Process data link intelligent construction method based on large model

Abstract

The invention discloses a process data link intelligent construction method based on a large model. The method comprises the following steps: setting a data structure of a construction process data link; constructing a structured prompt word library; establishing a construction technology professional large model; outputting a process step list according to the professional large model; outputting input and output data of the process steps through the professional large model; the professional large model establishes a tight-before-tight relationship between the process steps according to the input and output data output by the professional large model; judging whether the process data chain is correct or not; if yes, the professional large model outputs a structured process data chain; and otherwise, correcting the structured prompt word bank. The method has the advantage of high construction process data link construction efficiency, overcomes the problem of low efficiency of manual construction of the process data link, realizes intelligentization of construction of the construction process data link, and provides a basis for digital construction software development and intelligent construction equipment research and development.

Description

Technical Field

[0001] The present invention relates to the field of intelligent construction, and in particular to a method for intelligently constructing a process data chain based on a large model. Background Art

[0002] In the field of intelligent buildings, the process data chain is a digital representation of the construction process, and the construction of the construction process data chain is the basis for promoting intelligent construction. Traditional construction process data are mainly non-standardized data such as construction plans and construction specifications. Manual reading of construction process data to summarize, extract, and mark the data structure of the construction process is labor-intensive, difficult, and time-consuming. In addition, there are problems such as "multiple words for one meaning" in the construction process data, which affects the accuracy of the data structure. Therefore, how to improve the construction efficiency of the process data chain has become a technical problem that needs to be solved urgently in this field. Summary of the invention

[0003] The purpose of the present invention is to provide a method for intelligently constructing a process data chain based on a large model to solve the problem of low efficiency in constructing a construction process data chain.

[0004] In order to solve the above technical problems, the present invention provides a technical solution: a method for intelligently constructing a process data chain based on a large model, comprising: Set the data structure of the construction process data chain; Build a structured prompt vocabulary; Establish a professional large model of construction technology; Output the process step list according to the professional large model; Output the input and output data of the process steps through professional large models; The professional large model establishes the close relationship between process steps based on the input and output data it outputs; Determine whether the process data chain is correct; if correct, the professional large model outputs a structured process data chain; otherwise, the structured prompt word library is corrected.

[0005] Furthermore, the method for intelligently constructing a process data chain based on a large model provided by the present invention includes: setting a construction process step set P = {step p i =(Step Name pn i , input data set Ip i , output dataset Op i )} and step relation set PR = (relation pr ij ), where i and j are positive integers, p i For the i-th step, pn i is the name of the ith step, Ip iis the input data set of the ith step, Op i is the output data set of the i-th step, pr ij is the relationship between the i-th step and the j-th step.

[0006] Furthermore, the present invention provides a method for intelligently constructing a process data chain based on a large model, and the method for constructing a structured prompt word library includes: using a thinking chain method to write structured prompt words of the data structure of the process data chain, and constructing a structured prompt word library Pro={role definition word pro, process step list output prompt word ppn, process step output data prompt word pop, process step input data prompt word pip}.

[0007] Furthermore, the method for intelligently constructing a process data chain based on a large model provided by the present invention includes: If there is an error in the step, enter the correction prompt word mppn of the corresponding step; if the correction prompt word mppn is not empty, append the correction prompt word after the prompt word; If the input or output data of a step is wrong, enter the correction prompt word mpop corresponding to the input or output data i and mpip i , forming a modified prompt word set Pop={mpop i} and Pip={mpip i}. If the correction prompt words Pop and Pip are not empty, the correction prompt words are appended to the prompt words.

[0008] Furthermore, the method for intelligently constructing a process data chain based on a large model provided by the present invention includes: inputting a role definition word pro and a process step list output prompt word ppn into the professional large model Mp i Automatically output a list of process steps for the construction process represented by the role definition word.

[0009] Furthermore, the method for intelligently constructing a process data chain based on a large model provided by the present invention includes: inputting the final process step input data prompt word pip and the process step output data prompt word pop into the professional large model Mp respectively. i Output step pn i Input data list Ip i and output data list Op i .

[0010] Furthermore, the method for intelligently constructing a process data chain based on a large model provided by the present invention, wherein the method for establishing a close relationship between process steps according to the input and output data output by the professional large model comprises: traversing each step p in P i If p i The input dataset Ip i There is data element Ip ij Other steps p in P k The output dataset Op k , then establish a close before and after relationship r ik = (immediately preceding step p k , the next step p i ), and add it to PR; if p i The output dataset Op i There is data element Op in ij Other steps p in P k The input set Op k , then establish a close before and after relationship r ik = (the next step p i , the previous step p k ), and added to the step relation set PR, where k is a positive integer.

[0011] Furthermore, the present invention provides an intelligent construction method for a process data chain based on a large model, and the method for judging whether the process data chain is correct includes: judging whether the step name, step sequence, and input and output data of each step of the construction process are accurate and complete according to the output of the professional large model, and whether they meet the construction requirements; if so, it is correct, otherwise it is incorrect.

[0012] Furthermore, the present invention provides a method for intelligently constructing a process data chain based on a large model, and the method for outputting a structured process data chain using a professional large model includes: fine-tuning the process step name, sequence, and input and output data of each step; and outputting a step list and a step sequence list in a standard format.

[0013] Furthermore, the present invention provides a method for intelligently constructing a process data chain based on a large model, and the method for establishing a professional large model of construction technology includes: inputting professional corpus of each construction technology of the construction node into the basic large model for fine-tuning to form a large process model containing construction technology data.

[0014] Compared with the prior art, the present invention has the following beneficial effects: The present invention provides a method for intelligently constructing a process data chain based on a large model. The method automatically constructs a structured process data chain consisting of a list of construction process steps, input and output data, and establishes a close relationship between the input and output data according to the data structure of the process data chain and a structured prompt vocabulary through a professional large model. The method has the advantage of high efficiency in constructing a construction process data chain, overcomes the problem of low efficiency in manually constructing a process data chain, realizes the intelligent construction of the construction process data chain, provides a basis for the development of digital construction software and the research and development of intelligent construction equipment, reduces the workload of manual processing, reduces the difficulty of constructing the process data chain, and shortens the construction period of the process data chain.

[0015] The large-model-based intelligent construction method of the process data chain provided by the present invention draws on the thinking chain method to guide the use of a professional large model to complete the construction of a structured prompt vocabulary of the construction process data chain through process step list output and process step input and output data output decomposition analysis, thereby greatly improving the efficiency and accuracy of the construction of the structured prompt vocabulary.

[0016] The present invention provides an intelligent construction method for a process data chain based on a large model. Aiming at the problem of "hallucination" existing in professional large models, an expert mode is adopted to manually review and judge and correct the structured prompt vocabulary, so as to continuously optimize the output results of the process data chain of the professional large model and improve the accuracy, completeness and comprehensiveness of the constructed process data chain.

[0017] The present invention provides a method for intelligently constructing a process data chain based on a large model. Aiming at the problem that the process corpus lacks description of the close relationship between construction process steps, a professional large model is used to establish the close relationship between process steps according to input and output data, thereby ensuring the accuracy of the process data chain and the correlation between the steps.

[0018] The present invention provides a method for intelligently constructing a process data chain based on a large model. The method outputs a structured process data chain of a construction process through an intelligent professional large model. The method is suitable for quickly establishing a data chain for a construction process with a large amount of data, thereby improving the work efficiency of digital transformation. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a flow chart of the intelligent construction method of process data chain based on large model. DETAILED DESCRIPTION

[0020] The present invention is described in detail below in conjunction with the accompanying drawings: The advantages and features of the present invention will become more apparent from the following description. It should be noted that the accompanying drawings are in very simplified form and in non-precise proportions, and are only used to conveniently and clearly assist in explaining the purpose of the embodiments of the present invention.

[0021] Please refer to Figure 1, an embodiment of the present invention provides a method for intelligently constructing a process data chain based on a large model, comprising: Step 1: Set the data structure of the process data chain. This includes setting the set of construction process steps P = {step p i =(Step Name pn i , input data set Ip i , output dataset Op i )} and step relation set PR = (relation pr ij ), where i and j are positive integers, p i For the i-th step, pn i is the name of the ith step, Ip i is the input data set of the ith step, Op i is the output data set of the i-th step, pr ij is the relationship between the i-th step and the j-th step. For example: concrete preparation process P = {p1 = (mix design, Ip1, Op1), p2 = (raw material preparation and testing, Ip2, Op2), p3 = (weighing and batching, Ip3, Op3), p4 = (mixing control, Ip4, Op4), p5 = (discharging detection, Ip5, Op5), p6 = (concrete transportation, Ip6, Op6)}. Among them, "mix design", "raw material preparation and testing", "weighing and batching", "mixing control", "discharging detection", and "concrete transportation" are the names of the steps in the concrete preparation process.

[0022] Step 2, build a structured prompt word library. Use the thinking chain method to write the structured prompt words of the data structure of the process data chain, and build a structured prompt word library Pro={role definition word pro, process step list output prompt word ppn, process step output data prompt word pop, process step input data prompt word pip}. Specifically, it may include: Step 2.1, write the role definition word pro={"You are an expert in [construction technology]"}; the variable [construction technology] is input by the user during actual use, that is, the name of P; for example: [construction technology] is the concrete preparation process.

[0023] Step 2.2, write the process step list output prompt word ppn={"Please output the names of all steps of [construction process] in list form"}.

[0024] Step 2.3, write the process step output data prompt pop = {"Please output the input data of [step name] of [construction process] in list form"}; variable [step name] is the step name pn in P i .

[0025] For example, the prompt for the first step of the concrete preparation process is: "Please output the input data for the mix design of the concrete preparation process in a list format."

[0026] Step 2.4, write the process step input data prompt word pip={"Please output the output data of [step name] of [construction process] in list form"}. When constructing the structured prompt word, "multiple words with the same meaning" can be unified or marked to improve the accuracy of the data structure.

[0027] Step 3: Establish a large model of construction technology. This may include: Step 3.1: According to the name of the construction process P of the process data chain to be constructed, collect the electronic version of professional corpus of the construction process such as the relevant standards, construction plans, and construction briefings of the construction process P; perform format conversion, content cleaning, and key information annotation on the electronic version of the professional corpus to generate a standard corpus set Y containing the construction process P. i The standard corpus set Y i It is not limited to the construction process P, but can include the construction processes of all construction nodes. The format conversion of professional corpus is to meet the fine-tuning requirements of the basic large model M0 to be selected. Content cleaning can improve the accuracy of the standard corpus set, and key information annotation can improve the analysis of the corpus set by the basic large model. The basic large model can be a general large model or other commercial large models.

[0028] Step 3.2, through the standard corpus set Y i Fine-tune the basic large model M0 to form a professional large model Mp containing corresponding construction process knowledge i For example: a professional large model containing construction process P. The professional large model is not limited to only containing construction process P, and each construction process of each construction node can be integrated into the professional model. Step 3 is in parallel with steps 1 and 2, and the order can be interchanged. By establishing a professional large model, the accuracy of the output results can be improved.

[0029] Step 4: Output the process step list through the professional large model. Specifically, the role definition word pro and the process step list output prompt word ppn are input into the professional large model Mp i Automatically output the process step list of the construction process and store it in the construction process step set P={p i ={Step Name pn i For example, ppn = "You are an expert in concrete preparation technology. Please use a list to output the names of the process steps in concrete preparation technology."

[0030] Step 5: Output the input and output data of the process steps through a professional large model.

[0031] Step 5.1, traverse the names of each process step in the construction process P i ; For example, the mix design steps.

[0032] Step 5.2, use pn i Replace the [step name] parameter in the process step output data prompt word pop and the process step input data prompt word pip to form the final process step output data prompt word pop and process step input data prompt word pip. For example: pip = "Please output the input data of the mix design step of the concrete preparation process in list form."

[0033] Step 5.3, enter the final process step data prompt word pip into the professional large model Mp i , professional large model output steps pn i Input data list Ip i , stored in P={p i ={Step Name pn i , input data set Ip i}}. For example, the output result p i ={Mix design, {concrete strength, expansion, impermeability grade}}.

[0034] Step 5.4, input the final process step output data prompt word pop into the professional large model Mp i , professional large model analysis output steps pn i Output data list Op i , stored in P={p i ={Step Name pn i , input data set Ip i , output dataset Op i}}. For example, the output result p i ={Mix ratio design, {water-cement ratio, sand ratio, admixture dosage, cement content, sand content, gravel content, admixture content}}.

[0035] The order of step 5.3 and step 5.4 can be interchanged.

[0036] Step 6: The professional large model establishes the close relationship between the process steps based on the input and output data it outputs: Step 6.1, traverse each step p in P i .

[0037] Step 6.2, if p i The input dataset Ip i There is data element Ip ij Other steps p in P kThe output dataset Op k , then establish a close before and after relationship r ik = (immediately preceding step p k , the next step p i ), and add it to PR. For example, the input data "cement content" of the weighing batching step is the output of the mix design step, then the relationship r ik = (mix design, weighing ingredients) added to PR.

[0038] Step 6.3, if p i The output dataset Op i There is data element Op in ij Other steps p in P k The input set Op k , then establish a close before and after relationship r ik = (the next step p i , the previous step p k ), and added to PR, where k is a positive integer.

[0039] Step 7: Determine whether the process data chain is correct. According to the output of the professional large model, determine whether the step name, step sequence, input and output data of each step of the construction process are accurate and complete, and whether they meet the construction requirements; if so, it is correct, and step 8 is executed. The professional large model outputs the structured process data chain of the relevant construction process; otherwise, it is incorrect, and the structured prompt word library constructed in step 2 is corrected. This step requires manual judgment by industry experts.

[0040] Step 8: The professional large model outputs a structured process data chain, which may include: Step 8.1, fine-tune the process step name, sequence and input and output data of each step. This step 8.1 is manually adjusted by experts.

[0041] Step 8.2: Output the step list P and step sequence list RP in a standard format.

[0042] Fine-tuning and standard format output enable structured output of process data chains.

[0043] In order to improve the accuracy, completeness and comprehensiveness of the structured process data chain output by the professional large model, the embodiment of the present invention provides a method for intelligently constructing a process data chain based on the large model. The method for correcting the structured prompt word library may include: Step 2.5, if there is an error in the step, enter the correction prompt word mppn of the corresponding step; if the correction prompt word mppn is not empty, append the correction prompt word to the prompt word, that is, ppn=ppn+mppn; for example: before correction, the output is {raw material preparation and testing, weighing and batching, mixing control, concrete transportation}; after correction, the output is {mix design, raw material preparation and testing, weighing and batching, mixing control, discharge inspection, concrete transportation}.

[0044] Step 2.6: If the input or output data of a step is wrong, enter the correction prompt word mpop for the corresponding input or output data. i and mpip i , forming a modified prompt word set Pop={mpop i} and Pip={mpip i}. If the correction prompt words Pop and Pip are not empty, the correction prompt word is appended to the prompt word, that is, pop = pop + mpop i ,pip=pip+mpip i ; For example, mpip i =Please consider parameters such as admixture content.

[0045] The embodiment of the present invention provides a method for intelligently constructing a process data chain based on a large model. Through a professional large model, a process data chain in a structured form consisting of a list of construction process steps, input and output data, and a close relationship between the input and output data is automatically constructed according to the data structure of the process data chain and a structured prompt vocabulary. The method has the advantage of high efficiency in constructing a construction process data chain, overcomes the problem of low efficiency in manually constructing a process data chain, realizes the intelligent construction of the construction process data chain, provides a foundation for the development of digital construction software and the research and development of intelligent construction equipment, reduces the workload of manual processing, reduces the difficulty of constructing the process data chain, and shortens the construction cycle of the process data chain.

[0046] The embodiment of the present invention provides an intelligent construction method for a process data chain based on a large model. By drawing on the thinking chain method, it guides the use of a professional large model to complete the construction of a structured prompt vocabulary of a construction process data chain through process step list output and process step input and output data output decomposition analysis, thereby greatly improving the efficiency and accuracy of the construction of the structured prompt vocabulary.

[0047] The embodiment of the present invention provides a method for intelligently constructing a process data chain based on a large model. In view of the problem of "hallucination" in professional large models, an expert mode is used to manually review and judge and correct the structured prompt word library to continuously optimize the process data chain output results of the professional large model, thereby improving the accuracy, completeness and comprehensiveness of the constructed process data chain. The hallucination of the large model refers to the phenomenon that the content generated by the large model is inconsistent with the facts in the real world or the user input.

[0048] An embodiment of the present invention provides an intelligent construction method for a process data chain based on a large model. To address the problem that the process corpus lacks description of the immediate relationship between construction process steps, a professional large model is used to establish the immediate relationship between process steps according to input and output data, thereby ensuring the accuracy of the process data chain and the correlation between the steps.

[0049] The embodiment of the present invention provides a method for intelligently constructing a process data chain based on a large model. It intelligently outputs a structured process data chain of a construction process through a professional large model. It is suitable for quickly establishing a data chain construction of a construction process with a large amount of data, thereby improving the work efficiency of digital transformation.

[0050] The present invention is not limited to the above-mentioned specific implementation modes. Obviously, the above-mentioned embodiments are only some embodiments of the embodiments of the present invention, but not all embodiments. Based on the described embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field belong to the scope of protection of the present invention. Those skilled in the art can make other levels of modifications and changes to the present invention. In this way, if these modifications and changes of the present invention fall within the scope of the claims of the present invention, the present invention is also intended to include these changes and changes.

Claims

1. A method for intelligently constructing a process data chain based on a large model, characterized in that: include: Set the data structure of the construction process data chain; Build a structured prompt vocabulary; Establish a professional large model of construction technology; Output the process step list according to the professional large model; Output the input and output data of the process steps through professional large models; The professional large model establishes the close relationship between process steps based on the input and output data it outputs; Determine whether the process data chain is correct; if correct, the professional large model outputs a structured process data chain; otherwise, the structured prompt word library is modified.

2. The method for intelligently constructing a process data chain based on a large model according to claim 1, characterized in that: The method for setting the data structure of the construction process data chain includes: setting the construction process step set P = {step p i =(Step Name pn i , input data set Ip i , output dataset Op i )} and step relation set PR = (relation pr ij ), where i and j are positive integers, p i For the i-th step, pn i is the name of the ith step, Ip i is the input data set of the ith step, Op i is the output data set of the i-th step, pr ij is the relationship between the i-th step and the j-th step.

3. The method for intelligently constructing a process data chain based on a large model according to claim 2, characterized in that: The method for constructing a structured prompt word library includes: using a thinking chain method to write structured prompt words of the data structure of the process data chain, and constructing a structured prompt word library Pro={role definition word pro, process step list output prompt word ppn, process step output data prompt word pop, process step input data prompt word pip}.

4. The method for intelligently constructing a process data chain based on a large model according to claim 3 is characterized in that: The method for modifying the structured prompt word library includes: If there is an error in the step, enter the correction prompt word mppn of the corresponding step; if the correction prompt word mppn is not empty, append the correction prompt word after the prompt word; If the input or output data of a step is wrong, enter the correction prompt word mpop corresponding to the input or output data i and mpip i , forming a modified prompt word set Pop={mpop i } and Pip={mpip i }. If the correction prompt words Pop and Pip are not empty, the correction prompt words are appended to the prompt words.

5. The method for intelligently constructing a process data chain based on a large model according to claim 3, characterized in that: The method for outputting a process step list according to a professional large model comprises: inputting a role definition word pro and a process step list output prompt word ppn into the professional large model Mp i Automatically output a list of process steps for the construction process represented by the role definition word.

6. The method for intelligently constructing a process data chain based on a large model according to claim 5, characterized in that: The method for outputting the input and output data of the process step through the professional large model comprises: inputting the final process step input data prompt word pip and the process step output data prompt word pop into the professional large model Mp respectively i Output step pn i Input data list Ip i and output data list Op i .

7. The method for intelligently constructing a process data chain based on a large model according to claim 6, characterized in that: The method for establishing the close relationship between process steps according to the input and output data output by the professional large model includes: traversing each step p in P i If p i The input dataset Ip i There is data element Ip ij Other steps p in P k The output dataset Op k , then establish a close before and after relationship r ik = (immediately preceding step p k , the next step p i ), and add it to PR; if p i The output dataset Op i There is data element Op in ij Other steps p in P k The input set Op k , then establish a close before and after relationship r ik = (the next step p i , the previous step p k ), and added to the step relation set PR, where k is a positive integer.

8. The method for intelligently constructing a process data chain based on a large model according to claim 1, characterized in that: The method for judging whether the process data chain is correct includes: judging whether the step name, step sequence, input and output data of each step of the construction process are accurate and complete according to the output of the professional large model, and whether they meet the construction requirements; if so, it is correct, otherwise it is incorrect.

9. The method for intelligently constructing a process data chain based on a large model according to claim 1, characterized in that: The method for outputting a structured process data chain using a professional large model includes: fine-tuning the process step name, sequence and input and output data of each step; and outputting a step list and a step sequence list in a standard format.

10. The method for intelligently constructing a process data chain based on a large model according to claim 1, characterized in that: The method for establishing a professional large model of construction technology includes: inputting professional corpus of each construction technology of a building construction node into a basic large model for fine-tuning to form a large model of technology containing building construction technology data.

Images