Project management online auxiliary auditing method and system based on large model

Abstract

The invention discloses a project management online auxiliary auditing method and system based on a large model, and belongs to the technical field of data processing, and the system comprises a data storage layer which comprises a core index classification extraction module and an index storage module; the content extraction layer comprises an input construction scheme overall content preprocessing module and a core index content abstract extraction module; the content calculation layer comprises a similarity algorithm management module and a similarity calculation module; the automatic auditing layer comprises a primary auditing module and a joint auditing automatic distribution module, utilizes an optimized distributed editing distance algorithm to quickly calculate the similarity between pre-index rules of the to-be-audited indexes to meet the requirement of quick auditing, and utilizes the optimized distributed editing distance algorithm to calculate the similarity between pre-index rules of the to-be-audited indexes to meet the requirement of quick auditing; and quickly calculating the similarity between the construction scheme related content abstracts and the joint auditing indexes, and dynamically selecting related departments needing to participate in joint auditing.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing, and in particular relates to an online auxiliary audit method and system for project management based on a large model. Background Art

[0002] A project management system is a tool or software used to help organize, plan, execute and monitor projects. It mainly includes five parts: project planning, project application, project approval, project implementation and project supervision.

[0003] The project approval process mainly includes project letter of intent review, project preliminary review, construction plan review, etc. The construction plan review includes two stages: preliminary review and joint review. The project preliminary review refers to the competent authority first reviewing the feasibility of the construction plan; the project joint review refers to the dynamic selection of the unit information for the joint review based on the content of the construction plan. After all units have passed the review, the construction plan is jointly reviewed and approved. However, the content of a construction plan includes ten chapters, including overview, current situation and demand analysis, overall design, main construction content, operation and maintenance cost estimation, security assurance, performance target analysis, information and innovation credibility analysis, and technical implementation route, data resource plan, project budget details and implementation plan, with more than 100 summaries totaling more than 200 pages. Therefore, a complete review of a construction plan requires a lot of manpower.

[0004] The present invention utilizes big model technology and data processing technology to construct an online auxiliary review method and system for the project management system. By automatically summarizing and scoring the core indicators involved in the construction plan, such as project name, project goals, digital infrastructure application status, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis, the method and system can assist in the review of the project construction plan and improve the approval efficiency of the construction plan. Summary of the invention

[0005] To achieve the above object, the technical solution of the present invention is as follows: an online auxiliary audit system for project management based on a large model, the system comprising:

[0006] Data storage layer, including the classification and extraction module of core indicators and the indicator storage module;

[0007] Content extraction layer, including the input construction plan overall content preprocessing module and the core indicator content summary extraction module;

[0008] Content calculation layer, including similarity algorithm management module and similarity calculation module;

[0009] The automatic review layer includes a preliminary review module and a joint review automatic allocation module.

[0010] As an improvement of the present invention, an online auxiliary audit method for project management based on a large model is implemented by the audit system. The method comprises the following steps:

[0011] Step S1, defining the core indicators of the initial review of the construction plan and the core indicators of the joint review;

[0012] Step S2, collecting the content of the construction plan, pre-training the neural network large model, and building a large model related to the construction plan;

[0013] Step S3, defining the evaluation rules for the core indicators of the initial audit and the core indicators of the joint audit in step S1;

[0014] Step S4, using the large model trained in step S2 to extract the summary of the initial review of the core indicators in step S1;

[0015] Step S5, using an optimized edit distance algorithm, calculate the similarity between the index content summary extracted in step S4 and the preliminary review index evaluation rule in step S3, and give a preliminary review score;

[0016] Step S6, using the large model trained in step S2 to extract the content summary of the joint audit core indicators in step S1;

[0017] Step S7, using the optimized edit distance algorithm, calculate the similarity between the indicator content summary extracted in step S6 and the joint audit indicator evaluation rules in step S3, and give joint audit department recommendations.

[0018] As an improvement of the present invention, in step S1, the core indicators of the initial review of the construction plan include project name, project objectives, digital infrastructure application status, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance and performance goal analysis, and the core indicators of the joint review include data security indicators, network security indicators and system budget core indicators.

[0019] As an improvement of the present invention, in step S2, by collecting the construction plan contents in recent years, the neural network large model is pre-trained to construct a large model related to the construction plan, as follows:

[0020] Step S2.1, preprocessing the data in the construction plan to obtain the text data of the construction plan, wherein the preprocessing at least includes: denoising, word segmentation and part-of-speech tagging;

[0021] Step S2.2, classify and store the preprocessed data, including classifying, extracting and storing the preliminary review core indicators such as project name, project goal, digital infrastructure application status, project status, business demand analysis, security demand analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis;

[0022] Step S2.3, classify, extract and store data security indicators, network security indicators and system budget core indicators;

[0023] Step S2.4, using the training data extracted in steps S2.1 and S2.2, train the neural network model to obtain the preliminary review indicator model and the joint review model.

[0024] As an improvement of the present invention, in step S4, the construction plan preliminary review large model obtained in step S2 is used to extract the key indicator content summary of the input construction plan content. Since the extraction process of each indicator is similar, the extraction process is described by taking the project target indicator as an example, as follows:

[0025] Step S4.1, obtaining the project target content of the construction plan to be reviewed and removing invalid content;

[0026] Step S4.2, the project goal content preprocessed in the above steps is sent to the preliminary review model trained in step S3 to obtain the summary information of the current project goal content.

[0027] As an improvement of the present invention, the summary information of the preliminary review index content obtained in step S4 is used to calculate the similarity between the summary information and the preliminary review index evaluation rule in step S3 using the optimized edit distance algorithm. A high similarity indicates that the current index meets the preliminary review requirements. Otherwise, the specific calculation process is as follows:

[0028] Step S5.1, the edit distance between texts refers to the minimum number of edit operations required to change a policy in units of words into a text in units of words. There are three types of edit operations: insertion, deletion, and replacement.

[0029] Step S5.2, for the two indicators summary content Q 1 Q 2 …Q m and P 1 P 2 …P n , where m and n represent the number of Chinese characters in the content, respectively, and Q k Indicates the content Q 1 Q 2 …Q m The kth Chinese character in P l Indicates the content P 1 P2 …P n For the lth word, define the function edit(i,j) to represent the content Q 1 Q 2 …Q m To content P 1 P 2 …P n The edit distance of , the following dynamic programming formula is obtained,

[0030] edit(i,j)=0,if i=j=0; (1),

[0031] edit(i,j)=j,if i=0and 1≤j≤n(2),

[0032] edit(i,j)=i,if j=0and 1≤i≤m(3),

[0033] edit(i,j)=min(edit(i-1,j)+1,edit(i,j-1)+1,edit(i-1,j-1)+f(i,j)),if 1≤i≤

[0034]

[0035] Step S5.3, making a two-dimensional graph of the text content using the formula in step S5.2;

[0036] Step S5.4: Use the similarity algorithm to retrieve the information with the highest similarity for each computing unit, which is k 1 ,k 2 ,...k m , where m is the m computing units in step S5.2.

[0037] As an improvement of the present invention, the steps S6 and S7 extract key indicator content summaries from the input construction plan content through the joint audit large model obtained through step S2.

[0038] As an improvement of the present invention, since the extraction process of each indicator is similar, the extraction process of the network security indicator is described as follows:

[0039] Step S6.1, obtaining the contents of the network security level protection plan in the construction plan to be reviewed, and removing invalid contents;

[0040] Step S6.2, sending the target content preprocessed in the above steps to the joint large model trained in step S2 to obtain the current network security indicator summary information;

[0041] Step S6.3, using the optimized edit distance algorithm to calculate the similarity with the network security indicator evaluation rule in the above step, a high similarity means that the current joint audit requires the network security department to participate in the joint audit, and vice versa.

[0042] Compared with the prior art, the beneficial effects of the present invention are:

[0043] 1) Define a set of project preliminary review indicator systems and joint audit indicator systems, through which it is possible to quickly determine whether the current project construction plan meets the preliminary review requirements and dynamically generate information on the departments participating in the joint audit;

[0044] 2) Use the optimized distributed edit distance algorithm to quickly calculate the similarity between the pre-indicator rules of the indicators to be reviewed to meet the requirements of rapid review;

[0045] 3) Use the optimized distributed edit distance algorithm to quickly calculate the similarity between the summary of the relevant content of the construction plan and the joint audit indicators, and dynamically select the relevant departments that need to participate in the joint audit. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a technical architecture diagram of the system of the present invention;

[0047] Figure 2 The present invention adopts the optimized edit distance algorithm flow chart;

[0048] Figure 3 This is a flow chart of the process of extracting network security indicators in the present invention. DETAILED DESCRIPTION

[0049] The present invention will be further explained below in conjunction with the accompanying drawings and specific embodiments. It should be understood that the following specific embodiments are only used to illustrate the present invention and are not used to limit the scope of the present invention.

[0050] Example:

[0051] Technical Problem Solved (Business Value)

[0052] The initial review of the construction plan takes a long time. A complete and detailed construction plan contains more than 200 pages of content, including text, graphic symbols, tables, pictures and other formats. Therefore, it takes too long to rely entirely on manual detailed review. The present invention automatically extracts the core content summary of the construction plan through a well-trained large model, and uses the optimized edit distance algorithm to quickly calculate the similarity between the content summary and the review points, thereby improving the review efficiency of the construction plan.

[0053] Joint review of construction plans. Joint review refers to the collaborative review of construction plans by multiple departments. Only after all departments have reviewed and approved the construction plan, the construction plan is considered to have been jointly reviewed and approved. However, the number of departments involved in the joint review varies for different construction plan contents. For example, if the construction plan involves data security, the security department is required to participate in the joint review; if the construction plan is related to network security, the network security department is required to participate in the joint review. Therefore, it is entirely based on manual configuration of the departments participating in the joint review through careful reading and analysis of the contents of the construction plan. Due to the length of the construction plan, the departments for the joint review may be missed or misselected, resulting in repeated modifications to the joint review process, resulting in low efficiency of the joint review. The present invention extracts key point information in the construction plan through large model technology, and automatically selects the departments participating in the joint review, thereby improving the joint review process and improving the efficiency of the joint review.

[0054] In summary, the present invention uses large models and data processing technology to construct an online auxiliary audit method and system for project management systems. The traditional construction plan is long, which leads to the problem of time-consuming and error-prone manual audit. The audit of the construction plan is assisted by large model technology, thereby improving the audit efficiency.

[0055] Obtain project construction plans for the past five years, preprocess the data in the construction plans, and obtain construction plan text data, wherein the preprocessing includes at least: denoising, word segmentation, and part-of-speech tagging; secondly, classify and store the preprocessed data, wherein the classified storage includes classifying, extracting, and storing the core indicators involved in the construction plan, such as project name, project goal, digital infrastructure application status, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis; then use the constructed big model to train the preprocessed training data to obtain the trained big model, and finally train.

[0056] The complete technical solution of this embodiment: a project management system online auxiliary audit method and system, including the following units:

[0057] The data storage layer includes a classification extraction module and an indicator storage module for core indicators involved in the construction plan, including project name, project goals, current status of digital infrastructure applications, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance, performance target analysis, etc.;

[0058] Content extraction layer, including the overall content preprocessing module of the input construction plan and the core indicator content summary extraction module;

[0059] Content calculation layer, including similarity algorithm management module and similarity calculation module;

[0060] Automatic review layer, including preliminary review module and joint review automatic allocation module;

[0061] Specific technical architecture such as Figure 1 (System Technical Architecture) shown.

[0062] The methods and technical implementation solutions include:

[0063] Step 1: Define the core indicators of the initial review of the construction plan and the core indicators of the joint review;

[0064] Step 2: Collect the construction plan contents of the past five years, pre-train the neural network model, and build a large model related to the construction plan;

[0065] Step 3: Define the evaluation rules for the core indicators of the initial audit and the core indicators of the joint audit in Step 1;

[0066] Step 4: Use the large model trained in Step 2 to extract the summary of the initial review indicators in Step 1;

[0067] Step 5: Use the optimized edit distance algorithm to calculate the similarity between the index content summary extracted in Step 4 and the preliminary review index evaluation rules in Step 3, and give a preliminary review score;

[0068] Step 6: Use the large model trained in Step 2 to extract the summary of the core indicators of the joint audit in Step 1;

[0069] Step 7: Use the optimized edit distance algorithm to calculate the similarity between the indicator content summary extracted in Step 6 and the joint audit indicator evaluation rules in Step 3, and give recommendations for the joint audit department.

[0070] In step one, the core indicators of the initial review of the construction plan include 11 core indicators, including project name, project objectives, digital infrastructure application status, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis. The core indicators of the joint review include three core indicators, including data security indicators, network security indicators, and system budget, as shown in Table 1.

[0071] Step 2: By collecting the construction plan contents of the past five years, pre-training the neural network model, and building a large model related to the construction plan, the details are as follows:

[0072] Preprocessing the data in the construction plan to obtain the text data of the construction plan, wherein the preprocessing at least includes: denoising, word segmentation and part-of-speech tagging;

[0073] The pre-processed data is classified and stored, and the classified storage includes the classified extraction and storage of the preliminary review core indicators such as project name, project goal, digital infrastructure application status, project status, business demand analysis, security demand analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis;

[0074] Classify, extract and store three core indicators, including data security indicators, network security indicators and system budget;

[0075] The training data extracted in the first and second steps are used to train the neural network model to obtain the preliminary review indicator model and the joint review model.

[0076] Step 3: Define indicator evaluation rules for the preliminary review core indicators and the joint review core indicators. The preliminary review core indicators and the joint review core indicators are shown in Table 1. Table 1 is:

[0077]

[0078] The evaluation rules for the initial review of the core indicators are shown in Table 2. Table 2 is:

[0079]

[0080]

[0081] The evaluation rules for the core indicators of the joint audit are shown in Table 3. Table 3 is:

[0082] Indicator name Indicator rules Data security indicators The construction content mentions data collection, data aggregation, etc. Cybersecurity Indicators The cybersecurity level protection plan mentions the cybersecurity level description content System Budget Project budget details exceed fixed threshold

[0083] Step 4: extract key indicator content summaries from the input construction plan content using the construction plan preliminary review model obtained in step 2. Since the extraction process of each indicator is similar, this embodiment uses the project target indicator as an example to illustrate the extraction process, as follows:

[0084] Obtain the project objectives of the construction plan to be reviewed, and remove invalid content such as punctuation marks and stop words;

[0085] The project goal content preprocessed in the above steps is sent to the preliminary review model trained in step 3 to obtain the summary information of the current project goal content.

[0086] Step 5: Using the summary information of the preliminary review indicator content obtained in step 4, the optimized edit distance algorithm is used to calculate the similarity between it and the preliminary review indicator evaluation rule in step 3. A high similarity means that the current indicator meets the preliminary review requirements, otherwise it still does not. The specific calculation process is as follows (such as Figure 2 ):

[0087] 1) The edit distance between texts refers to the minimum number of edit operations required to change a policy in units of words to another text in units of words. There are three types of edit operations: insert, delete, and replace;

[0088] 2) Summary content for two indicators Q 1 Q 2 …Q m and P 1 P 2 …P n , where m and n represent the number of Chinese characters in the content, Q k Indicates the content Q 1 Q 2 …Q m The kth Chinese character in P l Indicates the content P 1 P 2 …P n For the lth word, define the function edit(i,j) to represent the content Q 1 Q 2 …Q m To content P 1 P 2 …P n The edit distance of , the following dynamic programming formula is obtained;

[0089] (1)edit(i,j)=0,if i=j=0;

[0090] (2)edit(i,j)=j,if i=0and 1≤j≤n;

[0091] (3)edit(i,j)=i,if j=0and 1≤i≤m;

[0092] (4)edit(i,j)=min(edit(i-1,j)+1,edit(i,j-1)+1,edit(i-1,j-1)+f(i,j)),if 1≤i≤

[0093]

[0094] 3) Let’s take an example to illustrate how to calculate edit(i,j). Assuming there are two contents, “Nanjing’s measures to support the development of green finance” and “Wuxi’s work to promote the development of green finance”, we can make the following two-dimensional chart from the formula in step 2, as shown in Table 4. Table 4 is:

[0095] none tin city close At green color gold melt hair exhibition of like Dry Measures Shi 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 South 1 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 Beijing 2 2 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 city 3 3 3 2 3 4 5 6 7 8 9 10 11 12 13 14 15 close 4 4 4 3 2 3 4 5 6 7 8 9 10 11 12 13 14 At 5 5 5 4 3 2 3 4 5 6 7 8 9 10 11 12 13 branch 6 6 6 5 4 3 3 4 5 6 7 8 9 10 11 12 13 hold 7 7 7 6 5 4 4 4 5 6 7 8 9 10 11 12 13 green 8 8 8 7 6 5 4 5 5 6 7 8 9 10 11 12 13 color 9 9 9 8 7 6 5 4 5 6 7 8 9 10 11 12 13 gold 10 10 10 9 8 7 6 5 4 5 6 7 8 9 10 11 12 melt 11 11 11 10 9 8 7 6 5 4 5 6 7 8 9 10 11 hair 12 12 12 11 10 9 8 7 6 5 4 5 6 7 8 9 10 exhibition 13 13 13 12 11 10 9 8 7 6 5 4 5 6 7 8 9 of 14 14 14 13 12 11 10 9 8 7 6 5 4 5 6 7 8 like 15 15 15 14 13 12 11 10 9 8 7 6 5 4 5 6 7 Dry 16 16 16 15 14 13 12 11 10 9 8 7 6 5 4 5 6 Measures 17 17 17 16 15 14 13 12 11 10 9 8 7 6 5 4 5 Shi 18 18 18 17 16 15 14 13 12 11 10 9 8 7 6 5 4

[0096] Using the similarity algorithm in step 3, retrieve the information with the highest similarity for each calculation unit, which is k1 ,k 2 ,...k m , where m is the m computing units in step 2.

[0097] Steps 6 and 7, through the joint audit model obtained in step 2, extract the key indicator content summary of the input construction plan content. Since the extraction process of each indicator is similar, this embodiment takes the network security indicator as an example to illustrate the extraction process, as follows (such as Figure 3 ):

[0098] Obtain the contents of the cybersecurity level protection plan in the construction plan to be reviewed, and remove invalid contents such as punctuation marks and stop words;

[0099] The target content preprocessed in the above step is sent to the joint large model trained in step 2 to obtain the summary information of the current network security indicators;

[0100] The optimized edit distance algorithm is used to calculate the similarity with the network security indicator evaluation rules in the above steps. A high similarity means that the current joint audit requires the network security department to participate in the joint audit, and vice versa.

[0101] The present invention proposes a distributed retrieval scheme for similar content retrieval efficiency. Given the number K contained in the retrieval space S and the number of nodes N, the content in the retrieval space is divided into N equal parts, and the number of each node is m. 1 ,m 2 ,…m N , the sub-search space of each node is where i = 1,…,N, and The distributed retrieval algorithm is as follows:

[0102] (1) Input content L, which is L 1 L 2 …L |L| , where |L| represents the number of Chinese characters;

[0103] (2) For each sub-search space Where i = 1, ..., N, For each subspace The maximum similarity with L, and the content of the subspace at this time is marked as index

[0104] for q=1,…,m i :

[0105] remember for The qth content in |P q | indicates the number of Chinese characters

[0106]

[0107] S(|L|,|P q |) is L to The similarity of

[0108]

[0109] if edit(|L|-k,|P q+1 |-k)>T,

[0110] where 0≤K≤max(|L|,|P q+1 |)

[0111] From this we can know

[0112] continue

[0113] else

[0114]

[0115]

[0116] (3) This is the maximum similarity between sentence L and retrieval space S, that is, the content is P at this time t ;

[0117] (4) Output and P t .

[0118] It should be noted that the above content only illustrates the technical idea of ​​the present invention and cannot be used to limit the protection scope of the present invention. For ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principle of the present invention. These improvements and modifications all fall within the protection scope of the claims of the present invention.

Claims

1. A project management online auxiliary audit system based on a large model, characterized in that: The system comprises: Data storage layer, including the classification and extraction module of core indicators and the indicator storage module; Content extraction layer, including the input construction plan overall content preprocessing module and the core indicator content summary extraction module; Content calculation layer, including similarity algorithm management module and similarity calculation module; The automatic review layer includes a preliminary review module and a joint review automatic allocation module.

2. A method for online auxiliary audit of project management based on a large model implemented by the audit system of claim 1, characterized in that: The method consists in the following steps: Step S1, defining the core indicators of the initial review of the construction plan and the core indicators of the joint review; Step S2, collecting the content of the construction plan, pre-training the neural network large model, and building a large model related to the construction plan; Step S3, defining the evaluation rules for the core indicators of the initial audit and the evaluation rules for the core indicators of the joint audit in step S1; Step S4, using the large model trained in step S2 to extract the summary of the initial review of the core indicators in step S1; Step S5, using an optimized edit distance algorithm, calculate the similarity between the index content summary extracted in step S4 and the preliminary review index evaluation rule in step S3, and give a preliminary review score; Step S6, using the large model trained in step S2 to extract the content summary of the joint audit core indicators in step S1; Step S7, using the optimized edit distance algorithm, calculate the similarity between the indicator content summary extracted in step S6 and the joint audit indicator evaluation rules in step S3, and give joint audit department recommendations.

3. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 2 is characterized in that: In step S1, the core indicators of the initial review of the construction plan include project name, project objectives, digital infrastructure application status, project status, business needs analysis, security needs analysis, application architecture design, security assurance, data architecture design, security assurance and performance target analysis. The core indicators of the joint review include data security indicators, network security indicators and system budget core indicators.

4. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 2 is characterized in that: In step S2, by collecting the construction plan contents in recent years, the neural network large model is pre-trained to build a large model related to the construction plan, as follows: Step S2.1, preprocessing the data in the construction plan to obtain the text data of the construction plan, wherein the preprocessing at least includes: denoising, word segmentation and part-of-speech tagging; Step S2.2, classify and store the preprocessed data, including classifying, extracting and storing the preliminary review core indicators such as project name, project goal, digital infrastructure application status, project status, business demand analysis, security demand analysis, application architecture design, security assurance, data architecture design, security assurance, and performance target analysis; Step S2.3, classify, extract and store data security indicators, network security indicators and system budget core indicators; Step S2.4, using the training data extracted in step S2.1 and step S2.2, train the neural network big model to obtain the preliminary review indicator big model and the joint review big model.

5. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 2 is characterized in that: In step S4, the construction plan preliminary review model obtained in step S2 is used to extract the key indicator content summary of the input construction plan content. Since the extraction process of each indicator is similar, the extraction process is described by taking the project target indicator as an example, as follows: Step S4.1, obtaining the project target content of the construction plan to be reviewed and removing invalid content; Step S4.2, the project goal content preprocessed in the above steps is sent to the preliminary review model trained in step S3 to obtain the summary information of the current project goal content.

6. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 2 is characterized in that: Step S5 uses the summary information of the preliminary review index content obtained in step S4 to calculate the similarity between it and the preliminary review index evaluation rule in step S3 using the optimized edit distance algorithm. A high similarity indicates that the current index meets the preliminary review requirements. Otherwise, the specific calculation process is as follows: Step S5.1, the edit distance between texts refers to the minimum number of edit operations required to change a policy in units of words to another text in units of words. There are three types of edit operations: insertion, deletion, and replacement. Step S5.2, summarize the content of two indicators Q1Q2…Q m and P1P2…P n , where m and n represent the number of Chinese characters in the content, respectively, and Q k Indicates content Q1Q2…Q m The kth Chinese character in P l Indicates content P1P2…P n For the lth word, define the function edit(i,j), which represents the content Q1Q2…Q m To content P1P2…P n The edit distance of , the following dynamic programming formula is obtained, edit(i,j)=0,if i=j=0; (1), edit(i,j)=j,if i=0and 1≤j≤n(2), edit(i,j)=i,if j=0and 1≤i≤m(3), Step S5.3, making a two-dimensional graph of the text content using the formula in step S5.2; Step S5.4, using the similarity algorithm, retrieve the information with the highest similarity for each computing unit, that is, k1, k2, ...k m , where m is the m computing units in step S5.

2.

7. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 2 is characterized in that: The steps S6 and S7 extract the key indicator content summary of the input construction plan content through the joint review large model obtained through step S2.

8. The online auxiliary audit method for project management based on a large model implemented by the audit system according to claim 7 is characterized in that: Since the extraction process of each indicator is similar, the extraction process of network security indicators is described as follows: Step S6.1, obtaining the contents of the network security level protection plan in the construction plan to be reviewed, and removing invalid contents; Step S6.2, sending the target content preprocessed in the above steps to the joint large model trained in step S2 to obtain the current network security indicator summary information; Step S6.3, using the optimized edit distance algorithm to calculate the similarity with the network security indicator evaluation rule in the above step, a high similarity means that the current joint audit requires the network security department to participate in the joint audit, and vice versa.

Images