Construction scheme auditing method and device based on short text

Abstract

The invention provides a construction scheme auditing method and device based on a short text, and relates to the technical field of building construction engineering, and the method comprises the steps: obtaining a to-be-audited construction scheme and a corresponding auditing demand; analyzing the auditing demand, determining a plurality of auditing points contained in the auditing demand, and extracting a to-be-audited text corresponding to each auditing point from the to-be-audited construction scheme; based on the similarity between each short text and the corresponding auditing content text, determining an auditing result of each auditing point; and generating a final audit result of the to-be-audited construction scheme based on the audit result corresponding to each audit point. According to the construction scheme auditing method and device based on the short text, the construction scheme can be automatically audited according to the auditing requirement, the construction scheme auditing efficiency is greatly improved, and the auditing cost is reduced.

Description

TECHNICAL FIELD

[0001] The application relates to the technical field of building construction engineering, and in particular to a construction scheme auditing method and device based on short texts. BACKGROUND

[0002] In recent years, the construction industry and related fields have actively promoted the research and development of digital and intelligent technologies, accelerated enterprise transformation, and promoted the industry towards intelligent construction. Under this background, the development of a building digital management system based on intelligent construction scheme auditing not only conforms to the development trend of the industry, but also has significant market application value. Such a system can effectively improve engineering efficiency and reduce operating costs by optimizing construction process management, providing important support for the upgrading of the industry.

[0003] In related technologies, the auditing and management of construction technical documents in the construction industry are mostly completed manually, with low digitalization, strong dependence on the experience of auditors, unstable output of audit results, and a large amount of repetitive labor in auditing the same type of construction scheme, resulting in high work intensity.

[0004] Therefore, there is an urgent need for an intelligent construction scheme auditing solution to improve the efficiency of construction scheme auditing and shorten the auditing cycle. SUMMARY

[0005] The purpose of the present application is to provide a construction scheme auditing method and device based on short texts, which can automatically audit construction schemes according to auditing requirements, greatly improving the auditing efficiency of construction schemes and reducing auditing costs.

[0006] The application provides a construction scheme auditing method based on short texts, comprising: Obtaining a construction scheme to be audited and corresponding auditing requirements; analyzing the auditing requirements to determine a plurality of auditing points contained in the auditing requirements, and extracting a to-be-audited text corresponding to each auditing point from the to-be-audited construction scheme; each to-be-audited text contains a plurality of short texts; determining the auditing result of each auditing point based on the similarity between each short text and a corresponding auditing content text; the auditing content text includes the auditing content corresponding to an auditing sub-item; one auditing point includes a plurality of auditing sub-items; and generating a final auditing result of the to-be-audited construction scheme based on the auditing result corresponding to each auditing point.

[0007] Optionally, the extracting of the to-be-audited text corresponding to each auditing point from the to-be-audited construction scheme comprises: performing text extraction on the text content in the to-be-audited construction scheme based on the auditing content corresponding to each auditing point to obtain the to-be-audited text corresponding to each auditing point.

[0008] Optionally, each audit point comprises a plurality of audit sub-items; the determining of the audit result of each audit point based on the similarity between each short text and the corresponding audit content text comprises: respectively matching the audit content text corresponding to each audit sub-item in the target audit point with the target text to be audited by using a regular expression, and obtaining a matching result; the target audit point is any one of the plurality of audit points; the target text to be audited is the text to be audited corresponding to the target audit point; in the case where the matching result indicates that the text content related to the target audit sub-item cannot be matched from the target text to be audited, the audit result of the target audit sub-item is determined based on the similarity between each short text included in the target text to be audited and the audit content text of the target audit sub-item; wherein the audit result of the target audit point is determined based on the audit result of each audit sub-item in the target audit point.

[0009] Optionally, the determining of the audit result of each audit point based on the similarity between each short text and the corresponding audit content text comprises: calculating the text vector of each short text included in the target text to be audited, and the text vector of the audit content corresponding to the target audit sub-item; respectively calculating the similarity value between the text vector of the audit content corresponding to the target audit sub-item and the text vector of each short text, to obtain a plurality of similarity values; and determining the audit result of the target audit sub-item based on the maximum similarity value in the plurality of similarity values.

[0010] Optionally, the calculating of the text vector of each short text included in the target audit point, and the text vector of the audit content corresponding to the target audit point comprises: performing word segmentation processing on the target short text to obtain a plurality of segmented words included in the target short text; respectively matching each segmented word in the plurality of segmented words with the segmented word corpus in the word vector library to determine the short text vector of each segmented word and the corresponding construction type label; calculating the segmented word text vector of each segmented word based on the product of the short text vector of each segmented word and the weight value corresponding to the label type, and adding the segmented word text vector of each segmented word in the plurality of segmented words to obtain the short text vector; and determining the ratio of the short text vector to the number of segmented words corresponding to the plurality of segmented words as the text vector of the target short text; wherein the target short text is any one of the plurality of short texts, or the audit content corresponding to the target audit point.

[0011] Optionally, the audit result corresponding to the audit point comprises pass, partial pass and fail; the determining of the audit result of the target audit sub-item based on the maximum similarity value in the plurality of similarity values comprises: in a case where the maximum similarity value is greater than or equal to a first preset threshold, determining that the audit result corresponding to the target audit sub-item is pass; or in a case where the maximum similarity value is less than the first preset threshold and greater than or equal to a second preset threshold, determining that the audit result corresponding to the target audit sub-item is partial pass; or in a case where the maximum similarity value is less than the second preset threshold, determining that the audit result corresponding to the target audit sub-item is fail; wherein the first preset threshold is greater than the second preset threshold.

[0012] Optionally, before the determining of the audit result of each audit point based on the similarity between each short text and the corresponding audit content text, the method further comprises: constructing a text training dataset containing relevant professional vocabulary based on historical construction scheme texts; training a word vector model using training samples in the text training dataset to obtain a target model; inputting a plurality of segmented corpora into the target model to obtain a text vector of each segmented corpus, and storing the construction type label and the text vector of each segmented corpus in association in a database to obtain the word vector library.

[0013] The application further provides a construction scheme auditing device based on short texts, comprising: An acquisition module is configured to acquire a construction scheme to be audited and corresponding audit requirements; an extraction module is configured to analyze the audit requirements, determine a plurality of audit points contained in the audit requirements, and extract a text to be audited corresponding to each audit point from the construction scheme to be audited; each text to be audited comprises a plurality of short texts; an auditing module is configured to determine an audit result of each audit point based on the similarity between each short text and a corresponding audit content text; the audit content text comprises audit content corresponding to an audit sub-item; one audit point comprises a plurality of audit sub-items; and a generation module is configured to generate a final audit result of the construction scheme to be audited based on the audit result corresponding to each audit point.

[0014] Optionally, the extraction module is specifically configured to perform text extraction on text content in the construction scheme to be audited based on the audit content corresponding to each audit point to obtain the text to be audited corresponding to each audit point.

[0015] Optionally, each audit point comprises a plurality of audit sub-items; and the auditing module is specifically configured to The target audit point is any one of the plurality of audit points; the target to-be-audited text is a to-be-audited text corresponding to the target audit point; the audit module is specifically further configured to determine an audit result of the target audit subitem based on a similarity between each short text included in the target to-be-audited text and the audit content text of the target audit subitem, in a case where the matching result indicates that the target to-be-audited text fails to match the text content related to the target audit subitem; and the audit result of the target audit point is determined based on the audit result of each audit subitem in the target audit point.

[0016] Optionally, the audit module is specifically configured to calculate a text vector of each short text included in the target to-be-audited text and a text vector of the audit content corresponding to the target audit subitem; the audit module is specifically further configured to calculate a similarity value between the text vector of the audit content corresponding to the target audit subitem and the text vector of each short text respectively, to obtain a plurality of similarity values; and the audit module is specifically further configured to determine the audit result of the target audit subitem based on a maximum similarity value in the plurality of similarity values.

[0017] Optionally, the audit module is specifically configured to perform word segmentation processing on a target short text to obtain a plurality of segmented words included in the target short text; the audit module is specifically further configured to match each segmented word in the plurality of segmented words with a segmented word corpus in a word vector library respectively, to determine a short text vector and a corresponding construction type label of each segmented word; the audit module is specifically further configured to calculate a segmented word text vector of each segmented word based on a product of the short text vector of each segmented word and a weight value corresponding to the label type, and to obtain a short text vector by adding the segmented word text vectors of each segmented word in the plurality of segmented words; and the audit module is specifically further configured to determine a ratio of the short text vector to a number of segmented words corresponding to the plurality of segmented words as a text vector of the target short text; wherein the target short text is any one of the plurality of short texts or the audit content corresponding to the target audit point.

[0018] Optionally, the audit result corresponding to the audit point comprises: pass, partial pass and fail; the audit module is specifically configured to determine that the audit result corresponding to the target audit sub-item is pass when the maximum similarity value is greater than or equal to a first preset threshold; the audit module is further configured to determine that the audit result corresponding to the target audit sub-item is partial pass when the maximum similarity value is less than the first preset threshold and greater than or equal to a second preset threshold; the audit module is further configured to determine that the audit result corresponding to the target audit sub-item is fail when the maximum similarity value is less than the second preset threshold; wherein the first preset threshold is greater than the second preset threshold.

[0019] Optionally, the device further comprises a construction module and a model training module; the construction module is configured to construct a text training data set containing relevant professional vocabulary based on historical construction scheme texts; the model training module is configured to train a word vector model using training samples in the text training data set to obtain a target model; the construction module is further configured to input a plurality of segmented word corpora into the target model to obtain a text vector of each segmented word corpus, and store the construction type label and the text vector of each segmented word corpus in a database to obtain the word vector library.

[0020] The application further provides a computer program product comprising computer programs / instructions which, when executed by a processor, implement the steps of any of the above-described short text-based construction scheme auditing methods.

[0021] The application further provides an electronic device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the steps of any of the above-described short text-based construction scheme auditing methods when executing the program.

[0022] The application further provides a computer-readable storage medium having a computer program stored thereon, wherein the computer program, when executed by a processor, implements the steps of any of the above-described short text-based construction scheme auditing methods.

[0023] The construction plan review method and apparatus based on short text provided in this application first obtain the construction plan to be reviewed and the corresponding review requirements; then, analyze the review requirements to determine multiple review points included in the review requirements, and extract the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed contains multiple short texts; and based on the similarity between each short text and the corresponding review content text, determine the review result for each review point; the review content text includes: the review content corresponding to the review sub-item; one review point includes multiple review sub-items; finally, based on the review result corresponding to each review point, generate the final review result for the construction plan to be reviewed. In this way, the construction plan can be automatically reviewed according to the review requirements, greatly improving the review efficiency of construction plans and reducing review costs. Attached Figure Description

[0024] To more clearly illustrate the technical solutions in this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram of the architecture of the construction plan review system based on short text provided in this application; Figure 2 This is a schematic diagram of the architecture of the construction scheme understanding engine provided in this application; Figure 3 This is a schematic diagram of the architecture of the intelligent review engine for construction plans provided in this application; Figure 4 This is a flowchart illustrating the short text-based construction plan review method provided in this application. Figure 5 This is a schematic diagram of the construction plan review device based on short text provided in this application; Figure 6 This is a schematic diagram of the structure of the electronic device provided in this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of this application clearer, the technical solutions of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0027] The terms "first," "second," etc., used in the specification and claims of this application are used to distinguish similar objects and not to describe a specific order or sequence. It should be understood that such use of data can be interchanged where appropriate so that embodiments of this application can be implemented in orders other than those illustrated or described herein, and the objects distinguished by "first," "second," etc., are generally of the same class and the number of objects is not limited; for example, a first object can be one or more. Furthermore, in the specification and claims, "and / or" indicates at least one of the connected objects, and the character " / " generally indicates that the preceding and following objects are in an "or" relationship.

[0028] To address the technical problems of long review cycles and low efficiency in manual review of construction plans, we have developed an intelligent review robot for construction plans. This product aims to solve the aforementioned pain points in the construction industry. Through intelligent review of construction plans, one-click uploading and rapid review can be achieved, producing stable review results in 1-2 minutes. This saves reviewers time and energy, and while maintaining a high review accuracy rate, it significantly improves review efficiency. The accumulation of review data is also a process of accumulating knowledge wealth for enterprises.

[0029] The intelligent construction plan review robot product is based on technologies such as natural language processing, deep learning language models, and image text recognition. It has already achieved intelligent review of typical sub-construction plans, including those for reinforced concrete, concrete (general and large-volume), high-formwork, scaffolding (cantilever, attached, ground-mounted, and basket-like), formwork engineering, secondary structures, ventilation and air conditioning, roofing engineering, and deep foundation pit monitoring. It has deeply developed three core functional engines: a plan understanding engine, a plan review engine, and a report generation engine, and expanded its development to include four review auxiliary functional modules: architectural synonyms, review opinion collection, review logs, and review key point mining.

[0030] like Figure 1 The diagram shown illustrates the architecture of a short text-based construction plan review system provided in this application. The construction plan understanding engine utilizes both grammar-based rule-based extraction and deep learning algorithm-based extraction. The grammar-based rule-based knowledge extraction employs fuzzy matching; the deep learning-based knowledge extraction is implemented by labeling and training the data before deployment. Currently, over 16,000 data entries have been labeled, focusing on content such as the basis for compilation and process procedures.

[0031] based on Figure 1 ,like Figure 2The diagram shows the architecture of the construction scheme understanding engine provided in this application embodiment. The intelligent review engine includes rule-based review, semantic similarity review based on sentence vector models, and text quality analysis based on deep learning, efficiently reviewing various manual text compilation modes. Rule-based review rigorously screens all elements of the content. For content that fails the rule review, such as short text content like titles, semantic similarity review based on sentence vector models will be performed. For short text content with diverse and complex descriptions, such as processes and methods, text quality analysis based on deep learning language models will be performed, and the review results will be provided.

[0032] like Figure 3 The diagram shows the architecture of the intelligent review engine for construction schemes provided in this application embodiment. In the review report generation engine, for the content of the report cover page, the scheme understanding engine extracts the project name and cover image content of the construction scheme and fills them into the specified position of the report cover page. For the main body of the review report, the pre-made report table lists all the review requirements items under this type of scheme. The review results, chapter positions, and review comments field information are obtained by parsing the structured set of review results from the scheme review engine, and then the corresponding review point key value names are filled into the specified position of the table.

[0033] For the audit points that fail the review, relevant standard sample texts will be displayed, as shown in the image on the right. Clicking "Standard Specifications or Examples" will bring up the relevant standard sample page, which mainly focuses on the audit points with clear formats, methods and requirements. The examples are displayed in a variety of formats, including rich text, images and other data formats, to provide a more intuitive and detailed reference.

[0034] The following description, in conjunction with the accompanying drawings, details the short text-based construction plan review method provided in this application through specific embodiments and application scenarios.

[0035] like Figure 4 As shown in the embodiment of this application, a construction plan review method based on short text is provided. This method may include the following steps 401 to 404: Step 401: Obtain the construction plan to be reviewed and the corresponding review requirements.

[0036] For example, the system can receive construction plans to be reviewed uploaded by users, as well as review requirements uploaded or set by reviewers. These review requirements are used to characterize which items need to be reviewed in the construction plan to be reviewed, i.e., the review points in this application. An review point can contain multiple review sub-items.

[0037] For example, the audit point in the audit requirements could be: completeness of the chapter and table of contents; the audit sub-items included in this audit point could be: 1. Basis for compilation; 2. Project overview; 3. Construction deployment; 4. Construction preparation; 5. Construction technology; 6. Safety and environmental protection measures, etc. The audit requirements mainly verify whether the chapter and table of contents contains the content required by each audit sub-item.

[0038] Step 402: Analyze the audit requirements, determine the multiple audit points included in the audit requirements, and extract the text to be audited corresponding to each audit point from the construction plan to be audited.

[0039] For example, after obtaining the construction plan to be reviewed and the corresponding review requirements, it is necessary to extract the text of the construction plan according to the review content of each review point in the review requirements, so as to obtain the text to be reviewed corresponding to each review point.

[0040] Specifically, step 402 above may also include step 402a: Step 402a: Based on the audit content corresponding to each audit point, extract the text content in the construction plan to be audited to obtain the text to be audited corresponding to each audit point.

[0041] For example, in this embodiment of the application, semantic recognition, keyword matching, or other methods are used to determine which parts of the construction plan to be reviewed are related to the review content of the review point. Then, this part of the text content can be used as the text to be reviewed corresponding to the review point. The text to be reviewed can be multiple text segments or paragraphs containing multiple sentences.

[0042] For example, if the review point is the completeness of the chapter and table of contents, then the corresponding text to be reviewed is the table of contents of the construction plan.

[0043] Step 403: Based on the similarity between each short text and the corresponding review content text, determine the review result for each review point.

[0044] For example, after extracting the text to be reviewed for each review point, regular expressions can be used to match each review sub-item. If the corresponding content can be matched, it can be determined that the review sub-item meets the requirements; otherwise, it cannot be determined whether the review sub-item meets the requirements and further matching is required.

[0045] Specifically, step 403 above may also include the following steps 403a and 403b: Step 403a: Use regular expressions to match the audit content text corresponding to each audit sub-item in the target audit point with the target text to be audited, and obtain the matching results.

[0046] Wherein, the target review point is any one of the plurality of review points; the target text to be reviewed is the text to be reviewed corresponding to the target review point.

[0047] Step 403b: If the matching result indicates that no text content related to the target review item was matched from the target text to be reviewed, the review result of the target review item is determined based on the similarity between each short text in the target text to be reviewed and the review content text of the target review item.

[0048] The audit result of the target audit point is determined based on the audit result of each audit sub-item in the target audit point.

[0049] For example, in this embodiment of the application, the similarity value between each short text and the review content text can be calculated using cosine similarity.

[0050] For example, when it is impossible to determine whether the construction plan catalog contains the content required by the audit sub-item "6. Safety and Environmental Protection Measures" through regular expression matching, the text to be audited corresponding to the audit point can be divided into multiple short texts, and the similarity value between each short text and "Safety and Environmental Protection Measures" (i.e., the audit content text mentioned above) can be calculated.

[0051] Specifically, step 403b above may also include steps 403b1 to 403b3: Step 403b1: Calculate the text vector of each short text in the multiple short texts contained in the target text to be reviewed, and the text vector of the review content corresponding to the target review sub-item.

[0052] Step 403b2: Calculate the similarity value between the text vector of the review content corresponding to the target review sub-item and the text vector of each short text to obtain multiple similarity values.

[0053] Step 403b3: Determine the review result of the target review sub-item based on the maximum similarity value among the multiple similarity values.

[0054] Specifically, in step 403 above, the method for calculating the text vector of any short text may further include the following steps 403c1 to 403c4: Step 403c1: Perform word segmentation on the target short text to obtain multiple words contained in the target short text.

[0055] Step 403c2: Match each of the multiple word segments with the word segment corpus in the word vector library to determine the short text vector of each word segment and the corresponding construction type label.

[0056] Step 403c3: Calculate the segmented text vector of each segmented word based on the product of the short text vector of each segmented word and the weight value corresponding to the tag type, and then add the segmented text vectors of each segmented word in the multiple segments to obtain the short text vector.

[0057] Step 403c4: Determine the text vector of the target short text by the ratio of the short text vector to the number of words corresponding to the multiple word segments.

[0058] The target short text is any one of the plurality of short texts, or the review content corresponding to the target review point.

[0059] For example, in this embodiment of the application, the text vector of each short text and the text vector of the review content text can be calculated using the following formula: (Formula 1) in, sen_vec Represents a text vector. m The total number of words contained in the short text. This represents the segmented text vector of each word in the word vector library. For the first i The weight value corresponding to the type label of each word segmentation.

[0060] For example, the weight values ​​of different types of tag pairs are not exactly the same. In this embodiment, the weight value of each word segment can be referred to Table 1 below:

[0061] Table 1 For example, after calculating the text vector of each short text and the text vector of the review content text, the similarity value between each short text and the review content text can be calculated, the maximum similarity value can be selected, and the review result of the review sub-item can be determined based on the maximum similarity value.

[0062] Specifically, the audit results corresponding to the audit points include: pass, partially pass, and fail. Based on this, the above step 403c may also include the following steps 403c1, 403c2, or step 403c3: Step 403c1: If the maximum similarity value is greater than or equal to the first preset threshold, determine that the review result corresponding to the target review sub-item is passed.

[0063] Step 403c2: If the maximum similarity value is less than the first preset threshold and greater than or equal to the second preset threshold, determine that the review result corresponding to the target review sub-item is partially passed.

[0064] Step 403c3: If the maximum similarity value is less than the second preset threshold, determine that the review result corresponding to the target review sub-item is not passed.

[0065] Wherein, the first preset threshold is greater than the second preset threshold.

[0066] For example, the above audit results can be expressed using the following Formula 2: (Formula 2) in, The maximum similarity score is 0.85. When the maximum similarity score is greater than or equal to 0.85, the sub-item can be directly determined to have passed the review.

[0067] For example, after obtaining the audit results for each audit sub-item, the audit results for each audit point can be generated, and finally the audit results for the audit scheme can be generated.

[0068] Step 404: Based on the audit results corresponding to each audit point, generate the final audit result of the construction plan to be audited.

[0069] For example, after determining the audit result for each audit point, the final audit result of the construction plan to be audited can be generated and output. This final audit result includes the audit result for each audit point. Users can understand what problems exist in the construction plan based on this final audit result.

[0070] Optionally, in this embodiment of the application, a word vector library can be constructed using a pre-trained word vector model.

[0071] For example, prior to step 403 above, the construction plan review method based on short text provided in this application embodiment may further include steps 405 to 407: Step 405: Construct a text training dataset containing relevant professional terms based on historical construction scheme texts.

[0072] For example, in this embodiment of the application, a vocabulary list for architectural professionals is first constructed and incorporated into the text segmentation process before pre-training to improve the quality of the model's word segmentation; then, a general stop word list is added to remove noise from the corpus and improve the model's word vector training effect; finally, the above training dataset, vocabulary list and stop word list are input into the model for training.

[0073] For example, in the embodiments of this application, a word vector model based on Word2Vec can be used to iteratively train the objective function through a skip-gram method to obtain the final target model.

[0074] For example, the objective function can be expressed by the following formula three: (Formula 3) Where T is the training document set; c is the size of the vocabulary sliding window, such as c=8, which means an abstract sliding window containing 8 words is created, traversing from the first word of the corpus; j takes values ​​sequentially in the range [-8, 8]. Skip-gram adopts the idea of ​​the central word, that is, it focuses on the central word in each window. j guarantees that it can iterate through every word to the left and right of the new word in the window; To find the conditional probability expression, that is, in the central word In certain circumstances, words The probability of occurrence is calculated, and logarithm serves as a generalization function. The iterative formula is used to train the word vector embedding representation.

[0075] Step 406: Train the word vector model using the training samples in the text training dataset to obtain the target model.

[0076] Step 407: Input multiple word segmentation corpora into the target model to obtain the text vector of each word segmentation corpus, and store the construction type label and text vector of each word segmentation corpus in the database to obtain the word vector library.

[0077] For example, after obtaining the target model, all professional terms related to the field of architecture can be converted into word vectors, and the word vector library can be constructed accordingly.

[0078] The short text-based construction plan review method provided in this application first obtains the construction plan to be reviewed and its corresponding review requirements. Then, it analyzes the review requirements to determine multiple review points, and extracts the text to be reviewed corresponding to each review point from the construction plan. Each text to be reviewed contains multiple short texts. Based on the similarity between each short text and its corresponding review content text, the review result for each review point is determined. The review content text includes: review content corresponding to a review sub-item; one review point includes multiple review sub-items. Finally, based on the review result corresponding to each review point, the final review result for the construction plan to be reviewed is generated. This method enables automated review of construction plans according to review requirements, greatly improving the efficiency of construction plan review and reducing review costs.

[0079] It should be noted that the execution entity of the short text-based construction plan review method provided in this application embodiment can be a short text-based construction plan review device, or a control module in the short text-based construction plan review device for executing the short text-based construction plan review method. This application embodiment uses the execution of the short text-based construction plan review method by a short text-based construction plan review device as an example to illustrate the short text-based construction plan review device provided in this application embodiment.

[0080] It should be noted that the methods illustrated in the accompanying drawings of this application are all illustrative examples of the construction plan review methods based on short text, using one accompanying drawing from one of the embodiments of this application as an example. In specific implementation, the construction plan review methods based on short text illustrated in the accompanying drawings of the above methods can also be implemented in conjunction with any other accompanying drawings shown in the above embodiments, which will not be elaborated here.

[0081] The short text-based construction plan review device provided in this application is described below. The short text-based construction plan review method described below can be referred to in correspondence with the short text-based construction plan review method described above.

[0082] Figure 5 This is a schematic diagram of the structure of the construction plan review device based on short text provided in the embodiments of this application, such as... Figure 5 As shown, it specifically includes: The acquisition module 501 is used to acquire the construction plan to be reviewed and the corresponding review requirements; the extraction module 502 is used to analyze the review requirements, determine the multiple review points included in the review requirements, and extract the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed contains multiple short texts; the review module 503 is used to determine the review result of each review point based on the similarity between each short text and the corresponding review content text; the review content text includes: the review content corresponding to the review sub-item; one review point includes multiple review sub-items; the generation module 504 is used to generate the final review result of the construction plan to be reviewed based on the review result corresponding to each review point.

[0083] Optionally, the extraction module 502 is specifically used to extract text from the text content of the construction plan to be reviewed based on the review content corresponding to each review point, so as to obtain the text to be reviewed corresponding to each review point.

[0084] Optionally, the review module 503 is specifically used to determine the tag to be matched based on the review content corresponding to the target review point, and to filter out multiple candidate corpora corresponding to the tag to be matched from the corpus based on the tag to be matched; the review module 503 is further used to calculate the similarity between the target text corresponding to the target review point and each candidate corpus in the multiple candidate corpora, and to filter out the target corpus with the largest similarity value to the target text from the multiple candidate corpora; the review module 503 is further used to determine the review result corresponding to the target review point based on the maximum similarity value; wherein, the target review point is any one of the multiple review points.

[0085] Optionally, the review results corresponding to the review points include: pass, partially pass, and fail; the review module 503 is specifically used to determine that the review result corresponding to the target review point is pass when the maximum similarity value is greater than or equal to a first preset threshold; the review module 503 is further used to determine that the review result corresponding to the target review point is partially pass when the maximum similarity value is less than the first preset threshold and greater than or equal to a second preset threshold; the review module 503 is further used to determine that the review result corresponding to the target review point is fail when the maximum similarity value is less than the second preset threshold; wherein, the first preset threshold is greater than the second preset threshold.

[0086] Optionally, the device further includes: a construction module; the acquisition module 501 is further configured to acquire a pre-constructed positive and negative sample database and multiple pre-annotated corpus samples; the construction module is configured to use the multiple corpus samples and the corresponding positive and negative corpus in the positive and negative sample database to train the corpus annotation model through contrastive learning to obtain a trained target model; wherein the target model is able to distinguish between semantically similar positive corpus and semantically unrelated negative corpus.

[0087] Optionally, the construction module is further configured to annotate multiple unannotated corpora using the target model to obtain the corpus.

[0088] Optionally, the construction module is further configured to, upon receiving feedback on objections to the target review point, determine, based on the expert review results, whether the corpus contains relevant corpus of the text to be reviewed corresponding to the target review point; the construction module is further configured to, if the corpus does not contain relevant corpus of the text to be reviewed corresponding to the target review point, use the target model to annotate the text to be reviewed corresponding to the target review point, and store the annotated text to be reviewed in the corpus.

[0089] The short text-based construction plan review device provided in this application first acquires the construction plan to be reviewed and its corresponding review requirements; then, it analyzes the review requirements to determine multiple review points included in the review requirements, and extracts the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed contains multiple short texts; and based on the similarity between each short text and its corresponding review content text, it determines the review result for each review point; the review content text includes: the review content corresponding to the review sub-item; one review point includes multiple review sub-items; finally, based on the review result corresponding to each review point, it generates the final review result for the construction plan to be reviewed. In this way, the construction plan can be automatically reviewed according to the review requirements, greatly improving the review efficiency and reducing the review cost.

[0090] Figure 6 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 6 As shown, the electronic device may include a processor 610, a communications interface 620, a memory 630, and a communication bus 640, wherein the processor 610, communications interface 620, and memory 630 communicate with each other via the communication bus 640. The processor 610 can call logical instructions in the memory 630 to execute a short text-based construction plan review method. This method includes: first, obtaining the construction plan to be reviewed and its corresponding review requirements; then, analyzing the review requirements to determine multiple review points included in the review requirements, and extracting the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed contains multiple short texts; and determining the review result for each review point based on the similarity between each short text and its corresponding review content text; the review content text includes: review content corresponding to a review sub-item; one review point includes multiple review sub-items; finally, generating the final review result for the construction plan to be reviewed based on the review result corresponding to each review point. In this way, the construction plan can be automatically reviewed according to the review requirements, which greatly improves the efficiency of the review of the construction plan and reduces the review cost.

[0091] Furthermore, the logical instructions in the aforementioned memory 630 can be implemented as software functional units and, when sold or used as independent products, can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0092] On the other hand, this application also provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, and when the program instructions are executed by a computer, the computer can execute the short text-based construction plan review method provided by the above methods. This method includes: first, obtaining the construction plan to be reviewed and the corresponding review requirements; then, analyzing the review requirements, determining multiple review points included in the review requirements, and extracting the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed includes multiple short texts; and determining the review result for each review point based on the similarity between each short text and the corresponding review content text; the review content text includes: review content corresponding to a review sub-item; one review point includes multiple review sub-items; finally, generating the final review result of the construction plan to be reviewed based on the review result corresponding to each review point. In this way, construction plans can be automatically reviewed according to review requirements, greatly improving the review efficiency of construction plans and reducing review costs.

[0093] Furthermore, this application also provides a computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program performs the aforementioned methods for reviewing construction plans based on short text. This method includes: first, acquiring the construction plan to be reviewed and its corresponding review requirements; then, analyzing the review requirements to determine multiple review points included in the review requirements, and extracting the text to be reviewed corresponding to each review point from the construction plan to be reviewed; each text to be reviewed includes multiple short texts; and determining the review result for each review point based on the similarity between each short text and its corresponding review content text; the review content text includes: review content corresponding to a review sub-item; a review point includes multiple review sub-items; finally, generating the final review result for the construction plan to be reviewed based on the review result corresponding to each review point. In this way, construction plans can be automatically reviewed according to review requirements, greatly improving the efficiency of construction plan review and reducing review costs.

[0094] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0095] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0096] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application.

Claims

1. A method for reviewing construction plans based on short texts, characterized in that, include: Obtain the construction plans to be reviewed and the corresponding review requirements; The review requirements are analyzed to identify multiple review points included in the review requirements, and the text to be reviewed corresponding to each review point is extracted from the construction plan to be reviewed. The review result for each review point is determined based on the similarity between each short text and its corresponding review content text; the text to be reviewed for each review point contains multiple short texts; each review point includes multiple review sub-items, and each review sub-item has a corresponding review content text; Based on the audit results corresponding to each audit point, the final audit result of the construction plan to be audited is generated.

2. The method according to claim 1, characterized in that, The step of extracting the text to be reviewed for each review point from the construction plan to be reviewed includes: Based on the review content corresponding to each review point, the text content in the construction plan to be reviewed is extracted to obtain the text to be reviewed corresponding to each review point.

3. The method according to claim 1, characterized in that, Each audit point contains multiple audit sub-items; The process of determining the review result for each review point based on the similarity between each short text and its corresponding review content text includes: Regular expressions are used to match the review content text corresponding to each review sub-item in the target review point with the target text to be reviewed, and the matching results are obtained; the target review point is any one of the multiple review points; the target text to be reviewed is the text to be reviewed corresponding to the target review point; If the matching result indicates that no text content related to the target review item was matched from the target text to be reviewed, the review result of the target review item is determined based on the similarity between each of the multiple short texts contained in the target text to be reviewed and the review content text of the target review item. The audit result of the target audit point is determined based on the audit result of each audit sub-item in the target audit point.

4. The method according to claim 3, characterized in that, The step of determining the review result of the target review sub-item based on the similarity between each short text contained in the target text to be reviewed and the review content text of the target review sub-item includes: Calculate the text vector of each short text in the multiple short texts contained in the target text to be reviewed, and the text vector of the review content corresponding to the target review sub-item; Calculate the similarity value between the text vector of the review content corresponding to the target review sub-item and the text vector of each short text to obtain multiple similarity values; The review result of the target review sub-item is determined based on the maximum similarity value among the multiple similarity values.

5. The method according to claim 4, characterized in that, The calculation of the text vector of each short text in the multiple short texts contained in the target review point, and the text vector of the review content corresponding to the target review point, includes: The target short text is segmented into words to obtain multiple words contained in the target short text; Each of the multiple word segments is matched with the word segment corpus in the word vector library to determine the short text vector of each word segment and the corresponding construction type label; The word segmentation text vector of each word is calculated by multiplying the short text vector of each word by the weight value corresponding to the tag type. The short text vector is obtained by adding the word segmentation text vectors of each word in the multiple word segments. The ratio of the short text vector to the number of words corresponding to the multiple word segments is determined as the text vector of the target short text; The target short text is any one of the plurality of short texts, or the review content corresponding to the target review point.

6. The method according to claim 4 or 5, characterized in that, The audit results for each audit point include: passed, partially passed, and failed. Determining the review result of the target review sub-item based on the maximum similarity value among the plurality of similarity values ​​includes: If the maximum similarity value is greater than or equal to the first preset threshold, the review result corresponding to the target review sub-item is determined to be passed; or, If the maximum similarity value is less than the first preset threshold and greater than or equal to the second preset threshold, the review result corresponding to the target review sub-item is determined to be partially passed; or, If the maximum similarity value is less than the second preset threshold, the review result corresponding to the target review sub-item is determined to be unsuccessful. Wherein, the first preset threshold is greater than the second preset threshold.

7. The method according to claim 1, characterized in that, Before determining the review result for each review point based on the similarity between each short text and the corresponding review content text, the method further includes: A text training dataset containing relevant professional terms was constructed based on historical construction scheme texts; The word vector model is trained using training samples from the text training dataset to obtain the target model; Multiple word segment corpora are input into the target model to obtain the text vector of each word segment corpus. The construction type label and text vector of each word segment corpus are then associated and stored in the database to obtain the word vector library.

8. A construction plan review device based on short text, characterized in that, The device includes: The acquisition module is used to acquire the construction plan to be reviewed and the corresponding review requirements. The extraction module is used to analyze the audit requirements, determine the multiple audit points included in the audit requirements, and extract the text to be audited corresponding to each audit point from the construction plan to be audited; each text to be audited contains multiple short texts; The review module is used to determine the review result of each review point based on the similarity between each short text and the corresponding review content text; the review content text includes: the review content corresponding to the review sub-item; one review point includes multiple review sub-items; The generation module is used to generate the final review result of the construction plan to be reviewed based on the review result corresponding to each review point.

9. An electronic device, characterized in that, It includes a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the short text-based construction plan review method as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, It stores a computer program that, when executed by a processor, implements the steps of the short text-based construction plan review method as described in any one of claims 1 to 7.

Images