Abstract: In some aspects, the techniques described herein relate to a method including: embedding a graph with an embedding model, wherein the embedding generates node and edge vectors for each node and each edge, respectively, in the graph; generating peer node groups based on the node vectors and the edge vectors; generating edge predictions based on the node vectors and the edge vectors; training a machine learning model based on the node vectors and the edge vectors; inputting the node vectors and the edge vectors into the machine learning model; receiving, as output from the machine learning model, a prediction of a future action based on the node vectors and the edge vectors; generating a plurality of insights based on the peer node groups, the edge predictions, and the prediction of a future action; storing the plurality of insights in an insight database, and providing an interface to the insight database.

Abstract: In some aspects, the techniques described herein relate to a method including: providing a graph neural network; and configuring the graph neural network to predict a competitor, the method comprising: receiving at least one dataset of nodes of supply chain companies, competitor companies, and customers and associated attached nodes' attributes; applying a first-order proximity to denote a local connection structure of some supply chain companies, competitor companies, and customers; applying a Laplacian Eigenmap to the first-order proximity to identify at least two positive pairs and at least two negative pairs; applying a pairwise ranking loss function that reduces the distance between the at least two positive pairs and increasing the distance between the at least two negative pairs; and based on an input identification of one company, ranking competitor companies of the company based on their Euclidean distances in the graph.

Abstract: In some aspects, the techniques described herein relate to a method including: providing a graph neural network; and configuring the graph neural network to predict a competitor, the method comprising: receiving at least one dataset of nodes of supply chain companies, competitor companies, and customers and associated attached nodes' attributes; applying a first-order proximity to denote a local connection structure of some supply chain companies, competitor companies, and customers; applying a Laplacian Eigenmap to the first-order proximity to identify at least two positive pairs and at least two negative pairs; applying a pairwise ranking loss function that reduces the distance between the at least two positive pairs and increasing the distance between the at least two negative pairs; and based on an input identification of one company, ranking competitor companies of the company based on their Euclidean distances in the graph.

Abstract: In some aspects, the techniques described herein relate to a method including: providing a graph neural network; and configuring the graph neural network to predict a competitor, the method comprising: receiving at least one dataset of nodes of supply chain companies, competitor companies, and customers and associated attached nodes' attributes; applying a first-order proximity to denote a local connection structure of some supply chain companies, competitor companies, and customers; applying a Laplacian Eigenmap to the first-order proximity to identify at least two positive pairs and at least two negative pairs; applying a pairwise ranking loss function that reduces the distance between the at least two positive pairs and increasing the distance between the at least two negative pairs; and based on an input identification of one company, ranking competitor companies of the company based on their Euclidean distances in the graph.

Abstract: In some aspects, the techniques described herein relate to a method including: embedding a graph with an embedding model, wherein the embedding generates node and edge vectors for each node and each edge, respectively, in the graph; generating peer node groups based on the node vectors and the edge vectors; generating edge predictions based on the node vectors and the edge vectors; training a machine learning model based on the node vectors and the edge vectors; inputting the node vectors and the edge vectors into the machine learning model; receiving, as output from the machine learning model, a prediction of a future action based on the node vectors and the edge vectors; generating a plurality of insights based on the peer node groups, the edge predictions, and the prediction of a future action; storing the plurality of insights in an insight database, and providing an interface to the insight database.

Abstract: Various methods, apparatuses/systems, and media for end-to-end entity linking are disclosed. The system includes a processor; and a memory operatively connected to the processor via a communication interface, the memory storing computer readable instructions, when executed, causes the processor to: detect all named entity mentions from a plurality of data sources; compute, in response to detecting, entity embeddings in a knowledge graph by implementing context information and a margin-based loss function; validate the entity embeddings; deploy, in response to validating the entity embeddings, a machine learning model to match character and semantic information, respectively; and link, in response to deployment of the wide and deep learning model, the named mentions in text with corresponding entities in the knowledge graph.

Abstract: A method for scalable user intent mining is provided. The method includes detecting named entities from a plurality of query logs in a public query log dataset and generating features of the plurality of query logs based on the detected named entities. The method also includes applying a multimodal restricted boltzmann machine (RBM) on the generated features of the plurality of query logs to train a public multimodal RBM and generating a plurality of public query representations. Further, the method includes receiving a search query from a user, determining whether there are a plurality of history queries of the user. When there is no history query, user intent is predicted using the public multimodal RBM. When there are the history queries, the public multimodal RBM is applied on the plurality of history queries to train a personalized multimodal RBM, and the user intent is predicted using the personalized multimodal RBM.

Abstract: A unified attractiveness prediction method is provided. The method includes receiving a plurality of videos and extracting at least one of metadata and view data from each of the plurality of received videos, wherein the metadata is information for describing video contents, and the view data is a total number of users who watch the video. The method also includes obtaining potential view amounts of the plurality of received videos and calculating impact factor scores of the plurality of received videos based on the potential view amounts, each impact factor score including a numerical score for indicating a degree of effectiveness of a corresponding video. Further, the method includes providing a video with a highest impact factor score based on the calculated impact factor scores.

Abstract: The present invention provides a method for providing augmented product specifications based on user reviews. The method obtains input data of specifications and user reviews on a plurality of products, each specification including at least a pair of a feature and a feature-value of the product. The method concatenates the user reviews of the products to form product-documents, each product-document corresponding to the concatenated user reviews of a product. The method further employs a topic model to process the input data and learn topic distributions in the product-documents and word distributions in topics. The topics include specifications of the products. The topic model is a type of statistical model for discovering topics that occur in a collection of product-documents. Based on the topic model, the method can provide augmented specifications including one or more of relevant sentences of the feature-value, feature importance information, and product-specific words of the product.

Abstract: A unified attractiveness prediction method is provided. The method includes receiving a plurality of videos and extracting at least one of metadata and view data from each of the plurality of received videos, wherein the metadata is information for describing video contents, and the view data is a total number of users who watch the video. The method also includes obtaining potential view amounts of the plurality of received videos and calculating impact factor scores of the plurality of received videos based on the potential view amounts, each impact factor score including a numerical score for indicating a degree of effectiveness of a corresponding video. Further, the method includes providing a video with a highest impact factor score based on the calculated impact factor scores.

Abstract: A method for scalable user intent mining is provided. The method includes detecting named entities from a plurality of query logs in a public query log dataset and generating features of the plurality of query logs based on the detected named entities. The method also includes applying a multimodal restricted boltzmann machine (RBM) on the generated features of the plurality of query logs to train a public multimodal RBM and generating a plurality of public query representations. Further, the method includes receiving a search query from a user, determining whether there are a plurality of history queries of the user. When there is no history query, user intent is predicted using the public multimodal RBM. When there are the history queries, the public multimodal RBM is applied on the plurality of history queries to train a personalized multimodal RBM, and the user intent is predicted using the personalized multimodal RBM.

Abstract: The present invention provides a method for providing augmented product specifications based on user reviews. The method obtains input data of specifications and user reviews on a plurality of products, each specification including at least a pair of a feature and a feature-value of the product. The method concatenates the user reviews of the products to form product-documents, each product-document corresponding to the concatenated user reviews of a product. The method further employs a topic model to process the input data and learn topic distributions in the product-documents and word distributions in topics. The topics include specifications of the products. The topic model is a type of statistical model for discovering topics that occur in a collection of product-documents. Based on the topic model, the method can provide augmented specifications including one or more of relevant sentences of the feature-value, feature importance information, and product-specific words of the product.

Abstract: A method is provided for ranking media contents. The method includes receiving media contents through a network and extracting feature values of the received media contents. The method also includes implementing a parameter reinforcement learning process to obtain automatically distribution over relativeness and irrelativeness of the received media contents. Further, the method includes ranking the received media contents by a multi-armed bandit algorithm based on the obtained distribution over relativeness and irrelativeness of the received media contents.