Solution generation method and device based on TRIZ and large language model

Abstract

The invention relates to the technical field of artificial intelligence, and provides a solution generation method and device based on TRIZ and a large language model, electronic equipment, a computer program product and a computer readable storage medium. The method comprises the following steps: in response to a received question description text, inputting a preset first cue word into a large language model to indicate the large language model to determine a TRIZ contradictory parameter corresponding to the question description text; inputting a preset second cue word into the large language model to indicate the large language model to search a target innovation principle corresponding to the TRIZ contradictory parameter from the TRIZ contradictory matrix table; and inputting a preset third prompt word into the large language model to indicate the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle. According to the method, an AI large language model is introduced, so that an end-to-end integrated solution from TRIZ contradictory parameter identification to solution generation can be realized.

Description

Technical Field

[0001] The present application relates to the field of artificial intelligence technology, and in particular to a solution generation method, device, electronic device, computer program product, and computer-readable storage medium based on TRIZ and a large language model. Background Art

[0002] TRIZ (Teoriya Resheniya Izobretatelskikh Zadach) is a theory of inventive problem solving and a systematic innovation approach designed to help people solve innovative problems more effectively. TRIZ provides a systematic set of tools and methods that help people identify the essence of problems and find reasonable solutions through processes such as identifying contradictions, looking up innovative principles, and translating innovative principles into solutions. However, existing tools typically only cover a subset of TRIZ steps, such as contradiction definition or principle recommendation, lacking an end-to-end, integrated approach from contradiction identification to solution generation. Summary of the Invention

[0003] In view of this, the embodiments of the present application provide a solution generation method, device, electronic device, computer program product and computer-readable storage medium based on TRIZ and a large language model. By introducing an AI large language model, an end-to-end integrated solution from TRIZ conflict parameter identification to solution generation can be achieved.

[0004] A first aspect of an embodiment of the present application provides a solution generation method based on TRIZ and a large language model, comprising: In response to receiving the problem description text, inputting a preset first prompt word into the large language model to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text; Inputting a preset second prompt word into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table; A preset third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

[0005] After receiving the problem description text, the technical solution of the embodiment of the present application will input a first prompt word into the large language model to instruct the large language model to determine the TRIZ contradiction parameters corresponding to the problem description text; then, a second prompt word is input into the large language model to instruct the large language model to find the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table; finally, a third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle. In the above process, the first prompt word, the second prompt word, and the third prompt word are all pre-cured and designed, so the user only needs to input the problem description text. Subsequently, the AI ​​large language model automatically performs the steps of TRIZ contradiction parameter identification, innovation principle table lookup, and conversion of innovation principle into solution. It can be seen that an end-to-end integrated solution from TRIZ contradiction parameter identification to solution generation is achieved.

[0006] In one implementation of the embodiment of the present application, a preset first prompt word is input into the large language model to instruct the large language model to determine the TRIZ contradiction parameter corresponding to the problem description text, including: The first prompt word is input into the large language model to instruct the large language model to parse the problem description text, identify multiple groups of potential candidate contradiction parameters, and score each group of candidate contradiction parameters according to the preset scoring rules. The group of candidate contradiction parameters with the highest score value is determined as the TRIZ contradiction parameters corresponding to the problem description text.

[0007] The large language model incorporates a multi-dimensional, comprehensive scoring mechanism that comprehensively scores each set of candidate contradiction parameters based on multiple dimensions, including relevance, applicability, and solvability. Each set of candidate contradiction parameters is assigned a corresponding score, and the set with the highest score is ultimately selected as the TRIZ contradiction parameter corresponding to the problem description. By introducing this comprehensive scoring mechanism, the accuracy of identifying TRIZ contradiction parameters can be effectively improved.

[0008] In one implementation of the embodiment of the present application, the second prompt word includes at least two different table lookup prompt words; the preset second prompt word is input into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for the target innovation principle corresponding to the TRIZ contradiction parameter, including: Traversing each lookup prompt word, inputting the lookup prompt word into the large language model, and instructing the large language model to search the TRIZ contradiction matrix table for the innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word; If the innovation principle results corresponding to the various lookup prompt words are exactly the same, the innovation principle results corresponding to the various lookup prompt words are determined as the target innovation principle.

[0009] To improve table lookup accuracy and obtain more precise results for innovative principles, a cross-validation approach using multiple table lookups can be introduced. By performing multiple table lookups using different lookup methods, the resulting innovative principle is confirmed only when the results of these multiple lookups are consistent. This effectively reduces table search errors and improves the accuracy of the target innovative principle.

[0010] In one implementation of the embodiment of the present application, after traversing each lookup prompt word and inputting the lookup prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word, the method further includes: If the innovation principle results corresponding to each lookup prompt word are different, then return and execute again to traverse each lookup prompt word, input the lookup prompt word into the large language model, and instruct the large language model to search the innovation principle results corresponding to the TRIZ contradiction parameters from the TRIZ contradiction matrix table according to the lookup method corresponding to the lookup prompt word, until the innovation principle results corresponding to each lookup prompt word are exactly the same.

[0011] If the results of the innovative principles corresponding to the various lookup prompts are different, that is, if there are inconsistencies in the results of multiple lookups, a fault-tolerant processing mechanism needs to be introduced. One fault-tolerant processing mechanism is a cyclic lookup, which returns and performs the multiple lookup operations of traversing all the lookup prompts again, and then determines whether the multiple lookup results obtained again are consistent. If there are still differences, the next loop is entered, and the multiple lookup operations of traversing all the lookup prompts are performed again. This is repeated until the multiple lookup results obtained are completely consistent, that is, the innovative principle results corresponding to the various lookup prompts are exactly the same. By introducing a fault-tolerant processing mechanism for cyclic lookup, the completeness and practicality of the lookup operation can be further improved.

[0012] In another implementation of the embodiment of the present application, after traversing each lookup prompt word and inputting the lookup prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word, the method further includes: If the innovation principle results corresponding to the various lookup prompt words are different, the target innovation principle is determined from the innovation principle results corresponding to the various lookup prompt words by majority voting.

[0013] When different innovation principle results are found for each lookup prompt, another fault-tolerant mechanism is majority voting. This involves determining the target innovation principle from the innovation principle results for each lookup prompt. This fault-tolerant mechanism also improves the completeness and practicality of the lookup operation.

[0014] In one implementation of the embodiment of the present application, a TRIZ contradiction matrix table is divided into multiple sub-tables according to deterioration parameters; a preset second prompt word is input into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for a target innovation principle corresponding to the TRIZ contradiction parameter, including: A second prompt word is input into the large language model to instruct the large language model to use the TRIZ contradiction parameter as an index to search for a target innovation principle from multiple sub-tables.

[0015] A matrix block retrieval mechanism can be introduced to divide the TRIZ contradiction matrix table into multiple sub-tables according to the deterioration parameters. This can decompose the complex large-scale matrix retrieval into multiple simple small and medium-sized matrix retrievals, thereby reducing the difficulty of data retrieval and improving the accuracy of table lookup.

[0016] In one implementation of the embodiment of the present application, a preset second prompt word is input into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table, including: A second prompt word is input into the large language model to instruct the large language model to call a TRIZ lookup tool, search the TRIZ contradiction matrix table for the innovation principle number corresponding to the number of the TRIZ contradiction parameter, and determine each innovation principle corresponding to the innovation principle number as a target innovation principle.

[0017] The large language model can call upon various existing TRIZ lookup tools to search the TRIZ contradiction matrix for the innovation principle numbers corresponding to the TRIZ contradiction parameter numbers. It can then retrieve the individual innovation principles corresponding to these innovation principle numbers as the target innovation principles corresponding to the TRIZ contradiction parameters. Using TRIZ lookup tools, using the TRIZ contradiction parameter numbers as indices to query the TRIZ contradiction matrix, it is possible to conveniently and accurately obtain the corresponding innovation principle results.

[0018] In one implementation of the embodiment of the present application, a preset third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle, including: A third prompt word is input into the large language model to instruct the large language model to use the retrieval enhancement generation technology to call the target knowledge base to generate various solutions corresponding to the problem description text based on the target innovation principle as the target solution.

[0019] After receiving the third prompt word, the large language model can use retrieval enhancement generation technology to call the knowledge base, query and integrate multiple solutions related to the target innovation principle through the target knowledge base, and finally obtain a more comprehensive and accurate problem solution.

[0020] In one implementation of the embodiment of the present application, the target knowledge base is determined by: Determine the technical field to which the target innovation principle belongs; At least one knowledge base corresponding to the technical field is searched from multiple preset knowledge bases as a target knowledge base.

[0021] To improve the accuracy and rationality of solution generation, we can first determine the technical field to which the target innovation principle belongs. Then, we search for at least one knowledge base corresponding to that technical field from among multiple pre-set knowledge bases, using this as the target knowledge base. This way, when searching for solutions using a large language model, we only call upon the target knowledge bases that correspond to the technical field to which the target innovation principle belongs. This allows us to filter out knowledge bases that are irrelevant to the technical field, thereby improving the accuracy of solution search and reducing search time.

[0022] In one implementation of the embodiment of the present application, each solution includes a non-cross-principle solution based on a single innovative principle, and a cross-principle solution based on at least two innovative principles.

[0023] With the improvement of intelligence, the large language model has broken through the limitations of the application of a single innovation principle. It can not only generate non-cross-principle solutions based on a single innovation principle, but also automatically identify and integrate the advantages and characteristics of multiple innovation principles to generate systematic cross-principle solutions.

[0024] In one implementation of the embodiment of the present application, after inputting a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle, the method further includes: Generate a structured report based on the problem description text, TRIZ contradiction parameters, target innovation principles and target solutions; Outputs a structured report.

[0025] By integrating problem description text, TRIZ contradiction parameters, target innovation principles, and target solutions, a corresponding structured report can be generated and output for subsequent human-computer interaction and secondary editing operations.

[0026] A second aspect of the embodiments of the present application provides a solution generation device based on TRIZ and a large language model, comprising: a contradiction parameter determination module, configured to input a preset first prompt word into the large language model in response to receiving the problem description text, so as to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text; A table lookup module is used to input a preset second prompt word into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table; The solution generation module is used to input a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

[0027] A third aspect of an embodiment of the present application provides an electronic device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor. When the processor executes the computer program, the solution generation method based on TRIZ and a large language model as provided in the first aspect of the embodiment of the present application is implemented.

[0028] A fourth aspect of the embodiments of the present application provides a computer program product. When the computer program product runs on an electronic device, the electronic device executes the solution generation method based on TRIZ and a large language model as provided in the first aspect of the embodiments of the present application.

[0029] The fifth aspect of the embodiments of the present application provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the solution generation method based on TRIZ and a large language model as provided in the first aspect of the embodiments of the present application.

[0030] It can be understood that the beneficial effects of the second to fifth aspects mentioned above can be found in the relevant description of the first aspect mentioned above, and will not be repeated here. BRIEF DESCRIPTION OF THE DRAWINGS

[0031] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the embodiments or descriptions of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0032] Figure 1 This is a flow chart of a solution generation method based on TRIZ and a large language model provided in an embodiment of the present application; Figure 2 This is a schematic diagram of the operating principle of the solution generation method based on TRIZ and a large language model provided in an embodiment of the present application; Figure 3 This is a schematic diagram of the structure of a solution generation device based on TRIZ and a large language model provided in an embodiment of the present application; Figure 4 This is a schematic diagram of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0033] In the following description, for the purpose of illustration rather than limitation, specific details such as specific system structures and technologies are provided to provide a thorough understanding of the embodiments of the present application. However, it should be clear to those skilled in the art that the present application can also be implemented in other embodiments without these specific details. In other cases, detailed descriptions of well-known systems, devices, circuits, and methods are omitted to avoid obstructing the description of the present application with unnecessary details. In addition, in the description of the present application specification and the appended claims, the terms "first," "second," "third," etc. are only used to distinguish descriptions and are not to be understood as indicating or implying relative importance.

[0034] TRIZ is a powerful innovation tool that helps people solve innovation problems more effectively. By drawing on past experience and knowledge and applying TRIZ's theories and methods, the innovation process can be accelerated and product quality improved. The basic TRIZ processing steps generally include identifying conflict parameters, looking up innovation principles in a table, and translating innovation principles into solutions. However, traditional manual TRIZ analysis methods are inefficient. The conflict matrix and material field model require experts to manually identify conflicts, look up tables, and match innovation principles. This process is cumbersome, highly dependent on expert experience, and time-consuming. Existing TRIZ analysis tools typically only cover a subset of TRIZ steps and fail to achieve an end-to-end integrated solution from identifying conflict parameters to generating solutions.

[0035] To address the above technical issues, the present invention proposes a solution generation method, apparatus, electronic device, and computer program product based on TRIZ and a large language model. By introducing an AI large language model and predefined model prompts, this method achieves an end-to-end integrated solution from identifying TRIZ conflict parameters to generating solutions. For more specific technical implementation details of the present invention, please refer to the various method embodiments described below.

[0036] It should be understood that the execution subjects of the various method embodiments proposed in the present application can be various types of electronic devices, such as mobile phones, tablet computers, desktop computers, wearable devices, augmented reality (AR) / virtual reality (VR) devices, laptop computers, ultra-mobile personal computers (UMPCs), netbooks, personal digital assistants (PDAs), large-screen TVs, etc. The embodiments of the present application do not impose any restrictions on the specific types of the electronic devices.

[0037] See also Figure 1 , shows a solution generation method based on TRIZ and a large language model provided by an embodiment of the present application, including: 101. In response to receiving the problem description text, input a preset first prompt word into the large language model to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text; The technical solution of the embodiments of this application aims to build an end-to-end TRIZ process engine, which mainly includes multiple aspects such as conflict parameter identification, precise table lookup, principle transformation, and solution generation. Using the problem input interface of the electronic device, users can enter corresponding problem description text based on the actual problem they encounter. In another implementation method, the problem description text can also be automatically generated by the electronic device. For example, when the user operates the electronic device, some problem information may be entered, or some problem information may be generated during the user interaction. The electronic device can automatically generate corresponding problem description text based on this problem information. In response to receiving the problem description text, the electronic device automatically obtains a pre-set first prompt word and inputs it into an AI large language model (LLM). The embodiments of this application do not impose any restrictions on the specific type of large language model used. The first prompt word is used to instruct the large language model to identify the corresponding TRIZ conflict parameters based on the problem description text. After receiving the first prompt word, the large language model will parse and extract semantics from the problem description text to identify TRIZ conflict parameters including improvement parameters and deterioration parameters.

[0038] In one implementation of the embodiment of the present application, a preset first prompt word is input into the large language model to instruct the large language model to determine the TRIZ contradiction parameter corresponding to the problem description text, including: The first prompt word is input into the large language model to instruct the large language model to parse the problem description text, identify multiple groups of potential candidate contradiction parameters, and score each group of candidate contradiction parameters according to the preset scoring rules. The group of candidate contradiction parameters with the highest score value is determined as the TRIZ contradiction parameters corresponding to the problem description text.

[0039] After receiving the first prompt, the large language model first parses the user's input problem description. It then rephrases the problem description through semantic segmentation and synonym rewriting, thereby improving the accuracy of problem understanding. This semantic analysis of the problem description extracts the core content of the technical contradiction, thereby identifying multiple sets of potential candidate contradiction parameters. The optimal set of these candidate contradiction parameters is then selected as the final TRIZ contradiction parameters. The large language model incorporates a multi-dimensional comprehensive scoring mechanism that comprehensively scores each set of candidate contradiction parameters based on multiple dimensions, including relevance, applicability, and solvability. Each candidate contradiction parameter is assigned a numerical score. The large language model then identifies the set of candidate contradiction parameters with the highest score as the optimal contradiction parameter and outputs it as the TRIZ contradiction parameter corresponding to the input problem description. By introducing this comprehensive scoring mechanism, the accuracy of identifying TRIZ contradiction parameters can be effectively improved.

[0040] For example, suppose a user enters a problem description such as "Increasing the pore size of porous ceramics improves oil conductivity but increases the probability of oil leakage." The default first prompt is, "Please perform semantic analysis on the input problem description to identify multiple sets of candidate conflict parameters. Each set of candidate conflict parameters will be scored comprehensively across multiple dimensions, and the highest-scoring candidate conflict parameter will be output as the TRIZ conflict parameter." After receiving the first prompt, the large language model rephrases the problem description based on the terminology and context of the specific engineering field, parses and extracts the core content of the technical conflict contained in the text, and automatically generates multiple sets of candidate conflict parameters with varying degrees of abstraction, covering multiple levels from physical properties to system functionality. For example, the output of multiple sets of candidate conflict parameters may include: (1) Improvement parameter 9 [speed] (increased oil conductivity) and deterioration parameter 23 [material loss] (increased oil leakage); (2) Improvement parameter 22 [energy loss] (reduced flow assistance) and deterioration parameter 31 [harmful factors generated by objects] (leakage pollution); (3) Improvement parameter 35 [adaptability] (pore structure adjustment) and deterioration parameter 14 [strength] (decrease in structural integrity)... After identifying the above multiple sets of candidate conflict parameters, the large language model uses a multi-dimensional comprehensive scoring mechanism to select a set of candidate conflict parameters with the highest score from the multiple sets of candidate conflict parameters as the optimal TRIZ conflict parameter output, and at the same time, it can map a standardized TRIZ parameter number to it. Finally, the large language model outputs a set of standardized TRIZ conflict parameters with the highest score: Improvement parameter 9 [speed] and deterioration parameter 23 [material loss].

[0041] 102. Inputting a preset second prompt word into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table; After using the large language model to obtain the corresponding TRIZ contradiction parameters, the electronic device automatically retrieves a pre-set second prompt word and inputs it into the large language model. The second prompt word instructs the large language model to determine the target innovation principle corresponding to the TRIZ contradiction parameter through a table lookup. Specifically, after receiving the second prompt word, the large language model first reads the standardized TRIZ contradiction matrix. Then, using the TRIZ contradiction parameter as an index, it uses a specific table lookup method to locate the corresponding row (corresponding to the improvement parameter in the TRIZ contradiction parameter) and column (corresponding to the deterioration parameter in the TRIZ contradiction parameter) in the TRIZ contradiction matrix. It then reads the recorded innovation principle result from this cell as the target innovation principle.

[0042] In one implementation of the embodiment of the present application, a preset second prompt word is input into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table, including: A second prompt word is input into the large language model to instruct the large language model to call a TRIZ lookup tool, search the TRIZ contradiction matrix table for the innovation principle number corresponding to the number of the TRIZ contradiction parameter, and determine each innovation principle corresponding to the innovation principle number as a target innovation principle.

[0043] Once the TRIZ contradiction parameters are determined and their numbers are fixed, the large language model can call on various existing TRIZ lookup tools to search the TRIZ contradiction matrix for the innovation principle numbers corresponding to the TRIZ contradiction parameter numbers. It can then retrieve the individual innovation principles corresponding to these innovation principle numbers as the target innovation principles corresponding to the TRIZ contradiction parameters. Specifically, assuming the TRIZ contradiction parameters are the improvement parameter 9 [speed] and the deterioration parameter 23 [material loss], their numbers include the improvement parameter number 9 and the deterioration parameter number 23. The large language model can query by calling the lookup tool ts-TRIZchabiao-TRIZ, using the improvement parameter number 9 as the row index and the deterioration parameter number 23 as the column index to locate the corresponding cell in the TRIZ contradiction matrix. It then reads the individual innovation principle numbers from these cells and converts them into the corresponding innovation principles to obtain the target innovation principles corresponding to the TRIZ contradiction parameters. It can be seen that using TRIZ lookup tools and using the TRIZ contradiction parameter numbers as indexes to query the TRIZ contradiction matrix can conveniently and accurately obtain the corresponding innovation principle results.

[0044] Considering that the standardized TRIZ contradiction matrix is ​​a 39-row, 39-column matrix containing a total of 1,521 intersecting cells, large language models are prone to errors when locating row and column coordinates in one go. Internal testing has shown that the accuracy of a single lookup is less than 40%. To improve the accuracy of the lookup and obtain more accurate results for the innovation principle lookup, the present embodiment introduces a cross-validation method for multiple lookups to reduce table retrieval errors.

[0045] In one implementation of the embodiment of the present application, the second prompt word includes at least two different table lookup prompt words; the preset second prompt word is input into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for the target innovation principle corresponding to the TRIZ contradiction parameter, including: (1) Traverse each lookup prompt word and input the lookup prompt word into the large language model to instruct the large language model to search the innovation principle result corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table according to the lookup method corresponding to the lookup prompt word; (2) If the innovation principle results corresponding to each lookup prompt word are exactly the same, then the innovation principle results corresponding to each lookup prompt word are determined as the target innovation principle.

[0046] The second prompt word may include at least two different table lookup prompt words, each corresponding to a different table lookup method. For example, table lookup prompt word A corresponds to table lookup method A: first traverse and search all rows, then traverse and search all columns; table lookup prompt word B corresponds to table lookup method B: first traverse and search all columns, then traverse and search all rows; table lookup prompt word C corresponds to table lookup method C: rows are searched from left to right and columns are searched from top to bottom; table lookup prompt word D corresponds to table lookup method D: rows are searched from right to left and columns are searched from bottom to top, and so on. In actual operation, different table lookup prompt words can be generated by using different prompt word templates and combining them with random seeds. When the lookup operation begins, the first lookup prompt is input into the large language model for the first lookup. The large language model uses the lookup method corresponding to the lookup prompt to search the TRIZ contradiction matrix for the innovation principle corresponding to the TRIZ contradiction parameter, generating the result of the first lookup. Then, the second lookup prompt is input into the large language model for a second lookup, obtaining the second lookup result using a different lookup method. This continues in this manner until all lookup prompts have been traversed, resulting in multiple lookup results. Subsequently, methods such as comparing JSON results can be used to determine whether the innovation principle results corresponding to each lookup prompt are identical. If the innovation principle results corresponding to each lookup prompt are identical, meaning that the results of multiple lookups are consistent, the corresponding innovation principle results can be directly output as the target innovation principle corresponding to the TRIZ contradiction parameter, given the extremely low probability of errors in multiple lookups. However, if the innovation principle results corresponding to each lookup prompt differ, this indicates that at least one lookup result is incorrect, requiring the implementation of a subsequent fault-tolerant mechanism. See below for details.

[0047] In one implementation of the embodiment of the present application, after traversing each lookup prompt word and inputting the lookup prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word, the method further includes: If the innovation principle results corresponding to each lookup prompt word are different, then return and execute again to traverse each lookup prompt word, input the lookup prompt word into the large language model, and instruct the large language model to search the innovation principle result corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table according to the lookup method corresponding to the lookup prompt word, as well as subsequent steps, until the innovation principle results corresponding to each lookup prompt word are exactly the same.

[0048] If the innovation principle results corresponding to each lookup prompt word differ, that is, if the results of multiple lookups are inconsistent, a fault-tolerant mechanism is needed. One fault-tolerant mechanism involves looping through the lookup table, returning to perform multiple lookups across all the lookup prompt words, and then determining whether the results obtained from these multiple lookups are consistent. If there are still differences, the next loop is entered, and the multiple lookups across all the lookup prompt words are performed again. This cycle repeats until the multiple lookup results are completely consistent, meaning that the innovation principle results corresponding to each lookup prompt word are exactly the same. The consistent innovation principle results can then be output as the target innovation principle corresponding to the TRIZ contradiction parameter. For example, assuming a loop table lookup includes three lookup operations using different lookup methods, after the first loop table lookup, the results of the three lookup operations in the first loop table lookup are determined to be consistent. If they are consistent, the consistent innovative principle lookup result is output. If not, a second loop table lookup is performed, and the results of the three lookup operations in the second loop table lookup are determined to be consistent. If they are consistent, the consistent innovative principle lookup result is output. If not, a third loop table lookup is performed, and so on, until the results of the three lookup operations in a certain loop table lookup are consistent, then the loop ends. By introducing a fault-tolerant processing mechanism for loop table lookup, the completeness and practicality of the table lookup operation can be further improved.

[0049] In another implementation of the embodiment of the present application, after traversing each lookup prompt word and inputting the lookup prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word, the method further includes: If the innovation principle results corresponding to the various lookup prompt words are different, the target innovation principle is determined from the innovation principle results corresponding to the various lookup prompt words by majority voting.

[0050] When the results for the innovation principles corresponding to each lookup prompt differ, another fault-tolerant mechanism is majority voting. This involves determining the target innovation principle from the innovation principle results corresponding to each lookup prompt using a majority vote. Majority voting involves determining the value that appears most frequently in a set of data. The number of occurrences of each innovation principle is then counted based on all lookup results. The innovation principles are then ranked in descending order of occurrence. Finally, a certain number of innovation principles ranked at the top of the list are selected as the target innovation principles corresponding to the TRIZ contradiction parameter. For example, suppose a table lookup is performed 10 times. Statistics show that Innovation Principle 10 (Pre-Action Principle) appears 10 times, Innovation Principle 13 (Reverse Path Principle) appears 10 times, Innovation Principle 28 (Mechanical System Substitution Principle) appears 9 times, Innovation Principle 38 (Accelerated Oxidation Principle) appears 8 times, and Innovation Principle 4 (Asymmetry Principle) appears 2 times. The top four innovation principles can be selected, and the target innovation principles corresponding to the TRIZ contradiction parameters are determined as: Innovation Principle 10 (Pre-Action Principle), Innovation Principle 13 (Reverse Path Principle), Innovation Principle 28 (Mechanical System Substitution Principle), and Innovation Principle 38 (Accelerated Oxidation Principle). By introducing a majority voting fault-tolerant mechanism, the completeness and practicality of the table lookup can also be improved.

[0051] On the other hand, considering that the TRIZ contradiction matrix table is a large matrix, there is a problem of high retrieval difficulty. In order to reduce the retrieval difficulty and thus improve the accuracy of table lookup, the embodiment of the present application introduces a matrix block retrieval mechanism. Specifically, the TRIZ contradiction matrix table can be divided into multiple sub-tables according to the deterioration parameter, thereby dividing a large matrix into multiple small and medium-sized matrices. This can reduce the difficulty of data retrieval and thus improve the accuracy of table lookup. For example, the 39 columns of the TRIZ contradiction matrix table can be divided into four sub-tables according to the number of deterioration parameters: the first sub-table is columns 1 to 10, the second sub-table is columns 11 to 20, the third sub-table is columns 21 to 30, and the fourth sub-table is columns 31 to 39. By setting it up in this way, each prompt word table lookup only needs to load about 1 / 4 of the number of columns, which can significantly compress the length of the Chinese context. It should be noted that the embodiment of the present application does not limit the division method of the table and the number of divided sub-tables. For example, according to actual needs, the 39 columns of the TRIZ contradiction matrix table can be divided into 5 or more sub-tables, or divided according to binary coding, etc.

[0052] In one implementation of the embodiment of the present application, a TRIZ contradiction matrix table is divided into multiple sub-tables according to deterioration parameters; a preset second prompt word is input into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for a target innovation principle corresponding to the TRIZ contradiction parameter, including: A second prompt word is input into the large language model to instruct the large language model to use the TRIZ contradiction parameter as an index to search for a target innovation principle from multiple sub-tables.

[0053] Since the numbers of the improvement parameters and the deterioration parameters contained in the TRIZ contradiction parameters are known, after receiving the second prompt word, the large language model can sequentially search the multiple subtables divided in the above manner based on the TRIZ contradiction parameter numbers as indexes (the improvement parameter numbers as row indexes, the deterioration parameter numbers as column indexes), thereby finding the target innovation principle corresponding to the TRIZ contradiction parameters. By indexing the TRIZ contradiction matrix into blocks based on the deterioration parameters, the complex large matrix search can be decomposed into multiple simple small and medium-sized matrix searches, significantly improving the accuracy of the large language model in structured data retrieval to 99%. Furthermore, explicit row and column annotations can be used in matrix queries to address the structural bias problem of the large language model when accurately locating numerical values.

[0054] As an example, suppose the determined TRIZ contradiction parameters are: improvement parameter 9 [speed] and deterioration parameter 23 [material loss]. First, perform the first table lookup, inputting the table lookup prompt A, "Please search the TRIZ contradiction matrix table according to table lookup method A, and find and output the innovation principle result with the contradiction parameter of improvement parameter 9 [speed] and deterioration parameter 23 [material loss]." The large language model will search the TRIZ contradiction matrix table according to table lookup method A and obtain the corresponding innovation principle result A. Then, perform the second table lookup, inputting the table lookup prompt B, "Please search the TRIZ contradiction matrix table according to table lookup method B, and find and output the innovation principle result with the contradiction parameter of improvement parameter 9 [speed] and deterioration parameter 23 [material loss]." The large language model will search the TRIZ contradiction matrix table according to table lookup method B and obtain the corresponding innovation principle result B. Afterward, the consistency of the two lookup results—that is, whether innovation principle result A and innovation principle result B are identical—is determined. If they are, either innovation principle result A or innovation principle result B is directly output as the target innovation principle corresponding to the TRIZ contradiction parameter "improvement parameter 9 [speed] and deterioration parameter 23 [material loss]." If innovation principle result A and innovation principle result B differ, a fault-tolerant mechanism is required. One approach is to loop through the lookup operations corresponding to lookup prompt word A and lookup prompt word B multiple times, continuously updating innovation principle result A and innovation principle result B during this process until, at the end of a certain loop, innovation principle result A and innovation principle result B are identical. At this point, the consistent innovation principle result can be output as the target innovation principle corresponding to the TRIZ contradiction parameter "improvement parameter 9 [speed] and deterioration parameter 23 [material loss]." Another approach is to use majority voting to count the innovation principles that appear the most frequently among all innovation principle results and use them as the target innovation principle corresponding to the TRIZ contradiction parameter "improvement parameter 9 [speed] and deterioration parameter 23 [material loss]." After the above table lookup processing, the target innovation principles finally output by the large language model are: Innovation Principle 10-Pre-action Principle, Innovation Principle 13-Reverse Path Principle, Innovation Principle 28-Mechanical System Replacement Principle and Innovation Principle 38-Accelerated Oxidation Principle.

[0055] The experimental results show that the combination of the above-mentioned cross-validation method of multiple table lookups and the matrix block retrieval mechanism can reduce the independent error probability of table lookups from p to p. n ,When the number of table lookups n is large enough, the table lookup accuracy can be increased from less than 40% to over 99%, thereby obtaining a high-precision innovative principle result.

[0056] 103. Input a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

[0057] After using the large language model to obtain the target innovation principle corresponding to the TRIZ contradiction parameter, the electronic device automatically obtains a pre-set third prompt word and inputs it into the large language model. The third prompt word is used to instruct the large language model to convert the innovation principle into a corresponding solution. It can prompt the large language model to explain the meaning of each innovation principle and the thinking direction it contains, provide typical application cases that are not limited to technical fields, and propose multiple solutions or thinking directions for specific contradictions through methods such as brainstorming. It can also briefly describe the technical difficulties that may be encountered in resolving the contradiction and provide possible breakthrough directions. Specifically, after receiving the third prompt word, the large language model can generate various solutions corresponding to the problem description text based on the target innovation principle by calling the knowledge base, and record them as target solutions.

[0058] In one implementation of the embodiment of the present application, a preset third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle, including: A third prompt word is input into the large language model to instruct the large language model to use the retrieval enhancement generation technology to call the target knowledge base to generate various solutions corresponding to the problem description text based on the target innovation principle as the target solution.

[0059] Retrieval Augmented Generation (RAG) technology retrieves relevant information from its own vertical domain database and combines it into a prompt template, which the large language model uses to polish and generate the corresponding response information. After receiving the third prompt word, the large language model uses RAG to access the knowledge base. It then searches the target knowledge base for multiple solutions related to the target innovation principle and integrates them to ultimately obtain the target solution corresponding to the problem description, completing the final step of the TRIZ process. By using RAG to access the knowledge base, the large language model can generate relatively comprehensive and accurate problem solutions, providing practical assistance to users.

[0060] In one implementation of the embodiment of the present application, the target knowledge base is determined by: (1) Determine the technical field to which the target innovation principle belongs; (2) From multiple preset knowledge bases, search for at least one knowledge base corresponding to the technical field as the target knowledge base.

[0061] There are a vast number of pre-existing knowledge bases, each of which can belong to a different technical field. To improve the accuracy and rationality of solution generation, we can first determine the technical field to which the target innovation principle belongs. Then, we can search for at least one knowledge base corresponding to that technical field from among the multiple pre-existing knowledge bases to serve as the target knowledge base. When searching for solutions using a large language model, we only call upon the target knowledge base corresponding to the technical field to which the target innovation principle belongs. This can filter out knowledge bases that are irrelevant to the technical field, thereby improving the accuracy of solution search and reducing search time.

[0062] In one implementation of the embodiment of the present application, each solution includes a non-cross-principle solution based on a single innovative principle, and a cross-principle solution based on at least two innovative principles.

[0063] Typically, a target innovation principle encompasses multiple innovation principles. The solutions generated by the large language model can include both non-cross-principle solutions based on a single innovation principle and cross-principle solutions based on at least two innovation principles. With increasing intelligence, the large language model has transcended the limitations of applying a single innovation principle and can automatically identify and integrate the strengths of multiple innovation principles to generate systematic cross-principle solutions. The large language model can output both non-cross-principle solutions and cross-principle solutions for user reference.

[0064] As an example, let's assume the target innovation principles are: Innovation Principle 10 - Pre-Action Principle, Innovation Principle 13 - Reverse Path Principle, Innovation Principle 28 - Mechanical System Replacement Principle, and Innovation Principle 38 - Accelerated Oxidation Principle. The default third prompt is "Please search the knowledge base and output a solution to the problem based on the input target innovation principle." After receiving the third prompt, the large language model uses RAG to access the knowledge base, searching the target knowledge base for various solutions related to the target innovation principle, including both non-cross-principle solutions and systematic cross-principle solutions, and outputs these solutions. The target solutions ultimately output by the large language model include: gradient pore structure (a non-cross-principle solution based on Innovation Principle 10), asymmetric pore size + pressure replenishment (a non-cross-principle solution based on Innovation Principle 13), electric field pore control (a non-cross-principle solution based on Innovation Principle 28), surface controllable oxidation (a non-cross-principle solution based on Innovation Principle 38), and intelligent response system and biomimetic multilayer structure (a cross-principle solution based on Innovation Principles 10 and 28).

[0065] On the other hand, the large language model can also attach corresponding experimental design and verification route suggestions to each set of solutions generated, providing users with clearer executable solutions and helping users shorten the conversion cycle from theoretical solutions to practical applications.

[0066] In one implementation of the embodiment of the present application, after inputting a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle, the method further includes: Generate a structured report based on the problem description text, TRIZ contradiction parameters, target innovation principles and target solutions; Outputs a structured report.

[0067] To facilitate human-computer interaction and secondary editing, the electronic device can integrate information such as the problem description text, TRIZ contradiction parameters, target innovation principles, and target solution, and generate and output a structured report through HTML rendering and other methods. By outputting and displaying this structured report, users can intuitively view the question being asked, the corresponding TRIZ contradiction parameters, which innovation principles can be utilized, and the technical solutions available to solve the problem.

[0068] As an example, Figure 2 This is a schematic diagram of an operation flow of a solution generation method based on TRIZ and a large language model provided in an embodiment of the present application. Figure 2 In the process, users enter a problem description based on the technical problem they want to solve through the client. After receiving the first prompt word of the fixed design, the AI ​​large language model automatically identifies the TRIZ contradiction parameters corresponding to the problem description text. Then, guided by the second prompt word of the fixed design, the AI ​​large language model uses the TRIZ contradiction parameters as an index to query the TRIZ contradiction matrix table to obtain the corresponding target innovation principle. Finally, guided by the third prompt word of the fixed design, the AI ​​large language model uses RAG to call the knowledge base to generate and output multiple problem solutions based on the target innovation principle. From the user's perspective, simply entering the problem description text will automatically output the problem solution through the large language model, clearly achieving an end-to-end integrated solution.

[0069] After receiving the problem description text, the technical solution of the embodiment of the present application will input a first prompt word into the large language model to instruct the large language model to determine the TRIZ contradiction parameters corresponding to the problem description text; then, a second prompt word is input into the large language model to instruct the large language model to find the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table; finally, a third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle. In the above process, the first prompt word, the second prompt word, and the third prompt word are all pre-cured and designed, so the user only needs to input the problem description text. Subsequently, the AI ​​large language model automatically performs the steps of TRIZ contradiction parameter identification, innovation principle table lookup, and conversion of innovation principle into solution. It can be seen that an end-to-end integrated solution from TRIZ contradiction parameter identification to solution generation is achieved.

[0070] In summary, the embodiments of this application utilize a large language model and fixed-design prompt words to provide an end-to-end integrated TRIZ solution encompassing conflict parameter identification, innovative principle table lookup, innovative principle transformation, and solution generation. Furthermore, by introducing a multi-table lookup cross-validation approach and a matrix block search mechanism, the table lookup accuracy rate is significantly improved, resulting in highly accurate innovative principle results.

[0071] It should be understood that the size of the serial numbers of the steps in the above embodiments does not mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of this application.

[0072] The above mainly describes a solution generation method based on TRIZ and a large language model. The following will describe a solution generation device based on TRIZ and a large language model.

[0073] See also Figure 3 , shows a solution generation device based on TRIZ and a large language model provided by an embodiment of the present application, comprising: The contradiction parameter determination module 301 is configured to input a preset first prompt word into the large language model in response to receiving the problem description text, so as to instruct the large language model to determine the TRIZ contradiction parameter corresponding to the problem description text; A table lookup module 302 is configured to input a preset second prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for a target innovation principle corresponding to the TRIZ contradiction parameter; The solution generation module 303 is used to input a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

[0074] In one implementation of the embodiment of the present application, the conflict parameter determination module includes: The contradiction parameter determination unit is used to input a first prompt word into the large language model to instruct the large language model to parse the problem description text, identify multiple groups of potential candidate contradiction parameters, and score each group of candidate contradiction parameters according to preset scoring rules, and determine the group of candidate contradiction parameters with the highest score value as the TRIZ contradiction parameters corresponding to the problem description text.

[0075] In one implementation of the embodiment of the present application, the second prompt word includes at least two different lookup prompt words; the lookup module includes: a multiple table lookup unit for traversing each table lookup prompt word and inputting the table lookup prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for the innovation principle result corresponding to the TRIZ contradiction parameter according to the table lookup method corresponding to the table lookup prompt word; The first table lookup result determination unit is configured to determine the innovation principle results corresponding to the table lookup prompt words as target innovation principles if the innovation principle results corresponding to the table lookup prompt words are exactly the same.

[0076] In one implementation of the embodiment of the present application, the table lookup module further includes: The loop table lookup unit is used for returning to execute the step of traversing each table lookup prompt word again if there are differences in the innovation principle results corresponding to each table lookup prompt word, inputting the table lookup prompt word into the large language model, and instructing the large language model to search the TRIZ contradiction matrix table for the innovation principle result corresponding to the TRIZ contradiction parameter according to the table lookup method corresponding to the table lookup prompt word, until the innovation principle results corresponding to each table lookup prompt word are exactly the same.

[0077] In another implementation of the embodiment of the present application, the table lookup module further includes: The second table lookup result determination unit is configured to determine a target innovation principle from the innovation principle results corresponding to the table lookup prompt words by majority voting if the innovation principle results corresponding to the table lookup prompt words are different.

[0078] In one implementation of the embodiment of the present application, the TRIZ contradiction matrix table is divided into multiple sub-tables according to the deterioration parameter; the table lookup module includes: The block indexing unit is used to input a second prompt word into the large language model to instruct the large language model to use the TRIZ contradiction parameter as an index to search for a target innovation principle from multiple sub-tables.

[0079] In one implementation of the embodiment of the present application, the table lookup module includes: The tool table lookup unit is used to input a second prompt word into the large language model to instruct the large language model to call the TRIZ table lookup tool, search the TRIZ contradiction matrix table for the innovation principle number corresponding to the number of the TRIZ contradiction parameter, and determine each innovation principle corresponding to the innovation principle number as the target innovation principle.

[0080] In one implementation of the embodiment of the present application, the solution generation module includes: The solution generation unit is used to input a third prompt word into the large language model to instruct the large language model to use the retrieval enhancement generation technology to call the target knowledge base to generate various solutions corresponding to the problem description text based on the target innovation principle as the target solution.

[0081] In one implementation of the embodiment of the present application, the solution generation module further includes: Technical field determination unit, used to determine the technical field to which the target innovation principle belongs; The knowledge base search unit is used to search for at least one knowledge base corresponding to the technical field from a plurality of preset knowledge bases as a target knowledge base.

[0082] In one implementation of the embodiment of the present application, each solution includes a non-cross-principle solution based on a single innovative principle, and a cross-principle solution based on at least two innovative principles.

[0083] In one implementation of the embodiment of the present application, the solution generation device based on TRIZ and a large language model further includes: Structured report generation module, used to generate structured reports based on problem description text, TRIZ contradiction parameters, target innovation principles and target solutions; Structured report output module, used to output structured reports.

[0084] An embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, it implements the solution generation method based on TRIZ and a large language model described in any of the above embodiments.

[0085] An embodiment of the present application also provides a computer program product. When the computer program product is run on an electronic device, the electronic device executes the solution generation method based on TRIZ and a large language model as described in any of the above embodiments.

[0086] Figure 4Schematic diagram of an electronic device provided by an embodiment of the present application. Figure 4 As shown, the electronic device 4 of this embodiment includes: a processor 40, a memory 41, and a computer program 42 stored in the memory 41 and executable on the processor 40. When the processor 40 executes the computer program 42, the steps in the above-mentioned embodiments of the solution generation method based on TRIZ and a large language model are implemented, for example Figure 1 Alternatively, when the processor 40 executes the computer program 42, the functions of the modules / units in the above-mentioned device embodiments are realized, for example, Figure 3 Functions of modules 301 to 303 of the illustrated apparatus.

[0087] The computer program 42 may be divided into one or more modules / units, which are stored in the memory 41 and executed by the processor 40 to implement the present application. The one or more modules / units may be a series of computer program instruction segments capable of performing specific functions, and the instruction segments are used to describe the execution process of the computer program 42 in the electronic device 4.

[0088] The processor 40 may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor.

[0089] The memory 41 can be an internal storage unit of the electronic device 4, such as a hard drive or memory of the electronic device 4. The memory 41 can also be an external storage device of the electronic device 4, such as a plug-in hard drive, a Smart Media Card (SMC), a Secure Digital (SD) card, a flash memory card, etc. equipped on the electronic device 4. Furthermore, the memory 41 can include both an internal storage unit of the electronic device 4 and an external storage device. The memory 41 is used to store the computer program and other programs and data required by the electronic device. The memory 41 can also be used to temporarily store data that has been output or is about to be output.

[0090] Those skilled in the art can clearly understand that, for the convenience and brevity of description, only the division of the above-mentioned functional units and modules is used as an example for illustration. In actual applications, the above-mentioned functions can be distributed and completed by different functional units and modules as needed, that is, the internal structure of the device can be divided into different functional units or modules to complete all or part of the functions described above. The functional units and modules in the embodiment can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above-mentioned integrated unit can be implemented in the form of hardware or in the form of software functional units. In addition, the specific names of the functional units and modules are only for the convenience of distinguishing each other, and are not used to limit the scope of protection of this application. The specific working process of the units and modules in the above-mentioned system can refer to the corresponding process in the aforementioned method embodiment, and will not be repeated here.

[0091] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0092] In the above embodiments, the description of each embodiment has its own focus. For parts that are not described or recorded in detail in a certain embodiment, reference can be made to the relevant description of other embodiments.

[0093] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0094] In the embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the system embodiments described above are merely schematic. For example, the division of the modules or units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0095] The units described as separate components may or may not be physically separate, and 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 units may be selected according to actual needs to achieve the purpose of the embodiments of the present application.

[0096] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0097] If the integrated unit is implemented as a software functional unit and sold or used as a standalone product, it can be stored in a computer-readable storage medium. Based on this understanding, the present application can implement all or part of the process steps in the above-mentioned method embodiments by instructing the relevant hardware through a computer program. The computer program can be stored in a computer-readable storage medium. When executed by a processor, the computer program can implement the steps of each of the above-mentioned method embodiments. The computer program includes computer program code, which can be in source code form, object code form, executable file, or some intermediate form. The computer-readable medium can include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard drive, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal, and software distribution medium. It should be noted that the content of the computer-readable medium can be appropriately increased or decreased based on the requirements of legislation and patent practice in a jurisdiction. For example, in some jurisdictions, based on legislation and patent practice, computer-readable media does not include electric carrier signals and telecommunication signals.

[0098] The above-described embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. These modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the various embodiments of the present application, and should all be included in the scope of protection of the present application.

Claims

1. A solution generation method based on TRIZ and a large language model, characterized in that: include: In response to receiving the problem description text, inputting a preset first prompt word into the large language model to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text; Inputting a preset second prompt word into the large language model to instruct the large language model to search a target innovation principle corresponding to the TRIZ contradiction parameter from a TRIZ contradiction matrix table; A preset third prompt word is input into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

2. The method according to claim 1, wherein The step of inputting a preset first prompt word into the large language model to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text includes: The first prompt word is input into the large language model to instruct the large language model to parse the problem description text, identify multiple groups of potential candidate contradiction parameters, and score each group of the candidate contradiction parameters according to a preset scoring rule, and determine the group of candidate contradiction parameters with the highest score value as the TRIZ contradiction parameters corresponding to the problem description text.

3. The method according to claim 1, wherein The second prompt word includes at least two different table lookup prompt words; the inputting of the preset second prompt word into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table includes: Traversing each of the table lookup prompt words, inputting the table lookup prompt words into the large language model, so as to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the table lookup method corresponding to the table lookup prompt word; If the innovation principle results corresponding to the table lookup prompt words are exactly the same, the innovation principle results corresponding to the table lookup prompt words are determined as the target innovation principle.

4. The method according to claim 3, wherein After traversing each of the table lookup prompt words and inputting the table lookup prompt words into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the table lookup method corresponding to the table lookup prompt word, the method further includes: If the innovation principle results corresponding to each of the lookup prompt words are different, then return to the step of traversing each of the lookup prompt words again, inputting the lookup prompt word into the large language model, and instructing the large language model to search the TRIZ contradiction matrix table for the innovation principle result corresponding to the TRIZ contradiction parameter according to the lookup method corresponding to the lookup prompt word, until the innovation principle results corresponding to each of the lookup prompt words are exactly the same.

5. The method according to claim 3, wherein After traversing each of the table lookup prompt words and inputting the table lookup prompt words into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for an innovation principle result corresponding to the TRIZ contradiction parameter according to the table lookup method corresponding to the table lookup prompt word, the method further includes: If the innovation principle results corresponding to the table lookup prompt words are different, the target innovation principle is determined from the innovation principle results corresponding to the table lookup prompt words by majority voting.

6. The method according to claim 1, wherein The TRIZ contradiction matrix table is divided into a plurality of sub-tables according to the deterioration parameter; and the step of inputting a preset second prompt word into the large language model to instruct the large language model to search the TRIZ contradiction matrix table for a target innovation principle corresponding to the TRIZ contradiction parameter includes: The second prompt word is input into the large language model to instruct the large language model to use the TRIZ contradiction parameter as an index to search for the target innovation principle from the multiple sub-tables.

7. The method according to claim 1, wherein Inputting a preset second prompt word into the large language model to instruct the large language model to search the target innovation principle corresponding to the TRIZ contradiction parameter from the TRIZ contradiction matrix table includes: The second prompt word is input into the large language model to instruct the large language model to call a TRIZ lookup tool, search for an innovation principle number corresponding to the number of the TRIZ contradiction parameter from a TRIZ contradiction matrix table, and determine each innovation principle corresponding to the innovation principle number as the target innovation principle.

8. The method according to claim 1, wherein Inputting a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle includes: The third prompt word is input into the large language model to instruct the large language model to use the retrieval enhancement generation technology to call the target knowledge base to generate various solutions corresponding to the problem description text based on the target innovation principle as the target solution.

9. The method according to claim 8, wherein The target knowledge base is determined by: Determine the technical field to which the target innovation principle belongs; At least one knowledge base corresponding to the technical field is searched from a plurality of preset knowledge bases as the target knowledge base.

10. The method according to claim 8, wherein The various solutions include non-cross-principle solutions based on a single innovative principle, and cross-principle solutions based on at least two innovative principles.

11. The method according to any one of claims 1 to 10, characterized in that After inputting a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle, the method further includes: Generate a structured report based on the problem description text, the TRIZ contradiction parameters, the target innovation principle and the target solution; The structured report is output.

12. A solution generation device based on TRIZ and a large language model, characterized in that: include: a contradiction parameter determination module, configured to input a preset first prompt word into the large language model in response to receiving the problem description text, so as to instruct the large language model to determine a TRIZ contradiction parameter corresponding to the problem description text; a table lookup module, configured to input a preset second prompt word into the large language model to instruct the large language model to search a target innovation principle corresponding to the TRIZ contradiction parameter from a TRIZ contradiction matrix table; A solution generation module is used to input a preset third prompt word into the large language model to instruct the large language model to generate a target solution corresponding to the problem description text based on the target innovation principle.

13. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein: When the processor executes the computer program, the solution generation method based on TRIZ and a large language model is implemented as described in any one of claims 1 to 11.

14. A computer program product, characterized in that When the computer program product is run on an electronic device, the electronic device is enabled to execute the solution generation method based on TRIZ and a large language model according to any one of claims 1 to 11.

15. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the solution generation method based on TRIZ and a large language model is implemented as claimed in any one of claims 1 to 11.

Images