Method and related device for constructing vertical domain knowledge graph
Through technologies such as preset information extraction task programs, data analysis engines and automatic labeling models, the problem of data fragmentation and knowledge extraction in vertical knowledge graph construction is solved, and efficient and accurate knowledge graph construction is achieved to adapt to rapid development. Real-time and accuracy requirements in the field.
Patent Information
- Application Number
- CN202410931749.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-07-12
- Publication Date
- 2024-11-08
- Estimated Expiration
- 2044-07-12
AI Technical Summary
When building vertical knowledge graphs, the existing technology faces the problems of data fragmentation, difficulty in knowledge extraction and integration, and difficult to ensure real-time and accuracy of knowledge. Especially in the rapidly developing field, the existing methods have poor scalability and good performance. Rule maintenance requirements are high, requires a large amount of labeled data, and lacks accuracy and real-timeness when processing fine-grained knowledge.
Using preset information extraction task programs, data analysis engines, automatic labeling models, similarity calculation and pattern matching technology, clustering technology, etc., the target vertical domain is determined through an interactive interface, data preprocessing, preliminary information labeling and correction are performed, and the propagation is carried out. Label information, perform data clustering, and finally generate a vertical knowledge graph.
The vertical knowledge graph construction process is optimized, and efficient and accurate knowledge graph construction is achieved in specific fields, improving the efficiency and accuracy of information extraction, adaptability and scalability, and ensuring real-time update and accuracy of knowledge.
Smart Images

Figure CN118504679B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field, and in particular to a method for constructing a vertical domain knowledge graph and related devices. Background Art
[0002] With the rapid development of information technology, the amount of data has exploded, especially in specific vertical fields, where unstructured data (such as text, pictures, videos, etc.) has taken a dominant position. However, the fragmented nature of these data and their inherent complexity have brought huge challenges to the extraction and integration of knowledge. First, how to accurately extract valuable information from massive, scattered unstructured data and transform it into structured knowledge is a technical problem; secondly, the real-time update and accuracy of knowledge are also challenges that must be faced when building vertical domain knowledge graphs, especially in some rapidly developing fields, such as medical health, where new knowledge and information are continuously generated.
[0003] Among the related technologies, there are currently a variety of methods and technologies used to construct vertical domain knowledge graphs, including rule-based information extraction, statistical learning methods, and deep learning models. These methods have their own advantages, but they also have shortcomings. For example, although rule-based methods can achieve high accuracy in specific tasks, they have poor scalability and high requirements for rule maintenance and updating; although statistical learning and deep learning methods can process large-scale data, they require a large amount of labeled data for training, and when faced with fine-grained knowledge in the field, accuracy and real-time performance are still difficult to overcome. To this end, how to effectively integrate these technologies to solve problems such as data fragmentation, difficulty in knowledge extraction and integration, and ensuring the real-time and accuracy of knowledge has become a key technical challenge that needs to be urgently solved in the field of vertical domain knowledge graph construction. Summary of the invention
[0004] In order to overcome the problems existing in the related technologies, the present application provides a method for constructing a vertical domain knowledge graph and a related device thereof, so as to optimize the construction process of the vertical domain knowledge graph and efficiently and accurately construct a vertical domain knowledge graph in a specific field.
[0005] The first aspect of the present application provides a method for constructing a vertical domain knowledge graph, comprising:
[0006] Utilize a preset information extraction task program to perform information extraction and obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through a task interface.
[0007] Preprocessing the unstructured data from the diverse data sources through a data parsing engine to obtain processable unstructured data;
[0008] Performing preliminary information annotation on the processable unstructured data according to the automatic annotation model, and reviewing and correcting the structured preliminary information to obtain corrected structured information, the automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model;
[0009] Using similarity calculation and pattern matching technology, the existing annotation information is propagated among diverse data sources, and clustering technology is used to cluster data among diverse data sources to obtain target structured data;
[0010] Generate and export the vertical domain knowledge graph corresponding to the target vertical domain based on the target structured data.
[0011] A second aspect of the present application provides a device for constructing a vertical domain knowledge graph, including:
[0012] A first processing unit is used to perform information extraction using a preset information extraction task program to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts a target vertical domain of the knowledge graph and task information to be extracted through a task interface;
[0013] A second processing unit, configured to pre-process the unstructured data from the diversified data sources through a data parsing engine to obtain processable unstructured data;
[0014] A third processing unit is used to perform preliminary information annotation on the processable unstructured data according to the automatic annotation model, and to review and correct the structured preliminary information to obtain corrected structured information. The automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model.
[0015] A fourth processing unit is used to propagate the existing annotation information among the diverse data sources by using similarity calculation and pattern matching technology, and cluster the data among the diverse data sources by using clustering technology to obtain target structured data;
[0016] The fifth processing unit is used to generate and export the vertical domain knowledge graph corresponding to the target vertical domain based on the target structured data.
[0017] A third aspect of the present application provides an electronic device, including:
[0018] Processor; and
[0019] A memory having executable code stored thereon, which, when executed by the processor, enables the processor to execute the method for constructing a vertical domain knowledge graph as described above.
[0020] The fourth aspect of the present application provides a non-temporary machine-readable storage medium having executable code stored thereon. When the executable code is executed by a processor of an electronic device, the processor executes the method for constructing a vertical domain knowledge graph as described above.
[0021] The technical solution provided by this application may have the following beneficial effects:
[0022] The technical solution of the present application provides a method and related device for constructing a vertical domain knowledge graph. The present application uses a preset information extraction task program to perform information extraction to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through the task interface; the unstructured data of the diversified data source is preprocessed by the data parsing engine to obtain processable unstructured data; preliminary information annotation is performed on the processable unstructured data according to the automatic annotation model, and the structured preliminary information is reviewed and corrected to obtain corrected structured information, and the automatic annotation model is updated and optimized according to the user feedback model; similarity calculation and pattern matching technology are used to propagate existing annotation information among diversified data sources, and clustering technology is used to cluster data among diversified data sources to obtain target structured data; the vertical domain knowledge graph corresponding to the target vertical domain is generated and exported according to the target structured data. Through the present application, the construction process of the vertical domain knowledge graph can be optimized, and the vertical domain knowledge graph can be efficiently and accurately constructed in a specific field.
[0023] It should be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present application. BRIEF DESCRIPTION OF THE DRAWINGS
[0024] The above and other objects, features and advantages of the present application will become more apparent through a more detailed description of exemplary embodiments of the present application in conjunction with the accompanying drawings, wherein the same reference numerals generally represent the same components in the exemplary embodiments of the present application.
[0025] Figure 1 It is a schematic diagram of a method for constructing a vertical domain knowledge graph shown in an embodiment of the present application;
[0026] Figure 2 It is a structural schematic diagram of a device for constructing a vertical domain knowledge graph shown in an embodiment of the present application;
[0027] Figure 3 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application. DETAILED DESCRIPTION
[0028] The preferred embodiments of the present application will be described in more detail below with reference to the accompanying drawings. Although the preferred embodiments of the present application are shown in the accompanying drawings, it should be understood that the present application can be implemented in various forms and should not be limited by the embodiments described herein. On the contrary, these embodiments are provided to make the present application more thorough and complete, and to fully convey the scope of the present application to those skilled in the art.
[0029] The embodiments of the present application provide a method for constructing a vertical domain knowledge graph and related devices thereof, so as to optimize the construction process of the vertical domain knowledge graph and efficiently and accurately construct the vertical domain knowledge graph in a specific field, thereby opening up new possibilities for knowledge management and application in the specific field.
[0030] The technical solution of the embodiments of the present application is described in detail below with reference to the accompanying drawings.
[0031] Figure 1 It is a flow chart of the method for constructing a vertical domain knowledge graph shown in the embodiment of the present application. The embodiment of the present application can be applied to certain rapidly developing specific vertical fields. The method can be executed by the device for constructing a vertical domain knowledge graph provided in the embodiment of the present application, and the device can be integrated in various electronic devices such as computers and mobile phones.
[0032] See also Figure 1 , the embodiment of the present application provides a method for constructing a vertical domain knowledge graph, the method specifically comprising the following steps:
[0033] S101: Utilize a preset information extraction task program to perform information extraction to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through a task interface.
[0034] In the embodiment of the present application, the establishment of the above-mentioned preset information extraction task program includes the following specific steps:
[0035] Creating an information extraction project by obtaining user requirements, setting a preset information extraction task for the information extraction project, and dynamically adjusting the preset information extraction task;
[0036] Obtain user needs, and set relevant parameters for each information extraction task according to the user needs to adjust the information extraction strategy. The parameters include but are not limited to the supported languages of the information extraction task and the type of vertical domain knowledge graph corresponding to the target vertical domain.
[0037] It should be noted that a key feature of the embodiment of the present application is that it allows users to collaborate with the AI agent in the system through an interactive interface to clarify the vertical field of the knowledge graph they want to build, such as medical, legal or financial. This feature solves the limitations of the prior art in building knowledge graphs in specific fields and provides a flexible and efficient method for setting projects and tasks. Specific implementation method: In this system, the user first communicates with the AI agent through an interactive interface to determine the target vertical field of the knowledge graph. Subsequently, the user can create a new project and set a series of specific information extraction tasks for the project, such as named entity recognition, entity-relationship triple extraction, and event extraction. These tasks are the basis for the construction of the knowledge graph and cover all aspects of extracting key information from the text.
[0038] Furthermore, in order to adapt to the needs of different fields and users, the system allows users to customize a series of parameters for each task. This includes but is not limited to the supported languages of the task (such as English, Chinese, etc.), the type of target knowledge graph (such as open domain knowledge graph or specific domain knowledge graph), so as to ensure that the AI agent can adjust the information extraction strategy according to the specific needs of the user. In addition, the AI agent in the system not only acts as an intelligent agent to perform tasks, but also can interact deeply with users to provide suggestions and feedback on task configuration. This interaction is not only based on previous project experience, but also combined with the latest technological advances to help users optimize task parameters and improve the efficiency and quality of knowledge graph construction.
[0039] Furthermore, the system supports dynamic adjustment of task settings during project execution. Users can modify task parameters in real time based on intermediate results and AI agent feedback, such as adding new entity types or adjusting the accuracy requirements of relationship extraction. It should be noted that this function ensures the flexibility and adaptability of the knowledge graph construction process, so that the final result can meet the needs of users to the greatest extent.
[0040] S102: Preprocessing the unstructured data from the diverse data sources through a data parsing engine to obtain processable unstructured data.
[0041] In the embodiment of the present application, the above-mentioned preprocessing of the unstructured data of the diversified data sources by the data parsing engine to obtain processable unstructured data includes the following specific steps:
[0042] Obtaining unstructured data from the diverse data sources, wherein the unstructured data from the diverse data sources includes PDF documents, web page content, and social media posts;
[0043] The unstructured data from the diverse data sources are processed by the data parsing engine to obtain processable unstructured data.
[0044] In an embodiment of the present application, this step is a key step in the knowledge graph construction process, focusing on ensuring the quality and consistency of the input data. This step processes and prepares data in multiple formats and sources uploaded by users through a highly automated and intelligent data processing process, laying the foundation for efficient and accurate information extraction. This technical feature solves the challenges of processing multi-source and multi-format data and improves the efficiency and accuracy of data processing. Specific implementation method: The data uploaded by users may include PDF documents, web page content, social media posts and other formats. The system has designed a flexible data parsing engine that supports text data in formats including but not limited to PDF, HTML, TXT, etc. For non-text formats, such as images or text in PDF, the system uses OCR technology to convert them into processable text formats.
[0045] S103: Perform preliminary information annotation on the processable unstructured data according to the automatic annotation model, and review and correct the structured preliminary information to obtain corrected structured information. The automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model.
[0046] In the embodiments of the present application, the preliminary information annotation stage occupies a core position in the process of knowledge graph construction. The present invention uses advanced natural language processing technology and machine learning algorithms to automatically identify and mark entities, relationships and events in text. This technical feature significantly improves the efficiency and accuracy of information extraction, while providing structured preliminary information for manual verification and further construction of the knowledge graph. Specific implementation method: The system uses an improved version of a large language model, such as a model based on a cognitive model, for automatic annotation of zero-sample learning. This means that the system is able to perform tasks such as named entity recognition and relationship extraction based on the general knowledge and understanding ability of the language model without a model specially trained for a specific task.
[0047] It should be noted that the automatic annotation function of the system does not rely on training data for specific tasks and belongs to zero-sample learning. Even without pre-annotated samples, the system can directly extract effective information from new texts. This is achieved through the extensive knowledge and natural language understanding capabilities built into large language models, which greatly improves the scope of application and flexibility of the system.
[0048] In a specific embodiment, the system automatically generates pre-annotation results based on the set task parameters and pre-processed text data, which includes identifying entities in the text, extracting relationships between entities, and identifying related events and their attributes. The generated pre-annotation results will be presented in a format that is easy for manual review and modification, which is convenient for the next step of manual verification and annotation process. Furthermore, during the automatic annotation process, the system has designed a real-time feedback learning mechanism. The system will adjust the annotation strategy in real time based on the feedback from the manual verification stage, and continuously improve the accuracy and efficiency of automatic annotation through online learning.
[0049] In the embodiments of the present application, one of the core features is to use user feedback to continuously optimize and update the automatic annotation model within the system to improve the accuracy and efficiency of information extraction. This feedback-based model optimization and update mechanism ensures that the system can adapt to the ever-changing data environment and user needs, thereby maintaining high performance in long-term use. Specific implementation method: Through the online learning mechanism, the system can absorb the feedback provided by users in the data review and correction stage in real time, and dynamically adjust the automatic annotation algorithm. This mechanism enables the system to continuously learn and adapt, gradually reduce the number of incorrect annotations, and improve the accuracy of information extraction.
[0050] Furthermore, in this step, based on the specific feedback from users, the system can not only adjust the annotation rules, but also fine-tune the natural language processing model used internally. This includes adjusting model parameters, optimizing training data sets, etc. to better meet the needs of specific tasks or fields. This step is key to ensuring that the system can accurately match user expectations and actual needs. In addition, the system has designed a user feedback loop mechanism to continuously extract new knowledge and annotation criteria by collecting and analyzing user behavior and feedback during the verification and annotation process. These criteria are used to optimize the system's automatic annotation performance and ensure the consistency and accuracy of the annotation results with user expectations.
[0051] It should be noted that the system also has built-in performance monitoring and reporting functions, which can track key indicators such as the accuracy of automatic annotation and processing speed. The embodiment of the present application can self-adjust according to the monitoring results to achieve continuous improvement of performance. At the same time, the system will report the performance improvement to the user to enhance transparency and trust.
[0052] S104: Using similarity calculation and pattern matching technology, the existing annotation information is propagated among the diverse data sources, and clustering technology is used to cluster the data among the diverse data sources to obtain target structured data.
[0053] In the embodiments of the present application, the annotation sharing and text clustering stage improves annotation efficiency and management efficiency through intelligent technology, and is a key link in the process of building a high-quality knowledge graph. This stage uses similarity calculation and pattern matching technology to realize the automatic propagation of existing annotation information between different documents, and efficiently manages and processes large amounts of data through clustering technology. This technical feature effectively reduces duplication of work and improves the consistency and efficiency of annotation. Implementation method: The system uses advanced similarity calculation and pattern matching technology to automatically identify similar or identical entities and relationships in the text. Once an entity or relationship is annotated in a document, the system automatically propagates this annotation information to other documents containing the same or similar entities and relationships. This automated annotation sharing mechanism reduces the workload of repeated annotation and improves the efficiency and consistency of data processing.
[0054] It should be noted that through text analysis and machine learning algorithms, the system can cluster uploaded documents and organize them together according to the similarity of the content. This allows users to focus on labeling and verification work for specific topics or events, improving processing speed and accuracy. The implementation of the clustering function not only improves the management efficiency of labeling work, but also optimizes the way data is organized.
[0055] In this step, although the annotation propagation and document clustering process is mainly automatic, the system also provides sufficient user control options to allow users to adjust clustering parameters, reorganize document groups, or manually specify the annotation information propagation of specific documents. This design ensures the flexibility of the system and the adaptability to user needs. In addition, combined with the user's activity record and the completed annotation information, the system can intelligently recommend documents that may be related or unannotated entities and relationships. This intelligent recommendation system further improves work efficiency and the comprehensiveness of annotation.
[0056] In the embodiment of the present application, the data review and correction stage is an indispensable part of the process of building a high-quality knowledge graph. The embodiment of the present application enables users to directly participate in the process of information extraction, correct the results generated by automatic annotation, and supplement missing entities, relationships or events by providing an intuitive and easy-to-use interactive annotation interface. This technical feature solves the errors and omissions that may occur in the automatic annotation process and ensures the accuracy and completeness of information extraction. Specific implementation method: The interactive annotation interface provided by the system supports text highlighting, visual editing of entities and relationships, and detailed configuration of event attributes. This design greatly improves the efficiency and convenience of manual verification and annotation, allowing users to easily browse the results of automatic annotation and make corrections or supplements.
[0057] Furthermore, during the manual verification process, the system will provide intelligent suggestions based on existing data and the user's operation history, such as entity recognition that may be missed, relationship suggestions, etc. These intelligent suggestions help users complete the labeling work faster. At the same time, the system also has an automatic correction function that can identify and correct some common labeling errors, further improving the verification efficiency. In addition, the system not only records the user's verification and labeling operations, but also feeds these data back to the automatic labeling algorithm in real time for learning and optimization. This interactive learning mechanism ensures that the system can continuously adjust according to the user's specific needs and labeling habits, and gradually improve the accuracy and adaptability of automatic labeling.
[0058] It should be noted that for large projects, the system supports collaborative annotation mode, allowing multiple users to participate in the verification and annotation process of the same project at the same time. The system will synchronize the operations of different users in real time to ensure the consistency and accuracy of the data, and provide a conflict resolution mechanism to handle the annotation differences between different users.
[0059] S105: Generate and export a vertical domain knowledge graph corresponding to the target vertical domain based on the target structured data.
[0060] In the embodiment of the present application, in the final stage of knowledge graph construction, the graph generation link is a key step to ensure that the construction results meet high quality standards and can be effectively utilized.
[0061] Furthermore, after uploading the data, the system automatically performs preprocessing operations to ensure the cleanliness and consistency of the data, reduce noise interference in the subsequent information extraction stage, and improve processing efficiency. To this end, after obtaining processable unstructured data, it also includes:
[0062] The processable unstructured data is cleaned to reduce the interference of noise signals in the information extraction process, wherein the data cleaning includes: removing duplicate texts, converting character case, and removing specific symbols or interference information.
[0063] It should be noted that the removal of specific symbols or interference information may specifically be the removal of interference information such as advertising links and irrelevant format tags.
[0064] In a specific embodiment, for a multilingual project, the system automatically detects the language of the uploaded text and performs corresponding language conversion or a preprocessing process for a specific language according to the task requirements. It should be noted that this step ensures that language does not become an obstacle to information extraction and improves the applicability and flexibility of the system. In order to improve the accuracy of information extraction, the system will also standardize the text, which includes unifying the format of dates and numbers, parsing and standardizing place names and names, etc., to reduce the ambiguity of entity recognition. Through these meticulous preprocessing steps, the embodiment of the present application provides a solid foundation for subsequent automatic annotation and information extraction work.
[0065] Furthermore, in the final stage of knowledge graph construction, the verification phase is a key step to ensure that the construction results meet high quality standards and can be effectively used. Therefore, before generating and exporting the vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data, it also includes:
[0066] The target structured data is verified using a preset verification algorithm to ensure the efficiency and accuracy of the generated and exported vertical domain knowledge graph corresponding to the target vertical domain.
[0067] It should be noted that the embodiment of the present application, by introducing powerful result verification tools and diverse export format support, implements a detailed review of the annotated documents, verifies the accuracy and completeness of the information extraction, and exports the verified data as a knowledge graph. This technical feature ensures the practicality and wide application of the knowledge graph. Specific implementation method: The system provides an interactive result verification tool that allows users to review results through an intuitive interface. The tool supports highlighting annotation information, quickly locates possible errors or omissions, and provides intuitive editing functions, allowing users to easily modify or supplement annotation information. In addition, before the user submits for verification, the system automatically performs consistency and completeness checks. This includes ensuring that the annotation information in all documents is consistent with no contradictions or omissions, and complies with the preset knowledge graph structure and rules. This step significantly reduces the workload of manual review and improves the efficiency and accuracy of the verification process.
[0068] Furthermore, users can choose to export the annotation results to a variety of standard formats of the knowledge graph, including but not limited to RDF (Resource Description Framework), OWL (Web Ontology Language), etc. This ensures that the exported data is compatible with a wide range of applications and services, enhancing the practicality and application scope of the knowledge graph. In addition to data export, the system also provides a series of guidance resources to help users understand how to apply the exported knowledge graph to different scenarios, such as data analysis, machine learning model training, or online query services. This step maximizes the value of the knowledge graph and promotes the dissemination and utilization of knowledge.
[0069] Through the above description and illustration of the embodiments of the present application, the technical solution of the embodiments of the present application has the following main advantages:
[0070] 1) Efficiency and automation: Through the application of advanced automatic annotation technology and large language models, the efficiency of information extraction has been significantly improved, and the workload of manual annotation has been reduced. The automated data preprocessing and annotation propagation mechanism further improves the efficiency of the entire system.
[0071] 2) Accuracy and reliability: Combining manual verification and annotation with online learning mechanisms, the system can continuously optimize the accuracy of automatic annotation to ensure that the information extraction results are both accurate and reliable. In addition, the application of result verification tools ensures the high quality of the final knowledge graph.
[0072] 3) Flexibility and adaptability: The system supports a variety of information extraction tasks and document formats, and can adapt to the needs of different fields and languages. The adaptive language model optimization and model fine-tuning mechanism enable the system to automatically adjust according to specific project requirements, improving the system's adaptability.
[0073] 4) Data management and organization: Document clustering functions and intelligent recommendation systems help users effectively manage and organize large amounts of documents, improving the organization and targeting of data processing.
[0074] 5) Scalability and compatibility: The system supports the export of multiple knowledge graphs in standard formats, ensuring the wide application and compatibility of knowledge graphs and providing a flexible foundation for subsequent analysis and application.
[0075] The embodiment of the present application provides a method for constructing a vertical domain knowledge graph, which uses a preset information extraction task program to perform information extraction to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through the task interface; the unstructured data of the diversified data sources are preprocessed by the data parsing engine to obtain processable unstructured data; preliminary information annotation is performed on the processable unstructured data according to the automatic annotation model, and the structured preliminary information is reviewed and corrected to obtain corrected structured information, and the automatic annotation model is updated and optimized according to the user feedback model; similarity calculation and pattern matching technology are used to propagate existing annotation information among diversified data sources, and clustering technology is used to cluster data among diversified data sources to obtain target structured data; the vertical domain knowledge graph corresponding to the target vertical domain is generated and exported according to the target structured data. The embodiment of the present application can optimize the construction process of the vertical domain knowledge graph, and efficiently and accurately construct the vertical domain knowledge graph in a specific field.
[0076] The present application embodiment also conducts experimental verification on the automatic construction and update system of the vertical domain knowledge graph to demonstrate its significant improvement in information extraction efficiency and accuracy. The experimental design is as follows:
[0077] 1. Experimental design
[0078] Test dataset: 1,000 scientific papers in the medical field were selected as the test dataset, which contains entities and relationships such as diseases, symptoms, drugs, and treatment methods.
[0079] Comparison group: Use the traditional manual annotation method as the baseline comparison.
[0080] Experimental group: applied the AI-assisted knowledge graph construction technical solution proposed in this invention.
[0081] 2. Experimental results
[0082] Labeling efficiency:
[0083] The time required for the control group (traditional method) to complete the labeling was 3,000 hours.
[0084] The time required for the experimental group (technical solution of the present invention) to complete the labeling is 600 hours.
[0085] Improved efficiency: Through the technical solution of the present invention, the labeling efficiency is improved by 80%.
[0086] Accuracy:
[0087] The average labeling accuracy of the comparison group was 90%.
[0088] The average labeling accuracy of the experimental group reached 95%.
[0089] Accuracy improvement: The accuracy rate increased by 5 percentage points.
[0090] User feedback:
[0091] In the experimental group, user feedback showed that the interactive annotation interface and collaborative annotation function significantly improved the convenience of annotation and user satisfaction.
[0092] 3. Experimental conclusion
[0093] The technical solution of this application significantly improves the efficiency and accuracy of vertical domain knowledge graph construction through automated pre-labeling, efficient human-computer interaction verification process and real-time learning optimization mechanism. Experimental results show that compared with traditional manual labeling methods, this application can significantly reduce manpower and time costs while improving the quality of information extraction, proving the effectiveness and practical value of this application.
[0094] It should be further explained that there are other alternatives to the embodiments of the present application, for example:
[0095] 1. Semi-automated project and task setting
[0096] In the alternative solution, the system provides templates and guided interfaces to help users quickly start new projects based on predefined project types and task settings. In addition, knowledge bases or domain expert systems can be introduced to recommend task settings, increasing the flexibility of manual adjustments.
[0097] 2. Cloud data processing and collaborative preprocessing
[0098] After data is uploaded, the system supports collaborative preprocessing in the cloud, allowing multiple users to edit and clean data online at the same time, enhancing the ability of teamwork. Cloud processing can also utilize distributed computing resources to accelerate the preprocessing of large-scale data.
[0099] 3. Automatic labeling of incremental learning
[0100] Incremental learning techniques are used to gradually train and optimize the automatic labeling model, allowing the system to learn instantly when receiving new data or feedback, without waiting for large batches of data to accumulate. This approach is more suitable for dynamically changing data environments.
[0101] 4. Crowdsourcing verification and annotation
[0102] In order to further improve the efficiency and coverage of manual verification, crowdsourcing technology is introduced in the alternative solution, allowing a wide range of user groups to participate in the verification and annotation process. Crowdsourcing can not only speed up the annotation work, but also improve the diversity of data and the quality of annotation.
[0103] 5. Rule-based information extraction
[0104] In some specific domains or tasks, automatic annotation and information extraction can be supplemented by rule-based methods, especially in structured data or documents with fixed formats. This method relies on domain knowledge and expert experience to define effective extraction rules.
[0105] 6. Intelligent annotation suggestion system
[0106] Combining machine learning and artificial intelligence technologies, we developed an intelligent annotation suggestion system that can automatically generate annotation suggestions based on document content and existing annotation data, helping users to quickly complete annotation work.
[0107] Corresponding to the aforementioned application function implementation method embodiment, the present application also provides a device, electronic device and corresponding embodiments for constructing a vertical domain knowledge graph.
[0108] Figure 2 Schematic diagram of the structure of the device shown in the embodiment of the present application. Figure 2 , the embodiment of the present application provides a device for constructing a vertical domain knowledge graph, the device specifically comprising:
[0109] The first processing unit 201 is used to perform information extraction using a preset information extraction task program to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through a task interface;
[0110] A second processing unit 202 is used to pre-process the unstructured data from the diversified data sources through a data parsing engine to obtain processable unstructured data;
[0111] The third processing unit 203 is used to perform preliminary information annotation on the processable unstructured data according to the automatic annotation model, and to review and correct the structured preliminary information to obtain corrected structured information. The automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model.
[0112] The fourth processing unit 204 is used to propagate the existing annotation information among the diverse data sources by using similarity calculation and pattern matching technology, and cluster the data among the diverse data sources by using clustering technology to obtain target structured data;
[0113] The fifth processing unit 205 is used to generate and export the vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data.
[0114] The first processing unit 201 is further configured to:
[0115] Creating an information extraction project by obtaining user requirements, setting a preset information extraction task for the information extraction project, and dynamically adjusting the preset information extraction task;
[0116] Obtain user needs, and set relevant parameters for each information extraction task according to the user needs to adjust the information extraction strategy. The parameters include but are not limited to the supported languages of the information extraction task and the type of vertical domain knowledge graph corresponding to the target vertical domain.
[0117] Furthermore, the second processing unit 202 is specifically configured to:
[0118] Obtaining unstructured data from the diverse data sources, wherein the unstructured data from the diverse data sources includes PDF documents, web page content, and social media posts;
[0119] The unstructured data from the diverse data sources are processed by the data parsing engine to obtain processable unstructured data.
[0120] The embodiment of the present application provides a device for constructing a vertical domain knowledge graph, wherein the first processing unit uses a preset information extraction task program to perform information extraction to obtain unstructured data from a variety of data sources, and the preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through the task interface; the second processing unit preprocesses the unstructured data of the diversified data source through the data parsing engine to obtain processable unstructured data; the third processing unit performs preliminary information annotation on the processable unstructured data according to the automatic annotation model, and reviews and corrects the structured preliminary information to obtain corrected structured information, and the automatic annotation model is updated and optimized according to the user feedback model; the fourth processing unit uses similarity calculation and pattern matching technology to propagate existing annotation information among diversified data sources, and clusters the data among diversified data sources through clustering technology to obtain target structured data; the fifth processing unit generates and exports the vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data. Through the embodiment of the present application, the construction process of the vertical domain knowledge graph can be optimized, and the vertical domain knowledge graph can be efficiently and accurately constructed in a specific field.
[0121] Regarding the device in the above embodiment, the specific manner in which each module performs the operation has been described in detail in the embodiment of the method, and will not be elaborated again here.
[0122] Figure 3 It is a schematic diagram of the structure of an electronic device shown in an embodiment of the present application.
[0123] See also Figure 3 , the electronic device 300 includes a memory 310 and a processor 320.
[0124] The processor 320 may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. A general-purpose processor may be a microprocessor or the processor may be any conventional processor, etc.
[0125] The memory 310 may include various types of storage units, such as system memory, read-only memory (ROM), and permanent storage devices. Among them, ROM can store static data or instructions required by the processor 320 or other modules of the computer. The permanent storage device may be a readable and writable storage device. The permanent storage device may be a non-volatile storage device that does not lose the stored instructions and data even after the computer is powered off. In some embodiments, the permanent storage device uses a large-capacity storage device (such as a magnetic or optical disk, flash memory) as a permanent storage device. In some other embodiments, the permanent storage device may be a removable storage device (such as a floppy disk, optical drive). The system memory may be a readable and writable storage device or a volatile readable and writable storage device, such as a dynamic random access memory. The system memory may store some or all instructions and data required by the processor at runtime. In addition, the memory 310 may include any combination of computer-readable storage media, including various types of semiconductor memory chips (DRAM, SRAM, SDRAM, flash memory, programmable read-only memory), and disks and / or optical disks may also be used. In some embodiments, the memory 310 may include a readable and / or writable removable storage device, such as a laser disc (CD), a read-only digital versatile disc (such as a DVD-ROM, a double-layer DVD-ROM), a read-only Blu-ray disc, an ultra-density optical disc, a flash memory card (such as an SD card, a mini SD card, a Micro-SD card, etc.), a magnetic floppy disk, etc. The computer-readable storage medium does not contain carrier waves and transient electronic signals transmitted wirelessly or wired.
[0126] The memory 310 stores executable codes, and when the executable codes are processed by the processor 320 , the processor 320 can execute part or all of the methods described above.
[0127] The devices in this article can be servers, PCs, PADs, mobile phones, etc.
[0128] The present application also provides a computer program product, which, when executed on a data processing device, is suitable for executing a program for initializing the following method steps:
[0129] Utilize a preset information extraction task program to perform information extraction and obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through a task interface.
[0130] Preprocessing the unstructured data from the diverse data sources through a data parsing engine to obtain processable unstructured data;
[0131] Performing preliminary information annotation on the processable unstructured data according to the automatic annotation model, and reviewing and correcting the structured preliminary information to obtain corrected structured information, the automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model;
[0132] Using similarity calculation and pattern matching technology, the existing annotation information is propagated among diverse data sources, and clustering technology is used to cluster data among diverse data sources to obtain target structured data;
[0133] Generate and export the vertical domain knowledge graph corresponding to the target vertical domain based on the target structured data.
[0134] The establishment of the preset information extraction task program includes:
[0135] Creating an information extraction project by obtaining user requirements, setting a preset information extraction task for the information extraction project, and dynamically adjusting the preset information extraction task;
[0136] Obtain user needs, and set relevant parameters for each information extraction task according to the user needs to adjust the information extraction strategy. The parameters include but are not limited to the supported languages of the information extraction task and the type of vertical domain knowledge graph corresponding to the target vertical domain.
[0137] The preprocessing of the unstructured data from the diversified data sources by the data parsing engine to obtain processable unstructured data includes:
[0138] Obtaining unstructured data from the diverse data sources, wherein the unstructured data from the diverse data sources includes PDF documents, web page content, and social media posts;
[0139] The unstructured data from the diverse data sources are processed by the data parsing engine to obtain processable unstructured data.
[0140] Preferably, after obtaining the processable unstructured data, the method further includes:
[0141] The processable unstructured data is cleaned to reduce the interference of noise signals in the information extraction process, wherein the data cleaning includes: removing duplicate texts, converting character case, and removing specific symbols or interference information.
[0142] Preferably, before generating and exporting the vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data, it also includes:
[0143] The target structured data is verified using a preset verification algorithm to ensure the efficiency and accuracy of the generated and exported vertical domain knowledge graph corresponding to the target vertical domain.
[0144] The scheme of the present application has been described in detail above with reference to the accompanying drawings. In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments. Those skilled in the art should also be aware that the actions and modules involved in the description are not necessarily required for the present application. In addition, it can be understood that the steps in the method of the embodiment of the present application can be adjusted in order, merged and deleted according to actual needs, and the modules in the device of the embodiment of the present application can be merged, divided and deleted according to actual needs.
[0145] In addition, the method according to the present application may also be implemented as a computer program or a computer program product, which includes computer program code instructions for executing some or all of the steps in the above method of the present application.
[0146] Alternatively, the present application can also be implemented as a non-temporary machine-readable storage medium (or computer-readable storage medium, or machine-readable storage medium) on which executable code (or computer program, or computer instruction code) is stored. When the executable code (or computer program, or computer instruction code) is executed by a processor of an electronic device (or electronic device, server, etc.), the processor executes part or all of the steps of the above-mentioned method according to the present application.
[0147] Those skilled in the art will further appreciate that the various illustrative logical blocks, modules, circuits, and algorithm steps described in connection with the application herein may be implemented as electronic hardware, computer software, or combinations of both.
[0148] The flow chart and block diagram in the accompanying drawings show the possible architecture, function and operation of the system and method according to multiple embodiments of the present application. In this regard, each square box in the flow chart or block diagram can represent a part of a module, a program segment or a code, and the part of the module, the program segment or the code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the square box can also occur in a sequence different from that marked in the accompanying drawings. For example, two continuous square boxes can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each square box in the block diagram and / or the flow chart, and the combination of the square boxes in the block diagram and / or the flow chart can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.
[0149] The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. The singular forms of "a", "said" and "the" used in this application and the appended claims are also intended to include plural forms unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used herein refers to and includes any or all possible combinations of one or more associated listed items.
[0150] It should be understood that although the terms "first", "second", "third", etc. may be used in this application to describe various information, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of this application, the first information may also be referred to as the second information, and similarly, the second information may also be referred to as the first information. Thus, the features defined as "first" and "second" may explicitly or implicitly include one or more of the features. In the description of this application, the meaning of "multiple" is two or more, unless otherwise clearly and specifically defined.
[0151] The embodiments of the present application have been described above, and the above description is exemplary, not exhaustive, and is not limited to the disclosed embodiments. Many modifications and changes will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the described embodiments. The selection of terms used herein is intended to best explain the principles of the embodiments, practical applications, or improvements to the technology in the market, or to enable other persons of ordinary skill in the art to understand the embodiments disclosed herein.
Claims
1. A method for constructing a vertical domain knowledge graph, characterized in that: include: Utilize a preset information extraction task program to perform information extraction and obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts the target vertical domain of the knowledge graph and the task information to be extracted through a task interface. Preprocessing the unstructured data from the diverse data sources through a data parsing engine to obtain processable unstructured data; Performing preliminary information annotation on the processable unstructured data according to the automatic annotation model, and reviewing and correcting the structured preliminary information to obtain corrected structured information, the automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model; Using similarity calculation and pattern matching technology, the existing annotation information is propagated among diverse data sources, and clustering technology is used to cluster data among diverse data sources to obtain target structured data; Generate and export a vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data; The establishment of the preset information extraction task program includes: Creating an information extraction project by obtaining user requirements, setting a preset information extraction task for the information extraction project, and dynamically adjusting the preset information extraction task; Obtain user needs, and set relevant parameters for each information extraction task according to the user needs to adjust the information extraction strategy. The parameters include but are not limited to the supported languages of the information extraction task and the type of vertical domain knowledge graph corresponding to the target vertical domain.
2. The method according to claim 1, characterized in that The preprocessing of the unstructured data from the diversified data sources by the data parsing engine to obtain processable unstructured data includes: Obtaining unstructured data from the diverse data sources, wherein the unstructured data from the diverse data sources includes PDF documents, web page content, and social media posts; The unstructured data from the diverse data sources are processed by the data parsing engine to obtain processable unstructured data.
3. The method according to claim 1, characterized in that After obtaining processable unstructured data, it also includes: The processable unstructured data is cleaned to reduce the interference of noise signals in the information extraction process, wherein the data cleaning includes: removing duplicate texts, converting character case, and removing specific symbols or interference information.
4. The method according to claim 1, characterized in that: Before generating and exporting the vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data, the method further includes: A preset verification algorithm is used to verify the target structured data to ensure the efficiency and accuracy of the generated and exported vertical domain knowledge graph corresponding to the target vertical domain.
5. A device for constructing a vertical domain knowledge graph, characterized in that: include: A first processing unit is used to perform information extraction using a preset information extraction task program to obtain unstructured data from a variety of data sources. The preset information extraction task program determines and adjusts a target vertical domain of the knowledge graph and task information to be extracted through a task interface; A second processing unit, configured to pre-process the unstructured data from the diversified data sources through a data parsing engine to obtain processable unstructured data; A third processing unit is used to perform preliminary information annotation on the processable unstructured data according to the automatic annotation model, and to review and correct the structured preliminary information to obtain corrected structured information. The automatic annotation model is updated and optimized according to the user feedback model, and the user feedback model is used to update and optimize the automatic annotation model. A fourth processing unit is used to propagate the existing annotation information among the diverse data sources by using similarity calculation and pattern matching technology, and cluster the data among the diverse data sources by using clustering technology to obtain target structured data; A fifth processing unit, configured to generate and export a vertical domain knowledge graph corresponding to the target vertical domain according to the target structured data; Wherein, the first processing unit is further used for: Creating an information extraction project by obtaining user requirements, setting a preset information extraction task for the information extraction project, and dynamically adjusting the preset information extraction task; Obtain user needs, and set relevant parameters for each information extraction task according to the user needs to adjust the information extraction strategy. The parameters include but are not limited to the supported languages of the information extraction task and the type of vertical domain knowledge graph corresponding to the target vertical domain.
6. The device according to claim 5, characterized in that The second processing unit is used for: Obtaining unstructured data from the diverse data sources, wherein the unstructured data from the diverse data sources includes PDF documents, web page content, and social media posts; The unstructured data from the diverse data sources are processed by the data parsing engine to obtain processable unstructured data.
7. An electronic device, characterized in that: include: processor; and a memory having executable codes stored thereon, which, when executed by the processor, causes the processor to execute the method for constructing a vertical domain knowledge graph as described in any one of claims 1 to 4.
8. A non-transitory machine-readable storage medium, characterized in that: An executable code is stored thereon, and when the executable code is executed by a processor of an electronic device, the processor executes the method for constructing a vertical domain knowledge graph as described in any one of claims 1-4.
