Data processing method, electronic device, storage medium, and program product
By extracting feature information and filtering similar data through data processing methods, the problem of low retrieval accuracy in the system was solved, achieving high-precision fault root cause localization and diagnosis, and reducing irrelevant cases and computational redundancy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 欧摩威软件系统开发(重庆)有限公司
- Filing Date
- 2026-04-24
- Publication Date
- 2026-07-21
AI Technical Summary
When searching for similar historical fault cases in the system, many irrelevant cases are often recommended due to the similarity of the fault text descriptions, resulting in low search accuracy and computational redundancy.
By extracting feature information from the data to be retrieved, including unique identifiers and search intent, and combining the similarity of identifiers and intents of similar data in the database, target data that meets preset conditions is selected. When necessary, image information is filtered and OCR technology is applied to improve search accuracy.
It improves the similarity between the retrieved target data and the data to be retrieved, reduces irrelevant cases, avoids computational redundancy, saves computing power, and improves the efficiency of fault diagnosis and repair.
Smart Images
Figure CN122432329A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of computer technology, and in particular to a data processing method, electronic device, storage medium, and program product. Background Technology
[0002] In project development or testing phases, quickly identifying the root cause of software failures (also known as "root cause localization") is a core step in ensuring software quality and improving development efficiency. During root cause localization, the system (such as a project management system) can be searched for similar historical failures to quickly deduce the cause of the current failure and corresponding solutions.
[0003] Currently, when searching for similar historical failure cases in the system, the system typically ignores the underlying technical differences between different projects, often recommending more irrelevant cases due to the similarity of the failure text descriptions. For example, if the failure condition of a current project is "upgrade failed," the system only focuses on the failure description "upgrade failed" when searching for similar historical failure cases, ignoring the underlying technologies such as the software version and communication module of the current project. This results in recommending many historical failure cases of "upgrade failed" from other projects with significantly different underlying technologies from the current project, forcing users to spend a lot of time manually identifying usable cases from a large number of recommended cases.
[0004] In summary, the current retrieval accuracy is low when searching for historical faults similar to the current fault. Summary of the Invention
[0005] This application provides a data processing method, electronic device, storage medium, and program product for achieving high-precision data retrieval.
[0006] Firstly, this application provides a data processing method applied to an electronic device. The method includes: acquiring data to be retrieved and extracting feature information from the data to be retrieved, wherein the feature information includes unique identifier information and a search intent. Based on the feature information, one or more similar data are retrieved from a database, and the unique identifier information and search intent of each of the one or more similar data are obtained. Based on the identifier similarity between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the intent similarity between the search intent of each similar data and the search intent of the data to be retrieved, target data that meets preset conditions is selected from the one or more similar data.
[0007] Thus, by using the above method, when retrieving target data, electronic devices not only consider the retrieval intent (such as malfunctions) but also the unique identifier information of the item. This makes the similarity between the retrieved target data and the data to be retrieved higher, and there will be fewer irrelevant cases, thereby improving retrieval accuracy.
[0008] In one possible implementation of the first aspect above, the extraction of feature information from the data to be retrieved includes: extracting unique identifier information from the data to be retrieved based on regular expression validation, wherein each character in the unique identifier information exists in the data to be retrieved.
[0009] This avoids the possibility that artificial intelligence (AI) might create non-existent words when extracting keywords, thus preventing zero recall or illusion phenomena during the retrieval process.
[0010] In one possible implementation of the first aspect above, retrieving one or more similar data from the database based on feature information includes: retrieving first similar data from the database based on feature information; wherein the unique identifier information of the first similar data is the same as the unique identifier information of the data to be retrieved, and / or, the keywords in the search intent of the first similar data are the same as the keywords in the search intent of the data to be retrieved. This allows for a higher degree of similarity between the retrieved first similar data and the data to be retrieved.
[0011] In one possible implementation of the first aspect described above, the method further includes: when the number of first similar data is less than a first quantity threshold, changing the search intent of the data to be retrieved to obtain a variant intent; and retrieving second similar data from the database based on the variant intent. In this way, a sufficient amount of similar data can be retrieved to meet the user's data retrieval needs.
[0012] In one possible implementation of the first aspect described above, the method further includes: if the total amount of the first and second similar data is less than a first quantity threshold, retrieving third similar data from the database based on the full-text semantics of the data to be retrieved. This allows for the retrieval of a sufficient amount of similar data to meet the user's data retrieval needs.
[0013] In one possible implementation of the first aspect described above, selecting target data that meets preset conditions from one or more similar data based on the similarity between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the similarity between the retrieval intent of each similar data and the retrieval intent of the data to be retrieved, includes: determining a comprehensive score corresponding to each similar data based on the weights corresponding to the identifier similarity, the similarity between the identifier information of each similar data, the weights corresponding to the intent similarity, the intent similarity between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the weights corresponding to each conflict factor. Based on the comprehensive scores corresponding to each similar data, n target data are selected from one or more similar data, where n is a positive integer. The comprehensive score of each of the n target data is greater than the comprehensive score of all the similar data other than the n target data in the one or more similar data sets.
[0014] This results in a higher similarity between the retrieved target data and the data to be retrieved, with fewer irrelevant cases, thus improving retrieval accuracy.
[0015] In one possible implementation of the first aspect above, if the first target data among the n target data includes one or more images, the method further includes: extracting information from one or more images to obtain image information; if the number of images in the first target data is less than the image number threshold, injecting the extracted image information into the target result; or, if the number of images in the first target data is greater than or equal to the image number threshold, selecting target images whose image information meets the relevance condition from each image based on the feature information of the data to be retrieved, and injecting the image information of the target images into the target result.
[0016] Therefore, in this application, the electronic device does not directly use optical character recognition (OCR) technology to recognize the image content in the database. Instead, after filtering the target data, it only recognizes the image content within the target data. Furthermore, it can determine whether to inject all image information into the target result based on the number of images. This avoids generating excessive invalid noise text, does not pollute the search index, and saves computational resources.
[0017] In one possible implementation of the first aspect above, the data to be retrieved may be a software fault ticket; the unique identification information may include, but is not limited to, one or more of the following: negative response code, product number, software version number, underlying driver name, and hexadecimal code corresponding to the error message; the retrieval intent may include, but is not limited to, business actions and / or functional module names.
[0018] In a second aspect, this application also provides an electronic device, comprising: at least one memory and at least one processor, wherein the memory is coupled to the processor; the memory is used to store computer program code / instructions; when the computer program code / instructions are executed by the processor, the electronic device performs the data processing method mentioned in the first aspect and any possible implementation thereof.
[0019] Thirdly, this application also provides a readable storage medium storing instructions that, when executed on an electronic device, cause the electronic device to perform the data processing methods mentioned in the first aspect and any possible implementation thereof.
[0020] Fourthly, this application also provides a computer program product, comprising: computer instructions that, when executed on an electronic device, cause the electronic device to perform the data processing method mentioned in the first aspect and any possible implementation thereof.
[0021] The beneficial effects of the second to fourth aspects mentioned above can be referred to the relevant descriptions in the first aspect and any possible implementation of the first aspect, which will not be repeated here. Attached Figure Description
[0022] Figure 1 According to some embodiments, a schematic diagram of a fault ticket in a project management system is shown;
[0023] Figure 2 A flowchart illustrating a first data processing method is shown according to some embodiments of this application;
[0024] Figure 3 According to some embodiments of this application, a flowchart of a second data processing method is shown;
[0025] Figure 4 According to some embodiments of this application, a flowchart of a third data processing method is shown;
[0026] Figure 5 According to some embodiments of this application, a flowchart of a fourth data processing method is shown;
[0027] Figure 6 According to some embodiments of this application, a schematic diagram of the hardware structure of an electronic device is shown. Detailed Implementation
[0028] The illustrative embodiments of this application include, but are not limited to, data processing methods, electronic devices, storage media, and program products.
[0029] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions in the embodiments of this application will be described in detail below with reference to the accompanying drawings and specific implementation methods.
[0030] During project development or testing phases, if software malfunctions, users can submit a malfunction report in the project management system (such as Jira). For example, Figure 1 As shown, a fault report may include a report number (e.g., E190E5), fault details (e.g., SWITCH upgrade failure, which caused CCU upgrade failure), software version, problem priority, severity, relevant attachments (e.g., test failure report, screenshots of operation monitoring screens, etc.), and activity logs (e.g., discussion or comment content from relevant users, etc.).
[0031] In the process of root cause analysis of software failures, electronic devices can search the project management system for historical failure records based on current failure documents. This allows for quick deduction of the cause of the current failure and corresponding solutions. However, when searching for similar historical failure records in the system, the system often ignores the underlying technical differences between different projects and frequently recommends more irrelevant cases due to the similarity of the failure descriptions.
[0032] Furthermore, when electronic devices retrieve similar historical fault cases, if the documents contain images, the devices typically first use optical character recognition (OCR) technology to identify the image content within all documents, thereby retrieving historical fault documents with similar fault cases. However, if... Figure 1 As shown, user-submitted documents may contain numerous images, such as those uploaded in user comments. In such cases, directly using OCR technology to recognize all images in all documents could generate a significant amount of invalid, noisy text, polluting the search index and wasting computational resources. Therefore, the current method of using OCR technology to recognize all image content in all documents exhibits computational redundancy.
[0033] In summary, the current method of retrieving historical faults that are similar to the current fault situation suffers from low retrieval accuracy and computational redundancy.
[0034] To address the aforementioned problems, this application provides a data processing method. Specifically, after acquiring data to be retrieved (such as software fault documents), the electronic device can extract feature information from the data to be retrieved. This feature information may include unique identifiers (such as negative response codes, software version numbers, and / or underlying driver names) and retrieval intents (such as upgrade failures or timeouts). Then, the electronic device can retrieve one or more similar data sets from a database (such as the database of a project management system) based on the feature information, and obtain the unique identifiers and retrieval intents of each similar data set. Finally, the electronic device can select target data that meets preset conditions from the one or more similar data sets based on the similarity between the unique identifiers of each similar data set and the unique identifiers of the data to be retrieved, and the similarity between the retrieval intents of each similar data set and the retrieval intents of the data to be retrieved.
[0035] Thus, by using the above method, when retrieving target data, electronic devices not only consider the retrieval intent (such as malfunctions) but also the unique identifier information of the item. This makes the similarity between the retrieved target data and the data to be retrieved higher, and there will be fewer irrelevant cases, thereby improving retrieval accuracy.
[0036] It should be noted that the data processing method described in this application can be applied to the software defect analysis process during the R&D phase, automating the process to help developers locate the root cause of faults and improving debugging efficiency. Alternatively, the data processing method described in this application can also be integrated into VS Code. When a developer selects a Log snippet, the plugin executes the data processing method to immediately match the historical repair baseline. Furthermore, the data processing method described in this application can also run on a CI server. When a test fails, the electronic device executes the data processing method to determine whether it is a known system defect, reducing repetitive analysis work. Therefore, the data processing method provided in this application can be applied to multiple fields such as software diagnostics, AI-assisted R&D, and enhanced search generation, and this application does not limit its application to these areas.
[0037] It should also be noted that the data processing method described in this application can be applied to any electronic device. This electronic device includes, but is not limited to, mobile stations (MS) and mobile terminals (MT). For example, the electronic device can be a computer, tablet computer, desktop computer, laptop computer, a terminal in industrial control, a terminal in self-driving technology, a terminal in a smart grid, a terminal in transportation safety, a terminal in a smart city, a terminal in a smart home, etc. This application does not limit the specific form of the electronic device in its embodiments.
[0038] The following is based on Figure 2 The flowchart shown provides a brief overview of the data processing method mentioned in the embodiments of this application. This data processing method can be applied to electronic devices, such as any electronic device like the computer mentioned above. Figure 2 As shown, specifically, the method is as follows:
[0039] S21: Obtain the data to be retrieved and extract the feature information from the data to be retrieved. The feature information includes unique identifier information and retrieval intent.
[0040] In some embodiments, the data to be retrieved may be a software fault ticket. The unique identification information (also known as a hard fingerprint or deterministic decision factor) may include, but is not limited to, one or more of the following: Negative Response Code (NRC), specific product identifier, software version number, underlying driver name, and hexadecimal code corresponding to the error message (such as 0x7F). The retrieval intent (also known as the core intent or semantic auxiliary factor) may include, but is not limited to, business actions (such as switchover failure, timeout, etc.) and / or functional module names (such as autosar.com, dcm, uds, etc.).
[0041] In some embodiments, electronic devices can extract unique identifiers from the data to be retrieved based on regular expression validation, thereby ensuring that each character in the unique identifier exists in the data to be retrieved. For example, the electronic device can first use artificial intelligence (AI) algorithms to extract data, and then use regular expression validation to determine whether the data output by the AI algorithm conforms to the rules (such as whether the software version number extracted by the AI conforms to the version number format) and whether it exists character by character in the original data to be retrieved (such as Log or title). If a unique identifier output by the AI algorithm conforms to the rules and exists character by character in the data to be retrieved, then the unique identifier is retained; otherwise, if a keyword output by the AI algorithm does not conform to the rules or does not exist character by character in the data to be retrieved, then the keyword is discarded. In this way, it is possible to avoid the possibility that AI may create non-existent words when extracting keywords, thereby avoiding the occurrence of zero recall or illusion phenomena during the retrieval process.
[0042] S22: Based on feature information, retrieve one or more similar data from the database, and obtain the unique identifier information and retrieval intent of each similar data in the one or more similar data.
[0043] In some embodiments, the electronic device can directly retrieve one or more first similar data from the database based on feature information. Specifically, the unique identifier information of each retrieved first similar data is the same as the unique identifier information of the data to be retrieved, and / or, the keywords in the search intent of each retrieved first similar data (such as functional module names, business actions, or faults) are the same as the keywords in the search intent of the data to be retrieved. This allows for a higher degree of similarity between the retrieved first similar data and the data to be retrieved.
[0044] In some embodiments, if the number of first similar data retrieved is greater than or equal to a first quantity threshold, the retrieval process can be terminated to execute S23. Conversely, in other embodiments, if the number of first similar data retrieved is less than the first quantity threshold, the electronic device can further generalize the retrieval intent of the data to be retrieved to obtain variant intents, and continue to retrieve second similar data in the database based on the variant intents.
[0045] In some embodiments, if the total amount of the first and second similar data is greater than or equal to a first quantity threshold, the retrieval process can be terminated to execute S23. Conversely, in other embodiments, if the total amount of the first and second similar data is less than the first quantity threshold, the electronic device can continue semantic retrieval in the database based on the full-text semantics of the data to be retrieved, thereby obtaining third similar data. The first quantity threshold can be determined by the developers based on experiments or experience, and this application does not limit it.
[0046] In this way, based on the above method, a sufficient amount of similar data can be retrieved to meet the user's data retrieval needs.
[0047] S23: Based on the similarity of the unique identifier information between each similar data and the unique identifier information of the data to be retrieved, and the similarity of the search intent between each similar data and the search intent of the data to be retrieved, select the target data that meets the preset conditions from one or more similar data.
[0048] In some embodiments, among the similar data retrieved in S22 above, there may be a similar data A1 whose unique identifier information is exactly the same as the unique identifier information of the data to be retrieved, and the keywords in the retrieval intent of the similar data A1 (such as functional module name, business action, or fault, etc.) are also exactly the same as the keywords in the retrieval intent of the data to be retrieved. For example, if the same component in the project has had the same problem as the current fault, then it means that the identifier similarity and intent similarity of the similar data A1 are both 100%, and the electronic device can directly use the similar data A1 as the target data.
[0049] In other embodiments, the unique identifier information of the similar data retrieved in S22 may differ from the unique identifier information of the data to be retrieved, or the keywords in the search intent of the similar data may not be completely consistent with the keywords in the search intent of the data to be retrieved. In this case, the electronic device needs to determine the overall similarity of the similar data based on the similarity of the identifier information and the similarity of the intent, thereby determining whether the similar data is the target data required by the user.
[0050] For example, the unique identifier of similar data A2 may be the same as that of the data to be retrieved (e.g., the similarity is 100%), but the keywords in the search intent of similar data A2 differ from those in the search intent of the data to be retrieved (e.g., the similarity is 40%). Similarly, the unique identifier of similar data A3 may differ from that of the data to be retrieved (e.g., the similarity is 70%), but the keywords in the search intent of similar data A3 are exactly the same as those in the search intent of the data to be retrieved (e.g., the similarity is 100%). In this case, the electronic device can determine the target data that better meets the user's needs by assessing the overall similarity between similar data A2 and similar data A3.
[0051] The process of determining the overall similarity of each similar data point is described below.
[0052] It is understandable that, in the actual operation of the code, for similar data with the same unique identification information and retrieval intent, the electronic device can directly assign the similar data an overall similarity of 100%, or the electronic device can uniformly calculate the overall similarity of the similar data in the following manner, which is not limited in this application.
[0053] For example, an electronic device can determine the overall similarity of each similar data based on the weights corresponding to the similarity of identification information, the similarity of identification information corresponding to each similar data, the weights corresponding to the similarity of intent, the similarity of intent corresponding to each similar data, the conflict factors (which can be called fingerprint conflict factors) between the unique identification information of each similar data and the unique identification information of the data to be retrieved, and the weights corresponding to each fingerprint conflict factor. For example, the overall similarity can be represented by a comprehensive score. Specifically, the calculation process of the comprehensive score can be referred to as shown in the following formula (1):
[0054] (1);
[0055] In the above formula (1), a is the weight corresponding to the intent similarity. The intent similarity corresponding to each similar data can be determined by calculating the cosine similarity, etc.
[0056] In the above formula (1), b is the weight corresponding to the similarity of the identification information. The similarity of the identification information corresponding to each similar data can also be determined by calculating the cosine similarity, etc. Alternatively, the electronic device can also set a corresponding score (which can be called fingerprint matching score) for it according to the size of the similarity of the identification information corresponding to each similar data. In this case, b can also be the weight corresponding to the fingerprint matching score.
[0057] In the above formula (1), c is the weight corresponding to each fingerprint conflict factor. For example, if the software version of similar data is mutually exclusive with the software version of the data to be retrieved, then c can be the weight corresponding to the conflict factor of software version mutual exclusion. For another example, if the error code meaning of similar data is different from the error code meaning of the data to be retrieved, then c can be the weight corresponding to the conflict factor of different error code meaning.
[0058] It is understood that in the above formula (1), a, b, and c can be determined by the developers based on experiments or experience, and this application does not limit them. For example, in some embodiments, c can be assigned a large weight. In this case, even if a similar data has a high degree of similarity to the search intent of the data to be searched, if the fingerprint verification (that is, the unique identification information) conflicts (such as version mutual exclusion or different meanings of error codes), the similar data can also be forcibly downgraded according to the weight c corresponding to the conflict factor.
[0059] Thus, based on formula (1), the electronic device can calculate the comprehensive score corresponding to each similar data.
[0060] Furthermore, in some embodiments, after calculating the comprehensive score corresponding to each similar data point, the electronic device can select n (n is a positive integer) high-potential target data points (also called target similar data points) from the similar data. The comprehensive score of each of the selected n target data points is greater than the comprehensive score of the other unselected similar data points. This results in a higher similarity between the retrieved target data and the data to be retrieved, minimizing irrelevant cases and improving retrieval accuracy.
[0061] Furthermore, in some embodiments, if the first target data among the selected n target data contains one or more images, such as Figure 1 If the software fault document contains multiple images, the electronic device can also extract information from one or more images in the first target data, for example, by using OCR technology to extract the image information.
[0062] Then, if the number of images in the first target data is less than the image number threshold, the extracted image information can be directly injected into the target result (i.e., the Prompt). Conversely, if the number of images in the first target data is greater than or equal to the image number threshold, the electronic device can select target images whose image information meets the relevance conditions from each image based on the feature information of the data to be retrieved, and inject the image information of the target image into the target result. For example, if the image information of image A in the first target data has a poor relevance to the feature information of the data to be retrieved (such as not mentioning the keywords in the feature information), the image information of image A can be directly discarded. As another example, if the image information of image B in the first target data has a strong correlation with the feature information of the data to be retrieved, the image information of image B can be injected into the target result. Thus, in this application, the electronic device can determine whether to inject all image information into the target result based on the number of images. This avoids generating a lot of invalid noise text, does not pollute the search index, and saves computing power.
[0063] In summary, the method provided in this application achieves higher similarity between the retrieved target data and the data to be retrieved, minimizing irrelevant cases and improving retrieval accuracy. Furthermore, it avoids generating excessive invalid noise text and saves computational resources.
[0064] The following is combined with Figure 3 The flowchart shown illustrates the data processing method mentioned above, using a software fault ticket as the data to be retrieved as an example. This data processing method can be applied to electronic devices, such as any electronic device like the computer mentioned above. Figure 3 As shown, specifically, the method is as follows:
[0065] S31: New software fault report received.
[0066] S32: Extract the unique identifier information from the new software fault ticket.
[0067] For example, the unique identification information may include, but is not limited to, one or more of the following: negative response code, specific product number, software version number, underlying driver name, and hexadecimal code corresponding to the error message (such as 0x7F).
[0068] For example, electronic devices can use AI algorithms combined with regular expression validation to extract unique identifier information from new software fault documents, thereby ensuring that the extracted words exist character by character in the new software fault documents (such as Log or title). The specific method for extracting unique identifier information can be found above. Figure 2 As described in S21, it will not be repeated here.
[0069] S33: Extract the search intent from the new software fault report.
[0070] For example, the search intent may include, but is not limited to, business actions (such as switching failure, timeout, etc.) and / or functional module names (such as autosar com, dcm, uds, etc.).
[0071] For example, electronic devices can also use AI algorithms combined with regular expression validation to extract the search intent from new software fault documents, and this application does not limit this.
[0072] S34: Retrieve one or more candidate documents based on feature information.
[0073] For example, the unique identifier information of each retrieved candidate document is the same as the unique identifier information of the new software fault document, and / or, the keywords (such as functional module names) in the search intent of each retrieved candidate document are the same as the keywords in the search intent of the new software fault document. This increases the similarity between the retrieved candidate documents and the new software fault document.
[0074] S35: Determine the similarity of intent between the search intent of each candidate document and the search intent of the new software failure document.
[0075] For example, electronic devices can determine the intent similarity corresponding to each candidate document in a variety of ways. For instance, electronic devices can calculate the cosine similarity between the search intent of each candidate document and the search intent of the new software failure document, and use the cosine similarity as the intent similarity.
[0076] S36: Determine the similarity of the unique identifier information between each candidate document and the unique identifier information of the new software failure document.
[0077] For example, electronic devices can determine the similarity of the identification information corresponding to each candidate document in a variety of ways. For instance, electronic devices can calculate the cosine similarity between the unique information identifier of each candidate document and the unique information identifier of the new software failure document, and use the cosine similarity as the identification similarity.
[0078] In some embodiments, the electronic device may also assign a corresponding score (which may be called a fingerprint matching score) to each candidate document based on the similarity of the identification information corresponding to each document. For example, the greater the similarity of the identification information, the higher the corresponding fingerprint matching score. This application does not limit this.
[0079] S37: Identify the conflict factors between the unique identifier information of each candidate document and the unique identifier information of the new software failure document.
[0080] For example, for unique identification information of the same type in both candidate documents and new software failure documents, electronic devices can determine whether there is a conflict. For instance, whether the software version of the candidate document and the software version of the new software failure document are mutually exclusive, or whether the meaning of the error code in the candidate document differs from the meaning of the error code in the new software failure document.
[0081] S38: Determine the overall score for each candidate document.
[0082] In some embodiments, the electronic device can determine the comprehensive score corresponding to each candidate document based on the weight corresponding to the similarity of identification information, the similarity of identification information corresponding to each candidate document, the weight corresponding to the similarity of intent, the similarity of intent corresponding to each candidate document, the conflict factor between the unique identification information of each candidate document and the unique identification information of the data to be retrieved, and the weight corresponding to each fingerprint conflict factor. For details, please refer to the above. Figure 2 As described in S23, it will not be repeated here.
[0083] S39: Based on the comprehensive score of each candidate document, select n target documents from the candidate documents, where n is a positive integer.
[0084] For example, the overall score of each of the selected n target documents is greater than the overall score of the other unselected candidate documents.
[0085] S310: On-demand triggered OCR technology.
[0086] For example, if an image is detected in the target document, the electronic device can recognize the content in the image using OCR technology.
[0087] Understandably, compared to traditional technologies that directly use OCR technology to recognize the image content in all documents, in this application, the electronic device does not directly use OCR technology to recognize the image content in the documents. Instead, it filters out the target documents and only recognizes the image content in those documents. This avoids generating a lot of invalid noise text, does not pollute the search index, and saves computing power.
[0088] S311: Inject the target result.
[0089] For example, an electronic device can inject the identified text and image content from the target document as final evidence into the target result (i.e., the Prompt).
[0090] S312: Output diagnostic reports and repair suggestions.
[0091] For example, electronic devices can output diagnostic reports and repair suggestions for new software fault documents based on the causes of the fault and the corresponding solutions recorded in the target document.
[0092] In this way, the above method can make the similarity between the retrieved target documents and the new software fault documents higher, with fewer irrelevant cases, thus improving the retrieval accuracy and facilitating users in fault diagnosis and repair.
[0093] The following is combined with Figure 4 The flowchart shown illustrates the data processing methods mentioned above, using software fault tickets as an example. Specifically, it describes the data processing methods related to... Figure 3 Compared to the data processing methods shown, Figure 4 The data processing method shown can ensure a sufficient number of candidate documents are retrieved. This data processing method can be applied to electronic devices, such as any electronic device like the computer mentioned above. Figure 4 As shown, specifically, the method is as follows:
[0094] S41: New software fault report received.
[0095] S42: Perform precise searches.
[0096] For example, precise retrieval can refer to retrieving one or more candidate documents based on the characteristic information of the new software fault document, where the unique identifier information of each candidate document is the same as the unique identifier information of the new software fault document, and / or, the keywords (such as functional module names) in the search intent of each candidate document are the same as the keywords in the search intent of the new software fault document. For details, please refer to... Figure 3 As described in S32 to S34, they will not be repeated here.
[0097] S43: Determine if the search results are sufficient. If yes, proceed to S47; otherwise, proceed to S44.
[0098] For example, if the number of candidate documents retrieved by the precise search (as an example of first similar data) is greater than or equal to a first quantity threshold, the search process can be terminated to execute S47. Conversely, in some embodiments, if the number of candidate documents retrieved by the precise search is less than the first quantity threshold, S44 can be executed for further searching.
[0099] S44: Generalize the search intent to a variant and perform further searches based on the generalized intent.
[0100] For example, generalizing the search intent in variations can include synonyms for obtaining the search intent.
[0101] S45: Determine if the search results are sufficient. If yes, proceed to S47; otherwise, proceed to S46.
[0102] For example, if the total number of candidate documents retrieved through precise retrieval (as an example of first similar data) and candidate documents retrieved based on generalized intent (as an example of second similar data) is greater than or equal to a first quantity threshold, the retrieval process can be terminated to execute S47. Conversely, in some embodiments, if the total number of retrieved candidate documents is less than the first quantity threshold, S46 can be executed for further retrieval.
[0103] S46: Perform semantic retrieval.
[0104] For example, electronic devices can also continue semantic retrieval based on the full-text semantics of the new software fault document, thereby obtaining more candidate documents (as an example of third-party similar data).
[0105] S47: Determine the target document based on the retrieved candidate documents.
[0106] For details, please refer to the above. Figure 3 As described in S35 to S39, they will not be repeated here.
[0107] For example, in other embodiments, after identifying the target document, the electronic device can also determine whether OCR technology needs to be triggered, and the electronic device can also output a diagnostic report and repair suggestions. See the above for details. Figure 3 As described in S310 to S312, they will not be repeated here.
[0108] In this way, the above method can increase the similarity between the retrieved target documents and the new software fault documents, minimizing irrelevant cases and improving retrieval accuracy. Furthermore, it can obtain a larger number of search results, which is beneficial for users in fault diagnosis and repair.
[0109] The following is combined with Figure 5 The flowchart shown illustrates the data processing methods mentioned above, using software fault tickets as an example. Specifically, it describes the data processing methods related to... Figure 3 or Figure 4 Compared to the data processing methods shown, Figure 5 The data processing method shown can filter the image content in the target document, thus avoiding the injection of too many irrelevant images into the Prompt. This data processing method can be applied to electronic devices, such as any electronic device like the computer mentioned above. Figure 5 As shown, specifically, the method is as follows:
[0110] S51: New software fault report received.
[0111] S52: Extract unique identifier information and search intent from new software fault reports.
[0112] For details, please refer to the above. Figure 3 As described in S32 and S33, they will not be repeated here.
[0113] S53: Identify the target document.
[0114] For details, please refer to the above. Figure 3 S34 to S39, or the above Figure 4 As described in S42 to S47, they will not be repeated here.
[0115] S54: Determine whether OCR technology is triggered. If not, proceed to S55; if yes, proceed to S56.
[0116] For example, if the target document does not contain an image, the electronic device does not need to perform OCR technology. Conversely, if the target document contains an image, the electronic device needs to use OCR technology to recognize the image content within the target document.
[0117] S55: Extract the text information from the target document.
[0118] For example, an electronic device can inject the extracted text information from the target document into Prompt.
[0119] S56: Perform OCR extraction.
[0120] For example, if the target document contains an image, the electronic device can extract the image information using OCR technology. For instance, such as... Figure 1 The document shown contains many images (such as images in attachments and comments). Electronic devices can use OCR technology to extract the image information from these images.
[0121] S57: Determine the correlation between image information and feature information as needed, based on the number of images.
[0122] In some embodiments, if the number of images in the first target document (as an example of the first target data) is less than the image number threshold, S58 can be directly executed to inject the extracted image information into Prompt. Conversely, if the number of images in the first target document is greater than or equal to the image number threshold, the electronic device can filter out target images whose image information meets the relevance condition from each image based on the feature information of the new software fault document, and inject the image information of the target image into the target result. For example, if the image information of image A in the first target document has a poor relevance to the feature information of the new software fault document (e.g., it does not mention the keywords in the feature information), the image information of image A can be directly discarded. As another example, if the image information of image B in the first target document has a strong correlation with the feature information of the new software fault document, S58 can be executed to inject the image information of image B into the target result. Thus, in this application, the electronic device can determine whether to inject all image information into the target result based on the number of images. This avoids generating a lot of invalid noise text, does not pollute the search index, and saves computing power.
[0123] S58: Inject Prompt.
[0124] It is understandable that the Prompt can contain not only the image information mentioned in S57 above, but also the text information extracted from the target document.
[0125] S59: Output the target result.
[0126] For example, electronic devices can output diagnostic reports and repair suggestions for new software fault documents based on the causes of the fault and the corresponding solutions recorded in the target document.
[0127] Thus, by using the above method, the similarity between the retrieved target documents and the new software fault documents can be increased, improving retrieval accuracy and facilitating users in fault diagnosis and repair. Furthermore, it avoids generating excessive invalid noise text and saves computational resources.
[0128] Furthermore, in some embodiments, this application also provides a readable storage medium storing instructions. When executed on an electronic device, the instructions cause the electronic device to implement the data processing method mentioned in this application.
[0129] Furthermore, in some embodiments, this application also provides a computer program product, including computer instructions. These computer instructions, when executed on an electronic device, cause the electronic device to implement the data processing method mentioned in this application.
[0130] Furthermore, in some embodiments, this application also provides an electronic device. The electronic device includes at least one memory and at least one processor, with the memory coupled to the processor. The memory stores computer program code / instructions, which, when executed by the processor, enable the electronic device to implement the data processing methods mentioned in this application.
[0131] For example, such as Figure 6 As shown, this application provides a schematic diagram of the structure of an electronic device. Figure 6 As shown, the electronic device 600 includes a processor 601, a communication interface 602, a memory 603, and a communication bus 604.
[0132] The memory 603 stores a computer program that can run on the processor 601. The memory 603 and the processor 601 communicate through the communication interface 602 and the communication bus 604.
[0133] Processor 601 may include general-purpose processors, including central processing units, neural network processors, etc.; it may also be digital signal processing (DSP), application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components.
[0134] The memory 603 may include random access memory (RAM) or non-volatile memory, etc. Optionally, the memory 603 may also be at least one storage device located remotely from the processor 601.
[0135] The communication bus 604 can be a peripheral component interconnect (PCI) bus or an extended industry standard architecture (EISA) bus, etc. This communication bus 604 can be divided into an address bus, a data bus, a control bus, etc.
[0136] It should be noted that the connection method between units / modules in the accompanying drawings provided in this application is only one example. In other embodiments, units / modules can be directly connected via buses, signal lines, or connectors, or they can be indirectly connected.
[0137] Furthermore, some structural or methodological features may be shown in a specific arrangement and / or order in the accompanying drawings provided in this application. However, it should be understood that such a specific arrangement and / or order may not be necessary. For example, in other embodiments, these features may be arranged in a manner and / or order different from that shown in the illustrative drawings. Additionally, including structural or methodological features in a particular figure does not imply that such features are required in all embodiments; for example, in other embodiments, these features may be omitted or may be combined with other features.
[0138] It should be noted that all units / modules mentioned in the device embodiments of this application are logical units / modules. Physically, a logical unit / module can be a physical unit / module, a part of a physical unit / module, or a combination of multiple physical units / modules. The physical implementation of these logical units / modules themselves is not the most important factor; the combination of functions implemented by these logical units / modules is the key to solving the technical problems proposed in this application. Furthermore, to highlight the innovative aspects of this application, the above-described device embodiments of this application have not introduced units / modules that are not closely related to solving the technical problems proposed in this application. This does not mean that the above-described device embodiments do not contain other units / modules.
[0139] The embodiments disclosed in this application can be implemented in hardware, software, firmware, or a combination of these implementation methods. Embodiments of this application can be implemented as computer programs or program code executable on a programmable system, the programmable system including at least one processor, a storage system (including volatile and non-volatile memory and / or storage elements), at least one input device, and at least one output device.
[0140] In some cases, the disclosed embodiments may be implemented in hardware, firmware, software, or any combination thereof. The disclosed embodiments may also be implemented as instructions carried on or stored thereon by one or more transient or non-transitory machine-readable (e.g., computer-readable) storage media, which may be read and executed by one or more processors.
[0141] It should be noted that in the examples and description of this application, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising one" does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes said element.
[0142] Although this application has been illustrated and described with reference to certain preferred embodiments thereof, those skilled in the art will understand that various changes in form and detail may be made thereto without departing from the scope of this application.
Claims
1. A data processing method, characterized in that, include: The data to be retrieved is obtained, and the feature information in the data to be retrieved is extracted. The feature information includes unique identification information and retrieval intent. Based on the aforementioned feature information, one or more similar data are retrieved from the database, and the unique identifier information and retrieval intent of each of the one or more similar data are obtained. Based on the similarity between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the similarity between the retrieval intent of each similar data and the retrieval intent of the data to be retrieved, target data that meets preset conditions is selected from the one or more similar data.
2. The method according to claim 1, characterized in that, The extraction of feature information from the data to be retrieved includes: The unique identifier information is extracted from the data to be retrieved based on regular expression validation, and each character in the unique identifier information exists in the data to be retrieved.
3. The method according to claim 1, characterized in that, The process of retrieving one or more similar data from the database based on the feature information includes: Based on the aforementioned feature information, first similar data was retrieved from the database; wherein... The unique identifier information of the first similar data is the same as the unique identifier information of the data to be retrieved, and / or the keywords in the retrieval intent of the first similar data are the same as the keywords in the retrieval intent of the data to be retrieved.
4. The method according to claim 3, characterized in that, The method further includes: When the number of similar data is less than a first quantity threshold, the search intent of the data to be searched is changed to obtain a variant intent; Based on the aforementioned variant, the intention is to retrieve second similar data from the database.
5. The method according to claim 4, characterized in that, The method further includes: If the total amount of the first and second similar data is less than the first quantity threshold, the third similar data is retrieved from the database based on the full-text semantics of the data to be retrieved.
6. The method according to claim 1, characterized in that, The method of selecting target data that meets preset conditions from one or more similar data based on the similarity between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the similarity between the retrieval intent of each similar data and the retrieval intent of the data to be retrieved, includes: Based on the weights corresponding to the similarity of the identifier information, the similarity of the identifier information corresponding to each similar data, the weights corresponding to the similar intent, the similar intent corresponding to each similar data, the conflict factors between the unique identifier information of each similar data and the unique identifier information of the data to be retrieved, and the weights corresponding to each conflict factor, the comprehensive score corresponding to each similar data is determined. Based on the comprehensive score corresponding to each of the similar data, n target data are selected from the one or more similar data, where n is a positive integer; where... The overall score of each target data in the n target data is greater than the overall score of each similar data in the one or more similar data, excluding the n target data.
7. The method according to claim 6, characterized in that, The first target data among the n target data includes one or more images, and the method further includes: Information is extracted from one or more images in the first target data to obtain image information; If the number of images in the first target data is less than the image number threshold, the extracted image information will be injected into the target result; or, If the number of images in the first target data is greater than or equal to the image number threshold, target images whose image information meets the relevance condition are selected from each image based on the feature information of the data to be retrieved, and the image information of the target images is injected into the target result.
8. The method according to claim 1, characterized in that, The data to be retrieved is a software fault ticket. The unique identification information includes one or more of the following: negative response code, product number, software version number, underlying driver name, and hexadecimal code corresponding to the error message. The retrieval intent includes business actions and / or functional module names.
9. An electronic device, characterized in that, include: At least one memory and at least one processor, the memory being coupled to the processor; the memory being used to store computer program code / instructions; when the computer program code / instructions are executed by the processor, causing the electronic device to implement the data processing method as described in any one of claims 1 to 8.
10. A readable storage medium, characterized in that, The readable storage medium stores instructions that, when executed on an electronic device, cause the electronic device to perform the data processing method as described in any one of claims 1 to 8.
11. A computer program product, characterized in that, include: Computer instructions that, when executed on an electronic device, cause the electronic device to perform the data processing method as described in any one of claims 1 to 8.