Abstract: A system, method, and computer program product for training a generative language model (GLM) is provided. A plurality of sampled rationales for various question-answer pairs are generated using the GLM. A gradient estimate of parameters of neurons in the GLM is determined based on these sampled rationales to maximize the learning objective of the GLM. The parameters of the GLM are modified using the gradient estimate over multiple iterations, ultimately providing a trained GLM.

Abstract: Systems, devices, and techniques are disclosed for efficient knowledge graph indexing and retrieval. Document chunks may be generated from documents. Summarizations may be generated from document chunks. Entity types, entity properties, relations, and relation properties may be generated from a subset of the summarizations. A schema including entity types, entity properties, relations, and relation properties may be generated. Entity property triplets and entity relation triplets may be generated from the summarizations based on the schema and linked to the document chunks. A knowledge graph including nodes representing entities from the entity property triplets and entity relation triplets and edges representing the entity property triplets and the entity relation triplets may be generated. A search query may be received.

Abstract: Embodiments described herein provide knowledge graph synthesis pipeline to generate a knowledge graph from long documents so as to serve a retrieval augmented generation (RAG) large language model (LLM) based AI chat agent. Specifically, each document is decontextualized by substituting entity references with their explicit mentions. Subsequently, to enhance coverage, the document is segmented into chunks and entities and relations are extracted from each chunk independently, e.g., by an LLM. The extracted entities and relations are then synthesized into a knowledge graph for the document. Therefore, the retrieval component may search the knowledge graph based on a received user query to retrieve entities and relations, which are in turn input to an LLM to generate a response.

Abstract: Methods, systems, apparatuses, devices, and computer program products are described. A system may obtain a set of documents associated with a knowledge base for retrieval-augmented generation (RAG). The system may generate multiple representations of the information included in the documents using multiple knowledge extraction pipelines. For example, the system may generate a set of metadata-based vector embeddings based on the documents, a set of knowledge graphs based on the documents, and a set of hierarchical tree representations based on the documents. The system may receive a user query and may retrieve contextual information from the set of vector embeddings, the set of knowledge graphs, and the set of hierarchical tree representations to augment the user query for a large language model (LLM) prompt. The system may input the prompt to the LLM, and the LLM may output a response based on the user query and the contextual information.

Abstract: Methods, apparatuses, and computer program products are disclosed. The method may include receiving a first request to ingest a document. The method may include generating, using a large language model (LLM), a knowledge graph including a plurality of graph triples, each graph triple including a first node, a second node, and an edge connecting the first node and the second node, where each first node corresponds to a first element type of the document, where each second node corresponds to a second element type of the document, and where each edge corresponds to a third element type of the document. The method may include receiving a second request to generate a generative response with the LLM. The method may include presenting a response to the second request, the response generated by the LLM based at least in part on the knowledge graph.

Abstract: A retrieval augmented generation (RAG) based query reformulation pipeline for a Query and Answer (QA) system is described. This pipeline leverages a Directed Acyclic Graph (DAG) and involves several operations, including retrieval of documents and knowledge graph triplets based on the initial query, reranking of retrieved elements based on relevance, refinement and summarization of relevant document chunks and knowledge triplets, reformulation of the initial query, and generation of a natural language response. The response is generated using a large language model (LLM) and is grounded in the knowledge base, which supports factual accuracy and consistency.

Abstract: A system and method for recommending demand-supply agent pairs for transactions uses a deep neural network on data of demand agents to produce a demand agent vector, which is used to select supply agents based on their likelihood of future transaction and to find k nearest neighbor demand agents for each of the demand agents. The candidate supply agents and the k nearest neighbor demand agents are then combined to produce candidate demand-supply agent pairs, which are used to find recommended demand-supply agent pairs by applying modeling using machine learning.

Abstract: A system and method for recommending demand-supply agent pairs for transactions uses a deep neural network on data of demand agents to produce a demand agent vector, which is used to select supply agents based on their likelihood of future transaction and to find k nearest neighbor demand agents for each of the demand agents. The candidate supply agents and the k nearest neighbor demand agents are then combined to produce candidate demand-supply agent pairs, which are used to find recommended demand-supply agent pairs by applying modeling using machine learning.

Abstract: A method is disclosed for high-speed processing of structured application messages in a network device. According to one aspect, a network device receives a structured application layer message and identifies, in message classification requirements at the network device, a reference to a classification portion of the structured application layer message and an operation portion of the structured application layer message. The system extracts, based on one or more expressions, a portion of the message for classifying the structured application layer message and classifies the message using the extracted portion and according to the message classification requirements. At least in part by accessing information indicated by one or more location identifiers, at least one operation is performed on the classified structured application layer message.

Abstract: An application framework may include a cloud-independent application manager, a cloud-independent content services manager, and cloud-independent content services servers. The framework may dynamically install and manage scalable, multi-tenant applications in a cloud, and may scale the applications, as needed. The application manager may receive and handle requests to install application domains and configurations thereof, and may receive and respond to requests for information about servers on which installed methods are available. The content services servers may execute installed methods, using underlying resources of the cloud, through a cloud-specific SPI. The content services manager and application manager may work together using shared cloud storage to provide scalable content services at a very large scale.

Abstract: A system is provided to manage cloud deployment configuration of a computing application. The system comprises a request detector, a retrieving module, a manager loader, a configuration change request detector, and a configuration module. The request detector may be configured to detect a request to install a manager agent on an instance of a virtual machine executing a computing application within a virtualization service. The retrieving module may be configured to obtain a manager agent object for loading the manager agent, and install the manager agent on the instance. The manager loader may be configured to invoke the manager agent to collect metrics for the computing application. The configuration change request detector may be configured to receive an instruction to alter cloud deployment configuration of the computing application. The configuration module may be configured to automatically alter the cloud deployment configuration of the computing application in response to the instruction.

Abstract: A system is provided to deploy a computing application in a cloud using a service provider interface. The system comprises a configuration request detector, a configuration object retriever, and a configuration installer. The configuration request detector receives, at a computer system provided within a virtualization service, a configuration request. The configuration object retriever retrieves a configuration object in response to the configuration request. The configuration installer uses the configuration object, to automatically install a configuration associated with the configuration object to produce a second version of the computing application. The second version of the computing application provides additional features or services according to the definition of the configuration object.

Abstract: A system is provided to deploy a computing application on a cloud using optimization criteria. The system comprises a launch request detector to detect a request to launch a computing application within a virtualization service, an application requirement detector to determine application requirement for the computing application, a cloud resource inventory detector to determine cloud resource inventory, an optimization requirement detector, a matching module, and a configuration module. The optimization requirement detector may be configured to determine an optimization requirement. The matching module may be configured to perform matching of the application requirement with the resource inventory based on the optimization requirement.

Abstract: A system is provided to deploy a computing application in a cloud using a service provider interface. The system comprises a configuration request detector, a configuration object retriever, and a configuration installer. The configuration request detector receives, at a computer system provided within a virtualization service, a configuration request. The configuration object retriever retrieves a configuration object in response to the configuration request. The configuration installer uses the configuration object, to automatically install a configuration associated with the configuration object to produce a second version of the computing application. The second version of the computing application provides additional features or services according to the definition of the configuration object.

Abstract: A system is provided to manage cloud deployment configuration of a computing application. The system comprises a request detector, a retrieving module, a manager loader, a configuration change request detector, and a configuration module. The request detector may be configured to detect a request to install a manager agent on an instance of a virtual machine executing a computing application within a virtualization service. The retrieving module may be configured to obtain a manager agent object for loading the manager agent, and install the manager agent on the instance. The manager loader may be configured to invoke the manager agent to collect metrics for the computing application. The configuration change request detector may be configured to receive an instruction to alter cloud deployment configuration of the computing application. The configuration module may be configured to automatically alter the cloud deployment configuration of the computing application in response to the instruction.

Abstract: A system is provided to deploy a computing application on a cloud using optimization criteria. The system comprises a launch request detector to detect a request to launch a computing application within a virtualization service, an application requirement detector to determine application requirement for the computing application, a cloud resource inventory detector to determine cloud resource inventory, an optimization requirement detector, a matching module, and a configuration module. The optimization requirement detector may be configured to determine an optimization requirement. The matching module may be configured to perform matching of the application requirement with the resource inventory based on the optimization requirement.

Abstract: A method is disclosed for high-speed processing of structured application messages in a network device. According to one aspect, a network device receives a structured application layer message and identifies, in message classification requirements at the network device, a reference to a classification portion of the structured application layer message and an operation portion of the structured application layer message. The system extracts, based on one or more expressions, a portion of the message for classifying the structured application layer message and classifies the message using the extracted portion and according to the message classification requirements. At least in part by accessing information indicated by one or more location identifiers, at least one operation is performed on the classified structured application layer message.

Abstract: A method is disclosed for high-speed processing of structured application messages in a network device. According to one aspect, a network device receives a set of message classification rules that have been prepared beforehand by a system administrator or customer. The system analyzes the message classification rules to determine what part(s) of the message are necessary to classify a message according to the message classification rules. This allows the system to consider only the relevant parts of the message and ignore the rest of the message. The system extracts the portion of the message necessary for classifying the message and classifies the message using the values of the extracted information and the message classification rules. A unique sequence of operations is implied by the message classification and those operations must then be applied to the message.

Abstract: A system is provided to deploy a computing application on a cloud using a service provider interface. The system comprises a launch request detector, a service provider interface selector, and a launcher. The launch request detector may be configured to detect a request to launch a server computer system within a virtualization service. The server computer system, in one example embodiment, comprises a computing application and a service provider interface layer. The service provider interface selector may be configured to determine an implementation of the virtualization service and access a service provider interface associated with the implementation of the virtualization service. The launcher may be configured to complete launching of the server computer system within the virtualization service, utilizing the selected service provider interface.

Abstract: A method is disclosed for performing message payload processing functions in a network element on behalf of an application. According to one aspect, a network element intercepts data packets comprising network layer or transport layer headers having an address of a destination which destination differs from the network element. The network element determines whether information contained in layer 2-4 headers of the data packet satisfies specified criteria. If the information satisfies the specified criteria, the network element directs the data packets to a blade of the network element that performs processing based on an application layer message at least partially contained in the data packets. If the information does not satisfy the specified criteria, the network element forwards the data packets towards the destination without sending them to the blade.