Electric power standard scene application matching method and system based on knowledge calculation and service engine

The method and system leverage deep learning and natural language processing to enhance scene matching by parsing user requests, retrieving relevant knowledge, and applying hybrid reasoning for personalized matching, addressing the limitations of existing systems in understanding complex queries and user preferences, thereby improving accuracy and adaptability.

CN119066148APending Publication Date: 2024-12-03ELECTRIC POWER RES INST CHINA SOUTHERN POWER GRID CO LTD +1

Patent Information

Application Number
CN202410988246.8
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-07-23
Publication Date
2024-12-03

AI Technical Summary

Technical Problem

Existing intelligent services fail to accurately understand complex user queries, multi-layer logical relationships, and personalized preferences due to reliance on simple keyword matching and static recommendation strategies, ignoring contextual information and user behavior patterns.

Method used

A method and system utilizing deep learning-based natural language processing to parse user requests, retrieve relevant knowledge from structured and unstructured data, apply hybrid knowledge reasoning algorithms, and perform personalized matching based on user preferences and contextual information, with a feedback loop for optimization.

Benefits of technology

Improves the accuracy and adaptability of scene matching by understanding complex requests and user needs, enhancing user satisfaction in fields like information retrieval, intelligent customer service, and personalized recommendation.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119066148A_ABST
    Figure CN119066148A_ABST
Patent Text Reader

Abstract

The invention provides a scene matching method and system based on knowledge calculation and a service engine, and aims to match corresponding services or recommended contents by automatically analyzing and utilizing situational information of a user. The method involves collecting context information of a user, and performing entity linking and relation reasoning by using a knowledge graph, so as to determine services or recommended contents matched with the user context. In addition, the system further comprises a user interaction module which is used for receiving and processing feedback information of a user so as to continuously optimize the scene matching effect.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to artificial intelligence (AI) and RAG enhanced retrieval and information processing technology, and specifically, belongs to an intelligent service system based on knowledge computing, and especially focuses on a method and system for scene matching using deep learning and natural language processing technology. Background Art

[0002] In existing intelligent services, the processing of user queries usually relies on simple keyword matching and static recommendation strategies. However, this approach often fails to fully understand the real needs of users, especially in scenarios dealing with complex queries, multi-layer logical relationships, and personalized preferences. In addition, existing systems often ignore the impact of contextual information and user behavior patterns, resulting in low accuracy and satisfaction of matching results. Therefore, there is a need to develop an intelligent scenario matching method and system that can understand complex requests and consider context and user preferences. Summary of the invention

[0003] In order to solve the above problems, the present invention provides a method for matching power standard scenario applications based on knowledge computing and service engine, comprising the following steps: a. Obtain user requests, taking into account query content, contextual information, and user preferences; b. Use deep learning-based natural language processing technology to parse requests to extract key information; c. Retrieve relevant knowledge from a knowledge base containing structured and unstructured data; d. Apply hybrid knowledge reasoning algorithms to process complex queries and logical relationships; e. Perform personalized matching based on user preferences and contextual information; f. Output matching results and optimize the matching algorithm based on user feedback.

[0004] At the same time, the present invention also proposes a scene matching system, including a data interface module, a knowledge base management module, an information retrieval module, a knowledge reasoning module, a personalized matching module, an output and feedback module, and a learning and optimization module to realize the functions of the above method.

[0005] By combining deep learning, natural language processing and knowledge reasoning, the present invention improves the accuracy and adaptability of scene matching, can better meet user needs, and is particularly suitable for fields such as information retrieval, intelligent customer service, knowledge question and answer, and personalized recommendation. BRIEF DESCRIPTION OF THE DRAWINGS

[0006] In order to show the specific implementation of the present invention in detail, this article will be accompanied by a series of diagrams to assist in the explanation. The following are some possible drawings, which are not exhaustive, but are sufficient to help readers understand the key components and technical points of the present invention.

[0007] Figure 1 The overall solution of the present invention is shown.

[0008] Figure 2 The technical framework of RAG enhanced retrieval of the present invention is demonstrated.

[0009] Figure 3 : Demonstrates the query understanding logic of the information retrieval module Figure 4 : Demonstrates the information retrieval module algorithm logic of the present invention.

[0010] Figure 5 : Demonstrates the knowledge reasoning module solution of the present invention.

[0011] Figure 6 : A schematic diagram of user knowledge base retrieval interaction, which shows how users can search for relevant knowledge bases through graphical interface operations Figure 7 :Application example of knowledge reasoning module - recommendation based on user behavior, and recommendation of relevant standard data that users are interested in based on reasoning.

[0012] Figure 8 :Example of application of knowledge reasoning module - based on standard literature recommendation, relevant standard data is recommended based on standard content reasoning analysis.

[0013] Fig. 9 :Application example of knowledge reasoning module - based on user subscription recommendations, recommend relevant standard data that users are interested in based on reasoning.

[0014] These drawings are only examples of embodiments of the present invention, not all possible implementations. The actual invention may include more or different drawings to adapt to different application scenarios and technical details. All of these drawings are intended to help readers better understand the technical connotation and application effects of the present invention. DETAILED DESCRIPTION

[0015] In order to make the above-mentioned purposes, features and advantages of the present invention more obvious and easy to understand, the specific implementation methods of the present invention are described in detail below in conjunction with the drawings of the specification. Obviously, the described embodiments are part of the embodiments of the present invention, but not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary persons in the art without creative work should fall within the scope of protection of the present invention.

[0016] The power standard scenario application matching method and system based on knowledge computing and service engine is based on the standard knowledge base and combined with RAG enhanced retrieval and other technologies to serve the digital transformation of the entire process of enterprise standardization work, and build a digital ecosystem covering the entire life cycle of standard management, query, use and compilation. It manages standards well, uses standards well, and compiles standards well, and comprehensively serves technological innovation and digital power grid construction.

[0017] In addition to providing a scientific and technological project evaluation method, the present invention also designs a corresponding system to implement this method. The system mainly consists of the following modules: Standard search module: Based on RAG enhanced search, document service and knowledge service mode, it provides rich standard search and fragment search functions to meet the basic search needs of different users in different scenarios. Standard reading module: By combining technologies such as semantic analysis, it recommends related standards while reading the standards, and recommends related knowledge such as terminology based on user behavior. Knowledge subscription module: By clustering, analyzing, mining, and labeling relevant standard data, it supports users to make personalized subscriptions based on their personal interests, and the system recommends relevant knowledge based on knowledge service reasoning.

[0018] Knowledge recommendation module: Based on the constructed standard portrait (standard tag), user portrait and user subscription tag, the user portrait is updated in combination with the user log, and personalized recommendations based on user behavior, popularity, quality, deduplication and other logic are comprehensively implemented.

Claims

1. A method for matching power standard scenarios based on knowledge computing and service engine, characterized in that: The following steps are involved: Obtain user requests, including but not limited to voice input, text input, historical interaction records, and user profile information, to comprehensively consider query content, contextual information, and user preferences; Use deep learning-based natural language processing techniques, including word embedding, syntactic analysis, and semantic understanding, to parse requests to extract key information and implicit requirements; Retrieve the knowledge related to the request from a comprehensive knowledge base containing metadata, entity relationships and semantic structures using semantic indexing and approximate search strategies; Apply hybrid knowledge reasoning algorithms that combine rule-based reasoning, probability-based reasoning, and deep learning-based reasoning to handle multi-level and multi-dimensional queries and logical relationships; Perform personalized matching using multi-objective optimization algorithms based on user preferences, real-time context, historical behavior, etc. Output matching results, including priority ranking, explanations, and actionable recommendations, and use user feedback and behavior data to dynamically optimize the matching algorithm using reinforcement learning strategies.

2. The method according to claim 1, wherein the knowledge base also includes real-time data sources, such as social media, news sources, and online forums, to update rules and models in real time to adapt to changes in the information environment.

3. According to the method of claim 1, the knowledge reasoning algorithm further includes knowledge distillation and knowledge graph reasoning to improve the efficiency and accuracy of reasoning.

4. A scene matching system based on knowledge computing, comprising: The data interface module integrates multiple input methods, such as voice recognition, keyboard input, and sensor data, to receive and process user requests; The knowledge base management module is responsible for the storage, updating, cleaning and maintenance of knowledge to ensure the freshness and consistency of knowledge; The information retrieval module uses distributed retrieval technology and semantic similarity calculation to quickly locate relevant information from massive knowledge; The knowledge reasoning module combines the rule engine and the neural network model to handle complex query and reasoning tasks; Personalized matching module, using feature engineering and collaborative filtering technology to achieve personalized matching; Output and feedback module, which provides a multimodal interactive interface and collects user feedback and satisfaction data; The learning and optimization module adopts online learning, transfer learning and meta-learning strategies to continuously improve the matching algorithm and system performance.

5. The system according to claim 4, wherein the learning and optimization module also includes anomaly detection and self-healing functions to ensure system stability and reliability.

6. Application of the scene matching method and system described in any one of claims 4-5, especially in the fields of information retrieval, intelligent customer service system, knowledge question and answer platform, personalized recommendation system, smart home and autonomous driving.

Citation Information

Patent Citations

  • Personal intelligent assistant system and data processing method

    CN111177330A

  • Intelligent question and answer interaction system based on semantic analysis engine

    CN117056479A

  • Medical literature intelligent question answering system and method based on RAG and LLM technologies

    CN118364088A

Cited By

  • Platform content identification method and system based on AIGC

    CN120493035A