An artificial intelligence-based digital government project consulting system

Abstract

The application discloses a digital government project consultation system based on artificial intelligence, and belongs to the technical field of project consultation, which comprises a data acquisition module, a data processing module, a knowledge graph construction module and a project consultation model construction module.The data acquisition module is used for collecting multi-source heterogeneous project information related to historical government projects.The data processing module is used for processing the data collected by the data acquisition module to obtain standardized project files.The knowledge graph construction module is used for extracting triplets in the standardized project files to construct a project knowledge graph.The project consultation model construction module is used for training a deep learning model based on the constructed knowledge graph to obtain a project consultation model.The online project consultation module is used for consulting a project to be consulted, including project scoring, project risk analysis and project legal regulation consultation.The system is used to improve the consultation efficiency through intelligent analysis, and meanwhile, the credibility of the consultation reply is improved.

Description

Technical Field

[0001] This invention relates to the field of project consulting technology, and in particular to a digital government project consulting system based on artificial intelligence. Background Technology

[0002] As the construction of digital government continues to advance, the demand for government project application, review, supervision, and consultation is increasing. Traditional project consultations often rely on manual policy analysis, regulation review, experience-based scoring, and risk assessment, which suffers from low efficiency, inconsistent standards, and poor knowledge reusability. With the continuous accumulation of government data, multi-source and heterogeneous project information is difficult to process quickly and in a structured manner. The correlation between policy clauses, review standards, risk points, and project content cannot be presented intuitively, easily leading to incomplete consultations and inaccurate basis.

[0003] While existing technologies have been applied in areas such as government information retrieval and text processing, most lack in-depth integration and intelligent reasoning of information throughout the entire project process, thus failing to achieve accurate matching between consultation intent and business tasks. Summary of the Invention

[0004] The purpose of this invention is to provide an artificial intelligence-based digital government project consulting system to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides an artificial intelligence-based digital government project consulting system, including a data acquisition module, a data processing module, a knowledge graph construction module, a project consulting model construction module, and an online project consulting module; The data acquisition module collects multi-source heterogeneous project information related to historical government projects, including project application forms, project planning documents, project scoring documents, and legal and regulatory documents. The data processing module processes the data collected by the data acquisition module to obtain standardized project files; The knowledge graph construction module extracts triples from normalized project files and constructs a project knowledge graph. The project consulting model building module trains a deep learning model based on the constructed knowledge graph to obtain the project consulting model; The online project consultation module provides consultation on projects, including project scoring, project risk analysis, and legal and regulatory consultation.

[0006] Preferably, the project application documents include an application form and a feasibility report; the project planning documents include annual plans, special plans, regional development plans and policy documents; the project evaluation documents include expert review opinions and scoring sheets; and the legal and regulatory documents include national or local administrative regulations and industry standards.

[0007] Preferably, the data processing module processes the data collected by the data acquisition module, including data cleaning, text preprocessing, and data conversion, wherein the data conversion transforms unstructured content into structured content.

[0008] Preferably, the knowledge graph construction module includes a triple extraction submodule and a knowledge base submodule; The triple extraction submodule identifies entities in the normalized project file, generates triples based on entity relationships, and constructs a project knowledge graph based on the triples. The knowledge base module uses the Neo4j database to store the project's knowledge graph.

[0009] Preferably, the project consulting model construction module includes a knowledge graph embedding submodule, a project scoring and prediction model, a legal and regulatory consulting model, a project risk analysis model, and a general project model; The knowledge graph embedding submodule is used to vectorize the entities, relations and attributes in the project knowledge graph, generate entity vectors and relation vectors, and provide semantic feature input. The project scoring prediction model takes knowledge graph embedding vectors, project text features, and historical project indicator features as inputs, and historical project expert review scores and sub-item scores as supervision labels. It uses a deep neural network for supervised training to learn the mapping relationship between project features and review scores, and obtains a scoring model that can predict the overall project score and sub-item scores. The legal and regulatory consultation model uses standardized legal and regulatory texts, policy clause texts, and entity relationships in a knowledge graph as training data. It is trained through semantic matching, intent recognition, and textual implication tasks to learn the compliance matching relationship between legal and regulatory clauses and project content, thereby achieving automatic legal provision retrieval, compliance judgment, and policy basis push. The project risk analysis model uses historical project problem cases, expert review opinions, and risk labeling data as training samples. It combines the relationships between risk points, violations, and failed projects in the knowledge graph and learns the rules for project risk identification and risk level determination through classification and regression tasks. It outputs project risk points, risk levels, and risk causes. The project general model integrates project text features, knowledge graph embedding features, business rule features, and multi-source project data. It is jointly trained through question-and-answer tasks, information extraction tasks, and material integrity verification tasks to achieve intelligent question-and-answer of project information, completeness check of application materials, prompts for key points of application, and feasibility auxiliary analysis.

[0010] Preferably, the online project consultation module includes a project information input submodule, a feature extraction submodule, an intelligent consultation scheduling submodule, and a result output submodule; The project information input submodule is used to receive project information and text inquiry instructions uploaded by users. The feature extraction submodule extracts features from the uploaded project information and text consultation instructions in the project input submodule. The intelligent consultation scheduling submodule extracts features from the feature extraction submodule and schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. The results output submodule outputs the reasoning results from the intelligent consultation and scheduling submodule.

[0011] Preferably, the intelligent consultation scheduling submodule includes an intent recognition unit, a task classification unit, a model scheduling unit, and a result fusion unit; The intent recognition unit performs semantic parsing on the project information features and text consultation instruction features extracted by the feature extraction submodule to identify the user's consultation intent, including project rating consultation, legal and regulatory consultation, project risk consultation, comprehensive consultation, and general Q&A consultation. The task classification unit categorizes consultation requests into scoring prediction tasks, compliance review tasks, risk analysis tasks, comprehensive assessment tasks, or general Q&A tasks based on the identified consultation intent. The model scheduling unit, based on the task classification results, schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. The results fusion unit normalizes, sorts, and structurally integrates the output results of a single model or the results of multi-model collaborative reasoning to form consultation results in a unified format, and then sends them to the results output submodule.

[0012] Preferably, the model scheduling unit specifically includes: establishing a mapping relationship between tasks and models based on task classification results; If it is a project rating consultation, then the project rating prediction model is scheduled to perform inference; If it is a legal consultation, then the legal consultation model will be scheduled to perform inference. If it is project risk consulting, then the project risk analysis model will be scheduled to perform inference. For comprehensive or multi-objective consulting, the parallel / serial scheduling project scoring prediction model, legal and regulatory consulting model, and project risk analysis model are used for collaborative reasoning. If the question is a general question and answer, information query, or application key point consultation, the project's general model will be scheduled to perform inference.

[0013] Therefore, the present invention employs the above-mentioned artificial intelligence-based digital government project consulting system, which has the following beneficial effects: (1) The data processing module realizes the automated cleaning and structured transformation of multi-source heterogeneous project data. Combined with the association retrieval capability of knowledge graph, it replaces the repetitive work of manually sorting out policies and review standards, thus improving the consultation efficiency. (2) By classifying consultation tasks, the credibility of consultation responses is improved.

[0014] The technical solution of the present invention will be further described in detail below with reference to the accompanying drawings and embodiments. Attached Figure Description

[0015] Figure 1 This is a system framework diagram of an embodiment of the present invention. Detailed Implementation

[0016] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can be arranged and designed in various different configurations, and therefore should not be construed as limiting the present invention.

[0017] Example like Figure 1 As shown, the present invention provides an artificial intelligence-based digital government project consulting system, including a data acquisition module, a data processing module, a knowledge graph construction module, a project consulting model construction module, and an online project consulting module.

[0018] The data acquisition module collects multi-source, heterogeneous project information related to historical government projects, including project application documents, project planning documents, project scoring documents, and legal and regulatory documents. Project application documents include application forms and feasibility reports; project planning documents include annual plans, special plans, regional development plans, and policy documents; project scoring documents include expert review opinions and scoring sheets; and legal and regulatory documents include national or local administrative regulations and industry standards.

[0019] The data processing module processes the data collected by the data acquisition module to obtain standardized project files. This includes data cleaning, text preprocessing, and data transformation. Data cleaning includes removing garbled characters, null values, duplicate records, invalid entries, and standardizing the format. Text preprocessing includes segmenting long texts, word segmentation, and punctuation correction. Data transformation includes converting unstructured content into structured content.

[0020] The knowledge graph construction module extracts triples from normalized project files and constructs a project knowledge graph. This module includes a triple extraction submodule, which identifies entities in the normalized project files, generates triples based on entity relationships, and constructs the project knowledge graph based on these triples.

[0021] The knowledge base module uses the Neo4j database to store the project knowledge graph, as well as standardized project files after data processing, model training data, and consultation history, providing data support for continuous system optimization. The Neo4j database has efficient graph query and association analysis capabilities, enabling rapid response to the retrieval and traversal needs of entities and relationships in the knowledge graph. This facilitates subsequent model access to semantic information within the knowledge graph and supports dynamic updates. When new project data or policy documents are added, the entities and relationships in the knowledge graph can be updated in real time, ensuring the timeliness of knowledge.

[0022] The project consulting model building module trains a deep learning model based on the constructed knowledge graph to obtain the project consulting model.

[0023] The project consulting model construction module includes a knowledge graph embedding sub-module, a project scoring and prediction model, a legal and regulatory consulting model, a project risk analysis model, and a general project model.

[0024] The knowledge graph embedding submodule is used to vectorize the entities, relations, and attributes in the project knowledge graph, generating entity vectors and relation vectors, and providing semantic feature input.

[0025] The project scoring prediction model takes knowledge graph embedding vectors, project text features, and historical project indicator features as inputs, and historical project expert review scores and sub-item scores as supervision labels. It uses a deep neural network for supervised training to learn the mapping relationship between project features and review scores, and obtains a scoring model that can predict the overall project score and sub-item scores.

[0026] The legal and regulatory consultation model uses standardized legal and regulatory texts, policy clause texts, and entity relationships in a knowledge graph as training data. It is fine-tuned using a BERT pre-trained model and trained through semantic matching, intent recognition, and textual implication tasks. It learns the compliance matching relationship between legal and regulatory clauses and project content, and realizes automatic legal provision retrieval, compliance judgment, and policy basis push.

[0027] The project risk analysis model uses historical project problem cases, expert review opinions, and risk-labeled data as training samples. It combines risk points, violations, and the correlation of failed projects in a knowledge graph. Using a combination of random forest algorithm and deep learning model, it learns the rules for project risk identification and risk level determination through classification and regression tasks. It outputs project risk points, risk levels, and risk causes, and provides targeted risk prevention and control suggestions.

[0028] The project's general model integrates project text features, knowledge graph embedding features, business rule features, and multi-source project data. It is jointly trained through question-answering tasks, information extraction tasks, and material integrity verification tasks to achieve intelligent question-answering of project information, completeness checks of application materials, prompts for key application points, and feasibility analysis assistance. The training model can employ a Seq2Seq model.

[0029] The online project consultation module provides consultation services for projects, including project scoring, project risk analysis, and legal and regulatory consultation. Specifically, the online project consultation module includes a project information input submodule, a feature extraction submodule, an intelligent consultation scheduling submodule, and a result output submodule.

[0030] The project information input submodule is used to receive project information and text inquiry instructions uploaded by users.

[0031] The feature extraction submodule extracts features from the uploaded project information and text consultation instructions in the project input submodule.

[0032] The intelligent consultation scheduling submodule extracts features from the feature extraction submodule and schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. The intelligent consultation scheduling submodule includes an intent recognition unit, a task classification unit, a model scheduling unit, and a result fusion unit. The intent recognition unit performs semantic parsing on the project information features and text consultation instruction features extracted by the feature extraction submodule to identify the user's consultation intent, including project rating consultation, legal and regulatory consultation, project risk consultation, comprehensive consultation, and general Q&A consultation. The task classification unit categorizes consultation requests into scoring prediction tasks, compliance review tasks, risk analysis tasks, comprehensive assessment tasks, or general Q&A tasks based on the identified consultation intent. The model scheduling unit, based on the task classification results, schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. Specifically, it establishes a mapping relationship between tasks and models based on the task classification results. If it is a project rating consultation, then the project rating prediction model is scheduled to perform inference; If it is a legal consultation, then the legal consultation model will be scheduled to perform inference. If it is project risk consulting, then the project risk analysis model will be scheduled to perform inference. For comprehensive or multi-objective consulting, the parallel / serial scheduling project scoring prediction model, legal and regulatory consulting model, and project risk analysis model are used for collaborative reasoning. If the question is a general question and answer, information query, or application key point consultation, the project's general model will be scheduled to perform inference.

[0033] The results fusion unit normalizes, sorts, and structurally integrates the output results of a single model or the results of multi-model collaborative reasoning to form consultation results in a unified format, and then sends them to the results output submodule.

[0034] The results output submodule outputs the reasoning results from the intelligent consultation and scheduling submodule.

[0035] Therefore, the present invention adopts the above-mentioned digital government project consultation system based on artificial intelligence, which improves consultation efficiency and enhances the credibility of consultation responses through intelligent analysis.

[0036] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit them. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the technical solutions of the present invention, and these modifications or equivalent substitutions cannot cause the modified technical solutions to deviate from the spirit and scope of the technical solutions of the present invention.

Claims

1. A digital government project consulting system based on artificial intelligence, characterized in that: It includes a data acquisition module, a data processing module, a knowledge graph construction module, a project consulting model construction module, and an online project consulting module; The data acquisition module collects multi-source heterogeneous project information related to historical government projects, including project application forms, project planning documents, project scoring documents, and legal and regulatory documents. The data processing module processes the data collected by the data acquisition module to obtain standardized project files; The knowledge graph construction module extracts triples from normalized project files and constructs a project knowledge graph. The project consulting model building module trains a deep learning model based on the constructed knowledge graph to obtain the project consulting model; The online project consultation module provides consultation on projects, including project scoring, project risk analysis, and legal and regulatory consultation.

2. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that: The project application documents include an application form and a feasibility report; the project planning documents include annual plans, special plans, regional development plans and policy documents; the project evaluation documents include expert review opinions and scoring sheets; and the legal and regulatory documents include national or local administrative regulations and industry standards.

3. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that: The data processing module processes the data collected by the data acquisition module, including data cleaning, text preprocessing, and data transformation. The data transformation converts unstructured content into structured content.

4. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that: The knowledge graph construction module includes a triplet extraction submodule and a knowledge base submodule; The triple extraction submodule identifies entities in the normalized project file, generates triples based on entity relationships, and constructs a project knowledge graph based on the triples. The knowledge base module uses the Neo4j database to store the project's knowledge graph.

5. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that: The project consulting model construction module includes a knowledge graph embedding submodule, a project scoring and prediction model, a legal and regulatory consulting model, a project risk analysis model, and a general project model; The knowledge graph embedding submodule is used to vectorize the entities, relations and attributes in the project knowledge graph, generate entity vectors and relation vectors, and provide semantic feature input. The project scoring prediction model takes knowledge graph embedding vectors, project text features, and historical project indicator features as inputs, and historical project expert review scores and sub-item scores as supervision labels. It uses a deep neural network for supervised training to learn the mapping relationship between project features and review scores, and obtains a scoring model that can predict the overall project score and sub-item scores. The legal and regulatory consultation model uses standardized legal and regulatory texts, policy clause texts, and entity relationships in a knowledge graph as training data. It is trained through semantic matching, intent recognition, and textual implication tasks to learn the compliance matching relationship between legal and regulatory clauses and project content, thereby achieving automatic legal provision retrieval, compliance judgment, and policy basis push. The project risk analysis model uses historical project problem cases, expert review opinions, and risk labeling data as training samples. It combines the relationships between risk points, violations, and failed projects in the knowledge graph and learns the rules for project risk identification and risk level determination through classification and regression tasks. It outputs project risk points, risk levels, and risk causes. The project general model integrates project text features, knowledge graph embedding features, business rule features, and multi-source project data. It is jointly trained through question-and-answer tasks, information extraction tasks, and material integrity verification tasks to achieve intelligent question-and-answer of project information, completeness check of application materials, prompts for key points of application, and feasibility auxiliary analysis.

6. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that: The online project consultation module includes a project information input submodule, a feature extraction submodule, an intelligent consultation scheduling submodule, and a result output submodule. The project information input submodule is used to receive project information and text inquiry instructions uploaded by users. The feature extraction submodule extracts features from the uploaded project information and text consultation instructions in the project input submodule. The intelligent consultation scheduling submodule extracts features from the feature extraction submodule and schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. The results output submodule outputs the reasoning results from the intelligent consultation and scheduling submodule.

7. The digital government project consulting system based on artificial intelligence according to claim 6, characterized in that: The intelligent consultation scheduling submodule includes an intent recognition unit, a task classification unit, a model scheduling unit, and a result fusion unit; The intent recognition unit performs semantic parsing on the project information features and text consultation instruction features extracted by the feature extraction submodule to identify the user's consultation intent, including project rating consultation, legal and regulatory consultation, project risk consultation, comprehensive consultation, and general Q&A consultation. The task classification unit categorizes consultation requests into scoring prediction tasks, compliance review tasks, risk analysis tasks, comprehensive assessment tasks, or general Q&A tasks based on the identified consultation intent. The model scheduling unit, based on the task classification results, schedules one or more of the following models for collaborative reasoning: project scoring prediction model, legal and regulatory consultation model, project risk analysis model, and general project model. The results fusion unit normalizes, sorts, and structurally integrates the output results of a single model or the results of multi-model collaborative reasoning to form consultation results in a unified format, and then sends them to the results output submodule.

8. The digital government project consulting system based on artificial intelligence according to claim 1, characterized in that, The model scheduling unit specifically includes: establishing a mapping relationship between tasks and models based on task classification results; If it is a project scoring consultation, schedule the project scoring prediction model to perform inference; If it is a legal or regulatory consultation, the legal or regulatory consultation model will be scheduled to perform inference. If it is project risk consulting, schedule the project risk analysis model to perform inference; For comprehensive or multi-objective consulting, collaborative reasoning is conducted based on project scoring prediction models, legal and regulatory consulting models, and project risk analysis models. If the question is a general question and answer, information query, or application key point consultation, the project's general model will be scheduled to perform inference.

Images