Construction scheme comparison and error correction system and method thereof

Abstract

The invention relates to a construction scheme comparison and error correction system and method, the system comprises a knowledge graph creation module, a similar scheme extraction module, a difference comparison module, an expected value filling module and a manual auditing module, the knowledge graph creation module completes intelligent information identification and standardization processing of unstructured documents; the similar scheme extraction module is used for quickly searching and matching entities of the construction schemes; the difference comparison module uses a text similarity algorithm to compare relation and event information extracted by different schemes, calculates difference weights, classifies comparison conditions, and sets different colors for the difference according to levels to display scheme texts; the expected value filling module predicts expected values of difference points marked in the scheme by using a regression model or a classification model, and performs filling; and the manual auditing module is used for auditing the difference labeling area of each scheme and the expected value of filling, feeding back the result to the expected value filling module, and optimizing the filling algorithm.

Description

Technical Field

[0001] The invention belongs to the technical field of building construction, and in particular relates to a construction scheme comparison and error correction system and method thereof. Background Art

[0002] In the process of preparing construction plans, the review process is a key step to ensure the rationality, safety, feasibility and compliance of the construction plan. Due to the wide variety of construction plans and increasingly complex review standards, traditional manual comparison and error correction not only consumes a lot of time and human resources, but is also prone to errors or omissions of key information. Only through intelligent information extraction and intelligent error correction, the review speed is improved, but the review of special working conditions and complex construction processes is prone to errors.

[0003] Therefore, how to provide a construction plan comparison and error correction system and method is a technical problem that technical personnel in this field urgently need to solve. Summary of the invention

[0004] In view of the problem that traditional construction plan comparison and error correction takes a long time and has low accuracy, the present invention proposes a construction plan comparison and error correction system and method, which combines manual and intelligent comparison and error correction, quickly matches similar construction plans in history through the intelligent system, and compares and analyzes them with approved plans, marks the differences, and gives expected values ​​to ensure the comprehensiveness and depth of the review. This not only helps to improve the accuracy and intuitiveness of the review, but also saves a lot of manpower and time costs.

[0005] In order to solve the above technical problems, the present invention includes the following technical solutions: A construction scheme comparison and error correction system, comprising: A knowledge graph creation module, which establishes a labeled database and a solution knowledge graph based on the information extraction algorithm of a large language model, and completes intelligent information recognition and standardized processing of unstructured documents; A similar scheme extraction module, which is based on a labeled construction scheme database and uses a deep semantic similarity model to quickly search and match entities of the construction scheme; A difference comparison module, which uses a text similarity algorithm to compare the relationship and event information extracted from different solutions, calculates the difference weights, classifies the comparison situations, and sets different colors to display the solution text according to the level of the differences; An expected value filling module, which uses a regression model or a classification model to predict the expected value of the difference points marked in the solution and fills the difference points; The manual review module is used to review the difference marked area and the expected value of filling of each solution, and feed the results back to the expected value filling module to improve the next filling accuracy module and optimize the filling algorithm.

[0006] Furthermore, the knowledge graph creation module includes named entity recognition, relationship extraction and event extraction. The named entity recognition includes using a large language model to perform named entity recognition, and extracting specific types of entities such as scheme categories, project names and construction practices from the scheme text; the relationship extraction includes identifying the relationship between entities through a deep learning model and generating relationship triples; the event extraction includes using an event extraction model to identify events in the text and their related information.

[0007] Furthermore, the similar solution extraction includes: for the solution knowledge graph, constructing a weighted similarity model including three elements: solution category, project name and solution main entity object, and presetting weighted weights of the three elements, and judging the similarity between solutions by calculating the weighted similarity value. If the weighted similarity value reaches or exceeds the preset matching threshold, it is judged as a similar solution.

[0008] Furthermore, the difference comparison module divides the comparison situations into three categories: comparison of schemes that have passed the review, comparison of schemes that have passed the review and those that have not passed the review, and comparison of schemes that have not passed the review.

[0009] Furthermore, the text similarity algorithm includes Jaccard similarity or TF-IDF.

[0010] The present invention also provides a construction scheme comparison and error correction method, the method comprising: Step S1, using the construction plan to compare and correct the system; Step S2: Establish a labeled database and a solution knowledge map through a knowledge map creation module to complete intelligent information recognition and standardized processing of the construction plan; Step S3: Using a similar solution extraction module to quickly search and match entities of the construction solution based on the labeled construction solution database and using a deep semantic similarity model; Step S4: Using the text similarity algorithm through the difference comparison module, compare the extracted relationship and event information of different schemes, calculate the difference weight, and set different colors to display the scheme text according to the three levels of low, medium and high; Step S5: using an expected value filling module to predict the expected values ​​of the difference points marked in the solution using a regression model or a classification model, and then filling them; Step S6: The manual review module reviews the difference marked areas and expected filling values ​​of each solution to optimize the filling algorithm.

[0011] Compared with the prior art, the present invention has the following advantages and beneficial effects: The present invention provides a construction plan comparison and error correction system and method thereof, the system comprising a knowledge graph creation module, a similar plan extraction module, a difference comparison module, an expected value filling module and a manual review module. The construction plan comparison and error correction system combines manual and intelligent comparison and error correction, quickly matches similar construction plans in history through an intelligent system, and compares and analyzes them with approved plans, marks the differences, and gives expected values, thereby ensuring the comprehensiveness and depth of the review.

[0012] Compared with the traditional method, the beneficial effects of the present invention are mainly as follows: (1) High efficiency of difference comparison: Rapidly search and match key information between documents and effectively compare differences, thereby improving audit efficiency.

[0013] (2) Optimization of expected value filling: The differences in the automatic filling scheme reduce manual operations and continuously optimize the filling accuracy through feedback from the manual review module.

[0014] (3) Assistance of manual review: The existence of the manual review module ensures the quality of the final solution, improves the accuracy of the system's intelligent filling through the feedback mechanism, and forms a closed-loop optimization process.

[0015] (4) Reduce human resource consumption: Automated knowledge extraction and solution review processes greatly reduce reliance on manual review and save human resources. This advantage is particularly evident when processing large-scale documents. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is an architecture diagram of a construction plan comparison and error correction system in one embodiment of the present invention. DETAILED DESCRIPTION

[0017] The following is a further detailed description of a construction scheme comparison and error correction system and method provided by the present invention in conjunction with the accompanying drawings and specific embodiments. The advantages and features of the present invention will become more apparent from the following description.

[0018] Embodiment 1

[0019] Combine the following Figure 1 , the construction scheme comparison and error correction system of the present invention is described in detail.

[0020] Please refer to Figure 1, a construction plan comparison and error correction system, including: a knowledge graph creation module, a similar plan extraction module, a difference comparison, an expected value filling module and a manual review module. In this embodiment, more preferably, the knowledge graph creation module is based on the information extraction algorithm of the large language model, establishes a labeled database and a plan knowledge graph, and completes the intelligent information recognition and standardized processing of unstructured documents such as construction plans. Specifically, it includes named entity recognition, relationship extraction and event extraction.

[0021] Specifically, named entity recognition (NER) includes: using large language models (such as BERT) for named entity recognition, extracting specific types of entities from the solution text, such as solution categories, project names, construction practices, etc. Relation extraction (RE) includes: identifying the relationship between entities through deep learning models (such as CNN, LSTM), and generating relationship triples (such as entity A-relationship-entity B). Event extraction (EE) includes: using event extraction models to identify events in text and related information, such as construction time, etc.

[0022] In this embodiment, more preferably, the similar scheme extraction module is based on the labeled construction scheme database and uses a deep semantic similarity model to quickly search and match the entities of the construction scheme. For the scheme knowledge graph, a weighted similarity model containing three elements is constructed. The three elements are the scheme category, the project name, and the main entity object of the scheme, and the weighted weights of the three elements are preset. The similarity between the schemes is judged by calculating the weighted similarity value. If the weighted similarity value reaches or exceeds the preset matching threshold, it is determined to be a similar scheme. For example: the weight of the scheme category is set to 0.6, the weight of the project name is set to 0.2, and the weight of the main entity object of the scheme is set to 0.2. The weighted similarity value = 0.6 × scheme category similarity value + 0.2 × project name similarity value + 0.2 × scheme main entity object similarity value; finally, the matching threshold is set to 0.9, and when the weighted similarity value ≥ 0.9, it is determined to be a similar scheme.

[0023] Among them, in the weighted similarity model, the calculation of similarity values ​​is usually based on the similarity measurement of different features. The following is the similarity value calculation method for solution category, project name and solution main entity object.

[0024] 1. Calculation method of similarity value of scheme category: The solution category is usually a predefined classification label, and the similarity value can be calculated by whether the categories are consistent. If the categories of the two solutions are exactly the same, the similarity value is 1; if the categories are different, the similarity value is 0.

[0025]

[0026] 2. Calculation method of project name similarity value: Project names are usually text, and the cosine similarity algorithm can be used to calculate similarity values. Convert the project names into vectors (such as TF-IDF or word embeddings), and then calculate the cosine similarity.

[0027]

[0028] Among them, A and B are the vector representations of the two project names.

[0029] 3. Calculation method of similarity values ​​of main entity objects of the scheme: The main entity object of the scheme can be a key entity in the scheme (such as a person's name, a place, an organization, etc.). The similarity value can be calculated by set similarity, which regards the entity objects in the two schemes as sets and calculates the Jaccard similarity coefficient.

[0030]

[0031] Among them, E1 and E2 are entity sets in two schemes.

[0032] In this embodiment, more preferably, the difference comparison module uses a text similarity algorithm (such as Jaccard similarity, TF-IDF) to compare the relationship and event information extracted from different solutions, calculate the difference weight, and divide the comparison into the following three categories: comparison of approved solutions, comparison of approved and unapproved solutions, and comparison of unapproved solutions. Set the differences to display the solution text in different colors according to the three levels of low, medium, and high.

[0033] In this embodiment, more preferably, the expected value filling module predicts the expected value of the difference points marked in the solution using a regression model or a classification model, and fills the difference points.

[0034] In this embodiment, more preferably, the manual review module is used to review the difference marked areas and the expected value of filling of each solution, and feed the results back to the expected value filling module to improve the next filling accuracy module and optimize the filling algorithm.

[0035] Please continue to refer to Figure 1 The present invention also provides a construction scheme comparison and error correction method, the method comprising: Step S1, using the construction plan to compare and correct the system; Step S2: Establish a labeled database and a solution knowledge map through a knowledge map creation module to complete intelligent information recognition and standardized processing of the construction plan; Step S3: Using a similar solution extraction module to quickly search and match entities of the construction solution based on the labeled construction solution database and using a deep semantic similarity model; Step S4: Using the text similarity algorithm through the difference comparison module, compare the extracted relationship and event information of different schemes, calculate the difference weight, and set different colors to display the scheme text according to the three levels of low, medium and high; Step S5: using an expected value filling module to predict the expected values ​​of the difference points marked in the solution using a regression model or a classification model, and then filling them; Step S6: The manual review module reviews the difference marked areas and expected filling values ​​of each solution to optimize the filling algorithm.

[0036] The above examples are preferred embodiments of the present invention, but the embodiments of the present invention are not limited to the above embodiments. The above embodiments only express several embodiments of the present invention, and the descriptions thereof are relatively specific and detailed, but they cannot be understood as limiting the scope of the invention patent. It should be pointed out that, for those of ordinary skill in the art, several variations and improvements can be made without departing from the concept of the present invention, and these all belong to the protection scope of the present invention. Therefore, the protection scope of the patent of the present invention shall be subject to the attached claims.

Claims

1. A construction plan comparison and error correction system, characterized in that: include: A knowledge graph creation module, which establishes a labeled database and a solution knowledge graph based on the information extraction algorithm of a large language model, and completes intelligent information recognition and standardized processing of unstructured documents; A similar scheme extraction module, which is based on a labeled construction scheme database and uses a deep semantic similarity model to quickly search and match entities of the construction scheme; A difference comparison module, which uses a text similarity algorithm to compare the relationship and event information extracted from different solutions, calculates the difference weights, classifies the comparison situations, and sets different colors to display the solution text according to the level of the differences; An expected value filling module, which uses a regression model or a classification model to predict the expected value of the difference points marked in the solution and fills the difference points; The manual review module is used to review the difference marked area and the expected value of filling of each solution, and feed the results back to the expected value filling module to improve the next filling accuracy module and optimize the filling algorithm.

2. The construction plan comparison and error correction system according to claim 1 is characterized in that: The knowledge graph creation module includes named entity recognition, relationship extraction and event extraction. The named entity recognition includes using a large language model to perform named entity recognition, and extracting specific types of entities such as solution categories, project names and construction practices from the solution text; the relationship extraction includes identifying the relationship between entities through a deep learning model and generating relationship triples; the event extraction includes using an event extraction model to identify events in the text and their related information.

3. The construction plan comparison and error correction system according to claim 1 is characterized in that: The similar solution extraction includes: for the solution knowledge graph, constructing a weighted similarity model including three elements: solution category, project name and solution main entity object, and presetting weighted weights of the three elements, and judging the similarity between solutions by calculating the weighted similarity value. If the weighted similarity value reaches or exceeds the preset matching threshold, it is judged as a similar solution.

4. The construction plan comparison and error correction system according to claim 1 is characterized in that: The difference comparison module divides the comparison situations into three categories: comparison of schemes that have passed the review, comparison of schemes that have passed the review and those that have not passed the review, and comparison of schemes that have not passed the review.

5. The construction plan comparison and error correction system according to claim 1 is characterized in that: The text similarity algorithm includes Jaccard similarity or TF-IDF.

6. A construction scheme comparison and error correction method, characterized in that: include: Step S1, using the construction plan comparison and error correction system described in any one of claims 1 to 5; Step S2: Establish a labeled database and a solution knowledge map through a knowledge map creation module to complete intelligent information recognition and standardized processing of the construction plan; Step S3: Using a similar solution extraction module to quickly search and match entities of the construction solution based on the labeled construction solution database and using a deep semantic similarity model; Step S4: Using the text similarity algorithm through the difference comparison module, compare the extracted relationship and event information of different schemes, calculate the difference weight, and set different colors to display the scheme text according to the three levels of low, medium and high; Step S5: using an expected value filling module to predict the expected values ​​of the difference points marked in the solution using a regression model or a classification model, and then filling them; Step S6: The manual review module reviews the difference marked areas and expected filling values ​​of each solution to optimize the filling algorithm.

Images