Intelligent checking method for aircraft engine control software requirement description specification driven by large language model

Abstract

The present invention discloses a method for intelligently checking the specification of aircraft engine control software requirement description driven by a large language model, including: defining specific specifications of software requirement description; using a large language model to complete word specification checking and sentence specification checking of software requirements; scanning all software requirements and recording all defined terms and symbols; scanning software requirements with a large language model to complete term and symbol usage specification checking; constructing partial semantic specification positive and negative examples and marking them to form a use case library; searching for use cases with semantic similarity to the requirements, constructing prompt words, and completing semantic specification checking with a large language model. The present invention concretizes the specification of aircraft engine control software requirement description, effectively improves the efficiency of the standardization checking of software requirement description, can automatically check the implementation of software requirement description specification, reduce the risk of inadequate manual analysis, and reduce the workload of developers in the requirement analysis stage.

Description

Technical Field

[0001] The invention belongs to the technical field of software requirement analysis, and in particular relates to an intelligent inspection method for aircraft engine control software requirement description specifications driven by a large language model. Background Art

[0002] Software requirements are the product of the requirements analysis phase and are the foundation and basis for software design, development, testing and verification. The accuracy and standardization of requirements descriptions are crucial in the software development process. Accurate requirements descriptions enable all team members to have a consistent understanding of the requirements and reduce erroneous implementations caused by ambiguity; standardized requirements descriptions provide clear language and terminology to ensure that the development team has a common understanding during communication, reduce the risk of misunderstanding, and ensure that the final product meets the expected requirements and quality standards. Aircraft engine control software is safety-critical software with its own unique characteristics such as high safety, continuous operation, and real-time response, and the standardization of requirements descriptions is more stringent.

[0003] The big language model is a natural language processing model obtained by training a large amount of text data. It has strong context understanding and semantic analysis capabilities and can understand and analyze software requirements. In the field of software engineering, the big language model has a good performance in code completion, test case generation, etc., showing great potential.

[0004] Currently, the inspection of the standardization of requirement descriptions basically relies on manual review, which places extremely high demands on the reviewer and requires checking the requirement items one by one, which is time-consuming and labor-intensive, and it is unlikely to check all non-conformities. Therefore, the present invention clarifies the requirement description specifications in the field of aircraft engine control software, and uses a large language model to automate the inspection process and free up manpower. Summary of the invention

[0005] In order to solve the problems existing in the above-mentioned prior art, the present invention is based on the characteristics of aircraft engine control software, the word specifications, sentence specifications, specifications for the use of terms and symbols, and semantic specifications of the itemized software requirement description, and on the basis of defining the software requirement description specifications, combined with a large language model, realizes the automation of the requirement description specification checking item by item; to ensure that the requirement description complies with the specifications, and to help the requirement analysis engineer complete accurate and standardized requirement items.

[0006] In order to achieve the above technical objectives, the present invention provides the following technical solutions:

[0007] A large language model driven aerospace engine control software requirement description specification intelligent checking method includes the following steps:

[0008] S1. Define the software requirements description specifications;

[0009] S2. Use a large language model to complete word and sentence standardization checks on software requirements;

[0010] S3. Scan all software requirements and record all defined terms and symbols;

[0011] S4. Use the large language model to scan software requirements and complete the terminology and symbol usage standard check;

[0012] S5. Construct some semantically standardized positive and negative examples and annotate them to form a use case library;

[0013] S6. Find use cases that are semantically close to the requirements based on semantic similarity, construct prompt words, and use a large language model to complete semantic specification checking.

[0014] Furthermore, the large language models used in the method proposed in the present invention include: doubao-lite model, doubao-embedding model and doubao-pro model.

[0015] Furthermore, step S1 specifically includes:

[0016] S11. Define the vocabulary standard for describing software requirements;

[0017] S12. Define the sentence structure for describing software requirements;

[0018] S13. Define the terminology and symbol usage specifications for software requirement description;

[0019] S14. Define the semantic specification of software requirement description.

[0020] Furthermore, step S2 specifically includes:

[0021] S21. Collect common uncertainty descriptions and construct prompt words for word standardization checking;

[0022] S22, combining the prompt word constructed in step S21 with the requirement to be checked and submitting it to the doubao-lite model to complete the word standardization check;

[0023] S23, constructing positive examples and negative examples according to the sentence structure norms and marking them respectively, and combining them as prompt words for checking the sentence structure norms;

[0024] S24. Combining the prompt words constructed in step S23 with the requirements to be checked, and submitting them to the doubao-lite model, completes the sentence standardization check.

[0025] Furthermore, step S3 specifically includes:

[0026] S31, traverse the symbol definition table and record all defined symbols;

[0027] S32. Traverse the requirement items, submit them to the doubao-lite model, check the term definitions, and record all defined terms.

[0028] Furthermore, step S4 specifically includes:

[0029] S41. Submit each requirement item to the doubao-lite model and record all the terms and symbols used;

[0030] S42. Check whether the terms and symbols used appear in the definition.

[0031] Furthermore, step S5 specifically includes:

[0032] S51. Randomly select a number of unambiguous software requirement descriptions;

[0033] S52. Reshape the requirement description to form an unambiguous use case, mark it as a positive example, and mark the original requirement item as a negative example; the marking content includes: whether there is ambiguity, ambiguous sentences, and modification suggestions;

[0034] S53, submitting the original requirement items and the transformed requirement items to the doubao-embedding model to obtain their semantic embedding;

[0035] S54. Use cases, annotations and semantics are embedded and saved to form a use case library.

[0036] Furthermore, step S6 specifically includes:

[0037] S61. Submit the requirements to be checked to the doubao-embedding model to calculate their semantic embedding;

[0038] S62, using cosine similarity to calculate the similarity between the requirements to be checked and the use cases in the use case library, taking the top five with the largest similarity as samples, and constructing prompt words for semantic specification checking;

[0039] S63. Combining the constructed prompt words with the software requirements to be checked, and submitting them to the doubao-pro model, completes the semantic specification check.

[0040] In addition, the present invention also provides a computer device, including 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 above-mentioned large language model-driven intelligent inspection method for aircraft engine control software requirement description specifications can be implemented.

[0041] Through the above technical solution, the present invention has the following technical effects:

[0042] The present invention refines the specifications of aircraft engine control software requirement description to ensure the accuracy and standardization of software requirement description; uses a large language model to check the standardization of software requirements, effectively improves the efficiency of checking the standardization of aircraft engine control software requirements, can automatically check the implementation of requirement specifications, reduces the risk of inadequate manual analysis, and effectively reduces the workload of developers. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] Figure 1 is a flow chart of the steps of the present invention;

[0044] Figure 2 A schematic diagram of the present invention using a large language model to complete word and sentence standardization checking;

[0045] Figure 3 A schematic diagram of checking the terminology and symbol usage specifications using a large language model for the present invention;

[0046] Figure 4 This is a schematic diagram of the present invention using a large language model to complete semantic standard checking. DETAILED DESCRIPTION

[0047] In order to make the purpose, technical solution and advantages of the present invention more clearly understood, the present invention is further described in detail below in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0048] Although the steps in the present invention are arranged with numbers, they are not used to limit the order of the steps. Unless the order of the steps is clearly stated or the execution of a certain step requires other steps as a basis, the relative order of the steps can be adjusted. It is understood that the term "and / or" used in this article involves and covers any and all possible combinations of one or more of the associated listed items.

[0049] like Figures 1 to 4 As shown, the large language model driven intelligent inspection method for aircraft engine control software requirement description specifications proposed in the present invention is given. The large language model used to implement the present invention includes: doubao-lite model, doubao-embedding model and doubao-pro model.

[0050] The method proposed by the present invention specifically comprises the following steps:

[0051] S1. Define the software requirements description specifications;

[0052] As a preferred implementation of step S1, it specifically includes:

[0053] S11. Define the vocabulary standard for describing software requirements;

[0054] In this embodiment, the defined word specifications are given in Table 1 below:

[0055] Table 1 Word Standards

[0056] ;

[0057] S12. Define the sentence structure for describing software requirements;

[0058] In this embodiment, the defined sentence structure is shown in Table 2 below:

[0059] Table 2 Sentence specifications

[0060] ;

[0061] S13. Define the terminology and symbol usage specifications for software requirement description;

[0062] In this embodiment, the defined terms and symbol usage specifications are given in the following Table 3:

[0063] Table 3 Terms and symbols used

[0064] ;

[0065] S14. Define the semantic specification of software requirement description;

[0066] In this embodiment, the semantic specification is defined as: the software requirement description must be unambiguous.

[0067] S2. Use a large language model to complete word and sentence standardization checks on software requirements;

[0068] As a preferred implementation of step S2, it specifically includes:

[0069] S21. Collect common uncertainty descriptions and construct prompt words for word standardization checking;

[0070] S22, combining the prompt word constructed in step S21 with the requirement to be checked and submitting it to the doubao-lite model to complete the word standardization check;

[0071] S23, constructing positive examples and negative examples according to the sentence structure norms and marking them respectively, and combining them as prompt words for checking the sentence structure norms;

[0072] S24. Combining the prompt words constructed in step S23 with the requirements to be checked, and submitting them to the doubao-lite model, completes the sentence standardization check.

[0073] In this embodiment, Figure 2As shown, the software requirement items are given to the doubao-lite model and checked from two aspects: word standardization and sentence standardization. The word standardization requires that the software requirement description should be accurate, definite and free of typos, while the sentence standardization reflects the functional description of the software requirements.

[0074] S3. Scan all software requirements and record all defined terms and symbols;

[0075] As a preferred implementation of step S3, it specifically includes:

[0076] S31, traverse the symbol definition table and record all defined symbols;

[0077] S32. Traverse the requirement items, submit them to the doubao-lite model, check the term definitions, and record all defined terms.

[0078] Step 3 mainly records all the terms and symbols defined in the software requirements as a basis for checking the use of terms and symbols in the next step.

[0079] S4. Use the large language model to scan software requirements and complete the terminology and symbol usage standard check;

[0080] As a preferred implementation of step S4, it specifically includes:

[0081] S41. Submit each requirement item to the doubao-lite model and record all the terms and symbols used;

[0082] S42. Check whether the terms and symbols used appear in the definition.

[0083] In this embodiment, Figure 3 As shown, by giving all the defined terms and symbols of the software requirement items and all the used terms and symbols to the doubao-lite model respectively, and comparing whether the two have a corresponding relationship, that is, whether all the used terms and symbols have been defined, the standardization of the software requirement description can be checked.

[0084] S5. Construct some semantically standardized positive and negative examples and annotate them to form a use case library;

[0085] As a preferred implementation of step S5, it specifically includes:

[0086] S51. Randomly select a number of unambiguous software requirement descriptions;

[0087] S52. Modify the requirement description to form an ambiguous use case as a positive example and mark it, and mark the original requirement item as a negative example; the marking content includes: whether there is ambiguity, ambiguous sentences, and modification suggestions;

[0088] S53, submitting the original requirement items and the transformed requirement items to the doubao-embedding model to obtain their semantic embedding;

[0089] In this embodiment, the sentence is encoded into a numerical vector, which contains the semantic information of the software requirement item. This numerical vector is the semantic embedding, which can be directly obtained by submitting the software requirement item to the doubao-embedding model.

[0090] S54. Use cases, annotations and semantics are embedded and saved to form a use case library.

[0091] In this embodiment, Figure 4 As shown, step S5 is the preparation for the semantic standard check in step S6.

[0092] S6. Find use cases with semantic similarity to the requirements, construct prompt words, and use a large language model to complete semantic specification checking;

[0093] As a preferred implementation of step S6, it specifically includes:

[0094] S61. Submit the requirements to be checked to the doubao-embedding model to calculate their semantic embedding;

[0095] S62, using cosine similarity to calculate the similarity between the requirements to be checked and the use cases in the use case library, taking the top five with the largest similarity as samples, and constructing prompt words for semantic specification checking;

[0096] S63. Combining the constructed prompt words with the software requirements to be checked, and submitting them to the doubao-pro model, completes the semantic specification check.

[0097] In this embodiment, Figure 4 As shown, first, based on the use case library obtained in step S5, similar use cases in the software requirement items are selected as prompt words, and then given to the large language model (LLM) together with the original software requirement items to complete the semantic specification check, mainly focusing on ensuring that the description of the software requirements is unambiguous.

[0098] In summary, the method proposed in the present invention gradually realizes intelligent inspection of the description of aerospace engine control software requirements from multiple aspects such as accuracy, standardization, and non-ambiguity; at the same time, the use of a large language model to check the standardization of software requirements effectively improves the inspection efficiency of the standardization of aerospace engine control software requirements, can automatically check the implementation of requirement specifications, reduce the risk of inadequate manual analysis, and effectively reduce the workload of developers.

[0099] It will be apparent to those skilled in the art that the invention is not limited to the details of the exemplary embodiments described above and that the invention can be implemented in other specific forms without departing from the spirit or essential features of the invention. Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description, and it is intended that all variations falling within the meaning and scope of the equivalent elements of the claims be included in the invention. Any reference numeral in a claim should not be considered as limiting the claim to which it relates.

[0100] In addition, it should be understood that although the present specification is described according to implementation modes, not every implementation mode contains only one independent technical solution. This description of the specification is only for the sake of clarity. Those skilled in the art should regard the specification as a whole. The technical solutions in each embodiment may also be appropriately combined to form other implementation modes that can be understood by those skilled in the art.

Claims

1. An intelligent inspection method for aircraft engine control software requirement description specification driven by a large language model, characterized in that: include: S1. Define the software requirements description specifications; S2. Use large language models to complete word and sentence standardization checks on software requirements; the large language models used include: doubao-lite model, doubao-embedding model and doubao-pro model; S3. Scan all software requirements and record all defined terms and symbols; S4. Use the large language model to scan software requirements and complete the terminology and symbol usage standard check; S5, constructing some semantic norm positive examples and negative examples and annotating them to form a use case library; step S5 specifically includes: S51. Randomly select a number of unambiguous software requirement descriptions; S52. Reshape the requirement description to form an unambiguous use case, which is used as a positive example and annotated. The original requirement item is used as a negative example and annotated. The annotation content includes: whether there is ambiguity, ambiguous sentences, and modification suggestions. S53, submitting the original requirement items and the transformed requirement items to the doubao-embedding model to obtain their semantic embedding; S54, embed and save the use cases, annotations and semantics to form a use case library; S6. Find use cases with semantic similarity to the requirements according to semantic similarity, construct prompt words, and use a large language model to complete semantic standard check; step S6 specifically includes: S61. Submit the requirements to be checked to the doubao-embedding model to obtain their semantic embedding; S62, using cosine similarity to calculate the similarity between the requirements to be checked and the use cases in the use case library, taking the top five with the largest similarity as samples, and constructing prompt words for semantic specification checking; S63. Combining the constructed prompt words with the software requirements to be checked, and submitting them to the doubao-pro model, completes the semantic specification check.

2. The large language model driven aircraft engine control software requirement description specification intelligent inspection method according to claim 1 is characterized in that: Step S1 specifically includes: S11. Define the vocabulary standard for describing software requirements; S12. Define the sentence structure for describing software requirements; S13. Define the terminology and symbol usage specifications for software requirement description; S14. Define the semantic specification of software requirement description.

3. The large language model driven aircraft engine control software requirement description specification intelligent inspection method according to claim 1 is characterized in that: Step S2 specifically includes: S21. Collect common uncertainty descriptions and construct prompt words for word standardization checking; S22, combining the prompt word constructed in step S21 with the requirement to be checked and submitting it to the doubao-lite model to complete the word standardization check; S23, constructing positive examples and negative examples according to the sentence structure norms and marking them respectively, and combining them as prompt words for checking the sentence structure norms; S24. Combining the prompt words constructed in step S23 with the requirements to be checked, and submitting them to the doubao-lite model, completes the sentence standardization check.

4. The large language model driven intelligent inspection method for aircraft engine control software requirement description specification according to claim 1 is characterized in that: Step S3 specifically includes: S31, traverse the symbol definition table and record all defined symbols; S32. Traverse the requirement items, submit them to the doubao-lite model, check the term definitions, and record all defined terms.

5. The large language model driven aircraft engine control software requirement description specification intelligent inspection method according to claim 1 is characterized in that: Step S4 specifically includes: S41. Submit each requirement item to the doubao-lite model and record all the terms and symbols used; S42. Check whether the terms and symbols used appear in the definition.

6. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the large language model-driven intelligent inspection method for aircraft engine control software requirement description specifications described in any one of claims 1 to 5 can be implemented.

Images