Knowledge question and answer rapid processing system based on artificial intelligence

Abstract

The invention discloses a knowledge question and answer rapid processing system based on artificial intelligence. The system comprises a knowledge graph construction module, a semantic entity analysis module, a user demand analysis module, a question and answer generation and interaction module, a back-end service interface module, a user feedback and optimization module and a security and privacy protection module. The knowledge graph construction module is used for constructing a knowledge graph; the semantic entity analysis module can identify semantic entities in questions of a user; the user demand analysis module can analyze the questioning demand of the user; the question and answer generation and interaction module is used for generating answers according to the recognized semantic entities and user requirements; the back-end service interface module provides an interface to realize load balancing and automatic capacity expansion and contraction; the user feedback and optimization module collects user evaluation and optimizes the model according to feedback; and the security and privacy protection module is used for protecting the privacy of user data and the security of the system. The user intention and entity information can be accurately recognized, and the response speed of the question answering system is remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a knowledge question and answer fast processing system based on artificial intelligence. Background Art

[0002] With the rapid development of the Internet, network data has exploded, and knowledge question-answering systems have been widely used as an important information retrieval tool. Traditional question-answering systems usually use a simple keyword matching model to retrieve relevant information from the database based on the keywords entered by the user and present the results to the user. However, this keyword matching-based question-answering system often cannot accurately understand the user's true intention, resulting in low accuracy and relevance of the retrieval results.

[0003] Although the existing technology has made certain progress, the traditional question-answering system still has many shortcomings. First, the question-answering system based on keyword matching often cannot accurately understand the user's true intention, resulting in low accuracy and relevance of the retrieval results. Secondly, when dealing with complex problems, the existing system lacks a deep understanding of the knowledge structure and has difficulty providing accurate answers. In addition, with the continuous increase in the amount of data, the existing system also faces challenges in query efficiency. Finally, the existing question-answering system has deficiencies in multimodal interaction, user privacy protection and system security, and it is difficult to meet the needs of modern users for efficient, secure and personalized services.

[0004] Prior art CN1821991A is a knowledge question and answer rapid processing system based on artificial intelligence. The system realizes the rapid processing of user questions and the push of answers by setting up components such as a communication platform, a knowledge base, a user base, and an artificial intelligence server. Among them, the artificial intelligence server has a built-in judgment engine, an intelligent classification engine, a similar question search engine, and an answer summary engine, which can classify user questions, search for similar questions, and summarize answers. The system also actively pushes questions and answers to users through a message push engine, thereby improving the real-time and initiative of information interaction. However, the system mainly focuses on information push through instant messaging tools and mobile communication platforms, and the intelligence level of the question and answer system and the in-depth mining of knowledge management still need to be improved.

[0005] Prior art CN111209379A proposes a knowledge question and answer fast processing system based on artificial intelligence, including a question processing module, an answer query module and a response processing module. Among them, the question processing module processes user questions through lexical analysis, sentence pattern recognition and content word extraction; the answer query module generates answers through answer extraction and reasoning; and the response processing module generates humanized answers through template matching. The system can use the types of questions that users often ask about objects as the attribute names of the objects, and use the text content that can answer the questions as the attribute values, so that the answers can be obtained efficiently and accurately. However, the system mainly focuses on the analysis of questions and the generation of answers through natural language processing technology, and the support for deep mining of knowledge management and multimodal interaction still needs to be improved.

[0006] At present, there is no public knowledge question and answer fast processing system that can deeply integrate artificial intelligence technology, knowledge graphs and multimodal interaction. Summary of the invention

[0007] The present invention provides a knowledge question and answer rapid processing system based on artificial intelligence to solve the existing technical problems of being unable to accurately understand user intentions, lacking in-depth understanding of knowledge structure, low query efficiency, multimodal interaction and insufficient security.

[0008] In order to solve the above technical problems, the present invention provides the following technical solutions: The present invention provides a knowledge question and answer fast processing system based on artificial intelligence, comprising: Knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module, backend service interface module, user feedback and optimization module, security and privacy protection module; The knowledge graph construction module is used to construct a knowledge graph containing entities, relationships and attributes, including ontology layer construction, data crawling, data preprocessing, knowledge extraction and data storage; The semantic entity parsing module uses the RoBERTa+Transformer+GlobalPointer model to process the user input text and can identify the semantic entities in the user's question; The user demand analysis module uses the RoBERTa+Transformer model to extract and classify the user input text, and can analyze the user's needs; The question-answer generation and interaction module is used to generate answers based on the identified semantic entities and user needs, including query statement generation, knowledge retrieval and reasoning, natural language generation, and multimodal interaction support; The backend service interface module provides a standardized RESTful interface and uses Kubernetes container orchestration technology to dynamically adjust resource allocation to achieve load balancing and automatic expansion and contraction; The user feedback and optimization module is used to collect user satisfaction evaluations on answers, and optimize the knowledge graph and question-answering model based on the feedback; The security and privacy protection module is used to protect the privacy of user data and the security of the system, including data encryption, access control, privacy protection, security auditing and system operation and maintenance.

[0009] Furthermore, the artificial intelligence-based knowledge question-answering rapid processing system adopts a layered architecture design, including a front-end interaction layer, a business logic layer, and a data management layer; The front-end interactive layer provides an intuitive user interface, supports seamless adaptation of desktop, mobile and web terminals, uses Bootstrap and jQuery frameworks to achieve a smooth interactive experience, and supports adaptive layout and multi-language display; The business logic layer is responsible for processing user requests, coordinating semantic entity resolution, user demand resolution, and question-answer generation and interaction modules, and supporting asynchronous processing and concurrent optimization; The data management layer interacts with the Neo4J graph database through a dedicated toolkit to support efficient data query and caching mechanisms.

[0010] The beneficial effects brought about by the technical solution provided by the present invention include at least: The artificial intelligence-based knowledge question-answering fast processing system of the present invention includes a knowledge graph construction module, a semantic entity resolution module, a user demand resolution module, a question-answering generation and interaction module, a back-end service interface module, a user feedback and optimization module, and a security and privacy protection module; the knowledge graph construction module is used to construct a knowledge graph containing entities, relationships, and attributes, including ontology layer construction, data crawling, data preprocessing, knowledge extraction, and data storage; the semantic entity resolution module adopts RoBERTa+Transformer+ The GlobalPointer model processes user input text and can identify semantic entities in user questions; the user demand analysis module uses the RoBERTa+Transformer model to extract and classify features of user input text, which can analyze the needs of user questions; the question and answer generation and interaction module is used to generate answers based on the identified semantic entities and user needs, including query statement generation, knowledge retrieval and reasoning, natural language generation, and multimodal interaction support; the back-end service interface module provides a standardized RESTful interface and uses Kubernetes container orchestration technology to dynamically adjust resource allocation to achieve load balancing and automatic scaling; the user feedback and optimization module is used to collect user satisfaction evaluations on answers, and optimize the knowledge graph and question and answer model based on feedback; the security and privacy protection module is used to protect the privacy of user data and the security of the system, including data encryption, access control, privacy protection, security auditing and system operation and maintenance.

[0011] Through deep learning models and knowledge graph technology, the present invention can accurately understand the semantics and intentions of user questions, thereby providing more accurate answers, significantly improving the accuracy and relevance of questions and answers, and reducing the time it takes for users to obtain effective information.

[0012] The system of the present invention supports multiple input methods such as text, picture, and voice, and can process multimodal data, which not only improves the user experience, but also meets the needs of different users in different scenarios, making the question-answering system more flexible and convenient.

[0013] The present invention supports efficient storage and query of large-scale data through the knowledge graph construction module, ensuring that the system can maintain efficient query performance when processing large amounts of data. The dynamic update and real-time synchronization functions of the knowledge graph ensure the timeliness and accuracy of knowledge.

[0014] The system of the present invention adopts a variety of security technologies to ensure the security and privacy of user data. Through differential privacy technology and security audit functions, user data is further protected from unauthorized access, thereby enhancing the credibility of the system.

[0015] The present invention adopts a layered architecture design and supports asynchronous processing and multi-threaded optimization, which not only improves the response speed of the system, but also facilitates the expansion and maintenance of the system. The system supports multi-terminal adaptation and can be seamlessly integrated into the existing IT environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0017] Figure 1 It is a system module structure diagram provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0018] In order to make the objectives, technical solutions and advantages of the present invention more clear, the embodiments of the present invention will be further described in detail below with reference to the accompanying drawings.

[0019] First embodiment This embodiment provides a knowledge question and answer fast processing system based on artificial intelligence, which is characterized by including: Knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module, backend service interface module, user feedback and optimization module, security and privacy protection module; The knowledge graph construction module is used to construct a knowledge graph containing entities, relationships and attributes, including ontology layer construction, data crawling, data preprocessing, knowledge extraction and data storage; The semantic entity parsing module uses the RoBERTa+Transformer+GlobalPointer model to process the user input text and can identify the semantic entities in the user's question; The user demand analysis module uses the RoBERTa+Transformer model to extract and classify the user input text, and can analyze the user's needs; The question-answer generation and interaction module is used to generate answers based on the identified semantic entities and user needs, including query statement generation, knowledge retrieval and reasoning, natural language generation, and multimodal interaction support; The backend service interface module provides a standardized RESTful interface and uses Kubernetes container orchestration technology to dynamically adjust resource allocation to achieve load balancing and automatic expansion and contraction; The user feedback and optimization module is used to collect user satisfaction evaluations on answers, and optimize the knowledge graph and question-answering model based on the feedback; The security and privacy protection module is used to protect the privacy of user data and the security of the system, including data encryption, access control, privacy protection, security auditing and system operation and maintenance.

[0020] Please refer to Figure 1 It is a diagram provided by an embodiment of the present invention.

[0021] For the AI-based knowledge question and answer fast processing system in this embodiment, its hardware deployment should include the knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module, backend service interface module, user feedback and optimization module, and security and privacy protection module: The knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module are core modules, and the back-end service interface module, user feedback and optimization module, and security and privacy protection module are auxiliary modules. The modules work together through data flow and interactive relationships to realize the overall function of the system.

[0022] 1. Knowledge graph construction module: Ontology layer construction refers to defining entities, attributes, concepts and the relationships between them, providing semantic representation and hierarchical constraints for the knowledge graph, and supporting dynamic expansion and updating; Data crawling supports multi-source data crawling, using a combination of request, Selenium and XPath to crawl the required structured or semi-structured data from the website. Multi-source data includes web pages, APIs, and files. Data preprocessing supports multilingual data processing, and can clean and preprocess crawled data. Cleaning includes removing useless tags, special characters, HTML entities, and processing missing values. Preprocessing includes word segmentation, part-of-speech tagging, and syntactic analysis of text data. Knowledge extraction can automatically extract entities, relationships, and attributes, support incremental updates and real-time synchronization, and convert pre-processed data into structured data of the knowledge graph; The data storage adopts a distributed storage architecture, storing structured data in the Neo4J graph database, supporting the storage and query of large-scale knowledge graphs, and having fault tolerance and backup mechanisms.

[0023] It should be noted that the ontology layer is the core framework of the knowledge graph. The ontology layer is designed as an extensible architecture that can dynamically add new entities, attributes, and relationships based on new knowledge fields or business needs, ensuring the flexibility and adaptability of the knowledge graph.

[0024] In the construction of the ontology layer of the knowledge graph: entities are specific knowledge objects; attributes are key-value pairs that describe entity characteristics and provide detailed information; concepts are the classification and abstraction of entities, forming a hierarchical structure; relationships are semantic connections between entities, representing the association between them.

[0025] An entity is a node in a knowledge graph, an object with a unique identifier, which can be a specific thing (person, place, organization, event) or an abstract concept (discipline, emotion). Each entity has a unique identifier and a set of attributes to describe its characteristics, and is connected to other entities through attributes and relationships to form a knowledge network.

[0026] Attributes are the characteristics or properties of an entity, which exist in the form of key-value pairs and can be simple data types (strings, numbers, dates) or complex data structures (lists, nested objects). Attributes provide a detailed description of the entity, enrich the semantic information of the knowledge graph, and can be used for query and reasoning to help the system better understand the characteristics of the entity.

[0027] Concepts are abstract classifications of entities, used to represent a group of entities with common characteristics. They can form a hierarchical structure and support semantic representation from concrete to abstract. Concepts provide a classification framework for entities, helping the system to better understand and organize knowledge. The concept hierarchy supports semantic reasoning and knowledge expansion.

[0028] A relationship represents a semantic association between entities, which can be direct or indirect, and has a type ("belongs to", "is located in", "associated with") and a direction ("from A to B"). Relationships are the core of the knowledge graph, forming a knowledge network by connecting entities and supporting semantic reasoning and knowledge query.

[0029] Combination of request, Selenium and XPath: request is used to send HTTP requests, Selenium is used to simulate browser behavior, and XPath is used to locate data in web pages; distributed crawler technology: improve crawling efficiency through distributed architecture and support rapid collection of large-scale data; anti-crawler strategy detection mechanism: can identify and respond to common anti-crawler technologies to ensure the stability of data crawling.

[0030] Remove useless tags and special characters: Use HTML parser (BeautifulSoup) to remove HTML tags, JavaScript codes and other useless special characters. After cleaning, only meaningful text content is retained.

[0031] Process HTML entities: Use the HTML decoding tool (HTMLParser) to convert HTML entities (&, <, >) into corresponding characters.

[0032] Handling missing values: Use decision tree algorithm to fill missing values. Mark or delete missing values ​​that cannot be filled to ensure data integrity.

[0033] Word segmentation: Use word segmentation tools (NLTK, spaCy, HanLP) to segment text, dividing continuous text into independent vocabulary units. It also supports word segmentation in multiple languages, including character-level segmentation in Chinese and word-level segmentation in English.

[0034] Part-of-speech tagging: Perform part-of-speech tagging on the text after word segmentation, mark the part of speech of each word (noun, verb, adjective), and use machine learning algorithm (hidden Markov model) for part-of-speech tagging.

[0035] Syntactic analysis: Perform syntactic analysis on the text to extract the grammatical structure of the sentence, including subject, predicate, object, and attributive, and use a deep learning model (Transformer) for syntactic analysis.

[0036] Automatic extraction of entities, relations, and attributes: Automatically identify and extract entities, relations, and attributes in text through deep learning models (BERT, BiLSTM). Incremental update and real-time synchronization: It supports dynamic update of knowledge graph and can synchronize newly extracted knowledge in real time to ensure the timeliness of knowledge graph.

[0037] Distributed storage architecture is a storage method that stores data in a dispersed manner on multiple nodes. It is used to process large-scale data sets and achieve efficient storage and query of data through the collaborative work of multiple nodes, while improving the scalability and reliability of the system.

[0038] Neo4J is a high-performance graph database designed specifically for storing and querying graph-structured data. It efficiently represents and queries complex relational networks through a model of nodes and relationships, making it very suitable for the storage and management of knowledge graphs.

[0039] Fault tolerance and backup mechanisms are used to ensure that the system can still operate normally and restore data when some nodes fail or data is lost. They are important components of distributed systems and ensure high availability of the system and reliability of data.

[0040] 2. Semantic Entity Parsing Module: The semantic entity resolution module uses the RoBERTa model as the word embedding layer, provides bidirectional semantic understanding capabilities, supports multi-language and domain adaptability, and obtains the semantic representation of the text; The semantic entity resolution module uses the Transformer model as the feature extraction layer to solve the long-distance dependency problem, capture contextual information, and support multi-layer stacking to enhance feature extraction capabilities; The semantic entity resolution module uses the GlobalPointer model as the feature classification layer, outputs annotations, implements named entity recognition, and supports custom annotation systems and domain-specific optimizations; The semantic entity resolution module has a data enhancement mechanism that improves the robustness and generalization ability of the model through random replacement, insertion, and deletion operations.

[0041] It should be noted that the semantic entity resolution module achieves high-performance named entity recognition by combining the advantages of RoBERTa, Transformer and GlobalPointer. RoBERTa provides powerful semantic representation, Transformer captures long-distance dependencies, GlobalPointer achieves accurate boundary detection, and the data enhancement mechanism improves the robustness and generalization ability of the model. This combination improves the accuracy of named entity recognition and ensures the efficiency and adaptability of the system.

[0042] The RoBERTa model is an improved version of the BERT model that can learn richer semantic representations through larger training data sets and longer training time.

[0043] Bidirectional semantic understanding: RoBERTa uses a bidirectional Transformer encoder that can simultaneously consider the previous and next information in the context, thereby providing more accurate semantic representation.

[0044] Multilingual support: RoBERTa provides a multilingual version (roberta-base-multilingual) that can process texts in multiple languages.

[0045] Domain adaptability: Through fine-tuning, RoBERTa can adapt to the named entity recognition task in a specific domain and improve the accuracy and adaptability of the model.

[0046] Transformer architecture: Based on the self-attention mechanism, it can capture long-distance dependencies in text.

[0047] Multi-layer stacking: Through multi-layer Transformer encoders, the model can gradually extract higher-level features and enhance the understanding of complex semantic structures.

[0048] Capturing contextual information: Transformer can simultaneously consider all positional information in the text and provide a more comprehensive contextual representation.

[0049] GlobalPointer is a span-based annotation method that recognizes named entities by predicting the start and end positions of the entities.

[0050] Customized annotation system: supports user-defined annotation systems that can adapt to the named entity recognition needs of different fields.

[0051] Domain-specific optimization: Through fine-tuning and domain data training, GlobalPointer can be optimized for specific domains to improve the accuracy and adaptability of the model.

[0052] Data augmentation mechanism is a method of artificially transforming training data or generating new data, which can increase data diversity and improve the robustness and generalization ability of the model.

[0053] Random Replacement: Randomly replace certain words in the text to simulate noisy data in real scenarios.

[0054] Random insertion: Randomly insert words into the text to increase the model’s robustness to inserted noise.

[0055] Random deletion: Randomly delete certain words in the text to improve the model's adaptability to missing data.

[0056] 3. User demand analysis module: The user demand parsing module uses the RoBERTa model as a feature extractor to learn the semantic representation of text, capture contextual semantic information, and support domain adaptability and multi-language processing; The user demand parsing module uses the Transformer model for further feature extraction and classification. It captures local and overall features of the text through a multi-head self-attention mechanism, extracts key features through pooling operations, and supports the capture of semantic features of different lengths. The user demand analysis module matches the demand analysis results with the entities and relationships in the knowledge graph to determine the user's specific needs for the entities.

[0057] It should be noted that the user demand parsing module achieves efficient and accurate user intent recognition by combining RoBERTa's high-performance semantic representation capabilities, Transformer's powerful feature extraction capabilities, and the semantic reasoning capabilities of the knowledge graph, thereby improving the accuracy of intent recognition and ensuring the efficiency and adaptability of the system, and can better meet user needs in different fields and language environments.

[0058] Multi-head self-attention mechanism: Multiple attention heads are used to process input data simultaneously, capture features in different subspaces, and enhance the model's ability to capture local and overall features.

[0059] Pooling operation: extract key features through maximum pooling or average pooling operations, reduce feature dimensions, and improve the generalization ability of the model.

[0060] Feature classification: The extracted features are input into the classifier (Softmax layer) and the classification results of user intent are output.

[0061] Knowledge graph matching: Match the demand analysis results with the entities and relationships in the knowledge graph, and further optimize the intent recognition results through semantic reasoning technology.

[0062] 4. Question and answer generation and interaction module: Query statement generation combines semantic entities and user needs to dynamically construct query instructions for the knowledge graph, supporting complex queries and multi-hop reasoning; Knowledge retrieval and reasoning is to execute queries in the knowledge graph, obtain relevant knowledge through the real-time reasoning engine, and support high-concurrency queries and cache optimization; Natural language generation refers to the use of natural language generation technology to convert query results into answers in natural language form, supporting multiple answer formats, including text, lists, and charts; Multimodal interaction support selects the answer form based on user preferences and input methods, and feeds the results back to the user through a standardized API. The answer forms include text, voice, and image.

[0063] It should be noted that semantic entities are combined with user needs: the entities (names, places, organizations) identified by the semantic entity resolution module and the intentions (queries, recommendations, explanations) identified by the user needs resolution module are used as key information for the query, and the entities and needs are converted into nodes and relationships in the graph database to generate precise query instructions.

[0064] Complex query support: supports multi-condition filtering, path finding, and subgraph matching operations.

[0065] Multi-hop reasoning: Through the depth-first search (DFS) or breadth-first search (BFS) algorithm of the graph database, it supports multi-hop relationship queries and can handle complex knowledge associations.

[0066] Dynamic construction: Dynamically generate query statements based on real-time user input and support real-time interaction.

[0067] Knowledge graph query: Use the query language (Cypher) of the graph database (Neo4J) to perform query operations.

[0068] Real-time reasoning engine: Through rule reasoning and machine learning methods, it automatically discovers implicit knowledge relationships and enriches query results.

[0069] High concurrency support: Adopts distributed architecture and load balancing technology to support high concurrency queries.

[0070] Cache optimization: Use cache technology (Redis) to cache hot data, reduce the overhead of repeated queries, and improve system performance; dynamically adjust the cache strategy according to the data update frequency and access pattern to ensure data consistency and efficiency.

[0071] Natural language generation: Use natural language generation technology to convert query results into answers in natural language. The specific steps are as follows: S1. Query result parsing: The query results obtained from the graph database (Neo4J) are usually structured data (nodes, relationships, attribute values). First, these structured data need to be parsed into an intermediate format for subsequent processing, specifically: Extract key information: Extract entities (nodes) and their attribute values ​​from the query results; extract the relationship between entities and their direction ("belongs to", "is located in", "is associated with"); for multi-hop query results, extract path information (entity A is connected to entity B through relationship X).

[0072] Data formatting: convert query results into key-value pairs or lists. Examples are: { "entity": "Entity name", "attributes": {"attribute1": "value1", "attribute2": "value2"}, "relations": [{"relation": "relation type", "target": "target entity"}] } S2. Natural language template matching: According to the type and structure of the query results, select the appropriate natural language template to convert the structured data into a natural language description, specifically: Template classification: select the template category according to the type of query result (single entity query, multi-entity relationship query, attribute query, etc.); Template filling: fill the parsed query result data into the template; Dynamic template generation: for complex query results (multi-hop relationships or list results), dynamically generate templates.

[0073] S3. Language optimization and polishing: Optimize the initially generated natural language answers to make them more natural, fluent, and in line with human language habits. Specifically: Grammar checking and adjustment: Check whether the generated sentences conform to grammatical norms and adjust the word order or word usage; Semantic enhancement: Supplement additional information or explanations based on context or domain knowledge; Multi-language support: For multi-language environments, translate the answers into the target language.

[0074] S4. Multimodal answer generation: In addition to text answers, it also supports presenting query results in the form of lists and charts to meet different user needs, specifically: Multiple answer formats: support text, list, and chart answer formats to meet the needs of different users. Text answers: directly display the answer content to users in natural language; list answers: for queries involving multiple results, the results are presented in list form; chart answers: for queries involving data comparison or trend analysis, generate charts (bar charts, line charts, pie charts), use chart generation tools (Matplotlib) to visualize data, and embed charts in answers.

[0075] User preference identification: Select the appropriate answer format based on the user's interaction history and preference settings.

[0076] Multimodal output: supports multiple answer forms such as text, voice, and image, and provides diverse interactive experiences through speech synthesis (TTS) and image generation technology. Text interaction: returns text answers in JSON format; voice interaction: converts text answers into voice answers through speech synthesis technology; image interaction: for chart answers, generates image files and returns them through API.

[0077] Standardized API: Feedback results to users through standardized interfaces such as RESTful API or GraphQL, supporting access from multiple clients.

[0078] 5. Backend service interface module: The backend service interface module provides a standardized RESTful interface, supports text input, image input, and voice input, meets multimodal interaction requirements, and can process multiple data formats and return corresponding results; The backend service interface module uses Kubernetes container orchestration technology to dynamically adjust resource allocation, achieve load balancing and automatic expansion and contraction, and support high-concurrency access and distributed deployment.

[0079] It should be noted that the RESTful interface adopts the REST architectural style, provides a standardized API interface through the HTTP protocol, and supports GET, POST, PUT, and DELETE operations.

[0080] Multimodal input support, including text input: receiving text data in JSON or XML format; image input: receiving image files (PNG, JPEG) or image URLs, and extracting text content through OCR technology; voice input: receiving voice files (MP3, WAV) or voice streams, and extracting text content through ASR (automatic speech recognition) technology.

[0081] Processing of multiple data formats: Supports processing of multiple input data formats and returns corresponding results. For output, it supports returning text answers in JSON or XML format; for voice answers, it supports returning voice streams through TTS (speech synthesis); it supports returning pictures or charts (bar charts, line charts) as answers.

[0082] Kubernetes container orchestration: Use Kubernetes to manage containerized applications and support the deployment, expansion, and management of containers.

[0083] Dynamic resource allocation: Dynamically adjust container resources according to the real-time load of the system to optimize resource utilization.

[0084] Load balancing: Through the load balancing mechanism of Kubernetes, user requests are evenly distributed to multiple backend service instances to ensure high availability of the system.

[0085] Automatic scaling: Automatically expand or shrink the number of container instances based on system load, ensuring that system performance is not affected under high load and saving resources under low load.

[0086] Distributed deployment: supports multi-node deployment, deploying the system on multiple physical or virtual nodes to improve the system's fault tolerance and scalability.

[0087] 6. User feedback and optimization module: The user feedback and optimization module collects users' satisfaction evaluations on answers, which include ratings, comments, and behavior data; The user feedback and optimization module uses reinforcement learning, supervised learning, and unsupervised learning algorithms to analyze user feedback and dynamically adjust the content of the knowledge graph and the parameters of the question-answering model; The user feedback and optimization module supports incremental learning and online updates, dynamically adjusts model parameters through real-time feedback mechanisms, optimizes system performance in real time, and ensures continuous system improvement; The user feedback and optimization module provides a visual interface to display optimization effects and user feedback statistics. The visual interface supports real-time data display, historical trend analysis and user interaction functions.

[0088] It should be noted that the data collection methods include front-end interaction, behavior analysis and data storage; front-end interaction: providing rating and comment functions in the user interface to collect user feedback; behavior analysis: collecting user behavior data through front-end tracking technology; data storage: storing the collected feedback data in the database to support subsequent analysis.

[0089] Rating: Users rate their satisfaction with the answer on a scale of 1 to 5, with 1 being the lowest and 5 being the highest.

[0090] Comment: A user's detailed evaluation of the answer, which can be a text description.

[0091] Behavioral data: user interaction behavior, including clicks, dwell time, and whether other answers are selected.

[0092] Reinforcement learning: Dynamically adjust the parameters of the question-answering model based on user satisfaction evaluation to optimize the quality of the answer.

[0093] Supervised learning: Use user feedback (ratings and reviews) as labels to retrain the question-answering model.

[0094] Unsupervised learning: Analyze user behavior data to discover potential user needs or problem patterns.

[0095] The specific steps of analyzing user feedback through reinforcement learning, supervised learning, and unsupervised learning algorithms and dynamically adjusting the content of the knowledge graph and the parameters of the question-answering model include: Step 1, data collection and preprocessing: collect user ratings, comments and behavior data, and clean and label them.

[0096] 1.1 Data collection: collecting user ratings, user comments, and behavior data; 1.2 Data preprocessing: remove invalid or abnormal data, convert user ratings and comments into labels that can be used for training, and mark high scores (4-5 points) as "satisfied" and low scores (1-2 points) as "unsatisfied".

[0097] 1.3 Feature extraction: Extract keywords and sentiment features from user comments; extract user interaction patterns from behavioral data.

[0098] Step 2: Apply supervised learning algorithms: Train models with labeled data to identify user satisfaction issues.

[0099] 2.1 Data Annotation,User feedback (ratings and comments) is used as labels and question-answer pairs as input features to,construct an annotated dataset.

[0100] 2.2 Model Training,Using the logistic regression algorithm and taking user satisfaction as the label, the model is trained to identify which answers are “satisfactory” and which are “unsatisfactory”.

[0101] 2.3 Application results, based on the results of the supervised learning model, identify "unsatisfactory" answers and analyze their causes (missing knowledge graph, inaccurate answers). For these problems, update the knowledge graph content or adjust the parameters of the question-answering model.

[0102] Step 3: Apply unsupervised learning algorithms: Discover potential problems and user needs through cluster analysis.

[0103] 3.1 Data Clustering,The unsupervised learning algorithm (K-Means) was used to perform cluster analysis on user feedback,,and cluster the user comments into the categories of “content relevance issue”, “knowledge missing”, and “unclear answer”.

[0104] 3.2 Pattern discovery: Analyze user behavior data to discover common user interaction patterns. Through clustering results, discover potential problems and needs in user feedback.

[0105] 3.3 Application Results Based on the results of clustering and pattern discovery, the content of the knowledge graph and the parameters of the question-answering model are dynamically adjusted to better handle the types of questions that users are concerned about.

[0106] Step 4: Apply reinforcement learning algorithm: dynamically adjust the question-answering strategy through the reward function.

[0107] 4.1 Environment definition, define the environment (Environment) of reinforcement learning, including state and action; state: user's query content, the state of the current knowledge graph, action: the answer generated by the question-answering model.

[0108] 4.2 Reward Function Design,Design a reward function (Reward Function), give rewards or punishments based on user feedback (ratings and behavior data), give positive rewards for high user ratings, and give negative rewards for user re-queries or low ratings.

[0109] 4.3 Model training,The question-answering model is trained using the reinforcement learning algorithm (Q-Learning), and the model dynamically adjusts the answer strategy based on user feedback to maximize the cumulative reward.

[0110] 4.4 Application Results,The reinforcement learning model dynamically adjusts the question-answering strategy,optimizes the accuracy of answers and user satisfaction,and updates the content of the knowledge graph based on the results of,reinforcement learning to make it more in line with user needs.

[0111] Step 5: Dynamic update and optimization: Use incremental learning and real-time feedback mechanisms to dynamically adjust the model and knowledge graph content.

[0112] 5.1 Incremental Learning,Using the incremental learning algorithm, the parameters of the question-answering model are updated in real time to avoid retraining the entire model.

[0113] 5.2 Knowledge graph update: Based on user feedback, the content of the knowledge graph is dynamically updated, including adding new entities or relationships, optimizing the description of existing entities, and deleting inaccurate or outdated knowledge.

[0114] 5.3 Real-time feedback mechanism, through the real-time feedback mechanism, the model parameters are dynamically adjusted to ensure that the system can quickly respond to changes in user needs.

[0115] Incremental learning: allows the model to dynamically update parameters based on new feedback data without retraining.

[0116] Online update: Receive user feedback in real time and dynamically adjust model parameters through real-time feedback mechanism to ensure continuous improvement of the system.

[0117] Real-time data display: Real-time display of user satisfaction evaluation (ratings, comments) and behavior data.

[0118] Historical trend analysis: Displays historical trends of user feedback to help analyze changes in system performance.

[0119] User interaction function: allows administrators to perform manual optimization through a visual interface, including adjusting knowledge graph content or model parameters.

[0120] 7. Security and Privacy Protection Module: Data encryption uses the AES encryption algorithm to encrypt and store user data, and uses the HTTPS protocol to encrypt during data transmission to prevent data leakage; Access control implements user authentication and authorization through the OAuth2.0 protocol, supporting fine-grained permission management; Privacy protection uses differential privacy technology to protect user privacy and prevent unauthorized access to user data; Security audit can record system operation logs, supporting real-time monitoring and post-event tracing; System operation and maintenance provides logging and monitoring functions, supporting real-time monitoring of system performance indicators, recording error logs, automatic alarm mechanisms, and log-based troubleshooting functions.

[0121] It should be noted that AES is a symmetric encryption algorithm with the characteristics of high security and fast encryption. Through AES encryption, even if the data is illegally accessed, the attacker cannot read the plaintext content. It is suitable for storage encryption of sensitive data such as user personal information and query records.

[0122] HTTPS encrypts data transmission through the SSL / TLS protocol to prevent data from being eavesdropped or tampered with during transmission. It is applicable to all interactions between users and the system, including query requests and answer returns.

[0123] OAuth 2.0 is an open standard authorization protocol that allows users to authorize third-party applications to access their resources without sharing passwords. It is used for user login and API access. It supports fine-grained permission management and can assign different access rights based on user roles or needs to ensure that users can only access authorized resources and prevent unauthorized access.

[0124] Differential privacy technology: Add noise during data processing and analysis to ensure that the impact of a single user's data on the analysis results is minimal. It is used to protect user privacy and prevent sensitive information of a single user from being inferred through data analysis.

[0125] Log recording: Records all operation logs of the system, including user operations, system operation status, and error information, which are used to monitor the system operation status in real time, facilitating subsequent tracing and troubleshooting.

[0126] Real-time monitoring and post-event tracing: Logs are monitored in real time through the log management system (ELK stack), which supports keyword search and time range query, so that abnormal behavior can be discovered in time and the cause of the problem can be quickly located.

[0127] System performance monitoring: Use monitoring tools (Prometheus, Grafana) to monitor system performance indicators in real time, including CPU usage, memory usage, and response time, to ensure that the system runs in an efficient state and to detect performance bottlenecks in a timely manner.

[0128] Automatic alarm mechanism: By setting monitoring thresholds, when system performance indicators exceed the normal range, alarms are automatically triggered (email, SMS notifications), which can promptly notify operation and maintenance personnel to handle abnormal situations and reduce system failure time.

[0129] Log-based troubleshooting: Log analysis tools (Kibana) can be used to quickly locate error logs and analyze the causes of failures, which can quickly solve problems and reduce system downtime.

[0130] The artificial intelligence-based knowledge question-answering rapid processing system in this embodiment adopts a layered architecture design, including a front-end interaction layer, a business logic layer, and a data management layer; The front-end interactive layer provides an intuitive user interface, supports seamless adaptation of desktop, mobile and web terminals, uses Bootstrap and jQuery frameworks to achieve a smooth interactive experience, and supports adaptive layout and multi-language display; The business logic layer is responsible for processing user requests, coordinating semantic entity resolution, user demand resolution, and question and answer generation and interaction modules, and supporting asynchronous processing and concurrency optimization; The data management layer interacts with the Neo4J graph database through a dedicated toolkit, supporting efficient data query and caching mechanisms.

[0131] In addition, it should be noted that the present invention can be provided as a method, an apparatus or a computer program product. Therefore, the embodiments of the present invention can take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, the embodiments of the present invention can take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0132] The embodiments of the present invention are described with reference to the flowcharts and / or block diagrams of the methods, terminal devices (systems), and computer program products according to the embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of the processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, an embedded processor, or other programmable data processing terminal device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing terminal device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0133] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing terminal device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured product including an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 These computer program instructions can also be loaded onto a computer or other programmable data processing terminal device, so that a series of operation steps are executed on the computer or other programmable terminal device to produce a computer-implemented process, so that the instructions executed on the computer or other programmable terminal device provide for implementing the process in the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0134] It should also be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. The terms "include", "comprises" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or terminal device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, article or terminal device. In the absence of further restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, article or terminal device including the elements.

[0135] Finally, it should be noted that the above is a preferred embodiment of the present invention. It should be pointed out that although the preferred embodiment of the present invention has been described, for those skilled in the art, once the basic creative concept of the present invention is known, several improvements and modifications can be made without departing from the principles of the present invention, and these improvements and modifications should also be regarded as the protection scope of the present invention. Therefore, the attached claims are intended to be interpreted as including the preferred embodiment and all changes and modifications that fall within the scope of the embodiments of the present invention.

Claims

1. A knowledge question and answer fast processing system based on artificial intelligence, characterized in that: include: Knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module, backend service interface module, user feedback and optimization module, security and privacy protection module; The knowledge graph construction module is used to construct a knowledge graph containing entities, relationships and attributes, including ontology layer construction, data crawling, data preprocessing, knowledge extraction and data storage; The semantic entity parsing module uses the RoBERTa+Transformer+GlobalPointer model to process the user input text and can identify the semantic entities in the user's question; The user demand analysis module uses the RoBERTa+Transformer model to extract and classify the user input text, and can analyze the user's needs; The question-answer generation and interaction module is used to generate answers based on the identified semantic entities and user needs, including query statement generation, knowledge retrieval and reasoning, natural language generation, and multimodal interaction support; The backend service interface module provides a standardized RESTful interface and uses Kubernetes container orchestration technology to dynamically adjust resource allocation to achieve load balancing and automatic expansion and contraction; The user feedback and optimization module is used to collect user satisfaction evaluations on answers, and optimize the knowledge graph and question-answering model based on the feedback; The security and privacy protection module is used to protect the privacy of user data and the security of the system, including data encryption, access control, privacy protection, security auditing and system operation and maintenance.

2. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module, backend service interface module, user feedback and optimization module, and security and privacy protection module include: The knowledge graph construction module, semantic entity parsing module, user demand parsing module, question and answer generation and interaction module are core modules, and the back-end service interface module, user feedback and optimization module, and security and privacy protection module are auxiliary modules. The modules work together through data flow and interactive relationships to realize the overall function of the system.

3. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The knowledge graph construction module is used to construct a knowledge graph containing entities, relationships and attributes, including ontology layer construction, data crawling, data preprocessing, knowledge extraction and data storage, where: The ontology layer construction refers to defining entities, attributes, concepts and the relationships between them, providing semantic representation and hierarchical constraints for the knowledge graph, and supporting dynamic expansion and updating; The data crawling supports multi-source data crawling, and uses a combination of request, Selenium and XPath to crawl the required structured or semi-structured data from the website. The multi-source data includes web pages, APIs, and files; The data preprocessing supports multilingual data processing and can clean and preprocess the crawled data. The cleaning includes removing useless tags, special characters, HTML entities, and processing missing values. The preprocessing includes word segmentation, part-of-speech tagging, and syntactic analysis of text data. The knowledge extraction can automatically extract entities, relationships and attributes, support incremental updates and real-time synchronization, and convert pre-processed data into structured data of the knowledge graph; The data storage adopts a distributed storage architecture, stores structured data in the Neo4J graph database, supports the storage and query of large-scale knowledge graphs, and has fault tolerance and backup mechanisms.

4. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The semantic entity parsing module uses the RoBERTa+Transformer+GlobalPointer model to process the user input text and can identify the semantic entities in the user's question, where: The semantic entity resolution module uses the RoBERTa model as a word embedding layer, provides bidirectional semantic understanding capabilities, supports multi-language and domain adaptability, and obtains the semantic representation of the text; The semantic entity resolution module uses the Transformer model as the feature extraction layer to solve the long-distance dependency problem, capture context information, and support multi-layer stacking to enhance feature extraction capabilities; The semantic entity resolution module uses the GlobalPointer model as the feature classification layer, outputs annotations, implements named entity recognition, and supports custom annotation systems and domain-specific optimizations; The semantic entity resolution module has a data enhancement mechanism, which improves the robustness and generalization ability of the model through random replacement, insertion, and deletion operations.

5. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The user demand analysis module uses the RoBERTa+Transformer model to extract and classify the user input text and analyze the user's needs, where: The user demand parsing module uses the RoBERTa model as a feature extractor to learn the semantic representation of text, capture contextual semantic information, and support domain adaptability and multilingual processing; The user demand parsing module uses the Transformer model for further feature extraction and classification, captures local and overall features of the text through a multi-head self-attention mechanism, extracts key features through a pooling operation, and supports semantic feature capture of different lengths; The user demand analysis module matches the demand analysis results with the entities and relationships in the knowledge graph to determine the user's specific needs for the entities.

6. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The question-answer generation and interaction module is used to generate answers based on the identified semantic entities and user needs, including query statement generation, knowledge retrieval and reasoning, natural language generation, and multimodal interaction support, wherein: The query statement generation is to combine semantic entities and user needs to dynamically construct query instructions for the knowledge graph, supporting complex queries and multi-hop reasoning; The knowledge retrieval and reasoning is to execute queries in the knowledge graph, obtain relevant knowledge through a real-time reasoning engine, and support high-concurrency queries and cache optimization; The natural language generation refers to using natural language generation technology to convert query results into answers in natural language form, supporting multiple answer formats, including text, lists, and charts; The multimodal interaction support is to select the answer form according to the user preference and input method, and feed back the result to the user through a standardized API. The answer form includes text, voice, and image.

7. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The backend service interface module provides a standardized RESTful interface and uses Kubernetes container orchestration technology to dynamically adjust resource allocation to achieve load balancing and automatic expansion and contraction, including: The backend service interface module provides a standardized RESTful interface, supports text input, image input and voice input, meets multi-modal interaction requirements, and can process multiple data formats and return corresponding results; The backend service interface module uses Kubernetes container orchestration technology to dynamically adjust resource allocation, achieve load balancing and automatic expansion and contraction, and support high concurrent access and distributed deployment.

8. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The user feedback and optimization module is used to collect user satisfaction evaluations on answers and optimize the knowledge graph and question-answering model based on the feedback, where: The user feedback and optimization module collects user satisfaction evaluations on answers, wherein the satisfaction evaluations include ratings, comments, and behavior data; The user feedback and optimization module uses reinforcement learning, supervised learning and unsupervised learning algorithms to analyze user feedback and dynamically adjust the content of the knowledge graph and the parameters of the question-answering model; The user feedback and optimization module supports incremental learning and online updating, dynamically adjusts model parameters through real-time feedback mechanism, optimizes system performance in real time, and ensures continuous improvement of the system; The user feedback and optimization module provides a visual interface to display optimization effects and user feedback statistics. The visual interface supports real-time data display, historical trend analysis and user interaction functions.

9. The artificial intelligence-based knowledge question and answer fast processing system as claimed in claim 1, characterized in that: The security and privacy protection module is used to protect the privacy of user data and the security of the system, including data encryption, access control, privacy protection, security auditing and system operation and maintenance, among which: The data encryption uses the AES encryption algorithm to encrypt and store user data, and uses the HTTPS protocol to encrypt during data transmission to prevent data leakage; The access control implements user identity authentication and authorization through the OAuth2.0 protocol and supports fine-grained permission management; The privacy protection adopts differential privacy technology to protect user privacy and prevent unauthorized access to user data; The security audit can record system operation logs to support real-time monitoring and post-event tracing; The system operation and maintenance provides log recording and monitoring functions, supports real-time monitoring of system performance indicators, recording error logs, automatic alarm mechanisms, and log-based troubleshooting functions.

10. The artificial intelligence-based knowledge question-answering rapid processing system according to claim 1, characterized in that: The artificial intelligence-based knowledge question-answering rapid processing system adopts a layered architecture design, including a front-end interaction layer, a business logic layer, and a data management layer; The front-end interactive layer provides an intuitive user interface, supports seamless adaptation of desktop, mobile and web terminals, uses Bootstrap and jQuery frameworks to achieve a smooth interactive experience, and supports adaptive layout and multi-language display; The business logic layer is responsible for processing user requests, coordinating semantic entity resolution, user demand resolution, and question and answer generation and interaction modules, and supporting asynchronous processing and concurrent optimization; the data management layer interacts with the Neo4J graph database through a dedicated toolkit, supporting efficient data query and caching mechanisms.

Images