Knowledge center summarization method and device, equipment, storage medium and program product
By obtaining collaborative discussion data from industrial software resource pools and generating knowledge summary documents using pre-trained big models, the problem of unstructured data management in weapon and missile design and development is solved, and efficient and accurate knowledge summary and storage is achieved.
Patent Information
- Application Number
- CN202411894654.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-12-20
- Publication Date
- 2025-04-15
AI Technical Summary
During the design and development of weapons and missiles, the accumulated unstructured data is difficult to systematically manage and utilize, resulting in waste of knowledge and repeated labor.
By obtaining collaborative discussion data from the industrial software resource pool and inputting it into the pre-trained model, a knowledge summary document is generated, and finally stored in the knowledge center.
It realizes automated sorting of unstructured data, improves the efficiency and accuracy of knowledge summary, reduces the workload of manual processing, and improves the accessibility and utilization efficiency of knowledge.
Smart Images
Figure CN119831018A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the technical field of data classification and grading, and in particular to a knowledge center summary method, device, equipment, storage medium and program product. Background Art
[0002] In the field of modern weapon missile research and design, with the rapid development of technology and the increasing complexity of the combat environment, the efficient use of information and the effective management of knowledge have become particularly important. At present, a large number of discussions, research reports and technical documents have been accumulated in the design and development of weapons and missiles. These unstructured data are scattered in different designer discussion groups and are difficult to manage and utilize in a systematic manner.
[0003] Traditional information processing methods often rely on manual collation and analysis, which is not only time-consuming and laborious, but also prone to missing important information. In addition, due to the lack of an effective knowledge sedimentation mechanism, many valuable experiences and insights have not been fully preserved and disseminated, resulting in a waste of knowledge and duplication of work. Summary of the invention
[0004] In order to solve the above technical problems or at least partially solve the above technical problems, the present application provides a knowledge center summary method, device, equipment, storage medium and program product.
[0005] In a first aspect, the present application provides a knowledge center summarizing method, which includes: obtaining collaborative discussion data from an industrial software resource pool; inputting the collaborative discussion data into a pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on collaborative discussion samples and their corresponding standard knowledge summary documents; and storing the knowledge summary document in a knowledge center.
[0006] As an optional implementation provided by the present application, collaborative discussion data is obtained from an industrial software resource pool, including: batch acquiring collaborative discussion data from the industrial software resource pool through crawler technology or an application programming interface.
[0007] As an optional implementation provided for the implementation of the present application, the training steps of the pre-trained large model include: constructing an initial model; inputting collaborative discussion samples into the initial model to obtain a predicted knowledge summary document output by the initial model; based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion samples, fine-tuning the model parameters of the initial model until the initial model converges to obtain the pre-trained large model.
[0008] As an optional implementation provided by the present application, based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, the model parameters of the initial model are fine-tuned until the initial model converges to obtain a pre-trained large model, including: using the predicted knowledge summary document output by the initial model as a negative sample for model training, and using the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization data set; identifying user preference features from the direct preference optimization data set through a direct preference optimization algorithm; based on the user preference features and a preset loss function, fine-tuning the model parameters of the initial model until the initial model converges to obtain a pre-trained large model.
[0009] As an optional implementation method provided for the implementation of the present application, the collaborative discussion data is text data, and accordingly, the pre-trained big model is a pre-trained text model; the collaborative discussion data is multimodal data, and accordingly, the pre-trained big model is a pre-trained multimodal model, and the standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for machines to understand the meaning of visual elements and their role in context.
[0010] As an optional implementation provided by the present application, the knowledge summary document is stored in the knowledge center, including: the knowledge summary document is stored in the knowledge center, the knowledge center assigns an identifier to each knowledge summary document, and establishes an association relationship with the collaborative discussion data.
[0011] In a second aspect, the present application provides a knowledge center summarizing device, the device comprising:
[0012] An acquisition module is used to obtain collaborative discussion data from the industrial software resource pool;
[0013] The knowledge summary document generation module is used to input the collaborative discussion data into the pre-trained large model to obtain the knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents;
[0014] The storage module is used to store the knowledge summary documents in the knowledge center.
[0015] As an optional implementation provided in the present application, the acquisition module is specifically used to: acquire collaborative discussion data in batches from the industrial software resource pool through crawler technology or application programming interface.
[0016] As an optional implementation provided by the present application, the device also includes a training module of a pre-trained large model, which is used to: construct an initial model; input the collaborative discussion sample into the initial model to obtain a predictive knowledge summary document output by the initial model; based on the predictive knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, fine-tune the model parameters of the initial model until the initial model converges to obtain the pre-trained large model.
[0017] As an optional implementation provided by the present application, the training module of the pre-trained large model is specifically used to: use the predicted knowledge summary document output by the initial model as a negative sample for model training, and use the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization data set; identify user preference features from the direct preference optimization data set through a direct preference optimization algorithm; and fine-tune the model parameters of the initial model based on the user preference features and a preset loss function until the initial model converges to obtain a pre-trained large model.
[0018] As an optional implementation method provided for the implementation of the present application, the collaborative discussion data is text data, and accordingly, the pre-trained big model is a pre-trained text model; the collaborative discussion data is multimodal data, and accordingly, the pre-trained big model is a pre-trained multimodal model, and the standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for machines to understand the meaning of visual elements and their role in context.
[0019] As an optional implementation provided by the present application, the storage module is specifically used to: store the knowledge summary document in the knowledge center, so that the knowledge center assigns an identifier to each knowledge summary document and establishes an association relationship with the collaborative discussion data.
[0020] In a third aspect, the present application provides an electronic device, comprising: a processor, a memory, and a computer program stored in the memory and executable on the processor, wherein when the computer program is executed by the processor, the knowledge center summarizing method as described in the first aspect or any one of its optional embodiments is implemented.
[0021] In a fourth aspect, the present application provides a computer-readable storage medium, comprising: a computer program is stored on the computer-readable storage medium, and when the computer program is executed by a processor, it implements the knowledge center summary method as described in the first aspect or any optional embodiment thereof.
[0022] In a fifth aspect, the present application provides a computer program product, including: the computer program product includes a computer program, and when the computer program runs on a computer, the computer implements the knowledge center summarizing method as described in the first aspect or any one of its optional embodiments.
[0023] Compared with the prior art, the technical solution provided by the embodiments of the present application has the following advantages:
[0024] The disclosed embodiment provides a knowledge center summary method, which first obtains collaborative discussion data from an industrial software resource pool; inputs the collaborative discussion data into a pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on collaborative discussion samples and their corresponding standard knowledge summary documents; and the knowledge summary document is stored in the knowledge center. In this way, the large model technology based on machine learning in this application can automatically extract key information from the collaborative discussion data in the industrial software resource pool, perform knowledge summary and document generation, realize the combing of non-structured data into structured data, improve the efficiency and accuracy of knowledge summary, reduce the workload of manual processing and improve the quality of documents; and store these high-quality knowledge summary documents in the knowledge center for subsequent query and use, thereby improving the accessibility and utilization efficiency of knowledge. BRIEF DESCRIPTION OF THE DRAWINGS
[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.
[0026] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.
[0027] Figure 1 A schematic diagram of a process flow of a knowledge center summary method provided in an embodiment of the present application;
[0028] Figure 2 A schematic diagram of the structure of a knowledge center summary device provided in an embodiment of the present application;
[0029] Figure 3 A schematic diagram of the structure of an electronic device described in an embodiment of the present application. DETAILED DESCRIPTION
[0030] In order to more clearly understand the above-mentioned purposes, features and advantages of the present application, the scheme of the present application will be further described below. It should be noted that the embodiments of the present application and the features in the embodiments can be combined with each other without conflict.
[0031] In the following description, many specific details are set forth to facilitate a full understanding of the present application, but the present application may also be implemented in other ways different from those described herein; obviously, the embodiments in the specification are only part of the embodiments of the present application, rather than all of the embodiments.
[0032] In order to improve the R&D efficiency and combat effectiveness of weapon systems, it is necessary to extract valuable information from a wide range of professional technologies and transform it into structured knowledge to support decision-making and technical improvements. At present, a large amount of discussions, research reports and technical documents have been accumulated in the design and development process of weapons and missiles. These non-institutional data are scattered in different designer discussion groups and are difficult to manage and utilize in a systematic manner. Traditional information processing methods often rely on manual collation and analysis, which is not only time-consuming and labor-intensive, but also prone to missing important information. In addition, due to the lack of an effective knowledge precipitation mechanism, many valuable experiences and insights have not been fully preserved and disseminated, resulting in a waste of knowledge and duplication of work.
[0033] In order to solve some or all of the technical problems existing in the related art, the embodiment of the present application provides a knowledge center summary method, device, equipment, storage medium and program product, wherein the method first obtains collaborative discussion data from the industrial software resource pool; inputs the collaborative discussion data into the pre-trained large model to obtain the knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; and the knowledge summary document is stored in the knowledge center. In this way, the large model technology based on machine learning in this application can automatically extract key information from the collaborative discussion data in the industrial software resource pool, perform knowledge summary and document generation, realize the combing of non-structured data into structured data, improve the efficiency and accuracy of knowledge summary, reduce the workload of manual processing and improve the quality of documents; and store these high-quality knowledge summary documents in the knowledge center for subsequent query and use, which improves the accessibility and utilization efficiency of knowledge.
[0034] A knowledge center summary method provided in the embodiments of the present application can be implemented by a knowledge center summary device or an electronic device, and the electronic device includes but is not limited to a vehicle terminal, a server, a personal computer, a laptop computer, a tablet computer, a smart phone, etc. The operating system of the electronic device may include Android, a mobile operating system (iOS) developed by Apple, an operating system (Windows) developed by Microsoft Corporation of the United States, etc., and the embodiments of the present application are not limited to this. The electronic device can be run alone to implement the present application, or it can be connected to the network and implement the present application through interactive operations with other computer devices in the network. Among them, the network in which the electronic device is located includes but is not limited to the Internet, a wide area network, a metropolitan area network, a local area network, a virtual private network (VPN) network, etc.
[0035] It should be noted that the protection scope of the knowledge center summary method described in the embodiment of the present application is not limited to the execution order of the steps listed in this embodiment. All solutions implemented by adding, reducing or replacing steps in the prior art based on the principles of the present application are included in the protection scope of the present application.
[0036] like Figure 1 As shown, Figure 1 1 is a flow chart of a knowledge center summarizing method according to an embodiment of the present application. The method can be executed by a knowledge center summarizing device, wherein the device can be implemented by software and / or hardware and can generally be integrated in an electronic device. Figure 1 As shown, the method mainly includes the following steps S101 to S104:
[0037] S101. Obtain collaborative discussion data from the industrial software resource pool.
[0038] Among them, the industrial software resource pool is a collection of industrial software related resources that are centrally stored and managed. These resources include industrial software itself, software components, data models, algorithm libraries, development tools, and related documents (including software user manuals, technical white papers, application program interface documents), etc. It is like a "software supermarket" that provides various required resources for industrial enterprises' software application, development, and maintenance activities. This application does not specifically limit the number of industrial software resource pools.
[0039] Collaborative discussion data includes, but is not limited to: technical exchange data, collaborative office data, problem-solving data, and research reports. Its content is difficult to understand and analyze directly. Collaborative discussion data can be any one or a combination of any multiple forms of text, voice, image, and video. It can be understood that collaborative discussion data can be single-modal text data or multimodal data. Collaborative discussion data can include code screenshots, technical diagrams, and other content. It should be noted that the industrial software resource pool in actual applications not only contains a large amount of text data, but also includes rich content such as pictures and videos. Especially in technical discussions and professional exchanges, experts and technicians often upload pictures to assist in illustrating their views or provide intuitive evidence. These visual information are crucial to a comprehensive understanding of the content of collaboration and discussion.
[0040] In some embodiments, collaborative discussion data is obtained in batches from an industrial software resource pool through crawler technology or an application programming interface (API). Collaborative discussion data is captured and sorted regularly through crawlers or APIs to ensure data diversity, integrity, and timeliness.
[0041] In some embodiments, after executing step S101, the collaborative discussion data is classified, formatted, and timestamped. Classifying the collaborative discussion data is helpful to improve data analysis efficiency. The unified format can avoid misunderstandings and errors caused by different data formats, facilitate operations such as time series analysis, and make batch processing easier. The timestamp can clarify the time when the data was generated, which is very important for data that needs to be processed in a timely manner. The timestamp can help trace the source and change history of the data. When there is a problem with the data, the timestamp can also be used to find out at which link the error occurred. The collaborative discussion data is preliminarily sorted to ensure the consistency and traceability of the collaborative discussion data.
[0042] In some embodiments, after executing step S101, the collaborative discussion data is preprocessed, including text cleaning, deduplication, format unification, etc. Text cleaning includes: removing irrelevant characters, punctuation marks, HTML tags, etc.
[0043] Collaborative discussion data may contain various noises, such as spelling errors, grammatical errors, extra punctuation, special characters, etc. By removing these noises through text cleaning, collaborative discussion data can be made more standardized and accurate, and data quality can be improved, thus providing a more reliable basis for subsequent data analysis and knowledge summarization. The cleaned text is also more conducive to the computer's understanding of its semantics. Collaborative discussion data from different sources may vary greatly in format and expression. Text cleaning can unify these texts into a relatively standard form, which is convenient for unified processing and analysis.
[0044] Deduplication can save a lot of storage resources, especially when processing massive amounts of collaborative discussion data, which can reduce storage costs. After deduplication, analysis can focus more on different text contents, improve analysis efficiency and accuracy, and obtain valuable analysis results more quickly. After deduplication, more realistic and objective word frequency statistics can be obtained, so as to more accurately understand the characteristics and patterns of the text.
[0045] After the unified format is adopted, these collaborative discussion data can be easily integrated together. The unified format also helps to avoid misunderstandings and mishandling caused by format differences. The unified format is easier to process in batches. The collaborative discussion data in the unified format can be stored quickly and accurately according to the pre-set database table structure, without spending a lot of time converting and adjusting data in different formats, thereby improving the overall efficiency of data processing.
[0046] S102. Input the collaborative discussion data into the pre-trained large model to obtain a knowledge summary document output by the pre-trained large model.
[0047] The knowledge summary documents output by the pre-trained large model are structured, making it easier for people to intuitively understand the meaning and relationship of the data, and facilitating subsequent analysis. The knowledge summary documents are more compatible with various data analysis tools and techniques to discover hidden patterns, trends, and associations from the knowledge summary documents.
[0048] Among them, the pre-trained large model is optimized and trained based on collaborative discussion samples and their corresponding standard knowledge summary documents. Collaborative discussion samples include text forms and multimodal forms. The standard knowledge summary document is a knowledge summary document that has been modified and improved by relevant personnel. Relevant personnel will check the key information, optimize the language expression, adjust the document structure, etc. to obtain a standard knowledge summary document. The content of the standard knowledge summary document is accurate and meets the format and quality requirements of the knowledge center (or knowledge management system). The knowledge center is a place or platform for centralized storage, management and dissemination of knowledge. It is a gathering place for knowledge resources. The knowledge management system (KMS) is a software system or platform that can collect, organize, store, share, innovate and apply internal and external knowledge of an enterprise or organization. The pre-trained large model is used to extract key information from unstructured collaborative discussion data, and then generate a structured knowledge summary document based on the key information. The pre-trained large model includes a pre-trained text model, or a pre-trained multimodal model.
[0049] In some embodiments, when executing step S102, if the collaborative discussion data is text data, the collaborative discussion data is input into a pre-trained text model;
[0050] If the collaborative discussion data is multimodal data, the collaborative discussion data is input into the pre-trained multimodal model.
[0051] In some embodiments, when training a pre-trained large model, an initial model is constructed, and collaborative discussion samples are input into the initial model to obtain a predicted knowledge summary document output by the initial model. Based on the loss value between the predicted knowledge summary document and the standard knowledge summary document, the model parameters of the initial model are fine-tuned until the initial model converges to obtain a pre-trained large model.
[0052] The loss value is based on the cross entropy loss function, as shown in formula (1):
[0053]
[0054] In formula (1), Indicates the generation of a standard knowledge summary document under the model parameter θ probability.
[0055] Based on the supervised fine-tuning mechanism (SFT), after the initial model has been unsupervisedly learned on a large-scale collaborative discussion sample, it is further trained using small-scale annotated data, so that the model can better adapt to the knowledge summarization task.
[0056] In some embodiments, the predicted knowledge summary documents output by the initial model are used as negative samples for model training, and the standard knowledge summary documents are used as positive samples for model training to construct a Direct Preference Optimization (DPO) dataset.
[0057] The direct preference optimization dataset is shown in formula (2):
[0058]
[0059] In formula (2), x i represents the i-th collaborative discussion sample, y i human represents the standard knowledge summary document, y i model Represents a predictive knowledge summary document.
[0060] Then, the user preference features are identified from the direct preference optimization data set through the direct preference optimization algorithm; based on the user preference features and the preset loss function, the model parameters of the initial model are fine-tuned until the initial model converges to obtain the pre-trained large model. The preset loss function is shown in formula (3):
[0061]
[0062] In formula (3), Is a binary function that represents the user's standard knowledge summary document and prediction knowledge summary document If the user prefers standard knowledge summary documents but otherwise,
[0063] DPO trains the model directly based on human preference information for model output, rather than relying on traditional clearly labeled data. This preference information can be obtained by manually sorting or comparing different outputs generated by the model.
[0064] The direct preference optimization algorithm compares the predicted knowledge summary document with the standard knowledge summary document to learn the user preference rules, identify the user preference characteristics, and clarify which content in the standard knowledge summary document the user prefers. Direct preference optimization can optimize the output results of the pre-trained large model, making the output knowledge summary document closer to the standard of manual editing and more in line with people's preferences, thereby improving the quality and accuracy of knowledge summaries.
[0065] Exemplarily, if the collaborative discussion data is text data, the pre-trained large model is a pre-trained text model. During model training, an initial text model is first constructed, and the collaborative discussion samples in text form are input into the initial text model to obtain a predicted knowledge summary document output by the initial model; then, the predicted knowledge summary document output by the initial text model is used as a negative sample for model training, and the standard knowledge summary document is used as a positive sample for model training to construct a direct preference optimization data set; user preference features are identified from the direct preference optimization data set through a direct preference optimization algorithm; based on user preference features and a preset loss function, the model parameters of the initial text model are fine-tuned until the initial text model converges to obtain the pre-trained large text model.
[0066] If the collaborative discussion data is multimodal data, relying solely on the text model cannot fully parse and utilize these non-textual data. To solve this problem, this application introduces a pre-trained multimodal model, which is pre-trained and fine-tuned on various types of collaborative discussion samples to achieve a comprehensive understanding of different modalities such as text, images, and videos. Collaborative discussion samples include but are not limited to: picture-text pairs, video-text pairs, and other related multimedia content.
[0067] Optionally, during model training, first construct an initial multimodal model, input the multimodal collaborative discussion samples into the initial multimodal model, and obtain the predicted knowledge summary document output by the initial model; then, use the predicted knowledge summary document output by the initial multimodal model as a negative sample for model training, and use the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization dataset; identify user preference features from the direct preference optimization dataset through a direct preference optimization algorithm; fine-tune the model parameters of the initial multimodal model based on the user preference features and a preset loss function until the initial multimodal model converges to obtain a pre-trained multimodal model. The preset function is shown in formula (4):
[0068]
[0069] In formula (4), S expert,i is the standard knowledge summary document corresponding to the i-th collaborative discussion sample, T i ,I i ,V i They are the corresponding text, picture and video inputs respectively.
[0070] By pre-training the model with multimodal collaborative discussion samples, we can learn the correlation between different modalities, for example, how to link the visual features in an image with the corresponding text description, or how to extract key frames from a video clip and match them with text information. After pre-training, in order to make the model better suited to specific tasks, it will be fine-tuned. For example, the initial multimodal model is asked to complete tasks such as image description and visual question answering, while understanding the input image and text instructions and outputting the corresponding content.
[0071] It should be noted that the standard knowledge summary document corresponding to the multimodal model includes visual element annotations, which are used by machines to understand the meaning of visual elements and their role in context. For example, for complex code screenshots or technical diagrams, experts will provide detailed explanations and annotations to help the model better understand the specific meaning of these images and their role in context, and guide the multimodal model to generate knowledge summary documents. Multimodal models are trained using multimodal collaborative discussion data and their corresponding standard knowledge summary documents (including visual element annotations) so that the model can learn how to more effectively integrate and understand different types of media information. Specifically, the model will learn how to extract key code snippets from code screenshots, how to parse important data points in technical diagrams, and how to combine these visual information with relevant text descriptions to generate more comprehensive and accurate knowledge summaries.
[0072] After this iterative training and optimization process, a converged pre-trained multimodal model is obtained. Applying this pre-trained multimodal model for knowledge summarization can improve the ability to understand multimodal information and generate richer and more accurate knowledge summaries.
[0073] The above embodiment organizes and standardizes collaborative discussion data through a pre-trained large model, and outputs a structured knowledge summary document with higher accuracy and reliability. The structured knowledge summary document is also easier to perform management operations such as data storage, retrieval and update.
[0074] S103. Store the knowledge summary document in the knowledge center.
[0075] In some embodiments, the knowledge summary document is stored in the knowledge center to achieve knowledge summary document storage management and quality control. The knowledge center assigns an identifier to each knowledge summary document and establishes an association with the collaborative discussion data to ensure data consistency and integrity.
[0076] Optionally, a database management system is used to efficiently manage structured knowledge summary documents, including backup, recovery, indexing, etc. of knowledge summary documents, thereby improving the security and availability of knowledge summary documents and reducing the cost and workload of knowledge summary document management.
[0077] The knowledge center will combine automated checks with manual review mechanisms to perform format checks, duplicate content detection, and factual consistency verification to ensure the quality of the knowledge summary documents stored in the database. Relevant personnel will regularly randomly check a certain proportion of knowledge summary documents to check their accuracy, completeness, and relevance. If any problems are found, such as inaccurate information or misleading content, they will feedback the problems through the review platform to optimize the pre-trained model, thereby improving the quality and reliability of the knowledge summary documents. Among them, the review platform provides users with a friendly interactive interface and supports functions such as adding comments, marking questions, and submitting feedback to improve review efficiency.
[0078] In summary, the embodiment of the present application provides a knowledge center summary method, which first obtains collaborative discussion data from the industrial software resource pool; inputs the collaborative discussion data into a pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on collaborative discussion samples and their corresponding standard knowledge summary documents; and the knowledge summary document is stored in the knowledge center. In this way, the large model technology based on machine learning in this application can automatically extract key information from the collaborative discussion data in the industrial software resource pool, perform knowledge summary and document generation, realize the combing of non-structured data into structured data, improve the efficiency and accuracy of knowledge summary, reduce the workload of manual processing and improve the quality of documents; and store these high-quality knowledge summary documents in the knowledge center for subsequent query and use, thereby improving the accessibility and utilization efficiency of knowledge.
[0079] like Figure 2 As shown, Figure 2 A schematic diagram of a knowledge center summary device provided in an embodiment of the present application, the device comprising:
[0080] An acquisition module 201 is used to acquire collaborative discussion data from an industrial software resource pool;
[0081] The knowledge summary document generation module 202 is used to input the collaborative discussion data into the pre-trained large model to obtain the knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents;
[0082] The storage module 203 is used to store the knowledge summary document in the knowledge center.
[0083] As an optional implementation provided in the present application, the acquisition module 201 is specifically used to: acquire collaborative discussion data in batches from an industrial software resource pool through crawler technology or an application programming interface.
[0084] As an optional implementation provided by the present application, the device also includes a training module of a pre-trained large model, which is used to: construct an initial model; input the collaborative discussion sample into the initial model to obtain a predictive knowledge summary document output by the initial model; based on the predictive knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, fine-tune the model parameters of the initial model until the initial model converges to obtain the pre-trained large model.
[0085] As an optional implementation provided by the present application, the training module of the pre-trained large model is specifically used to: use the predicted knowledge summary document output by the initial model as a negative sample for model training, and use the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization data set; identify user preference features from the direct preference optimization data set through a direct preference optimization algorithm; and fine-tune the model parameters of the initial model based on the user preference features and a preset loss function until the initial model converges to obtain a pre-trained large model.
[0086] As an optional implementation method provided for the implementation of the present application, the collaborative discussion data is text data, and accordingly, the pre-trained big model is a pre-trained text model; the collaborative discussion data is multimodal data, and accordingly, the pre-trained big model is a pre-trained multimodal model, and the standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for machines to understand the meaning of visual elements and their role in context.
[0087] As an optional implementation provided by the present application, the storage module 203 is specifically used to: store the knowledge summary document in the knowledge center, so that the knowledge center assigns an identifier to each knowledge summary document and establishes an association relationship with the collaborative discussion data.
[0088] For the specific definition of the knowledge center summary device, please refer to the definition of the knowledge center summary method above, which will not be repeated here. Each module in the above-mentioned knowledge center summary device can be implemented in whole or in part by software, hardware and their combination. The above-mentioned modules can be embedded in or independent of the processor in the computer device in the form of hardware, or can be stored in the memory of the computer device in the form of software, so that the processor can call and execute the operations corresponding to the above modules.
[0089] In one embodiment, the present application provides an electronic device, which may be a terminal, and its internal structure diagram may be as follows: Figure 3As shown. The electronic device includes a processor, a memory, a communication interface, a display screen and an input device connected via a system bus. Among them, the processor of the electronic device is used to provide computing and control capabilities. The memory of the electronic device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The communication interface of the electronic device is used to communicate with an external terminal in a wired or wireless manner, and the wireless manner can be achieved through WIFI, an operator network, near field communication (NFC) or other technologies. When the computer program is executed by the processor, a knowledge center summary method is implemented. The display screen of the electronic device can be a liquid crystal display screen or an electronic ink display screen, and the input device of the electronic device can be a touch layer covering the display screen, or a button, trackball or touchpad set on the housing of the electronic device, or an external keyboard, touchpad or mouse, etc.
[0090] Those skilled in the art will understand that Figure 3 The structure shown in the figure is merely a block diagram of a partial structure related to the scheme of the present application, and does not constitute a limitation on the electronic device to which the scheme of the present application is applied. Specifically, the electronic device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.
[0091] In one embodiment, the knowledge center summarizing device provided in the present application can be implemented in the form of a computer program. Figure 3 The memory of the electronic device may store various program modules constituting the knowledge center summary device, for example, Figure 2 The acquisition module 201, the knowledge summary document generation module 202 and the storage module 203 are shown. The computer program composed of various program modules enables the processor to execute the steps of the knowledge center summary method of each embodiment of the present application described in this specification.
[0092] For example, Figure 3 The electronic device shown can be Figure 2 In the knowledge center summary device shown, the acquisition module 201 executes to acquire collaborative discussion data from the industrial software resource pool; the electronic device can execute, through the knowledge summary document generation module 202, input the collaborative discussion data into the pre-trained large model to obtain the knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; the electronic device can execute, through the storage module 203, to store the knowledge summary document in the knowledge center.
[0093] In one embodiment, the present application provides an electronic device, including a memory and a processor, wherein the memory stores a computer program, and when the processor executes the computer program, the following steps are implemented:
[0094] Acquire collaborative discussion data from the industrial software resource pool; input the collaborative discussion data into the pre-trained big model to obtain the knowledge summary document output by the pre-trained big model; wherein the pre-trained big model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; store the knowledge summary document in the knowledge center.
[0095] In one embodiment, when the processor executes the computer program, it also implements the following steps: obtaining collaborative discussion data from the industrial software resource pool, including: obtaining collaborative discussion data in batches from the industrial software resource pool through crawler technology or an application programming interface.
[0096] In one embodiment, the processor also implements the following steps when executing the computer program: the training steps of the pre-trained large model include: constructing an initial model; inputting the collaborative discussion sample into the initial model to obtain a predicted knowledge summary document output by the initial model; based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, fine-tuning the model parameters of the initial model until the initial model converges to obtain the pre-trained large model.
[0097] In one embodiment, the processor also implements the following steps when executing the computer program: based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, fine-tuning the model parameters of the initial model until the initial model converges to obtain a pre-trained large model, including: using the predicted knowledge summary document output by the initial model as a negative sample for model training, and using the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization data set; identifying user preference features from the direct preference optimization data set through a direct preference optimization algorithm; and fine-tuning the model parameters of the initial model based on the user preference features and a preset loss function until the initial model converges to obtain a pre-trained large model.
[0098] In one embodiment, when the processor executes the computer program, the following steps are also implemented: the collaborative discussion data is text data, and accordingly, the pre-trained large model is a pre-trained text model; the collaborative discussion data is multimodal data, and accordingly, the pre-trained large model is a pre-trained multimodal model, and the standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for the machine to understand the meaning of the visual elements and their role in the context.
[0099] In one embodiment, when the processor executes the computer program, it also implements the following steps: storing the knowledge summary document in the knowledge center, including: storing the knowledge summary document in the knowledge center, assigning an identifier to each knowledge summary document by the knowledge center, and establishing an association relationship with the collaborative discussion data.
[0100] When the processor in the electronic device provided by the present application executes a computer program, it first obtains collaborative discussion data from the industrial software resource pool; inputs the collaborative discussion data into the pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; and stores the knowledge summary document in the knowledge center. In this way, the large model technology based on machine learning in the present application can automatically extract key information from the collaborative discussion data in the industrial software resource pool, perform knowledge summary and document generation, realize the combing of non-structured data into structured data, improve the efficiency and accuracy of knowledge summary, reduce the workload of manual processing and improve the quality of documents; and store these high-quality knowledge summary documents in the knowledge center for subsequent query and use, thereby improving the accessibility and utilization efficiency of knowledge.
[0101] In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and when the computer program is executed by the computer program, the following steps are implemented:
[0102] Acquire collaborative discussion data from the industrial software resource pool; input the collaborative discussion data into the pre-trained big model to obtain the knowledge summary document output by the pre-trained big model; wherein the pre-trained big model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; store the knowledge summary document in the knowledge center.
[0103] In one embodiment, when the computer program is executed by the computer program, the following steps are also implemented: obtaining collaborative discussion data from the industrial software resource pool, including: obtaining collaborative discussion data in batches from the industrial software resource pool through crawler technology or application program interface.
[0104] In one embodiment, the computer program also implements the following steps when executed by the computer program: the training steps of the pre-trained large model include: constructing an initial model; inputting collaborative discussion samples into the initial model to obtain a predicted knowledge summary document output by the initial model; based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion samples, fine-tuning the model parameters of the initial model until the initial model converges to obtain the pre-trained large model.
[0105] In one embodiment, the computer program also implements the following steps when executed by the computer program: based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, fine-tuning the model parameters of the initial model until the initial model converges to obtain a pre-trained large model, including: using the predicted knowledge summary document output by the initial model as a negative sample for model training, and using the standard knowledge summary document as a positive sample for model training to construct a direct preference optimization data set; identifying user preference features from the direct preference optimization data set through a direct preference optimization algorithm; based on the user preference features and a preset loss function, fine-tuning the model parameters of the initial model until the initial model converges to obtain a pre-trained large model.
[0106] In one embodiment, the computer program further implements the following steps when executed by the computer program: the collaborative discussion data is text data, and accordingly, the pre-trained large model is a pre-trained text model; the collaborative discussion data is multimodal data, and accordingly, the pre-trained large model is a pre-trained multimodal model, and the standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for the machine to understand the meaning of the visual elements and their role in the context.
[0107] In one embodiment, when the computer program is executed by the computer program, the following steps are also implemented: the knowledge summary document is stored in the knowledge center, including: the knowledge summary document is stored in the knowledge center, so that the knowledge center assigns an identifier to each knowledge summary document and establishes an association relationship with the collaborative discussion data.
[0108] When the computer program in the computer-readable storage medium provided by the present application executes the computer program, it first obtains collaborative discussion data from the industrial software resource pool; inputs the collaborative discussion data into the pre-trained large model to obtain the knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; and stores the knowledge summary document in the knowledge center. In this way, the large model technology based on machine learning in this application can automatically extract key information from the collaborative discussion data in the industrial software resource pool, perform knowledge summary and document generation, realize the combing of non-structured data into structured data, improve the efficiency and accuracy of knowledge summary, reduce the workload of manual processing and improve the quality of documents; and store these high-quality knowledge summary documents in the knowledge center for subsequent query and use, which improves the accessibility and utilization efficiency of knowledge.
[0109] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media that include computer-usable program code.
[0110] In several embodiments provided by the present application, it should be understood that the disclosed devices and methods can also be implemented in other ways. The device embodiments described above are merely schematic. For example, the flowcharts and block diagrams in the accompanying drawings show the possible architecture, functions and operations of the devices, methods and computer program products according to multiple embodiments of the present application. In this regard, each box in the flowchart or block diagram can represent a module, a program segment or a part of a code, and the module, a program segment or a part of a code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order from the order marked in the accompanying drawings. For example, two consecutive 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 box in the block diagram and / or flowchart, and the combination of boxes in the block diagram and / or flowchart can be implemented with a dedicated hardware-based system that performs a specified function or action, or can be implemented with a combination of dedicated hardware and computer instructions.
[0111] In the present application, the processor 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 also be any conventional processor, etc.
[0112] In this application, memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. Memory is an example of a computer-readable medium.
[0113] In this application, computer-readable media includes permanent and non-permanent, removable and non-removable storage media. Storage media can be implemented by any method or technology to store information, and the information can be computer-readable instructions, data structures, modules of programs or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, read-only compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. According to the definition in this article, computer-readable media does not include temporary computer-readable media (transitory media), such as modulated data signals and carrier waves.
[0114] It should be noted that, in this article, relational terms such as "first" and "second" 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. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article or 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 device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the existence of other identical elements in the process, method, article or device including the elements.
[0115] The above is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to these embodiments herein, but will conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A knowledge center summarization method, characterized in that: include: Acquire collaborative discussion data from the industrial software resource pool; Inputting the collaborative discussion data into the pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on the collaborative discussion samples and their corresponding standard knowledge summary documents; The knowledge summary document is stored in a knowledge center.
2. The method according to claim 1, characterized in that The obtaining of collaborative discussion data from the industrial software resource pool includes: obtaining the collaborative discussion data in batches from the industrial software resource pool through crawler technology or an application programming interface.
3. The method according to claim 1, characterized in that The training steps of the pre-trained large model include: Build an initial model; Inputting the collaborative discussion sample into the initial model to obtain a prediction knowledge summary document output by the initial model; Based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample, the model parameters of the initial model are fine-tuned until the initial model converges to obtain the pre-trained large model.
4. The method according to claim 3, characterized in that The fine-tuning of the model parameters of the initial model based on the predicted knowledge summary document and the standard knowledge summary document corresponding to the collaborative discussion sample until the initial model converges to obtain the pre-trained large model includes: The predicted knowledge summary documents output by the initial model are used as negative samples for model training, and the standard knowledge summary documents are used as positive samples for model training to construct a direct preference optimization dataset. identifying user preference features from the direct preference optimization data set by a direct preference optimization algorithm; Based on the user preference features and the preset loss function, the model parameters of the initial model are fine-tuned until the initial model converges to obtain the pre-trained large model.
5. The method according to claim 1, characterized in that The collaborative discussion data is text data, and accordingly, the pre-trained large model is a pre-trained text model; The collaborative discussion data is multimodal data, and accordingly, the pre-trained large model is a pre-trained multimodal model. The standard knowledge summary document corresponding to the pre-trained multimodal model includes visual element annotations, and the visual element annotations are used for machines to understand the meaning of visual elements and their role in context.
6. The method according to claim 1, characterized in that The storing the knowledge summary document in the knowledge center includes: The knowledge summary documents are stored in the knowledge center, so that the knowledge center assigns an identifier to each knowledge summary document and establishes an association relationship with the collaborative discussion data.
7. A knowledge center summarizing device, characterized in that: include: An acquisition module is used to obtain collaborative discussion data from the industrial software resource pool; A knowledge summary document generation module is used to input collaborative discussion data into a pre-trained large model to obtain a knowledge summary document output by the pre-trained large model; wherein the pre-trained large model is optimized and trained based on collaborative discussion samples and their corresponding standard knowledge summary documents; The storage module is used to store the knowledge summary document in the knowledge center.
8. An electronic device, characterized in that: include: A processor, a memory, and a computer program stored in the memory and executable on the processor, wherein the computer program, when executed by the processor, implements the knowledge center summarizing method according to any one of claims 1 to 6.
9. A computer-readable storage medium, characterized in that: include: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the knowledge center summarizing method according to any one of claims 1 to 6 is implemented.
10. A computer program product, characterized in that include: The computer program product comprises a computer program, and when the computer program is executed on a computer, the computer is enabled to implement the knowledge center summarizing method according to any one of claims 1 to 6.