence using a multi-label classifier, the multi-label classifier including a self-attended contextual word embedding backbone layer, a bank of trainable unigram convolutions, a bank of trainable bigram convolutions, and a fully connected layer the multi-label classifier trained using a weakly labeled data set; and labeling the document based on the set of labels. The various embodiments can target multiple use cases such as identifying related entities, trending related entities, creating ephemeral timeline of entities, and others using a single solution. Further, the various embodiments provide a weakly supervised framework to train a model when a labeled golden set does not contain a sufficient number of examples.

Abstract: The disclosed systems and methods provide a novel framework that enables cost-effective, accurate and scalable detection and recognition of key events in sporting or live events. The framework functions by creating a domain-specific video dataset with frame level annotations (i.e., deep domain datasets) and then training a lightweight camera view classifier to detect camera views for a given video. The disclosed framework uses pre-trained pose estimation and panoptic segmentation models along with geometric rules as labeling functions to define scene types and derive frame level classification training data. According to some embodiments, disclosed frameworks may be used to identify key persons or events, select a thumbnail corresponding to a key person or event, generate personalized highlights to enhance user experience and social media promotions for a team, sport or players, and predict and select the best camera view sequence for automatic highlights generation.

Abstract: The disclosed systems and methods provide a novel framework that provides mechanisms for performing cost-effective, accurate and scalable detection and recognition of fine-grained events. The framework functions by training high precision and high recall object/optical character recognition (OCR) models and aligning video frames to text commentaries of the videos (e.g., licensed play-by-play). The disclosed framework operates as a single algorithm that performs multimodal alignments between events/actions within videos and their prescribed text. Thus, the disclosed framework is able to scale to fine-grained action categories across different venues by delving into the key frames and key aspects of a video to identify particular actions performed by particular actors, thereby providing the novelty of fine-granted action detection and recognition.