Method and related device for recommending similar defects by integrating multiple features

By processing defective data and multiple recall, using technologies such as inverted indexing and keyword indexing, the problem of poor short text recommendations is solved, and efficient recommendation of similar defects is achieved.

CN114116977BActive Publication Date: 2025-08-01GUANGDONG OPPO MOBILE TELECOMMUNICATIONS CORP LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202111406223.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-11-24
Publication Date
2025-08-01
Estimated Expiration
2041-11-24

AI Technical Summary

Technical Problem

The existing recall recommendation methods do not work well for short texts, and the information redundancy problem leads to poor recommendation results.

Method used

By processing defect data, the input data of the multi-channel recall model is constructed, and multiple recalls are performed using inverse index, keyword index and synonym index to narrow the scope of defect comparison and improve the efficiency of recommendations for similar defects.

Benefits of technology

Improve the rate and efficiency of similar defect recommendations, and can quickly respond to similar defect recommendations and provide more accurate similarity results.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN114116977B_ABST
    Figure CN114116977B_ABST
Patent Text Reader

Abstract

An embodiment of the present application discloses a method and related device for recommending similar defects by integrating multiple features. The method includes: after obtaining defect data, performing data processing on the defect titles in the defect data to obtain input data for a multi-channel recall model; inputting the input data into the multi-channel recall model to obtain a recall data set; inputting the defect data and the recall data set into a recommendation model to perform defect similarity comparison, obtaining a similarity result arranged in reverse order; and recommending the similarity result to the user. By adopting the embodiment of the present application, the comparison range of defects can be narrowed, thereby improving the efficiency of similar defect comparison.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This application relates to the field of electronic technologies, and particularly to a method and related device for recommending similar defects by integrating multiple features. Background Art

[0002] Currently, in the information age, while users obtain a vast amount of information, they also face the problem of information redundancy. A recommendation system can quickly help users locate target information, establish a connection between the user's retrieval target and the database data through the processing of recommendation algorithms, and achieve the goal of providing personalized recommendations to users.

[0003] However, existing recall recommendation methods mostly process long texts or numerical type data. The small amount of available effective information and information redundancy in short texts result in poor recommendation effects. Summary of the Invention

[0004] Embodiments of this application provide a method and related device for recommending similar defects by integrating multiple features, which can achieve a quick response to similar defect recommendations, improve the rate of similar defect comparison, and the efficiency of defect recommendation.

[0005] In a first aspect, embodiments of this application provide a method for recommending similar defects by integrating multiple features, characterized in that the method includes:

[0006] Obtain defect data;

[0007] Perform data processing on the defect titles in the defect data to obtain input data for a multi-way recall model;

[0008] Input the input data into the multi-way recall model to obtain a recall data set;

[0009] Input the defect data and the recall data set into a recommendation model to perform defect similarity comparison, and obtain a similarity result in reverse order;

[0010] Recommend the similarity result to the user.

[0011] In a second aspect, embodiments of this application provide a device for recommending similar defects by integrating multiple features, characterized in that the device includes:

[0012] An obtaining unit, configured to obtain defect data;

[0013] A data processing unit, configured to perform data processing on the defect titles in the defect data to obtain input data for a multi-way recall model;

[0014] A recall unit, configured to input the input data into the multi-way recall model to obtain a recall data set;

[0015] A recommendation unit, configured to input the defect data and the recall data set into a recommendation model to perform defect similarity comparison, obtain a similarity result arranged in reverse order, and recommend the similarity result to a user.

[0016] In a third aspect, an embodiment of the present application provides an electronic device, including a processor, a memory, a communication interface, and one or more programs, wherein the one or more programs are stored in the memory and are configured to be executed by the processor, and the programs include instructions for performing the steps in any of the methods in the first aspect of the embodiments of the present application.

[0017] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores a computer program for electronic data exchange, and the computer program enables a computer to execute some or all of the steps described in any of the methods in the first aspect of the embodiments of the present application.

[0018] In a fifth aspect, an embodiment of the present application provides a computer program product, wherein the computer program product includes a non-transitory computer-readable storage medium storing a computer program, and the computer program is operable to enable a computer to execute some or all of the steps described in any of the methods in the first aspect of the embodiments of the present application. The computer program product may be a software installation package.

[0019] It can be seen that the embodiments of the present application disclose a method and related device for recommending similar defects by fusing multiple features. The method includes: after obtaining defect data, performing data processing on the defect title in the defect data to obtain input data for a multi-channel recall model; inputting the input data into the multi-channel recall model to obtain a recall data set; inputting the defect data and the recall data set into a recommendation model to perform defect similarity comparison, obtaining a similarity result arranged in reverse order; and recommending the similarity result to a user. In this way, through the method proposed in the embodiments of the present application, an inverted index can be established for defects, and at the same time, a keyword index, a synonym index, a category index, etc. are applied to perform multi-channel recall on the defect data set, narrowing the comparison range of defects and greatly improving the efficiency of similar defect comparison. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required for use in the embodiments or the description of the prior art. Obviously, the following drawings are only some embodiments of the present application. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0021] Figure 1AIt is the system structure diagram of a method for recommending similar defects by integrating multiple features provided by an embodiment of the present application;

[0022] Figure 1B It is the overall function flowchart of a method for recommending similar defects by integrating multiple features provided by an embodiment of the present application;

[0023] Figure 2 It is the process schematic diagram of a method for recommending similar defects by integrating multiple features provided by an embodiment of the present application;

[0024] Figure 3 It is the process schematic diagram of a defect data preprocessing provided by an embodiment of the present application;

[0025] Figure 4 It is the process schematic diagram of a defect title data processing provided by an embodiment of the present application;

[0026] Figure 5 It is the structural schematic diagram of a device for recommending similar defects by integrating multiple features provided by an embodiment of the present application;

[0027] Figure 6 It is the structural schematic diagram of an electronic device provided by an embodiment of the application. Detailed implementation manners

[0028] In order to enable those skilled in the art to better understand the solution of the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0029] The terms "first", "second", etc. in the specification and claims of the present application and the above-mentioned drawings are used to distinguish different objects, rather than to describe a specific order. In addition, the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products or devices.

[0030] References to "embodiments" in this specification mean that the specific features, structures, or characteristics described in connection with the embodiments can be included in at least one embodiment of the present application. The phrase appears in various places in the specification and does not necessarily refer to the same embodiment each time, nor are they independent or alternative embodiments mutually exclusive of other embodiments. Those skilled in the art will explicitly and implicitly understand that the embodiments described herein can be combined with other embodiments.

[0031] To better understand the solutions of the embodiments of the present application, the electronic devices, related algorithms, concepts, and backgrounds that the embodiments of the present application may involve will be introduced first below.

[0032] 1) The electronic devices involved in the embodiments of the present application may include various handheld devices, vehicle-mounted devices, wearable devices, computing devices, or other processing devices connected to a wireless modem with wireless communication functions, as well as various forms of user equipment (UE), mobile stations (MS), terminal devices, etc. For the convenience of description, the devices mentioned above are collectively referred to as electronic devices.

[0033] 2) Calculating the similarity of short texts is a difficult and hot topic in the field of natural language processing and even machine learning. The implementation process is as follows: For a given set of short text data, based on the study of the short text structure, several features contained in the text are extracted, such as word frequency, word order, semantic features, etc., and then quantified. Furthermore, the similarity between short texts is reflected by data. The more overlapping features there are between texts, the higher the similarity.

[0034] 3) Recall system: In essence, it is an information funnel, responsible for quickly screening out valuable information from a vast amount of information, narrowing the search scope of the sorting algorithm to solve the problem of information overload; at the same time, it is also equivalent to an information fusion device, used to solve problems such as single-channel recall with single features, small amount of information, and poor diversity. Specifically, the data from multiple channels of recall are fused to obtain a refined candidate set.

[0035] To better understand the solutions of the embodiments of the present application, the system structure diagram of the overall solution of the embodiments of the present application will be introduced first below.

[0036] Please refer to Figure 1A , Figure 1A which is the system structure diagram of a similar defect recommendation method that fuses multiple features provided by the embodiments of the present application: mainly including a data preprocessing module 10, a feature extraction module 20, a defect recall module 30, and a defect similarity comparison and sorting module 40.

[0037] Next, for the functions implemented by the above-mentioned each module, in combination with Figure 1BAs shown below, further description will be given. Please refer to Figure 1B , Figure 1B which is the overall functional flowchart of a similar defect recommendation method that integrates multiple features provided by an embodiment of the present application.

[0038] In the embodiment of the present application, the overall process is divided into two stages:

[0039] In the data preparation stage, it is necessary to process the defect data in the database and obtain an inverted index. The specific steps are as Figure 1B shown in the dashed box:

[0040] Input 101: This part is responsible for inputting the obtained defect data into the model;

[0041] Preprocessing 102: This part is used to perform word segmentation, stop word removal, rewriting, etc. on the titles of the above-mentioned defect data to obtain a preprocessed data set;

[0042] Feature extraction 103, which is used to extract features from the above-mentioned preprocessed data set, and the features include words and their positions;

[0043] Word index / recall 104. First, use the segmented phrases to recall defect titles; then obtain the corresponding identification ID of the above-mentioned defect titles in the defect data set; finally, construct an inverted index table using the phrases and their corresponding identification IDs to achieve defect recall.

[0044] In the similar defect recommendation stage, that is, Figure 1B the entire process shown in the figure. After data processing, it is trained by a recommendation model, and the defect data is sorted by similarity 105, and then the sorting result is output 106 as a defect recommendation to the user.

[0045] Next, the present application will be described in detail with reference to specific embodiments.

[0046] Please refer to Figure 2 , Figure 2 which is a schematic flowchart of a similar defect recommendation method that integrates multiple features provided by an embodiment of the present application. As Figure 2 shown, the similar defect recommendation method that integrates multiple features in the present application includes the following operating steps:

[0047] S201. Obtain defect data.

[0048] Specifically, the above-mentioned defect data refers to the product defect reports feedback during the actual use by users. These defect reports are submitted to the background and stored in the background database. The content of the defect report includes but is not limited to: defect title, specific defect description, defect product type, etc. Developers or after-sales service personnel obtain the defect data from the database and give corresponding solutions to the problems involved in the defect report.

[0049] Exemplarily, in actual applications, different types of problems will occur during the use of different products, resulting in different types of defect reports; during the use of the same product, different defect reports will be generated due to different defect problems, or the same or similar problems will be encountered when the same product is used by different people, resulting in the same or similar content of the defect reports obtained for the same or similar problems. In short, the redundancy of the defect data collected in the database is relatively high, which is not conducive to the efficiency of defect problem solving. Therefore, relevant processing needs to be carried out on the defect data.

[0050] S202. Perform data processing on the defect titles in the defect data to obtain input data for the multi-way recall model.

[0051] Specifically, "multi-way recall" means using different strategies, features or simple models to recall a part of the candidate set respectively, and then mixing the candidate sets together for subsequent sorting models to use. The above multi-way recall model is to use the processed defect data to recall similar defects for different strategies and features, and then as the subsequent training recommendation model to complete the sorting of similar defect data, and finally achieve the purpose of defect recommendation.

[0052] Specifically, as described in S201, the initial defect data in the database has a relatively high redundancy. If the original defect data is directly input into the model for recall processing without being processed, it will slow down the model training process and result in poor effects. Therefore, before inputting the data into the recall model, it is necessary to first perform data processing on the original defect data to obtain data that can be used for model training.

[0053] In a possible example, before performing data processing on the defect titles in the defect data, the method further includes the following steps: constructing a synonym dictionary using the defect data; updating the synonym dictionary according to the new word discovery mechanism to obtain a target synonym dictionary.

[0054] Specifically, in actual application scenarios, the defect reports feedback by users often describe defects according to their living habits. Therefore, there will inevitably be many colloquial words. Therefore, it is necessary to construct a thesaurus of near-synonyms for these colloquial words, so as to replace the colloquial words in the original defect report with more professional near-synonyms, in order to obtain a defect report with relatively professional descriptions.

[0055] Furthermore, the database collects user defect reports in an incremental manner, and new defect reports will continue to be added. Therefore, the above-mentioned thesaurus of near-synonyms also needs to be updated accordingly. Therefore, by introducing a new word discovery mechanism to update the thesaurus of near-synonyms, a target thesaurus of near-synonyms is obtained.

[0056] Specifically, it can be independent of any existing thesaurus. Only based on the common characteristics of words, extract the text fragments that may form words in the defect reports newly added to the database, whether they are new words or old words. Then, compare all the extracted words with the existing thesaurus of near-synonyms. If the current word fails to be retrieved in the existing thesaurus of near-synonyms, then add the word to the existing thesaurus of near-synonyms to achieve the purpose of updating the thesaurus of near-synonyms and obtain the target thesaurus of near-synonyms.

[0057] It can be seen that in the embodiments of the present application, in order to obtain a defect report with a standardized expression, a thesaurus of near-synonyms is introduced, and the update of the thesaurus is realized through a new word discovery mechanism. Using the obtained target thesaurus to optimize the current defect data can better provide high-quality data for the subsequent training of the model, thereby improving the speed and accuracy of model training.

[0058] In a possible example, please refer to Figure 3 as shown in Figure 3 is a schematic flowchart of a defect data preprocessing provided by the embodiments of the present application. Before performing data processing on the defect title in the defect data, the method further includes the following steps:

[0059] S301, classify the defect data according to a preset business type to obtain a classification result.

[0060] Specifically, as described in S201, in actual application scenarios, there will be different types of defect reports due to different product types, and there will also be differences in the types of defect reports due to different types of defects in the same product. Therefore, before processing the defect data, it is necessary to first classify the defect reports. The classification result includes the target business type and the target defect data corresponding to the target business type.

[0061] Exemplarily, different product types such as cameras, mobile phones, computers, etc. will have different types of defect reports during use. Therefore, the defect dataset can be classified according to different product types, and several defect data correspond to each product category.

[0062] S302. Determine the target defect category corresponding to the target business type according to the mapping relationship between the preset business type and the defect category.

[0063] Exemplarily, after classifying according to product types, different types of products will have different defect types. For example, the defects that a camera may have include: poor focusing effect, dazzling automatic backlight brightness, etc.; the defects that a mobile phone may have include: charging failure, black screen, etc.; the defects that a computer may have include: screen flickering, unable to boot, etc. Therefore, under different business types, it is necessary to classify the defects to obtain the mapping relationship between the business type and the defect category.

[0064] Exemplarily, under the mobile phone business type, the defect reports collected from user feedback include but are not limited to: "the mobile phone consumes power too quickly", "the mobile phone cannot be charged", "the mobile phone has a black screen", etc. Among them, mobile phone power consumption and mobile phone charging both belong to the defects of the "battery" category. At this time, there is a mapping relationship between the business type mobile phone and the defect category battery. And "the mobile phone has a black screen" belongs to the defects of the screen category. There are many such mapping relationships between business types and defect types, which are specifically identified and classified according to the actual application scenario and will not be listed one by one here.

[0065] Furthermore, according to the mapping relationship defined above, determine the target defect category corresponding to the target business type.

[0066] S303. Determine the keywords corresponding to the target defect category from the target defect data.

[0067] Specifically, there must be one or more words in the target defect data that can accurately characterize the defect, which we call keywords. For example, the defect title is: "dazzling automatic backlight brightness". For example, "backlight" reflects what the subject of the problem is, and "dazzling" reflects the phenomenon. These two words are relatively crucial for the definition of the defect. It should be noted that in the embodiments of the present application, the selection of keywords is carried out by means of manual proofreading. Manual identification of keywords can be more accurate, and the keywords determined for describing different defects need to be distinguishable. The effect of manual selection may be more accurate than that of machine or model selection.

[0068] S304. Determine the target weight corresponding to the keyword according to the mapping relationship between the preset keyword and the weight.

[0069] Exemplarily, in step S303, two or more manually determined keywords may appear in the same defect title, and the influence of the keywords on the defect title is also different. For example, in the sentence "The automatic backlight brightness is dazzling", the word "backlight" appears frequently because it represents the main body of a problem, but there are many problems caused by "backlight". In this defect title, the actual defect to be solved is "dazzling". Therefore, "dazzling" has a greater impact on the defect definition than "backlight". In this case, a higher weight can be assigned to "dazzling". In this way, the target weight corresponding to the keyword can be determined.

[0070] S305. Construct a keyword dictionary according to the keyword and the target weight.

[0071] It can be seen that in the embodiment of the present application, by classifying the obtained defect data, defining keywords and their weights, a keyword dictionary is obtained. The obtained keyword dictionary can be used in the subsequent recall training process to more quickly and accurately recall defect data containing keywords through the keywords and their weights, improving the recall efficiency and accuracy.

[0072] S203. Input the input data into the multi-channel recall model to obtain a recall data set.

[0073] Exemplarily, inputting the preprocessed defect data into the multi-channel recall model can quickly recall defect data related to the target defect, which is more conducive to developers or after-sales service personnel to solve defects.

[0074] In a possible example, the data processing of the defect title in the defect data to obtain the input data for the multi-channel recall model may include the following steps: performing data preprocessing on the defect title to obtain a defect data set; performing feature extraction on the defect data set to obtain a feature data set; constructing the input data according to the defect data set and the feature data set. The specific process is as Figure 4 shown Figure 4 FIG. is a schematic flowchart of defect title data processing provided by an embodiment of the present application, which specifically includes:

[0075] S401. Perform data preprocessing on the defect title to obtain a defect data set.

[0076] In a possible example, the data preprocessing of the defect title includes the following steps: performing word segmentation on the defect title to obtain a first data set; removing stop words in the first data set to obtain a second data set; replacing replaceable words in the second data set according to the target synonym dictionary to obtain the defect title that meets preset rules; constructing the defect data set based on the defect titles that meet the preset rules, the defect data set including each defect title that meets the preset rules and its corresponding identification ID.

[0077] Specifically, it should be noted that, for example, in recent years, with the increasing maturity of natural language processing technology, there are more and more open source word segmentation tools, such as Ansj, Pangu word segmentation, etc., which are not limited here. Among the methods of Chinese text processing, the more commonly used method is Jieba word segmentation. In this application, Jieba word segmentation is taken as an example, and the title of the defect data is mainly pre-processed. The characteristic of the text title is that it is relatively short. For example, the defect title is: "Is the automatic backlight brightness glaring?", then a possible effect after word segmentation processing will be: "Automatic / backlight / brightness / glaring / ? / ?" At this time, the description of the defect still contains words or punctuation marks such as "? / ?" that are irrelevant to the definition of the defect, so it is necessary to further process the data after word segmentation.

[0078] Furthermore, the above-mentioned first data set is subjected to a stop word removal operation. The specific steps are: using the existing stop word list, the unimportant components of the first data set that has been segmented are removed. The unimportant components include but are not limited to: modal particles, punctuation marks, etc. The stop word removal list contains a large number of non-essential dictionaries that have been repeatedly tested through experiments. As mentioned above, after the word segmentation of "automatic / backlight / brightness / glaring / ? / ?" is processed for stop words, the result is "automatic / backlight / brightness / glaring", and a similar data result set is obtained as the second data set. However, it can be seen that the word "glaring" in the title is a colloquial word commonly used by people in daily habits. Such similar colloquial words are not conducive to the model's recognition of defective words, so the second data set needs to be further processed.

[0079] Furthermore, the second data set is rewritten. The specific steps are: using the acquired target synonym dictionary to replace synonyms in the second data set. The defect title after replacement is more standardized in its description of the defect. For example, the "automatic / backlight / brightness / glaring" mentioned above may be processed as "automatic / backlight / brightness / glaring". After rewriting the second data set, a defect data set containing relatively standardized and non-redundant vocabulary is obtained. The defect data set is saved in the form of a pair of title data and identification ID. Specifically, each defect title will have a unique identification ID corresponding to it.

[0080] It can be seen that in the embodiments of the present application, by performing word segmentation, stop word removal, and rewriting operations on the defect titles of the originally obtained defect data, the information redundancy of the obtained defect data set is relatively low. The lower the redundancy of the defect data set, the more available key information there is, which is more beneficial for the subsequent model training results obtained by using it as the model input for training.

[0081] S402. Extract features from the defect data set to obtain a feature data set.

[0082] In a possible example, after constructing the defect data set according to the defect titles that meet the preset rules, the method further includes the following steps: determining the standardized defect title data corresponding to each phrase in the first data set; determining the identification ID corresponding to each word according to the defect title data; constructing an inverted index table according to each word and its corresponding identification ID, and the inverted index table is used to quickly recall the defects corresponding to the retrieval words during multi-way recall. Here, the role of recall is to narrow the range of defects for the next similarity comparison, improve the speed, that is, to screen out relevant defects (such as those containing the same words) from a large number of defects, and then perform a one-by-one similarity comparison with the input defects, and recommend them according to the similarity ranking.

[0083] Specifically, after performing word segmentation processing on the defect title, the format of each title data in the first data set is like the example defect title "Automatic / Backlight / Brightness / Dazzling / ?". At the same time, there are many defect titles in the first data set that can contain phrases such as "Automatic" and "Backlight" in the example. To recall all defect titles containing the above example phrases faster, the embodiments of the present application adopt: first, use the segmented phrases to recall defect titles; then obtain the identification ID corresponding to the defect title in the defect data set; finally, construct an inverted index table using the phrase and its corresponding identification ID. A specific example is shown in Table 1 below, which is a partial display content of the defect data set, and Table 2, which is an example table of constructing an inverted index using phrases:

[0084] Table 1

[0085] ID Defect Title 1 2 Backlight is too dazzling 3 Backlight cannot be automatically adjusted

[0086] Table 2

[0087] Screen is too dazzling Index IDset 1,2 Backlight 1,3 Dazzling 2 Adjustment 3

[0088] Specifically, for the partial content of the defect data set shown in Table 1, each defect title has a unique corresponding identification ID. And a phrase is included in at least one defect title, so a phrase corresponds to at least one defect identification ID.

[0089] It can be seen that in the embodiment of the present application, by constructing an inverted index table for the phrases in the defect title after word segmentation, the defect title ID containing the phrase can be quickly located, and then the defect title corresponding to the ID can be accurately located. According to the phrase recall, the defect titles containing the phrase can be quickly recalled, which is more conducive to performing the next similarity calculation on the recalled titles, thereby improving the efficiency and accuracy of similar defect recommendation.

[0090] In a possible example, the feature extraction of the defect data set may further include the following steps: retrieving the target defect data containing the keyword from the defect data set by using the keyword dictionary; obtaining the position of the keyword in the target defect data; and constructing the feature data set according to the keyword and the position.

[0091] Specifically, in the embodiment of the present application, a word pair topic model can be selected to perform feature extraction of short texts. Since the defect data obtained in the solution of the present application are all sparse data, and the word pair topic model can well solve the problem of data sparsity. For defect titles of the same type, the defect description subjects they target are all for this type; then, there will be corresponding keywords in each defect title. The target defect data containing the keyword can be retrieved by using these keywords, and this process is actually a recall process.

[0092] Furthermore, a word pair model is used to determine the position of each short text in the defect title, that is, the corresponding semantic space, so as to further analyze and judge the defect category according to the keyword and its corresponding semantic space. Since in the above-mentioned keyword dictionary, corresponding weights are assigned to different keywords, according to the keyword as a feature, the target defect data can be more accurately located. The keyword and its corresponding semantic space are called a feature, and this feature can reflect the closeness between the keyword and the defect title to a certain extent. Then, the feature data set is constructed according to the keyword and the position. It should be noted here that the selection of features is not limited to keywords only, and word frequency or other ways that can characterize defect features can also be used as a feature, which is not limited here.

[0093] It can be seen that in the embodiment of the present application, for the selection of features including keywords, semantic space, word frequency, etc. to construct the feature data set, and further using the feature data set as a way in the training process of the multi-way recall model can improve the recall rate of the recall model, and further improve the efficiency of defect recommendation.

[0094] S403, construct the input data according to the defect data set and the feature data set.

[0095] Specifically, the input data used in the multi-channel recall model in the embodiments of the present application includes the defect data set constructed by the above method and the feature data set. At the same time, various methods such as keyword dictionaries, synonym dictionaries, and categories are applied to help the index achieve multi-channel recall, narrow the scope of defect comparison, and thus improve the efficiency of similar defect comparison.

[0096] It can be seen that the embodiments of the present application disclose a method and related device for recommending similar defects by fusing multiple features. The method includes: after obtaining defect data, performing data processing on the defect titles in the defect data to obtain input data for the multi-channel recall model; inputting the input data into the multi-channel recall model to obtain a recall data set; inputting the defect data and the recall data set into the recommendation model to perform defect similarity comparison to obtain a similarity result arranged in reverse order; and recommending the similarity result to the user. In this way, through the method proposed in the embodiments of the present application, an inverted index can be established for defects, and at the same time, methods such as keyword indexing, synonym indexing, and category indexing are applied to perform multi-channel recall on the defect data set, narrow the comparison range of defects, and greatly improve the efficiency of similar defect comparison.

[0097] S204. Input the defect data and the recall data set into the recommendation model to perform defect similarity comparison to obtain a similarity result arranged in reverse order.

[0098] Specifically, the defect data to be solved is used as the target defect data, and the recall data set recalled through the above steps is used as input data and input into the recommendation model. By calculating the similarity between the target defect data and the defect titles in the recall data set, a similarity result arranged in reverse order is obtained.

[0099] S205. Recommend the similarity result to the user.

[0100] Specifically, the result arranged in reverse order obtained by the S204 model is recommended to the user. The user can be either a user who actually uses the product and encounters product defects, or a developer or after-sales personnel who needs to find a solution to similar defects for the target defect and thus refer to the solution to repair the target defect. No specific limitation is made here.

[0101] It can be seen that the embodiments of the present application disclose a method for recommending similar defects by fusing multiple features. The method includes: after obtaining defect data, performing data processing on the defect titles in the defect data to obtain input data for a multi-channel recall model; inputting the input data into the multi-channel recall model to obtain a recall data set; inputting the defect data and the recall data set into a recommendation model to perform defect similarity comparison, obtaining a similarity result arranged in reverse order; and recommending the similarity result to the user. In this way, through the method proposed in the embodiments of the present application, by establishing an inverted index for defects and applying methods such as keyword index, synonym index, and category index to perform multi-channel recall on the defect data set, the comparison range of defects is reduced, and the efficiency of similar defect comparison is greatly improved.

[0102] The above mainly introduces the solution of the embodiments of the present application from the perspective of the execution process on the method side. It can be understood that in order for an electronic device to implement the above functions, it includes corresponding hardware structures and / or software modules for executing each function. Those skilled in the art should easily realize that, in combination with the units and algorithm steps of each example described in the embodiments provided herein, the present application can be implemented in the form of hardware or a combination of hardware and computer software. Whether a certain function is executed in the form of hardware or computer software driving the hardware depends on the specific application and design constraints of the technical solution. Professional technicians can use different methods to implement the described functions for each specific application, but such implementation should not be considered to exceed the scope of the present application.

[0103] The embodiments of the present application can perform functional unit division on the electronic device according to the above method examples. For example, each functional unit can be divided corresponding to each function, or two or more functions can be integrated into one processing unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. It should be noted that the division of units in the embodiments of the present application is illustrative, only a logical functional division, and there can be other division methods in actual implementation.

[0104] Consistent with Screen the embodiments of Figure 2 , Figure 5 FIG. Figure 5 shows a structural schematic diagram of a device for recommending similar defects by fusing multiple features provided by the embodiments of the present application. As

[0105] shown, the above-mentioned device 500 for recommending similar defects by fusing multiple features includes:

[0106] An obtaining unit 501, configured to obtain defect data;

[0107] The recall unit 503 is configured to input the input data into the multi-channel recall model to obtain a recall data set;

[0108] The recommendation unit 504 is configured to input the defect data and the recall data set into the recommendation model to perform defect similarity comparison, obtain a similarity result arranged in reverse order, and recommend the similarity result to the user.

[0109] It can be seen that for a similar defect recommendation device integrating multiple features described in the embodiments of the present application, after obtaining defect data, data processing is performed on the defect titles in the defect data to obtain input data for the multi-channel recall model; the input data is input into the multi-channel recall model to obtain a recall data set; the defect data and the recall data set are input into the recommendation model to perform defect similarity comparison, obtain a similarity result arranged in reverse order; and the similarity result is recommended to the user. In this way, through the method proposed in the embodiments of the present application, an inverted index can be established for defects, and at the same time, methods such as keyword indexing, synonym indexing, and category indexing are applied to perform multi-channel recall on the defect data set, narrowing the comparison scope of defects and greatly improving the efficiency of similar defect comparison.

[0110] In a possible example, before performing data processing on the defect titles in the defect data, the data processing unit 502 is specifically configured to:

[0111] Classify the defect data according to a preset service type to obtain a classification result, where the classification result includes a target service type and target defect data corresponding to the target service type;

[0112] According to the mapping relationship between the preset service type and the defect category, determine the target defect category corresponding to the target service type; determine keywords corresponding to the target defect category from the target defect data;

[0113] According to the mapping relationship between the preset keyword and the weight, determine the target weight corresponding to the keyword;

[0114] Construct a keyword dictionary according to the keyword and the target weight.

[0115] In a possible example, when performing data processing on the defect titles in the defect data to obtain input data for the multi-channel recall model, the data processing unit 502 is specifically configured to:

[0116] Perform data preprocessing on the defect title to obtain a defect data set;

[0117] Perform feature extraction on the defect data set to obtain a feature data set;

[0118] Construct the input data according to the defect data set and the feature data set.

[0119] In a possible example, for the data preprocessing of the defect title, the data processing unit 402 is specifically configured to:

[0120] Perform word segmentation on the defect title to obtain a first data set;

[0121] Remove the stop words in the first data set to obtain a second data set;

[0122] According to the target thesaurus, replace the replaceable words in the second data set to obtain the defect title that meets the preset rules;

[0123] Construct the defect data set according to the defect title that meets the preset rules, where the defect data set includes each defect title that meets the preset rules and its corresponding identification ID.

[0124] Consistent with the above Figure 5 shown embodiment, please refer to Figure 2 , Figure 6 is a schematic structural diagram of an electronic device provided by an embodiment of the present application, as Figure 6 Figure 6 shown:

[0125] An electronic device includes:

[0126] One or more processors; one or more memories for storing programs, and one or more memories and programs are configured such that one or more processors control the device to execute some or all of the steps described in any method of the first aspect of the embodiments of the present application.

[0127] Among them, the memory can be a volatile memory such as a dynamic random access memory DRAM, or a non-volatile memory such as a mechanical hard disk. The above memory is used to store a set of executable program codes, and the above processor is used to call the executable program codes stored in the memory and can execute instructions for the following steps:

[0128] Obtain defect data;

[0129] Perform data processing on the defect title in the defect data to obtain input data for the multi-way recall model;

[0130] Input the input data into the multi-way recall model to obtain a recall data set;

[0131] Input the defect data and the recall data set into the recommendation model to perform defect similarity comparison, and obtain a similarity result in reverse order;

[0132] Recommend the similarity result to the user.

[0133] It can be seen that for the electronic device described in the embodiments of the present application, after obtaining defect data, it can perform data processing on the defect titles in the defect data to obtain input data for the multi-channel recall model; input the input data into the multi-channel recall model to obtain a recall data set; input the defect data and the recall data set into the recommendation model to perform defect similarity comparison to obtain a similarity result arranged in reverse order; and recommend the similarity result to the user. In this way, through the method proposed in the embodiments of the present application, by establishing an inverted index for defects and applying methods such as keyword indexing, synonym indexing, and category indexing to the defect data set for multi-channel recall, the comparison range of defects can be narrowed, and the efficiency of similar defect comparison can be greatly improved.

[0134] In a possible example, before performing data processing on the defect titles in the defect data, the above program includes instructions for performing the following steps:

[0135] Construct a synonym dictionary using the defect data;

[0136] Update the synonym dictionary according to the new word discovery mechanism to obtain a target synonym dictionary.

[0137] In a possible example, before performing data processing on the defect titles in the defect data, the above program includes instructions for performing the following steps:

[0138] Classify the defect data according to a preset business type to obtain a classification result, where the classification result includes a target business type and the target defect data corresponding to the target business type;

[0139] Determine the target defect category corresponding to the target business type according to the mapping relationship between the preset business type and the defect category;

[0140] Determine the keywords corresponding to the target defect category from the target defect data;

[0141] Determine the target weight corresponding to the keyword according to the mapping relationship between the preset keyword and the weight;

[0142] Construct a keyword dictionary according to the keyword and the target weight.

[0143] In a possible example, for performing data processing on the defect titles in the defect data to obtain input data for the multi-channel recall model, the above program includes instructions for performing the following steps:

[0144] Perform data preprocessing on the defect title to obtain a defect data set;

[0145] Extract features from the defect data set to obtain a feature data set;

[0146] Construct the input data according to the defect data set and the feature data set.

[0147] In a possible example, after constructing the defect data set according to the defect title that meets the preset rules, the above program includes instructions for performing the following steps:

[0148] Determine the corresponding standard defect title data for each phrase in the first data set;

[0149] Determine the corresponding identification ID for each word according to the defect title data;

[0150] Construct an inverted index table according to each word and its corresponding identification ID.

[0151] The inverted index table is used to quickly recall defects corresponding to the retrieved words during multi-way recall. Here, the role of recall is to narrow the range of defects for the next similarity comparison, improve the speed, that is, to screen out relevant defects (such as those containing the same words) from a large number of defects, and then compare the similarity with the input defects one by one and recommend them according to the similarity ranking.

[0152] In a possible example, for the feature extraction of the defect data set, the above program includes instructions for performing the following steps:

[0153] Use the keyword dictionary to retrieve the target defect data containing the keyword from the defect data set;

[0154] Obtain the position of the keyword in the target defect data;

[0155] Construct the feature data set according to the keyword and the position.

[0156] Some or all of the steps of any method for recommending similar defects by fusing multiple features as described in the method embodiments of recommending similar defects by fusing multiple features.

[0157] The embodiments of the present application provide a computer-readable storage medium. The computer-readable storage medium stores a computer program for electronic data exchange. The computer program includes execution instructions, and the execution instructions are used to execute some or all of the steps of any method for recommending similar defects by fusing multiple features as described in the method embodiments of recommending similar defects by fusing multiple features. The above computer includes an electronic terminal device.

[0158] An embodiment of the present application provides a computer program product. The computer program product includes a computer program, and the computer program is operable to cause a computer to perform some or all of the steps of any one of the methods for recommending similar defects by integrating multiple features as described in the foregoing method embodiments. The computer program product may be a software installation package.

[0159] It should be noted that, for any of the foregoing embodiments of the method for recommending similar defects by integrating multiple features, for the sake of simple description, they are all expressed as a series of action combinations. However, those skilled in the art should know that the present application is not limited by the described action sequence, because according to the present application, some steps may be performed in other sequences or simultaneously. Secondly, those skilled in the art should also know that the embodiments described in the specification are all preferred embodiments, and the actions involved are not necessarily essential to the present application.

[0160] The above has introduced the embodiments of the present application in detail. Specific examples are used in this article to elaborate on the principles and implementation manners of a method for recommending similar defects by integrating multiple features and related devices of the present application. The description of the above embodiments is only used to help understand the method and its core idea of the present application; at the same time, for those of ordinary skill in the art, according to the idea of a method for recommending similar defects by integrating multiple features and related devices of the present application, there will be changes in the specific implementation manners and application scopes. In summary, the content of this specification should not be construed as a limitation to the present application.

[0161] The present application is described with reference to the flowcharts and / or block diagrams of the methods, hardware products, and computer program products of the embodiments of the present application. It should be understood that each process and / or block in the flowchart and / or block diagram, and the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to the processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing devices to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing devices generate a device for implementing the functions specified in one or more processes in the flowchart and / or one or more blocks in the block diagram.

[0162] These computer program instructions can also be stored in a computer-readable memory capable of guiding a computer or other programmable data processing device to work in a specific manner, so that the instructions stored in the computer-readable memory produce a manufactured article including an instruction device, and the instruction device implements the functions specified in one or more processes of the flowchart and / or one or more blocks of the block diagram. The memory may include: a flash drive, a read-only memory (abbreviated as ROM), a random access memory (abbreviated as RAM), a magnetic disk or an optical disk, etc.

[0163] Although the present application has been described in conjunction with various embodiments herein, however, in the process of implementing the claimed present application, those skilled in the art can understand and implement other variations of the disclosed embodiments by viewing the drawings, the disclosure, and the appended claims. In the claims, the word "comprising" does not exclude other components or steps, and "a" or "an" does not exclude a plurality. Certain measures are recited in mutually different dependent claims, but this does not mean that these measures cannot be combined to produce good results.

[0164] Those of ordinary skill in the art can understand that all or part of the steps in the various methods of the above-described method embodiments for recommending similar defects by integrating multiple features can be completed by a program instructing related hardware, and the program can be stored in a computer-readable memory. The memory may include: a flash drive, a read-only memory (abbreviated as ROM), a random access memory (abbreviated as RAM), a magnetic disk or an optical disk, etc.

[0165] It can be understood that any product controlled or configured to execute the processing method of the flowchart described in the method embodiment for recommending similar defects by integrating multiple features of the present application, such as the device in the above flowchart and the computer program product, all belong to the scope of the related products described in the present application.

[0166] Obviously, those skilled in the art can make various changes and modifications to the method and device for recommending similar defects by integrating multiple features provided by the present application without departing from the spirit and scope of the present application. Thus, if these modifications and variations of the present application fall within the scope of the claims of the present application and their equivalent technologies, the present application also intends to include these changes and modifications.

Claims

1. A method for recommending similar defects by integrating multiple features, characterized in that The method includes: Obtaining defect data; Classifying the defect data according to a preset business type to obtain a classification result, where the classification result includes a target business type and target defect data corresponding to the target business type; Determining a target defect category corresponding to the target business type according to a mapping relationship between the preset business type and the defect category, where the mapping relationship between the business type and the defect category is identified and classified according to the actual application scenario; Determining keywords corresponding to the target defect category from the target defect data, where the keywords include one or more words for characterizing the defect, and the keywords reflect the subject and phenomenon of the problem; Determining a target weight corresponding to the keyword according to a mapping relationship between the preset keyword and the weight; Constructing a keyword dictionary according to the keyword and the target weight, where the keyword dictionary is used for recall training, and the keyword and the target weight are used for quickly and accurately recalling defect data including the keyword; Performing data processing on the defect titles in the defect data to obtain input data for a multi-channel recall model; Inputting the input data into the multi-channel recall model to obtain a recall data set; Inputting the defect data and the recall data set into a recommendation model to perform defect similarity comparison, and obtaining a similarity result arranged in reverse order; Recommending the similarity result to the user; Among them, the performing data processing on the defect titles in the defect data to obtain input data for a multi-channel recall model includes: performing data preprocessing on the defect titles to obtain a defect data set; using the keyword dictionary to retrieve the target defect data including the keyword from the defect data set; obtaining the position of the keyword in the target defect data; constructing a feature data set according to the keyword and the position; and constructing the input data according to the defect data set and the feature data set.

2. The method according to claim 1, characterized in that, Before the performing data processing on the defect titles in the defect data, the method further includes: Constructing a synonym dictionary using the defect data; Updating the synonym dictionary according to a new word discovery mechanism to obtain a target synonym dictionary.

3. The method according to claim 2, wherein The performing data preprocessing on the defect titles includes: Performing word segmentation processing on the defect titles to obtain a first data set; Removing stop words in the first data set to obtain a second data set; Replacing replaceable words in the second data set according to the target synonym dictionary to obtain the defect titles that meet the preset rules; Constructing the defect data set according to the defect titles that meet the preset rules, where the defect data set includes each defect title that meets the preset rules and its corresponding identification ID.

4. The method according to claim 3, wherein After the constructing the defect data set according to the defect titles that meet the preset rules, the method further includes: Determining a standardized defect title corresponding to each phrase in the first data set; Determining the identification ID corresponding to each phrase according to the defect title; Constructing an inverted index table according to each phrase and its corresponding identification ID.

5. A similar defect recommendation device integrating multiple features, characterized in that The apparatus includes: An obtaining unit, configured to obtain defect data; A data processing unit, configured to classify the defect data according to a preset service type to obtain a classification result, where the classification result includes a target service type and target defect data corresponding to the target service type; determine a target defect category corresponding to the target service type according to a mapping relationship between the preset service type and the defect category, where the mapping relationship between the service type and the defect category is identified and classified according to the actual application scenario; determine keywords corresponding to the target defect category from the target defect data, where the keywords include one or more words for characterizing the defect, and the keywords reflect the subject and phenomenon of the problem; determine a target weight corresponding to the keywords according to a mapping relationship between the preset keywords and the weights; construct a keyword dictionary according to the keywords and the target weights, where the keyword dictionary is used for recall training, and the keywords and the target weights are used for quickly and accurately recalling defect data including the keywords; perform data processing on the defect titles in the defect data to obtain input data for a multi-way recall model; where, the performing data processing on the defect titles in the defect data to obtain input data for a multi-way recall model includes: performing data preprocessing on the defect titles to obtain a defect data set; retrieving the target defect data including the keywords from the defect data set by using the keyword dictionary; obtaining the positions of the keywords in the target defect data; constructing a feature data set according to the keywords and the positions; constructing the input data according to the defect data set and the feature data set; A recall unit, configured to input the input data into the multi-way recall model to obtain a recall data set; A recommendation unit, configured to input the defect data and the recall data set into a recommendation model to perform defect similarity comparison, obtain a similarity result in reverse order, and recommend the similarity result to a user.

6. An electronic device, characterized in that, It includes a processor, a memory, a communication interface, and one or more programs, where the one or more programs are stored in the memory and are configured to be executed by the processor, and the programs include instructions for performing the steps in the method according to any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, Store a computer program for electronic data exchange, where the computer program causes a computer to execute the method according to any one of claims 1-4.

Citation Information

Patent Citations

  • Automobile product defect determination method and device based on case reasoning

    CN113326349A

  • Multi-path recall recommendation method for crowdsourcing software testing

    CN113360632A