Project process intelligent supervision platform based on large model
Through the project process intelligent supervision platform based on large models, the problems of low manual efficiency and difficulty in cross-domain collaboration in project management have been solved, real-time synchronization of multi-source data and decision optimization have been achieved, management efficiency has been improved and costs have been reduced.
Patent Information
- Application Number
- CN202510868816.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-26
- Publication Date
- 2025-10-21
AI Technical Summary
During the project management process, manual efficiency is low, cross-domain management collaboration is difficult, and cost balance is difficult.
A large-model-based intelligent project process supervision platform, including data layer, engine layer and application layer, is adopted to achieve real-time synchronization and decision optimization of heterogeneous data through multimodal data collection, understanding, risk prediction and decision optimization.
It achieves minute-level synchronization of multi-source data, improves the efficiency and accuracy of management decision-making, reduces transactional work, and reduces the complexity and cost of cross-domain collaboration.
Smart Images
Figure CN120822913A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of project management, and in particular to a project process intelligent supervision platform based on a large model. Background Art
[0002] In today's project management across various industries, the management complexity index is showing an exponential upward trend. For example, in terms of data dimension assurance, modern projects on average involve multimodal data such as requirement documents, meeting communications, code submission records, and hardware logs. Another example is the challenge of cross-domain collaboration. The time zone and cultural differences caused by distributed teams lead to a communication loss rate of up to 37%, accompanied by problems such as cost overruns, efficiency black holes, and innovation barriers.
[0003] With breakthroughs in big model capabilities, semantic understanding has undergone a qualitative leap. The big model's accuracy in requirements document parsing has reached 91.2%. Furthermore, the big model's multimodal retrieval capabilities across text, code, flowcharts, and other formats have been enhanced. Powered by agent technology, the system can autonomously complete the entire chain of tasks: "project analysis → risk identification → root cause analysis → solution generation." This represents a groundbreaking transformation for big models in project management.
[0004] Low manual efficiency, difficulty in cross-domain management collaboration, and difficulty in cost balance in the project management process are technical problems that need to be solved. Summary of the Invention
[0005] The technical task of the present invention is to address the above shortcomings and provide a project process intelligent supervision platform based on a large model to solve the technical problems of low manual efficiency, difficulty in cross-domain management collaboration, and difficulty in cost balance in the project management process.
[0006] The present invention provides a project process intelligent supervision platform based on a large model, comprising a data layer, an engine layer and an application layer;
[0007] The data layer is deployed with various data sources, including project management systems, code repositories, communication and collaboration platforms, and document knowledge bases;
[0008] The engine layer is used to perform data collection tasks and data analysis tasks. When performing data collection tasks, it obtains heterogeneous data from various data sources from the data layer, performs data preprocessing on the collected heterogeneous data, and stores the preprocessed data in the knowledge base. When performing data analysis tasks, it calls the large model based on the data stored in the knowledge base to perform risk prediction and decision optimization, and sends the risk prediction results and decision optimization results to the library table configured in the data layer;
[0009] The application layer is used to obtain risk prediction results and decision optimization results from the database table of the data layer, perform risk warnings and generate reports based on the risk prediction results and decision optimization results, and schedule data collection tasks and data analysis tasks.
[0010] Preferably, the engine layer includes a multimodal data acquisition module, a multimodal understanding engine, a knowledge base, a risk prediction module and a decision optimization engine;
[0011] The multimodal data acquisition module is used to perform data acquisition tasks, collect heterogeneous data from various data sources in the data layer, perform data preprocessing on the heterogeneous data, and store the processed data in the knowledge base; the multimodal understanding engine deploys multiple large models, which form a hybrid model architecture that integrates time series signals and semantic features;
[0012] The risk prediction module is used to obtain processed data from the knowledge base and call the adapted large model from the multimodal understanding engine, use the processed data as input, perform risk prediction based on the selected large model, and output a three-level risk warning with root cause analysis as the risk warning result;
[0013] The decision optimization engine is used to optimize decisions based on the risk warning results output by the risk prediction module, output decision optimization results, simulate and monitor the project through the digital twin environment, and calculate the optimal solution for project resource allocation under the current state through reinforcement learning.
[0014] Preferably, the large model deployed in the multimodal understanding engine includes an LSTM model for implementing temporal feature analysis, a Transformer model for semantic understanding, and a graph neural network model for establishing dependencies for each subtask in the project.
[0015] Preferably, the heterogeneous data collected by the multimodal data collection module includes task status change events collected from the project management system, code submission records collected from the code repository, meeting recordings collected from the communication and collaboration platform, and requirement documents collected from the document knowledge base;
[0016] Correspondingly, the multimodal data acquisition module is configured with a document parsing unit, a code analysis unit, a voice processing unit and a data processing unit;
[0017] The document parsing unit is used to parse the demand documents collected from the document knowledge base and extract tables, flow charts and version differences in PDF / Word documents based on the LayoutLMv3 model;
[0018] The code analysis unit is used to perform code analysis based on code submission records collected from the code repository, generate a cross-language AST using Tree-sitter, and calculate code quality indicators;
[0019] The speech processing unit is used to process the conference recordings collected from the communication and collaboration platform, convert the speech into text documents using speaker separation technology, and perform emotion recognition based on the text documents and speech intonation, calling a large model;
[0020] The data processing unit is used to take the text documents output by the document parsing unit, the code analysis unit and the speech processing unit as processed documents, and use the processed documents as input, extract document features through the HNSW algorithm, store the processed documents and document features in the knowledge base and build an index.
[0021] Preferably, the engine layer is used to store the pre-processed data in the knowledge base in an incremental update manner.
[0022] Preferably, the knowledge base includes a vector database and a graph database.
[0023] Preferably, the data layer, engine layer and application layer are encapsulated as pluggable modules, and the engine layer outputs risk prediction results and decision optimization results through WebSocket or grpc protocol.
[0024] Preferably, the application layer is configured with a risk warning dashboard, a report generation module and a task scheduling module;
[0025] The risk warning dashboard is used to display risk warning results and decision optimization results in the form of dynamic heat maps, and supports query services for risk warning results and decision optimization results;
[0026] The report generation module is used to generate reports based on risk warning results and decision optimization results, and supports viewing and downloading of reports;
[0027] The task scheduling module is used to schedule data collection tasks and data analysis tasks, so as to realize that the execution time of data collection tasks and data analysis tasks is in an idle state.
[0028] The project process intelligent supervision platform based on a large model of the present invention has the following advantages:
[0029] 1. It achieves real-time integration of multi-source data, enabling minute-level synchronization of heterogeneous data such as Jira tasks, GitLab code, and conference audio, eliminating the time-consuming and labor-intensive traditional manual methods.
[0030] 2. The basis for management's decision-making shifts from experience and intuition to quantitative analysis, allowing them to focus on high-value work (strategic coordination, stakeholder management) and reduce transactional work. BRIEF DESCRIPTION OF THE DRAWINGS
[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative work.
[0032] The present invention will be further described below with reference to the accompanying drawings.
[0033] Figure 1 This is a structural block diagram of a project process intelligent supervision platform based on a large model in Example 1. DETAILED DESCRIPTION
[0034] The present invention will be further described below with reference to the accompanying drawings and specific embodiments so that those skilled in the art can better understand the present invention and implement it. However, the embodiments given are not intended to limit the present invention. Unless there is a conflict, the embodiments of the present invention and the technical features in the embodiments may be combined with each other.
[0035] The embodiment of the present invention provides a project process intelligent supervision platform based on a large model, which is used to solve the technical problems of low manual efficiency, difficulty in cross-domain management collaboration, and difficulty in cost balance in the project management process.
[0036] Example:
[0037] The present invention provides a project process intelligent supervision platform based on a large model, comprising a data layer, an engine layer and an application layer.
[0038] Various data sources are deployed in the data layer, including project management systems, code repositories, communication and collaboration platforms, and document knowledge bases.
[0039] The engine layer is used to perform data collection tasks and data analysis tasks. When performing data collection tasks, it obtains heterogeneous data from various data sources from the data layer, preprocesses the collected heterogeneous data, and stores the preprocessed data in the knowledge base. When performing data analysis tasks, based on the data stored in the knowledge base, it calls the large model to perform risk prediction and decision optimization, and sends the risk prediction results and decision optimization results to the library table configured in the data layer.
[0040] In this embodiment, the engine layer includes a multimodal data acquisition module, a multimodal understanding engine, a knowledge base, a risk prediction module, and a decision optimization engine.
[0041] Among them, the multimodal data acquisition module is used to perform data acquisition tasks, collect heterogeneous data from various data sources in the data layer, preprocess the heterogeneous data, and store the processed data in the knowledge base; a variety of large models are deployed in the multimodal understanding engine, and the multiple large models form a hybrid model architecture that integrates time series signals and semantic features. In this embodiment, the large models deployed in the multimodal understanding engine include an LSTM model for implementing time series feature analysis, a Transformer model for semantic understanding, and a graph neural network model for establishing the dependencies of each subtask in the project; the risk prediction module is used to obtain processed data from the knowledge base and call the adapted large model from the multimodal understanding engine, using the processed data as input and performing risk prediction based on the selected large model, and outputting a three-level risk warning with root cause analysis as the risk warning result; the decision optimization engine is used to optimize decisions based on the risk warning results output by the risk prediction module, output decision optimization results, simulate and monitor the project through a digital twin environment, and calculate the optimal solution for project resource allocation in the current state through reinforcement learning.
[0042] As a specific implementation of the multimodal data collection module, the heterogeneous data it collects includes task status change events collected from the project management system, code submission records collected from the code repository, meeting recordings collected from the communication and collaboration platform, and requirement documents collected from the document knowledge base.
[0043] Correspondingly, the multimodal data acquisition module is configured with a document parsing unit, a code analysis unit, a speech processing unit and a data processing unit.
[0044] The document parsing unit is used to parse the requirement documents collected from the document knowledge base and extract tables, flow charts and version differences in PDF / Word documents based on the LayoutLMv3 model.
[0045] The code analysis unit is used to analyze code submission records collected from the code repository, generate cross-language ASTs using Tree-sitter, and calculate code quality indicators.
[0046] The speech processing unit is used to process the meeting recordings collected from the communication and collaboration platform, convert the speech into text documents using speaker separation technology, and perform emotion recognition based on the text documents and speech intonation, calling a large model.
[0047] The data processing unit is used to take the text documents output by the document parsing unit, the code analysis unit and the speech processing unit as processed documents, and use the processed documents as input to extract document features through the HNSW algorithm, store the processed documents and document features in the knowledge base in an incremental update manner and build an index.
[0048] The knowledge base includes a vector database and a graph database.
[0049] The application layer is used to obtain risk prediction results and decision optimization results from the database table of the data layer, issue risk warnings and generate reports based on the risk prediction results and decision optimization results, and schedule data collection tasks and data analysis tasks.
[0050] The application layer is configured with a risk warning dashboard, a report generation module, and a task scheduling module. The risk warning dashboard is used to display risk warning results and decision optimization results in the form of a dynamic heat map, and supports query services for risk warning results and decision optimization results; the report generation module is used to generate reports based on risk warning results and decision optimization results, and supports viewing and downloading of reports; the task scheduling module is used to schedule data collection tasks and data analysis tasks, so that the execution time of data collection tasks and data analysis tasks is in idle time.
[0051] In this embodiment, the data layer, engine layer and application layer are encapsulated as pluggable modules, and the engine layer outputs risk prediction results and decision optimization results through WebSocket or grpc protocol.
[0052] The platform disclosed in this embodiment utilizes a layered fusion architecture to process multi-source heterogeneous data, breaking through the bottlenecks of traditional single-modal processing. It utilizes technologies such as LayoutLMv3, Tree-sitter, and Whisper-X to achieve document understanding, code understanding, and speech processing capabilities. Compared to traditional single-modal systems, multimodal joint reasoning improves the completeness of requirements understanding by 65%, with outstanding performance in identifying discrepancies between document descriptions and code implementations (detection accuracy of 89%).
[0053] The platform uses incremental vectorization to achieve real-time vector data updates, integrating traditional BM25 with vector search to achieve context-aware retrieval capabilities. Compared to traditional daily batch updates, this approach offers the advantages of enhanced data timeliness and faster processing speeds. According to actual tests, incremental vector indexing supports processing over 2,000 knowledge unit updates per second.
[0054] The above is a detailed introduction to a project process intelligent supervision platform based on a large model provided by the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.
Claims
1. A project process intelligent supervision platform based on a large model, characterized by: Including data layer, engine layer and application layer; The data layer is deployed with various data sources, including project management systems, code repositories, communication and collaboration platforms, and document knowledge bases; The engine layer is used to perform data collection tasks and data analysis tasks. When performing data collection tasks, it obtains heterogeneous data from various data sources from the data layer, performs data preprocessing on the collected heterogeneous data, and stores the preprocessed data in the knowledge base. When performing data analysis tasks, it calls the large model based on the data stored in the knowledge base to perform risk prediction and decision optimization, and sends the risk prediction results and decision optimization results to the library table configured in the data layer; The application layer is used to obtain risk prediction results and decision optimization results from the database table of the data layer, perform risk warnings and generate reports based on the risk prediction results and decision optimization results, and schedule data collection tasks and data analysis tasks.
2. The project process intelligent supervision platform based on large model according to claim 1 is characterized in that: The engine layer includes a multimodal data acquisition module, a multimodal understanding engine, a knowledge base, a risk prediction module, and a decision optimization engine; The multimodal data acquisition module is used to perform data acquisition tasks, collect heterogeneous data from various data sources in the data layer, perform data preprocessing on the heterogeneous data, and store the processed data in the knowledge base; The multimodal understanding engine deploys a variety of large models, which form a hybrid model architecture that integrates time series signals and semantic features. The risk prediction module is used to obtain processed data from the knowledge base and call the adapted large model from the multimodal understanding engine, use the processed data as input, perform risk prediction based on the selected large model, and output a three-level risk warning with root cause analysis as the risk warning result; The decision optimization engine is used to optimize decisions based on the risk warning results output by the risk prediction module, output decision optimization results, simulate and monitor the project through the digital twin environment, and calculate the optimal solution for project resource allocation under the current state through reinforcement learning.
3. The project process intelligent supervision platform based on large model according to claim 2 is characterized in that: The large model deployed in the multimodal understanding engine includes an LSTM model for implementing time series feature analysis, a Transformer model for semantic understanding, and a graph neural network model for establishing the dependencies of each subtask in the project.
4. The project process intelligent supervision platform based on a large model according to any one of claims 1 to 3, characterized in that: The heterogeneous data collected by the multimodal data collection module include task status change events collected from the project management system, code submission records collected from the code repository, meeting recordings collected from the communication and collaboration platform, and requirement documents collected from the document knowledge base; Correspondingly, the multimodal data acquisition module is configured with a document parsing unit, a code analysis unit, a voice processing unit and a data processing unit; The document parsing unit is used to parse the demand documents collected from the document knowledge base and extract tables, flow charts and version differences in PDF / Word documents based on the LayoutLMv3 model; The code analysis unit is used to perform code analysis based on code submission records collected from the code repository, generate a cross-language AST using Tree-sitter, and calculate code quality indicators; The speech processing unit is used to process the conference recordings collected from the communication and collaboration platform, convert the speech into text documents using speaker separation technology, and perform emotion recognition based on the text documents and speech intonation, calling a large model; The data processing unit is used to take the text documents output by the document parsing unit, the code analysis unit and the speech processing unit as processed documents, and use the processed documents as input, extract document features through the HNSW algorithm, store the processed documents and document features in the knowledge base and build an index.
5. The project process intelligent supervision platform based on large model according to claim 1 is characterized in that: The engine layer is used to store the pre-processed data in the knowledge base in an incremental update manner.
6. The project process intelligent supervision platform based on large model according to claim 1 is characterized in that: The knowledge base includes a vector database and a graph database.
7. The project process intelligent supervision platform based on large model according to claim 1 is characterized in that: The data layer, engine layer and application layer are encapsulated as pluggable modules, and the engine layer outputs risk prediction results and decision optimization results through WebSocket or grpc protocol.
8. The project process intelligent supervision platform based on large model according to claim 1 is characterized in that: The application layer is configured with a risk warning dashboard, a report generation module, and a task scheduling module; The risk warning dashboard is used to display risk warning results and decision optimization results in the form of dynamic heat maps, and supports query services for risk warning results and decision optimization results; The report generation module is used to generate reports based on risk warning results and decision optimization results, and supports viewing and downloading of reports; The task scheduling module is used to schedule data collection tasks and data analysis tasks, so as to realize that the execution time of data collection tasks and data analysis tasks is in an idle state.