Method and device for analyzing and calculating formula in building field

Abstract

The invention relates to a building field formula analytical calculation method and device, which can obtain a building field industry specification file, extract formula segments from the file, and further construct a formula knowledge base; obtaining a formula query instruction input by a user, and performing retrieval enhancement generation recall in the formula knowledge base to obtain a correct formula fragment; combining the correct formula fragment and the formula query instruction into a first combination field, and performing intention identification and instruction routing on the first combination field; according to the intention recognition result and the instruction routing result, different parts of the first combination field are calculated or answered through the general large model and the special model, and an answer summary is generated. It can be understood that the technical scheme provided by the invention can carry out intention recognition and instruction routing, meanwhile, the advantages of a general large model and a special model are utilized to carry out instruction processing, respective limitations are avoided, and the accuracy, efficiency and applicability of formula analysis and calculation in the building field can be improved.

Description

Technical Field

[0001] The present invention relates to the technical field of model formula parsing, and specifically relates to a method and device for formula parsing and calculation in the field of architecture. Background Art

[0002] In the field of architecture, the parsing and calculation of formulas are of crucial significance for many aspects such as engineering design, construction calculation, and project evaluation. With the development of technology, large models have shown certain potential in processing various types of information, but still face many challenges in formula processing in the field of architecture.

[0003] Currently, in terms of formula parsing and calculation, there are mainly two methods: using a general large model for formula parsing and calculation, and using a dedicated large model for formula parsing and calculation. When using a general large model, its parsing and calculation capabilities in the field of formulas are weaker than those of the corresponding math model, and it cannot guarantee the accuracy of formula calculation results. General models usually have a large weight and consume a large amount of computing power resources. When using a dedicated large model, it is weak in semantic understanding, context understanding, and text summarization. If multi-domain problems are involved, a single-domain dedicated model may not be able to answer.

[0004] In summary, the existing general large models and dedicated models both have their respective limitations in formula parsing and calculation in the field of architecture, and it is difficult to meet the actual needs of efficiently and accurately processing formulas in the field of architecture. Therefore, there is an urgent need for a new method and device to solve these problems and improve the accuracy, efficiency, and applicability of formula parsing and calculation in the field of architecture. Summary of the Invention

[0005] In view of this, the purpose of the present invention is to provide a method, device, and equipment for formula parsing and calculation in the field of architecture, so as to solve the problem that the existing general large models and dedicated models both have their respective limitations in formula parsing and calculation in the field of architecture and it is difficult to meet the actual needs of efficiently and accurately processing formulas in the field of architecture.

[0006] According to the first aspect of the embodiments of the present invention, a method for formula parsing and calculation in the field of architecture is provided, including:

[0007] Obtain industry standard documents in the field of architecture, extract original formula fragments from the industry standard documents in the field of architecture, perform text and symbol error repair on the original formula fragments to obtain repaired formula fragments, and construct a formula knowledge base according to the repaired formula fragments according to classification rules and indexing methods. The industry standard documents in the field of architecture include various architecture-related documents containing formulas such as architectural design standards, construction specifications, and engineering calculation manuals;

[0008] Obtain the formula query instruction input by the user, parse the formula query instruction to obtain the retrieval keyword, perform retrieval enhancement in the formula knowledge base according to the retrieval keyword to generate a recall and obtain the query formula fragment, obtain the negative instruction or the confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, use the query formula fragment corresponding to the confirmation instruction as the correct formula fragment. If it is a negative instruction, parse the negative instruction to obtain the feedback information, adjust the formula query instruction according to the feedback information to obtain the adjusted formula query instruction, and execute retrieval enhancement in the formula knowledge base according to the adjusted formula query instruction to generate a recall and obtain a new query formula fragment until a confirmation instruction from the user for the new query formula fragment is received, and use the new query formula fragment corresponding to the confirmation instruction as the correct formula fragment;

[0009] Combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intent recognition and instruction routing on the first combined field to obtain the intent recognition result and instruction routing result corresponding to the first combined field. The intent recognition result and instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part;

[0010] Utilize the context understanding ability of the general large model to explain the formula explanation part, and enhance the specific domain functions of the dedicated model according to the explanation, and perform typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result. Combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, and further analyze and summarize the calculation result to generate an answer summary. The answer summary includes the first calculation result, a rationality analysis, and suggestions for subsequent engineering design or construction.

[0011] Preferably, construct a general large model based on prompt engineering according to the characteristics of formulas in the construction field, text and symbol errors, as well as construction field knowledge and mathematical rules, and input the original formula fragment into the general large model based on prompt engineering to obtain the repaired formula fragment.

[0012] Preferably, the factors for formulating the classification rules include the formula application field, formula type, and / or formula complexity, and the factors for formulating the indexing method include the formula source file, the location of the formula in the source file, and the formula-related parameter description.

[0013] Preferably, obtain construction-specific domain data to fine-tune and train the multi-classification model to obtain the trained multi-classification model. The trained multi-classification model performs intent recognition and instruction routing on the first combined field to obtain a formula explanation part, a typical calculation part, and a substitution calculation part.

[0014] Preferably, the building-specific field data includes actual engineering project cases in the building field, expert experience data, and a collection of common question answers.

[0015] According to the second aspect of the embodiments of the present invention, there is provided a device for parsing and calculating formulas in the building field, which adopts a method for parsing and calculating formulas in the building field as described above. The device includes:

[0016] A knowledge base construction module, configured to extract original formula fragments from the building field industry specification documents, perform text and symbol error repair on the original formula fragments to obtain repaired formula fragments, and construct a formula knowledge base according to the repaired formula fragments according to classification rules and indexing methods. The building field industry specification documents include various building field documents containing formulas such as building design standards, construction specifications, and engineering calculation manuals;

[0017] A retrieval module, configured to obtain a formula query instruction input by a user, parse the formula query instruction to obtain a retrieval keyword, perform retrieval enhancement in the formula knowledge base according to the retrieval keyword to generate a recall to obtain a query formula fragment, obtain a negative instruction or a confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, the query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment. If it is a negative instruction, the negative instruction is parsed to obtain feedback information, and the formula query instruction is adjusted according to the feedback information to obtain an adjusted formula query instruction. The adjusted formula query instruction is used to perform retrieval enhancement in the formula knowledge base to generate a new query formula fragment until a confirmation instruction of the user for the new query formula fragment is received, and the new query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment;

[0018] An intention recognition module, configured to combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intention recognition and instruction routing on the first combined field to obtain an intention recognition result and an instruction routing result corresponding to the first combined field. The intention recognition result and the instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part;

[0019] An answer summary module, configured to use the context understanding ability of a general large model to explain the formula explanation part, enhance the specific field function of a dedicated model according to the explanation, perform typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result, combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, further analyze and summarize the calculation result, and generate an answer summary. The answer summary includes a first calculation result, a rationality analysis, and suggestions for subsequent engineering design or construction.

[0020] The technical solution shown in the present invention can obtain industry standard documents in the construction field, extract formula fragments from them, and then construct a formula knowledge base; obtain a formula query instruction input by the user, perform retrieval enhancement generation recall in the formula knowledge base to obtain the correct formula fragment; combine the correct formula fragment and the formula query instruction into a first combined field, and perform intent recognition and instruction routing on it; according to the intent recognition result and the instruction routing result, use a general large model and a dedicated model to calculate or answer different parts of the first combined field respectively, and generate an answer summary. The technical solution provided by the present invention realizes the construction of a separate formula knowledge base to improve the RAG recall effect; can perform intent recognition and instruction routing, and at the same time utilizes the advantages of the general large model and the dedicated model for instruction processing, avoiding their respective limitations, and combines the outputs of the general large model and the dedicated model to generate an answer summary, improving the accuracy, efficiency and applicability of formula parsing and calculation in the construction field. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The accompanying drawings herein are incorporated into and constitute a part of this specification, showing embodiments consistent with the present invention and, together with the specification, are used to explain the principles of the present invention.

[0022] Figure 1 is a schematic diagram of the steps of a method for formula parsing and calculation in the construction field shown according to an exemplary embodiment;

[0023] Figure 2 is a schematic diagram of the knowledge base construction process shown according to an exemplary embodiment;

[0024] Figure 3 is a schematic diagram of the formula retrieval process shown according to an exemplary embodiment;

[0025] Figure 4 is a schematic diagram of model selection based on functional routing shown according to an exemplary embodiment;

[0026] Figure 5 is a schematic diagram of the process of generating an answer summary shown according to an exemplary embodiment. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0027] Here, the exemplary embodiments will be described in detail, and the examples are shown in the accompanying drawings. When the following description refers to the accompanying drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present invention. On the contrary, they are merely examples of devices and methods consistent with some aspects of the present invention as detailed in the appended claims.

[0028] In one embodiment, Figure 1It is a schematic diagram of the steps of a method for formula parsing and calculation in the construction field shown according to an exemplary embodiment. Refer to Figure 1 , a method for formula parsing and calculation in the construction field is provided, including:

[0029] Step 1: Obtain the industry standard documents in the construction field, extract the original formula fragments from the industry standard documents in the construction field, perform text and symbol error repair on the original formula fragments to obtain the repaired formula fragments, and construct a formula knowledge base according to the classification rules and indexing methods for the repaired formula fragments. Refer to Figure 2 . The original formula fragments are extracted according to a preset algorithm and pattern recognition rules.

[0030] Preferably, a general large model based on prompt engineering is constructed according to the characteristics of the formulas in the construction field, text and symbol errors, construction field knowledge, and mathematical rules. The original formula fragments are input into the general large model based on prompt engineering to obtain the repaired formula fragments, so as to guide the general large model to accurately identify and repair text and symbol errors in the formulas. For example, if there is a spelling error in the variable name in the formula, such as "Lengt" should be "Length", the general large model can recognize and correct this error under the guidance of the prompt. For symbol errors, such as mismatched parentheses, incorrect use of operators, etc., the general large model can also correct them.

[0031] Preferably, the industry standard documents in the construction field are scanned, parsed, and recognized using optical character recognition technology (OCR), and the paper documents or image format files are converted into an editable text format to extract the original formula fragments. For example, in the building structure design specifications, formula fragments such as "Bending moment calculation formula of beam: M = qL2 / 8" are recognized. In addition, for some complex formulas, there may be multiple representation forms or layout formats, and the system will perform normalization processing to convert them into a unified standard format for subsequent operations.

[0032] Preferably, the industry standard documents in the construction field include various documents containing formulas such as building design standards, construction specifications, and engineering calculation manuals.

[0033] Preferably, the classification rules are formulated based on the application fields, formula types, and / or formula complexity of the formulas. Among them, the application fields include: structural calculation, water supply and drainage calculation, electrical calculation, and the formula types include algebraic formulas, geometric formulas, and physical formulas. For example, all formulas related to building structural mechanics are classified into one category, and then further subdivided according to the specific mechanical principles involved in the formulas (such as beam bending, column stability, etc.). The indexing methods are formulated based on factors such as the source documents of the formulas, their positions in the documents, and relevant parameter descriptions, so as to be able to quickly and accurately locate and call the formulas during subsequent retrieval processes.

[0034] In this step, the industry specification documents in the construction field are processed through OCR technology, which can efficiently identify and extract formula fragments therein, convert the formulas in paper or image form into an editable text format, and lay a foundation for subsequent operations. Based on prompt engineering, the general large model is used to repair text and symbol errors in the formulas. With the powerful language understanding ability and extensive knowledge reserve of the general large model, various errors that may exist in the formulas, such as variable naming errors, symbol confusion, and syntax errors, can be accurately identified and corrected, ensuring the accuracy of the formulas. Compared with the traditional manual inspection and repair methods, this method greatly improves the efficiency, reduces human errors, and can process a large number of formula texts, ensuring the high quality of formula parsing and providing a reliable premise for subsequent calculations and applications.

[0035] Step 2: Obtain the formula query instruction input by the user, parse the formula query instruction to obtain the retrieval keyword, perform retrieval enhancement in the formula knowledge base according to the retrieval keyword to generate a recall and obtain the query formula fragment, obtain the negative instruction or confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, use the query formula fragment corresponding to the confirmation instruction as the correct formula fragment. If it is a negative instruction, parse the negative instruction to obtain the feedback information, adjust the formula query instruction according to the feedback information to obtain the adjusted formula query instruction, and execute retrieval enhancement in the formula knowledge base according to the adjusted formula query instruction to generate a recall and obtain a new query formula fragment until a confirmation instruction from the user for the new query formula fragment is received, and use the new query formula fragment corresponding to the confirmation instruction as the correct formula fragment. See Figure 3 。

[0036] Users can input formula query instructions through the man-machine interaction interface. The formula query instructions include the requirements for specific formulas, such as "query the deflection calculation formula of beams in building structures", and calculation problems involving formulas, such as "given that the span of a beam is 5 meters and the uniformly distributed load is 10 kN / m, find the maximum deflection of the beam". Preferably, they are sorted according to relevance and importance, and the top-ranked query formula fragments are output to the user.

[0037] Negative instructions include only pointing out output errors, the applicable conditions of the output formula not meeting the actual requirements, and formula type errors. Negative instructions are used to describe the user's confirmation of the content of the output formula. If it is found that the output formula does not match the user's expectations, such as the applicable conditions of the formula not meeting the actual requirements or the formula type being incorrect, an instruction issued through the man-machine interaction interface operation. For example, if the user believes that the previously retrieved formula is for simply supported beams, while the actual requirement is the deflection calculation formula for continuous beams, the system will further screen the formula fragments related to continuous beams in the knowledge base and output them to the user again.

[0038] This step constructs a formula knowledge base and performs formula recognition and extraction, enabling users to quickly and accurately obtain the required formulas. After the user inputs a formula query instruction, the system can quickly locate relevant formula fragments through retrieval and recall in the knowledge base. Since the knowledge base is constructed based on construction field specification documents, the formulas in it are highly professional and accurate, so the quality of the retrieval results is high.

[0039] Step 3: Combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intent recognition and instruction routing on the first combined field to obtain the intent recognition result and instruction routing result corresponding to the first combined field. The intent recognition result and instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part.

[0040] Obtain construction-specific domain data to fine-tune and train a multi-classification model to obtain a trained multi-classification model. The trained multi-classification model performs intent recognition and instruction routing on the first combined field to obtain a formula explanation part, a typical calculation part, and a substitution calculation part.

[0041] The construction-specific domain data includes actual engineering project cases in the construction field, expert experience data, common question and answer sets, etc., enabling the multi-classification model to accurately identify the intent in the first combined field. For example, it can determine which part of the user's input instruction requires formula explanation, which part requires typical calculation, or which part requires substitution calculation, etc.

[0042] In this step, through the user's confirmation of the output formula and the instruction adjustment mechanism, the system can continuously optimize the retrieval results until the correct formula is provided. This process improves the flexibility and accuracy of formula retrieval, meets the diverse needs of users for formulas in different scenarios, saves the time and effort of users in searching for formulas, and improves work efficiency.

[0043] Step 4: Use the context understanding ability of a general large model to explain the formula explanation part. Based on the explanation, use the specific domain functions of a dedicated model to enhance the typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result. Combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, further analyze and summarize the calculation result, and generate an answer summary. The answer summary includes the first calculation result, a rationality analysis, and suggestions for subsequent engineering design or construction. See Figure 4 and Figure 5 。

[0044] Preferably, a rationality analysis is obtained based on the first calculation result in combination with the knowledge in the construction field and the actual application scenario, and the rationality analysis is compared with relevant standards and specifications to obtain subsequent engineering design or construction suggestions.

[0045] The general large model can understand the knowledge system and context information of formulas in the construction field, and provide users with detailed explanations such as the meaning of formulas, scope of application, and meaning of variables. For example, for the seismic calculation formula in building structures, the general large model can explain how each parameter in the formula reflects factors such as seismic action and structural characteristics, as well as the application of this formula under different building types and seismic fortification intensities.

[0046] The special model has more in-depth professional knowledge and precise calculation capabilities in the construction field, and can perform accurate calculations according to the requirements of the formula and the parameters provided by the user. For example, when calculating the internal forces of a building structure, the special model can accurately calculate the internal force values of components such as beams and columns based on the geometric shape of the structure, material properties, load conditions, etc.

[0047] In this step, the formula query instruction, the correct formula fragment, and the calculation result are combined into a second combined field, and the second combined field is input into the general large model. The general large model uses its powerful text summarization ability to further analyze and summarize the calculation result, generating an answer summary that can ensure the integrity and relevance of each part of the information, so that the general large model can comprehensively understand the calculation background and the meaning of the result. For example, the answer summary not only includes the numerical value of the calculation result, but also combines the knowledge in the construction field and the actual application scenario to give content such as the rationality analysis of the result, the comparison with relevant standards and specifications, and the suggestions for subsequent engineering design or construction. For example, after calculating the settlement of a building foundation, the answer summary of the general large model may point out whether the calculation result is within the allowable range, the comparison with the settlement cases of building foundations under similar geological conditions, and the engineering measures that need to be taken if the settlement is large.

[0048] The model selection strategy based on functional routing in this step effectively solves the respective limitations of the general large model and the special model. The general large model has advantages in semantic understanding and context understanding. By using it to explain formulas, it can provide users with in-depth and comprehensive information such as the meaning of formulas and scope of application, helping users better understand and apply formulas. The special model has strong calculation capabilities in specific fields. By using it for typical calculations and substitution calculations, it can ensure the accuracy and professionalism of the calculation results. The collaborative work of the two models gives full play to their respective advantages, avoids the deficiencies of a single model in dealing with formulas in the construction field, and improves the ability and quality of the entire system to process formulas.

[0049] In practical applications, for example, during the building structure design process, designers need to calculate the internal forces of beams in a certain frame structure. The designer inputs the formula query instruction "Calculate the internal forces of the frame beam under vertical and horizontal loads" through the system interface. The system first retrieves relevant formulas in the formula knowledge base, and after confirmation by the designer, obtains the correct formula for calculating the internal forces of the beam. Then, through intention recognition and instruction routing, it is determined that a dedicated model is required for the calculation. The dedicated model calculates based on parameters such as the cross-sectional dimensions, material properties, and load distribution of the beam provided by the designer, and obtains internal force values such as the bending moment and shear force of the beam. Finally, the calculation results are combined with the original instruction and formula fragments and input into the general large model, and the general large model generates a response summary, including the specific values of the internal force values, their distribution in the structure, whether they meet the design requirements, and suggestions for the reinforcement design of the beam, etc.

[0050] The technical solution shown in the present invention can obtain industry standard documents in the construction field, extract formula fragments from them, and then construct a formula knowledge base; obtain the formula query instruction input by the user, perform retrieval enhancement generation recall in the formula knowledge base, and obtain the correct formula fragment; combine the correct formula fragment and the formula query instruction into a first combined field, and perform intention recognition and instruction routing on it; according to the intention recognition result and instruction routing result, use the general large model and the dedicated model to calculate or answer different parts of the first combined field respectively, and generate a response summary. It can be understood that the technical solution provided by the present invention realizes the construction of a separate formula knowledge base to improve the RAG recall effect; can perform intention recognition and instruction routing, and at the same time utilizes the advantages of the general large model and the dedicated model for instruction processing, avoids their respective limitations, and combines the outputs of the general large model and the dedicated model to generate a response summary. It improves the accuracy, efficiency, and applicability of formula parsing and calculation in the construction field.

[0051] According to the second aspect of the embodiments of the present invention, there is provided a device for formula parsing and calculation in the construction field, adopting the method for formula parsing and calculation in the construction field as described above. The device includes:

[0052] A knowledge base construction module, configured to extract original formula fragments from the industry standard documents in the construction field, perform text and symbol error repair on the original formula fragments to obtain repaired formula fragments, and construct a formula knowledge base according to the repaired formula fragments according to classification rules and indexing methods. The industry standard documents in the construction field include various construction field documents containing formulas such as building design standards, construction specifications, and engineering calculation manuals;

[0053] A retrieval module, which is used to obtain a formula query instruction input by a user, parse the formula query instruction to obtain a retrieval keyword, perform retrieval enhancement in a formula knowledge base according to the retrieval keyword to generate a recall and obtain a query formula fragment, obtain a negative instruction or a confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, the query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment. If it is a negative instruction, the negative instruction is parsed to obtain feedback information, the formula query instruction is adjusted according to the feedback information to obtain an adjusted formula query instruction, and retrieval enhancement is performed in the formula knowledge base according to the adjusted formula query instruction to generate a recall and obtain a new query formula fragment until a confirmation instruction of the user for the new query formula fragment is received, and the new query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment;

[0054] An intention recognition module, which is used to combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intention recognition and instruction routing on the first combined field to obtain an intention recognition result and an instruction routing result corresponding to the first combined field. The intention recognition result and the instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part;

[0055] An answer summarization module, which is used to utilize the context understanding ability of a general large model to explain the formula explanation part, enhance the specific domain function of a dedicated model according to the explanation, perform typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result, combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, further analyze and summarize the calculation result, and generate an answer summary. The answer summary includes the first calculation result, a rationality analysis, and subsequent engineering design or construction suggestions.

[0056] It can be understood that the same or similar parts in the above embodiments can be referred to each other, and the content not detailed in some embodiments can be referred to the same or similar content in other embodiments.

[0057] It should be noted that in the description of the present invention, terms such as "first" and "second" are only used for descriptive purposes and cannot be understood as indicating or implying relative importance. In addition, in the description of the present invention, unless otherwise specified, the meaning of "a plurality" refers to at least two.

[0058] Any process or method description depicted in a flowchart or described otherwise herein may be understood to represent a module, segment, or portion of code including one or more executable instructions for implementing a specific logical function or process. The scope of the preferred embodiments of the present invention includes additional implementations where functions may be executed in a manner that is not shown or discussed, including in a substantially simultaneous manner according to the involved functions or in a reverse order, which should be understood by those skilled in the art to which the embodiments of the present invention pertain.

[0059] It should be understood that various parts of the present invention may be implemented using hardware, software, firmware, or a combination thereof. In the above-described embodiments, multiple steps or methods may be implemented using software or firmware stored in a memory and executed by a suitable instruction execution system. For example, if implemented using hardware, as in another embodiment, any one or a combination of the following techniques well known in the art may be used: discrete logic circuits having logic gate circuits for implementing logical functions on data signals, application specific integrated circuits having appropriate combinational logic gate circuits, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

[0060] Those of ordinary skill in the art of this technology can understand that all or part of the steps carried by the method of implementing the above embodiments can be completed by instructing relevant hardware through a program. The said program can be stored in a computer-readable storage medium. When the program is executed, it includes one or a combination of the steps of the method embodiments.

[0061] In addition, in each embodiment of the present invention, the functional units can be integrated into one processing module, or each unit can exist physically alone, or two or more units can be integrated into one module. The above integrated module can be implemented in the form of hardware or in the form of a software functional module. When the above integrated module is implemented in the form of a software functional module and sold or used as an independent product, it can also be stored in a computer-readable storage medium.

[0062] The above-mentioned storage medium can be a read-only memory, a magnetic disk, an optical disk, etc.

[0063] In the description of this specification, the description with reference to terms such as "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials, or characteristics described in connection with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials, or characteristics described can be combined in a suitable manner in any one or more embodiments or examples.

[0064] Although the embodiments of the present invention have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limiting the present invention. Those of ordinary skill in the art can make changes, modifications, substitutions, and variations to the above embodiments within the scope of the present invention.

Claims

1. A method for parsing and calculating formulas in the field of architecture, characterized in that, Including: Obtain industry standard documents in the construction field, extract original formula fragments from the industry standard documents in the construction field, repair text and symbol errors according to the original formula fragments to obtain repaired formula fragments, and construct a formula knowledge base according to the repaired formula fragments according to classification rules and indexing methods. The industry standard documents in the construction field include various construction field documents containing formulas such as building design standards, construction specifications, and engineering calculation manuals; Obtain the formula query instruction input by the user, parse the formula query instruction to obtain a retrieval keyword, perform retrieval enhancement in the formula knowledge base according to the retrieval keyword to generate a recall to obtain a query formula fragment, and obtain the negative instruction or confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, use the query formula fragment corresponding to the confirmation instruction as the correct formula fragment. If it is a negative instruction, parse the negative instruction to obtain feedback information, adjust the formula query instruction according to the feedback information to obtain an adjusted formula query instruction, and perform retrieval enhancement in the formula knowledge base according to the adjusted formula query instruction to generate a recall to obtain a new query formula fragment until a confirmation instruction from the user for the new query formula fragment is received, and use the new query formula fragment corresponding to the confirmation instruction as the correct formula fragment; Combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intent recognition and instruction routing on the first combined field to obtain the intent recognition result and instruction routing result corresponding to the first combined field. The intent recognition result and instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part; Use the context understanding ability of the general large model to explain the formula explanation part, enhance the specific domain function of the dedicated model according to the explanation, perform typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result, combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, further analyze and summarize the calculation result, and generate an answer summary. The answer summary includes the first calculation result, a rationality analysis, and suggestions for subsequent engineering design or construction; 2. The method for formula parsing and calculation in the field of architecture according to claim 1, wherein Construct a general large model based on prompt engineering according to the characteristics of construction field formulas, text and symbol errors, construction field knowledge, and mathematical rules, and input the original formula fragment into the general large model based on prompt engineering to obtain a repaired formula fragment.

3. A method for parsing and calculating formulas in the field of architecture according to claim 1, characterized in that, The factors for formulating classification rules include formula application fields, formula types, and / or formula complexity levels. The factors for formulating indexing methods include formula source documents, the positions of formulas in source documents, and formula-related parameter descriptions.

4. A method for parsing and calculating formulas in the field of architecture according to claim 1, characterized in that, Obtain construction-specific domain data to fine-tune and train a multi-classification model to obtain a trained multi-classification model. The trained multi-classification model performs intent recognition and instruction routing on the first combined field to obtain a formula explanation part, a typical calculation part, and a substitution calculation part.

5. A method for parsing and calculating formulas in the field of architecture according to claim 4, characterized in that, The construction-specific domain data includes actual engineering project cases in the construction field, expert experience data, and common question and answer sets.

6. An apparatus for parsing and calculating formulas in the field of architecture, characterized in that, Adopt a method for formula parsing and calculation in the field of architecture as described in claim 1. The device includes: A knowledge base construction module, which is used to extract original formula fragments from the industry specification documents in the field of architecture, repair text and symbol errors according to the original formula fragments to obtain repaired formula fragments, and construct a formula knowledge base according to the repaired formula fragments according to classification rules and indexing methods. The industry specification documents in the field of architecture include various architecture documents containing formulas such as architectural design standards, construction specifications, and engineering calculation manuals; A retrieval module, which is used to obtain a formula query instruction input by the user, parse the formula query instruction to obtain a retrieval keyword, perform retrieval enhancement in the formula knowledge base according to the retrieval keyword to generate a recall to obtain a query formula fragment, and obtain a negative instruction or a confirmation instruction of the user for the query formula fragment. If it is a confirmation instruction, the query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment. If it is a negative instruction, the negative instruction is parsed to obtain feedback information, the formula query instruction is adjusted according to the feedback information to obtain an adjusted formula query instruction, and the adjusted formula query instruction is used to perform retrieval enhancement in the formula knowledge base to generate a recall to obtain a new query formula fragment until a confirmation instruction of the user for the new query formula fragment is received, and the new query formula fragment corresponding to the confirmation instruction is used as the correct formula fragment; An intention recognition module, which is used to combine the correct formula fragment and the corresponding formula query instruction or the adjusted formula query instruction into a first combined field, perform intention recognition and instruction routing on the first combined field to obtain the intention recognition result and instruction routing result corresponding to the first combined field. The intention recognition result and instruction routing result include a formula explanation part, a typical calculation part, and a substitution calculation part; An answer summary module, which is used to explain the formula explanation part by using the context understanding ability of a general large model, enhance the specific domain function of a dedicated model according to the explanation, perform typical calculation and substitution calculation on the typical calculation part and the substitution calculation part in sequence to generate a first calculation result, combine the formula query instruction, the correct formula fragment, and the calculation result into a second combined field, input the second combined field into the general large model, further analyze and summarize the calculation result, and generate an answer summary. The answer summary includes the first calculation result, a rationality analysis, and suggestions for subsequent engineering design or construction.

Images