95 results about "Specific knowledge" patented technology

Systems, methods, and computer-readable media provide a context-specific prompt to answer a user query. The systems, methods, and computer-readable media determine a context based on content of a natural language request and / or determine a role of a user who submitted the natural language request. Additionally or alternatively, templates or RAG sources that will be used for prompt generation may include financials domain-specific knowledge or other domain-specific knowledge or insights. Inclusion of this additional information in the prompt enhances the context to promote more accurate results from a large language model. In one embodiment, the prompt templates are created from various RAG sources, such as payables, general ledger, receivables, and asset management, containing structured data and information specific to the financial domain, enterprise, or other domain, which helps craft accurate prompts. A prompt is generated that identifies a subset of available fields and other selected information based on the role or other context. The prompt template may contain domain-specific knowledge uses a relevant domain or enterprise information to drive relevant results, and an executable query is generated by a large language model based on the prompt. The executable query causes data to be retrieved from a database to generate a result, and information is displayed based at least in part on the result.

The invention provides a complex scene-oriented end-to-end semantic extraction system, belongs to the technical field of artificial intelligence and natural language processing, and realizes cross-modal information association through a multi-source heterogeneous data fusion module to construct a dynamic semantic network model. A hierarchical attention mechanism is adopted to carry out context-aware coding on unstructured input, and unsupervised pre-training and a weak supervised fine tuning strategy are combined to optimize a feature representation space. And designing an adaptive inference engine, automatically switching semantic analysis paths based on scene complexity, and generating a structured output result. According to the method, the dependency on specific knowledge in the field is reduced, the semantic understanding generalization ability in a complex scene is remarkably improved, high-precision analysis performance can still be kept in a low-resource environment, meanwhile, calculation resource consumption is reduced, and the method is suitable for practical application scenes with multi-language mixing, serious noise interference and high real-time performance requirements.

The invention relates to the field of artificial intelligence, the technical scheme is applied to the financial field and the medical health field, and discloses a federated learning-based large model optimization method and device, computer equipment and a storage medium, and the method comprises the steps of receiving a global model from a server; performing local training on the global model through local data, keeping weight freezing of the basic model during local training, and updating the first low-rank matrix and the second low-rank matrix; sending the updated first low-rank matrix to a server, so that the server aggregates the updated first low-rank matrix of each client to obtain a global low-rank matrix; and receiving the global low-rank matrix from the server, generating a weight update quantity to update the weight of the local adapter of the global model, and obtaining an optimized global model. According to the scheme, the communication efficiency can jump; the second low-rank matrix solidifies client-side specific knowledge to prevent global knowledge pollution; and a weight merging mechanism in the reasoning stage ensures zero delay spread.

Techniques for constraining the results of a generative language model to valid information using knowledge-grounded documentation. A generative language model may generate invalid results, including compound entities and incorrect entity relations. The techniques include, for a given user inquiry, determining a set of documented information, from a particular knowledge base, that corresponds to the user inquiry. The techniques further include determining a subgraph from a knowledge graph representing the knowledge base, as well as determining a trie data structure representation of the set of documented information. The user inquiry and subgraph are provided as input to a trained generative language model for generating a response to the user inquiry. The techniques include using the trie data structure to validate that the generated response corresponds to real information from the set of documented information.

The invention discloses an interpretable clinical decision support system based on label generation and a knowledge graph. The method comprises the following steps: based on a breast cancer domain knowledge enhanced version Qwen-BrCaAdapt of a general large language model Qwen, analyzing an unstructured medical record text of a patient, and generating a structured result containing tags, values, evidences and explanations; calculating a reasoning label through a path matching engine by utilizing an editable structured path rule table, and matching a candidate treatment scheme according to the reasoning label; and taking the matched treatment scheme as a central node, calling a medical knowledge graph to bind entity information including clinical evidence, recommendation levels, medical insurance information, medication risks, usage and dosage, and generating a traceable JSON structure and a visual report. The method has the beneficial effects that the accuracy and efficiency of tag generation in the breast cancer field are improved, the rule maintenance cost is reduced, the interpretability and traceability of a clinical decision scheme are enhanced, and the acceptability of a doctor to a recommendation result is improved.

The invention belongs to the technical field of artificial intelligence, and particularly relates to an enterprise special agent rapid construction method based on large model fine tuning. The method comprises the following steps: S1, performing agent basic configuration; s2, outputting an enterprise special agent based on the agent basic configuration scheme and the general large model; s3, setting rule parameters of an agent response strategy, and generating an agent interaction strategy; s4, performing full-life-cycle management on enterprise knowledge entries to ensure the accuracy and traceability of agent answering; and S5, verifying the accuracy of the agent knowledge and continuously optimizing the agent knowledge. According to the invention, through low-code configuration, an enterprise can construct an agent without a technical team, and autonomous controllability is realized; a large model adaptation technology enables the intelligent agent to accurately understand proprietary knowledge, solves the problem of term misunderstanding, and achieves high-precision adaptation; the knowledge full-life-cycle management and rapid updating mechanism ensures the timeliness of the agent knowledge; customizable interaction rules enable the intelligent agent to adapt to multi-scene requirements, and the user experience is improved.

Context-aware content filtering adapted for a knowledge domain is provided. In certain examples, a classification confidence score by a classifier indicates a level of confidence that a prompt from a user is associated with the knowledge domain. The classification confidence score is compared with a threshold. When the score is below a threshold, a violation notice is provided to the user without submitting the prompt to a generative artificial intelligence (GAI) model. When the classification confidence score is above the threshold, the prompt is further processed to determine, according to rules, whether the prompt should be submitted to the GAI model. In various examples, the rules are applied to contextual information, safety score information, and intent information derived from the prompt.

The present disclosure relates to a technique for addressing an issue to be resolved associated with an electronic document. The method discloses accessing an actionable portion associated with a particular knowledge domain of the electronic document and associated context. Further, retrieve data from data sources to provide additional information related to the particular knowledge domain and the associated context. Then structuring the retrieved data to produce a subset of organized data and determine the issue to be resolved related to the electronic document. Further, generate data elements associated with the issue to be resolved and map dependency relationships between data elements. Also, determine test goals associated with the issue to be resolved based on the dependency relationships. Thereafter, determine corresponding test cases associated with resolution and determines actionable test steps related to the issue to be resolved based on the corresponding test cases associated with the electronic document.

Systems and methods for implementing domain-specific agent networks including configuring specialized agents optimized for retrieving information from a respective specific knowledge domain, receiving a user query, analyzing the user query to identify relevant knowledge domains, activating a subset of the specialized agents corresponding to the relevant knowledge domains, retrieving information by the subset of specialized agents, aggregating the retrieved information including information from a plurality of knowledge domains, providing the aggregated information to one or more h-LLMs, receiving a plurality of responses from the one or more h-LLMs, and generating a comprehensive response from the plurality of responses, the comprehensive response incorporating information from the plurality of knowledge domains.

An AI-based response generation chatbot system that acts as a digital replica of a person or expert, rather than being the expert itself, interacts with a user while being entirely guided by the information provided to it, without revealing its AI nature and the source of information. The AI-based response generation chatbot system includes a knowledge database initialized with knowledge documents containing expert knowledge with a specific viewpoint. The knowledge documents are compiled into a vector database through chunking and embedding techniques and are converted into unique topic-specific knowledge chunks in a machine-readable format. The compiled vector database further incorporates the Retrieval Augmented Generation (RAG) framework, enabling the retrieval of relevant information from the vector database and then using the retrieved information to frame accurate and contextually relevant responses aligned with user queries.

The invention discloses a personalized learning portrait construction method fusing multi-source data and state updating, and belongs to the field of student portrait construction. The method comprises the steps of firstly collecting learning behavior data of a student, calculating a mastering probability value of the student for each knowledge point in a knowledge graph based on a preset evaluation model, and mapping the mastering probability value into an explicit mastering state; active interaction behaviors of the students are monitored in real time, the self-evaluation mastering state is deduced, and whether correction of the explicit mastering state of the specific knowledge points is triggered or not is judged according to the state difference degree; and taking the corrected mastering state of the knowledge point as an updating trigger point, calculating an expected influence quantity of the mastering state change of the corrected knowledge point on the associated knowledge point based on the topological structure of the knowledge graph, and updating the mastering state of the associated knowledge point according to the expected influence quantity, thereby generating an updated personalized learning portrait. According to the method, the mastering state of the associated knowledge points is dynamically and intelligently updated, and a solid technical support is provided for self-adaptive learning path recommendation.

The application provides a kind of instruction driving machine task planning method and system based on knowledge model, and the specific knowledge of field is structuredly represented by knowledge model, and the object of class level, instance level of field knowledge and its logical relationship are represented;Further, the semantics of instruction is understood, and then the task planning problem represented by instruction is normalized, specifically including task planning object and its mutual constraint relationship;On this basis, further consider the space-time constraint between task objects and the measurement and evaluation of task efficiency.Finally, through atlas and visual graph display, give the pareto optimal scheme under multi-dimension, support intelligent or man-machine interactive decision, realize man-machine interaction in task planning, and output field-specific reliable scheme based on natural language.

Disclosed herein are systems and method for using machine learning to generate / provide supplemental information material for a presentation. In an aspect, presentation material associated with the presentation is analyzed using a knowledge model to identify one or more elements in the presentation material, the one or more elements being associated with one or more concepts related to the presentation and / or related to a specific knowledge area of the knowledge model. For each element of the one or more elements, supplemental information material related to a concept associated with that element is created. A table that links each element of the one or more elements with corresponding supplemental information material is generated. The table links each element in the one or more elements with a position of that element in the presentation material.

The invention discloses domain adaptation for automatic speech recognition systems using retrieval enhancement generation. The method presented herein provides for the generation of a textual transcript of speech represented in audio data. In particular, an Automatic Speech Recognition (ASR) model may be used with a Retrieval Enhanced Generation (RAG) pipeline to provide improvements in transcripts including terms related to or specific to a particular knowledge domain. A knowledge base of a given domain may include a number of a plurality of different formats of files or documents (e.g., documents, images, and web pages) that do not need to be cleaned, classified, or curated. When the ASR generates a transcript for which the confidence level of at least one word is below a confidence threshold, the transcript may be passed to a language model of the RAG pipeline that may use the retrieved domain-specific data to attempt to identify an appropriate word or term for replacing a word tagged as having a low confidence.

This invention provides a method, apparatus, and electronic device for multi-turn conversations. The method includes: generating a knowledge matrix comprising multiple items, where each element represents a knowledge point corresponding to a specific knowledge type within that item; extracting the target knowledge point from the current round of information; generating a response for the current round when the target item in the knowledge matrix cannot be located based on the target knowledge points from all current rounds, until the target item is located based on the target knowledge points from all current rounds; and outputting the content result corresponding to the target item. The multi-turn conversation method, apparatus, and electronic device provided by this invention can generate a knowledge matrix with a simple structure; it does not require accurate entity identification, nor does it require attention to the relationships between entities, making the generation process simple. Even in complex application scenarios involving a large number of items, multi-turn dialogues can still be effectively implemented based on this knowledge matrix.

An improved knowledge graph for augmenting queries in a retrieval augmented generation system for generative artificial intelligence is constructed using entity data comprising information regarding a first entity of a first entity type and a second entity of a second entity type, the first and second entity types being defined by a domain-specific ontology for a knowledge domain. Relationship data for the knowledge graph comprises information regarding relationships between the first and second entities using relationship definitions from the domain-specific ontology. The knowledge graph is constructed by adding nodes corresponding to the entities and edges corresponding to the relationships. In some examples, the graph is used to identify cybersecurity threats applicable to a specific context.

The application provides an enhanced structured federated graph learning method, and belongs to the technical field of deep learning.The method comprises customer contribution evaluation, parameter adjustment and knowledge forgetting;the customer contribution evaluation evaluates customer contribution by using a reputation theory;the parameter adjustment adjusts key parameters by using an attention mechanism and an entropy weight method, reduces aggregation errors and optimizes global model performance;the knowledge forgetting processes data forgetting requests by combining soft confusion and hard confusion loss, and ensures that specific knowledge is forgotten without affecting overall performance.The application solves the data isolation problem in large-scale graph data training, and solves the problem of non-independent and identically distributed data and diversified local model features;even if the reliability of participants is low, the accuracy can still be guaranteed;and specific knowledge is removed from the client and propagated to the global model, which well responds to the forgetting request proposed by the user.The application significantly improves the accuracy of the global model, and better maintains the model precision after meeting the forgetting request.

The embodiment of the invention relates to a context-aware and stateless deep learning auto-tuning framework. Systems and methods are provided for improving auto-tuning procedures using stateless processing with a remote key value store. For example, the system may implement task launchers, schedulers, and agents to launch, schedule, and execute decomposed auto-tuning phases, respectively. Scheduling policies implemented by the scheduler may perform operations beyond simple scheduling policies (e.g., FIFO-based scheduling policies), which may result in high queuing latency. By utilizing auto-tuning domain-specific knowledge, queuing latency is reduced and resource utilization is improved compared to conventional systems.

The invention relates to a bilingual network threat intelligence relation extraction-oriented large language model optimization method and system, and the method comprises the steps: generating an enhanced data set according to an example set, and combining the example set and the enhanced data set to form a training set; the performance of a plurality of LLMs models is tested under various conditions, task performance and instruction following ability are extracted based on the relation of the models on a Chinese and English data set, and a basic model is screened out. And carrying out fine tuning on the screened model by adopting LoRA, and correcting the output by the LoRA through two trainable weight matrixes and by introducing an increment updating item containing task specific knowledge. Generating a dynamic instruction with a cluster exclusive feature for each test sample based on cluster affiliation, retaining a part of candidate examples with the highest correlation, screening out a part of examples with the highest correlation from the training set, constructing a sample subset under redundancy constraints, and performing cluster classification; and finally, fusing the cluster exclusive instruction with the demonstration examples for enhancing diversity in the sample subset by the cue word.

Methods, systems, and computer storage media for providing domain-integrated contextual response management using a domain-integrated contextual response engine in an artificial intelligence (AI) system are described. Domain-integrated contextual response management is a systematic approach that combines specific industry knowledge with contextual understanding to generate accurate, relevant, and specific industry-tailored responses to user queries. Domain-integrated contextual response management further includes fine-tuning models for Retrieval-Augmented Generation (RAG) tasks using customer-specific data based on a two-fold approach involving skill distillation and knowledge distillation (i.e., skill distillation from a more powerful model like Large Language Model “LLM” and knowledge distillation from domain-specific data). Domain-integrated contextual response management also includes creating a synthetic dataset that enables smaller models (e.g., domain-integrated contextual response models) to effectively manage RAG tasks while incorporating domain-specific knowledge. Domain-integrated contextual response management further ensures that the domain-integrated contextual response models can retrieve relevant information, support citations, and decline out-of-domain (OOD) questions.

The invention relates to a tunnel fire intelligent decision support method and device and a medium. The method comprises the following steps: constructing an ontology layer comprising a concept knowledge graph, a fact knowledge graph and a standard knowledge graph; the method comprises the following steps: acquiring tunnel fire related multi-source heterogeneous data, and extracting a structured triple and text semantics; using a graph database as a core storage medium, instantiating the structured triad into nodes, relationships and attributes according to the constructed ontology layer, and using text semantics as attributes or nodes to mount in a knowledge graph corresponding to the ontology layer; the method comprises the following steps: acquiring a natural language question input by a user, identifying a core intention, converting the natural language question into a standard graph query statement aiming at a specific knowledge graph level based on the type of the core intention, performing retrieval in a corresponding knowledge graph, generating a fire decision support answer according to a retrieval result, and outputting the fire decision support answer. Compared with the prior art, the method has the advantages that the efficiency and scientificity of tunnel disaster prevention design and emergency decision making are remarkably improved, and the like.

The present invention relates to the field of artificial intelligence technology, and specifically discloses a mutual learning method and system between multi-view medical images and text reports. By constructing a mutual learning model and dividing the learning process of the mutual learning model into an image reconstruction task, a report reconstruction task, and a multi-view alignment task, the image reconstruction task obtains the feature representation of the multi-view medical image, and the report reconstruction task obtains the feature representation of the text report and cross-modally fuses it with the image feature representation to obtain a reconstructed text report. The integration of multimodal reconstruction tasks enables the model to learn richer and more detailed feature representations, thereby improving its robustness to lost or damaged data. In addition, the exchange-based multimodal fusion method in cross-modal text reconstruction aims to fully integrate visual features and enrich the semantic representation of domain-specific knowledge. Through pre-training and joint training, the model performance is optimized so that more accurate and comprehensive diagnostic results can be obtained after applying the mutual learning model.

The invention belongs to the technical field of information, and particularly relates to a method for learning specific industry knowledge in a large model pre-training stage. Comprising the following steps of 1, cleaning and processing industry data; 2, pre-training industry knowledge; and step 3, knowledge increment pre-training. The method has the beneficial effects that the knowledge expression of the constructed industry is about 4.6 T, and the effect score in a downstream system generation task is improved by about 10% when the method is compared with a standard NST learning method. The specific content is as follows: for pre-trained industry knowledge, about 4.6 T data is constructed through ocr, data cleaning, manual annotation and the like;

The present disclosure provides a measurement application control unit for a measurement application, comprising: a primary database interface configured to be coupled to a primary database, where the primary database is configured to store generic knowledge data about the measurement application; a secondary database interface configured to couple to at least one secondary database, where the at least one secondary database is configured to store specific knowledge data about the measurement application; the system includes a primary database, at least one secondary database, a text-based input interface configured to receive a text-based user request for a measurement application to be executed, a request processor coupled to the primary database, the at least one secondary database, and the text-based input interface, the request processor is configured to retrieve and output database entries from the primary database and the at least one secondary database, the similarity of the database entries to the text-based user requests being above a predetermined threshold.

Techniques for constraining the results of a generative language model as valid information using knowledge-based documents. The generative language model may generate invalid results, including composite entities and incorrect entity relationships. The techniques include, for a given user query, determining a set of document information corresponding to the user query from a particular knowledge base. The technique further includes determining a sub-graph from a knowledge graph representing the knowledge base, and determining a prefix tree data structure representation of the set of document information. The user query and the sub-graph are provided as input to a trained generative language model for generating a response to the user query. The technique includes using the prefix tree data structure to verify that the generated response corresponds to real information in the set of document information.

This invention discloses a method and system for constructing a knowledge graph for personalized health consultation, relating to the field of medical information processing technology. The method includes: acquiring user static profile data and generating user feature vectors; based on a global medical knowledge graph, calculating node relevance scores using user feature vectors to filter and generate a user-specific knowledge subgraph; acquiring user dynamic health data, converting it into temporary nodes, and establishing associations with the specific knowledge subgraph to generate a personalized fusion knowledge graph; receiving user consultation questions and performing multi-hop reasoning on the personalized fusion knowledge graph to obtain reasoning paths; generating personalized consultation answers based on the reasoning paths and answer templates; removing temporary nodes after the session ends, and updating the specific knowledge subgraph when the user's static profile changes. This invention constructs a personalized knowledge graph by fusing user static features and dynamic data, achieving accurate and efficient health consultation services, significantly improving the personalization of answers and reasoning efficiency.

The invention relates to a large model recognition method and system based on interaction characteristics, a terminal and a medium, and relates to the field of artificial intelligence, and the method comprises the steps: carrying out the multi-round interactive questioning of a to-be-tested large model, and collecting the answer statement of the to-be-tested large model; feature extraction is carried out on the basis of the answer statement, statement features are obtained, and the statement features comprise at least one of a semantic coherence score, the coverage rate of specific knowledge points and an evasion strategy for sensitive questions; obtaining high-dimensional features of the candidate large model; respectively calculating the similarity between the statement features and the high-dimensional features; and determining the type of the to-be-tested large model as the candidate large model corresponding to the maximum similarity. The method has the effects of realizing content security supervision, preventing false information propagation and guaranteeing digital ecology credible operation.

The invention provides a knowledge base semantic retrieval vector model suitability evaluation method and device based on reinforcement learning, and solves the problems disclosed in the background technology. A reinforcement learning mechanism is introduced, a query generation process is modeled into an intelligent agent, and the intelligent agent is enabled to interact with an evaluation environment and autonomously learn to generate high-value queries capable of maximumly distinguishing different model performances, so that deeper and more accurate evaluation of model suitability is realized. By introducing a reinforcement learning mechanism, the fundamental transformation from traditional static and open-loop evaluation to dynamic and closed-loop intelligent evaluation is realized. According to the method, a general test set is abandoned, and the evaluation query with the highest distinction degree can be adaptively generated aiming at a specific knowledge base, so that the performance difference between different vector models can be more efficiently and deeply revealed. The method is high in automation and objectivity, guarantees the fairness and reliability of an evaluation result, and finally provides unprecedented intelligent and data-driven decision support for a user to select an optimal model.

The application belongs to the field of natural language processing and parameter efficient fine-tuning, and particularly relates to a multi-domain question answering task fine-tuning method based on a low-rank adaptive matrix and a mixed expert, which comprises the following steps: loading a data set and dividing; extracting a sentence vector representation of the data set by using a sentence vector model, and obtaining a category number N by using a K-means clustering algorithm on the sentence vector; loading a pre-trained language model and freezing original model parameters; constructing N asymmetric low-rank expert modules and a routing module beside a specified structure in the model; inputting the sentence vector representation into the routing module, calculating expert weights by the routing module, and performing weighted summation on different experts; the multi-domain question answering task fine-tuning method comprises a low-rank adaptive matrix and a mixed expert module; the application significantly reduces the number of trainable parameters by using the low-rank adaptive matrix, and improves the training efficiency; the mixed expert module is used to learn the specific knowledge of different domain problems in a complex question answering task, and the generalization ability of the model is improved.

The application provides a ship design domain knowledge retrieval method and device based on a large language model, and belongs to the technical field of big data processing.The method comprises the following steps: performing feature extraction on a multi-modal ship data set to obtain a feature set, and constructing a first multi-modal knowledge graph based on the feature set; mapping data in the first multi-modal knowledge graph to an embedding space to obtain embedding vectors, obtaining semantic association relationships based on a similarity matrix between the embedding vectors of different modal data, integrating the semantic association relationships into the first multi-modal knowledge graph to obtain a second multi-modal knowledge graph; fusing the second multi-modal knowledge graph with a preset large language model to obtain a first large language model; adding specific knowledge in the ship design field to the first large language model to obtain a second large language model; and inputting a query sentence of a user into the second large language model to obtain a retrieval result.The application improves the accuracy of the retrieval result of the knowledge in the ship design field.