Data identification method, device, electronic device and storage medium

Abstract

The embodiments of the present application disclose a data identification method, device, electronic device and storage medium; including: obtaining multiple classifier sets, each classifier set includes at least two classifiers; for multiple historical time periods, obtaining the number of hits of the text data corresponding to each historical time period to each classifier set; if the hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold, then the first classifier set is determined to be an abnormal classifier set; and at least one abnormal classifier set is displayed. The embodiments of the present application can screen and display abnormal classifier sets, and the staff can quickly summarize the corresponding abnormal data after viewing, which can reduce the consumption time and reduce the consumption of human resources.

Description

Technical Field

[0001] The present application relates to the field of computers, and in particular to a data identification method, device, electronic device and storage medium. Background Art

[0002] In the prior art, with the rapid development of informatization, a large amount of text information is generated every day in all walks of life. For example, the convenience service industry often receives a large amount of feedback on people's livelihood issues; the medical service industry often receives a large amount of patient medical records, etc. In industries related to public services, such as the convenience service industry and the medical service industry, it is often necessary to analyze the text information generated in the industry in order to achieve early warning of emergencies.

[0003] However, in the prior art, text information is usually analyzed and processed by manual statistics and expert experience analysis, which is time-consuming and consumes a lot of human resources. Summary of the invention

[0004] The embodiments of the present application provide a data identification method, device, electronic device and storage medium, which can improve the problem that text information analysis in the prior art is time-consuming and labor-intensive.

[0005] The present application provides a data identification method, the method comprising:

[0006] Acquire multiple classifier sets, wherein each of the classifier sets includes at least two classifiers;

[0007] For multiple historical time periods, obtaining the number of hits of the text data corresponding to each historical time period to each of the classifier sets;

[0008] If the hit difference between the hit count of the first set of classifiers for the text data corresponding to the first historical time period and the hit count of the first set of classifiers for the text data corresponding to the second historical time period exceeds a preset threshold, the first set of classifiers is determined to be an abnormal set of classifiers; wherein the first historical time period is any one of the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first set of classifiers is any one of the multiple sets of classifiers;

[0009] Display at least one set of abnormal category words.

[0010] The present application also provides a data identification device, the device comprising:

[0011] A set acquisition unit, used to acquire a plurality of classifier sets, wherein each of the classifier sets includes at least two classifiers;

[0012] A hit count acquisition unit, used for acquiring, for multiple historical time periods, the hit count of the text data corresponding to each historical time period for each of the classifier sets;

[0013] an abnormality determination unit, configured to determine that the first classifier set is an abnormal classifier set if a hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold; wherein the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the multiple classifier sets;

[0014] The set display unit is used to display at least one abnormal category word set.

[0015] In some embodiments, the collection acquisition unit includes:

[0016] A classifier acquisition subunit is used to acquire multiple classifiers;

[0017] The set determination subunit is used to determine the plurality of classifier sets based on the plurality of classifiers.

[0018] In some embodiments, the set determining subunit includes:

[0019] The to-be-screened category sub-subunit is used to determine a to-be-screened category word combination consisting of any two category words among a plurality of category words;

[0020] The primary screening combination sub-subunit is used to screen multiple classifier combinations to be screened to obtain at least one primary screening classifier combination;

[0021] The classifier adding sub-subunit is used to add any remaining classifier to each of the preliminary screening classifier combinations with remaining classifiers to obtain a new classifier combination to be screened, wherein the remaining classifier is a classifier among the multiple classifiers except the classifiers included in the preliminary screening classifier combination;

[0022] The step jump sub-subunit is used to jump to the step: screen multiple classifier word combinations to be screened to obtain at least one preliminary screening classifier word combination until there are no remaining classifier words; wherein, when there are no remaining classifier words, the corresponding multiple preliminary screening classifier word combinations are the multiple classifier word sets.

[0023] In some embodiments, the primary screening combination sub-subunit is specifically used for:

[0024] Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed;

[0025] For each of the to-be-screened category word combinations, calculating the number of hits of the to-be-screened category word combination with respect to the plurality of to-be-processed category word groups;

[0026] For the plurality of to-be-screened category word combinations, the to-be-screened category word combinations with hit times higher than a first threshold are eliminated, and the to-be-screened category word combinations with hit times lower than a second threshold are eliminated, to obtain at least one preliminary screening category word combination.

[0027] In some embodiments, the primary screening combination sub-subunit is specifically used for:

[0028] Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed;

[0029] For each of the to-be-screened classifier word combinations, calculating the confidence of the to-be-screened classifier word combination with respect to the plurality of to-be-processed classifier word groups;

[0030] Sorting the plurality of to-be-screened classifier word combinations in descending order of confidence to obtain a sorted sequence of to-be-screened classifier word combinations;

[0031] The first n to-be-screened classifier word combinations of the to-be-screened classifier word combination sequence are intercepted to obtain at least one preliminary screening classifier word combination; wherein n is a positive integer.

[0032] In some embodiments, the hit count acquisition unit includes:

[0033] A keyword extraction subunit, configured to extract a plurality of keywords from each of the plurality of text data corresponding to the historical time period;

[0034] A classifier acquisition subunit is used to acquire a classifier corresponding to each of the keywords, thereby obtaining a plurality of classifiers corresponding to each piece of text data;

[0035] The hit determination subunit is used to determine whether the text data hits the classifier set when multiple classifiers corresponding to a text data are all included in a classifier set, and increase the hit count of the classifier set by one.

[0036] In some embodiments, the classifier acquisition subunit is specifically used to perform natural language processing on the keyword to obtain the classifier corresponding to the keyword.

[0037] In some embodiments, the classifier acquisition subunit is specifically used to calculate the similarity between the keyword and a plurality of classifiers in the classifier library;

[0038] The classifier with the greatest similarity is obtained, where the classifier with the greatest similarity is the classifier corresponding to the keyword.

[0039] In some embodiments, the collective display unit includes:

[0040] A score calculation subunit is used to calculate, for each abnormal classifier set in the plurality of abnormal classifier sets, an evaluation score of the abnormal classifier set based on the popularity, specificity, complexity and abnormality of the abnormal classifier set;

[0041] A sorting subunit, used to sort the abnormal classifier set corresponding to the evaluation score in descending order according to the evaluation score, to obtain an abnormal classifier set sequence;

[0042] The sequence display subunit is used to display the abnormal category word set sequence.

[0043] In the data identification method provided in the embodiment of the present application, multiple classifier sets can be obtained; for each historical time period in multiple historical time periods, the number of hits of multiple text data corresponding to the historical time period to each classifier set can be obtained. If there is a historical time period among multiple historical time periods, and the difference between the number of hits of the text data corresponding to the historical time period to a certain classifier set and the number of hits of the text data corresponding to other historical time periods to the above classifier set exceeds a preset threshold, then the above classifier set can be determined to be an abnormal classifier set. At least one abnormal classifier set is displayed so that the staff can identify the corresponding abnormal data based on the displayed abnormal classifier set.

[0044] In the present application, compared with the prior art of manually analyzing and processing a large amount of text data, the embodiments of the present application can screen and display a set of abnormal category words. After viewing the set of abnormal category words, the staff can more quickly and conveniently summarize the corresponding abnormal data and analyze it, which shortens the time consumption and reduces the consumption of human resources. BRIEF DESCRIPTION OF THE DRAWINGS

[0045] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For those skilled in the art, other drawings can be obtained based on these drawings without creative work.

[0046] Figure 1 It is a scenario diagram of the data identification method provided in the embodiment of the present application;

[0047] Figure 2It is a flowchart of a data identification method provided in an embodiment of the present application;

[0048] Figure 3 It is a structural schematic diagram of a data identification device provided in one embodiment of the present application;

[0049] Figure 4 It is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0050] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative work are within the scope of protection of this application.

[0051] Embodiments of the present application provide a data identification method, device, electronic device, and storage medium.

[0052] The data identification device can be integrated into an electronic device, which can be a terminal, a server, or other devices. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, or a personal computer (PC). The server can be a single server or a server cluster consisting of multiple servers.

[0053] In some embodiments, the data identification device may also be integrated into multiple electronic devices. For example, the data identification device may be integrated into multiple servers, and the data identification method of the present application may be implemented by multiple servers.

[0054] In some embodiments, the server may also be implemented in the form of a terminal.

[0055] For example, the electronic device described above may execute the following method: obtaining a plurality of classifier sets, wherein each classifier set includes at least two classifiers; for a plurality of historical time periods, obtaining the number of hits of the text data corresponding to each historical time period to each classifier set; if the hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold, then determining that the first classifier set is an abnormal classifier set; wherein the first historical time period is any historical time period among the plurality of historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the plurality of classifier sets; and displaying at least one abnormal classifier set so that the staff can identify the corresponding abnormal data based on the at least one abnormal classifier set.

[0056] For more information, please see Figure 1 , let's assume that there are k sets of classifiers and j historical time periods, where k and j are both positive integers; then we can calculate the number of hits of the text data corresponding to each historical time period in the j historical time periods to each set of classifiers in the k sets of classifiers. For example, see Figure 1 , for any classifier set i (i∈[1,k] and i is a positive integer) among the k classifier sets, the text data corresponding to the j historical time periods have their own corresponding hit times in the classifier set i, that is, there are j hit times. Each classifier set will get j hit times, so the k classifier sets will get k*j hit times in total.

[0057] For more information, please see Figure 1 , let's assume that historical time period 1 and historical time period j-1 meet the preset time relationship, then the hit difference between hit count 1 and hit count j-1 can be calculated. Among them, hit count 1 represents the hit count of multiple text data corresponding to historical time period 1 to category word set i; hit count j-1 represents the hit count of multiple text data corresponding to historical time period j-1 to category word set i. If the hit difference calculated above exceeds the preset threshold, the category word set i is determined to be an abnormal category word set.

[0058] If the hit difference exceeds the preset threshold, it means that the hit situation of the corresponding multiple text data in a historical time period in historical time period 1 or historical time period j-1 that hits the category word set i has changed dramatically, which means: either the category word combination that originally did not appear frequently suddenly appears in large numbers, or the category word combination that originally appeared frequently suddenly decreases greatly; no matter which of the above situations is the case, it can be considered that the category word combination that has changed dramatically is abnormal. Therefore, the category word set corresponding to the category word combination (i.e., category word set i) can be displayed so that the staff can identify all the category words included in the abnormal category word set and conduct further analysis. The method provided in the embodiment of the present application is conducive to reducing the time consumption of identifying abnormal data and saving human resources.

[0059] It should be noted that the serial numbers of the following embodiments are not intended to limit the preferred order of the embodiments.

[0060] In this embodiment, a data identification method is provided, such as Figure 2 As shown, the data identification method can be applied to a terminal, and the specific process of the method may include the following steps 110 to 140:

[0061] 110. Obtain multiple classifier sets, wherein each of the classifier sets includes at least two classifiers.

[0062] Classifiers are words that represent categories. Different industries often have different classifiers. For example, for the convenience service industry, classifiers include feedback, complaints, gatherings, disputes, verbal disputes, physical disputes, tickets, test certificates, invoices, education and training, preschool education, students, junior high school students, senior high school students, etc.; for the medical service industry, classifiers include patient gender, disease location, disease name, etc.

[0063] For the sake of convenience, let's take the category words of the convenience service industry as an example. Category words can be divided into different levels, and the category words of the lower level of a certain category word are the sub-categories of the category word.

[0064] For example, feedback is the first level, and complaints and meetings are the second level under the level of "feedback"; that is, complaints and meetings are both subordinate levels of feedback, and feedback is the superior level of complaints and meetings; in other words, complaints and meetings are both subcategories of "feedback". For another example, disputes are the first level, and verbal disputes and physical disputes are the second level under the level of "disputes"; that is, verbal disputes and physical disputes are both subordinate levels of disputes, and disputes are the superior level of verbal disputes and physical disputes; in other words, verbal disputes and physical disputes are both subcategories of "disputes". Similarly, "test certificates" and "invoices" are both subordinate levels of "tickets and receipts"; "preschool education" is the subordinate level of "education and training"; "junior high school students" and "high school students" are both subordinate levels of "students".

[0065] Category words can also be divided into different categories. For example, feedback, complaints, gatherings, disputes, verbal disputes, and physical disputes all belong to the "event" category; tickets, receipts, test certificates, and invoices all belong to the "item" category; education and training, preschool education all belong to the "industry" category; students, junior high school students, and high school students all belong to the "occupation" category.

[0066] For details, see Table 1, which shows several first-level classifiers and corresponding second-level classifiers under the category of "event":

[0067] Table 1

[0068]

[0069] For details, see Table 2, which shows several first-level classifiers and corresponding second-level classifiers under the category of "items":

[0070] Table 2

[0071]

[0072] For details, see Table 3, which shows several first-level classifiers and corresponding second-level classifiers under the category of "industry":

[0073] Table 3

[0074]

[0075] For details, see Table 4, which shows several first-level classifiers and corresponding second-level classifiers under the category of "occupation":

[0076] Table 4

[0077]

[0078] The hierarchical division and category division of classifiers can be carried out by staff based on industry experience. It should be understood that the above descriptions of the hierarchical division and category of classifiers are all examples, and classifiers can include more or fewer hierarchical divisions and more or fewer categories. The number of hierarchical divisions of classifiers and the number of categories to which classifiers belong should not be understood as limitations on this application.

[0079] The classifier set is a set including at least two classifiers. Optionally, in a specific implementation, step 110 may specifically include the following steps 111 to 112:

[0080] 111. Obtain multiple category words.

[0081] The above-mentioned multiple classifiers may be classifiers obtained by manual screening. The above-mentioned multiple classifiers may be obtained by the terminal in a manner of manual input, or the terminal may obtain the multiple classifiers by receiving classifiers sent by other terminals.

[0082] 112. Determine the multiple classifier sets based on the multiple classifiers.

[0083] Optionally, multiple classifier sets can be determined based on multiple classifiers obtained through manual screening. In the above embodiment, multiple classifier sets can be determined based on multiple classifiers obtained through manual screening, so that the multiple classifier sets obtained can be more targeted, which is conducive to improving the accuracy of abnormal category set screening.

[0084] In a specific implementation, step 112 may specifically include the following steps 1121 to 1124:

[0085] 1121. Determine a combination of classifiers to be screened consisting of any two classifiers from among the plurality of classifiers.

[0086] For multiple classifiers, any two of the multiple classifiers can be combined to obtain multiple classifier combinations to be screened.

[0087] 1122. Screen multiple to-be-screened category word combinations to obtain at least one preliminary screening category word combination.

[0088] The specific process of screening multiple combinations of category words to be screened will be described in detail below.

[0089] 1123. Determine whether there are any remaining classifiers. If so, execute step 1124; if not, execute step 1125.

[0090] The remaining classifiers are classifiers among the multiple classifiers except the classifiers included in the initial screening classifier combination.

[0091] After obtaining at least one preliminary screening classifier combination in step 1122, for each preliminary screening classifier combination, it can be determined whether there are any remaining classifiers. If there are no remaining classifiers in all preliminary screening classifier combinations, it is determined that there are no remaining classifiers, and step 1125 is executed; if there are any preliminary screening classifier combinations that still have remaining classifiers, it is determined that there are still remaining classifiers, and step 1124 is executed.

[0092] 1124. For each of the initially screened classifier combinations with remaining classifiers, add any remaining classifier to obtain a new classifier combination to be screened, and jump to step 1122.

[0093] For the initial screening category word combination with remaining category words, any remaining category word can be added to the initial screening category word combination, so that the initial screening category word combination with the number of category words increased by one becomes a new category word combination to be screened, and jump to step 1122 to perform a new round of screening of the category word combination to be screened.

[0094] 1125. Record at least one initially screened category word combination as a category word set.

[0095] If there are no remaining classifiers in all the initially screened classifier combinations, it indicates that the screening process of all the classifiers obtained in step 111 has been completed, and at least one initially screened classifier combination obtained at this time can be recorded as a classifier set.

[0096] Next, steps 1121 to 1125 are explained with reference to a specific example: suppose there are a plurality of classifiers including a1, a2, a3, a4, and a5, a total of 5 classifiers. Then, all the classifier combinations to be screened consisting of two classifiers can be determined first: a1 and a2, a1 and a3, a1 and a4, a1 and a5; a2 and a3, a2 and a4, a2 and a5; a3 and a4, a3 and a5; a4 and a5.

[0097] Then, all the above-mentioned combinations of classifiers to be screened are screened to obtain at least one preliminary screening combination of classifiers. The specific process of screening all the above-mentioned combinations of classifiers to be screened is described in detail below. Let's assume that after this screening, the preliminary screening combinations of classifiers obtained are the following four groups: a1 and a3, a1 and a4, a1 and a5, and a2 and a5.

[0098] Then, the judgment step is performed to determine whether there are any remaining classifiers. It is obvious that for the five classifiers a1, a2, a3, a4, and a5, the above-mentioned initial screening classifier combinations only include two of them, that is, there are remaining classifiers in each initial screening classifier combination. For example, for the initial screening classifier combination: a1 and a3, there are also remaining classifiers: a2, a4, and a5.

[0099] Let's take the initial screening classifier combination of a1 and a4 as an example. For this initial screening classifier combination, by adding any one of the remaining classifiers, we can get a classifier combination to be screened that includes three classifiers: a1, a4 and a2; a1, a4 and a3; a1, a4 and a5. Here, the initial screening classifier combination of a1 and a4 is used as an example for explanation; it should be understood that for a1 and a3, a1 and a5, a2 and a5, we can get a classifier combination to be screened that includes three classifiers by adding any one of the remaining classifiers. That is, each group of initial screening classifier combinations can get three new classifier combinations to be screened, and four groups of initial screening classifier combinations can get a total of twelve new classifier combinations to be screened.

[0100] A new round of screening is performed on the twelve new classification word combinations to be screened obtained above, so as to obtain preliminary screening classification word combinations. The preliminary screening classification word combinations currently obtained are all composed of three classification word combinations.

[0101] Then continue to perform the judgment step to determine whether there are remaining category words...

[0102] The same process is repeated until there are no remaining classifiers in all the initial screening classifier combinations, and then it is determined that there are no remaining classifiers. In the case where there are no remaining classifiers, the initial screening classifier combinations obtained are the multiple classifier sets determined in step 112.

[0103] In the above implementation, the existence of remaining classifiers is used as the loop termination condition to determine multiple classifier sets, and each time after a new combination of classifiers to be screened is determined, a screening operation is performed to obtain a preliminary screening classifier combination. In this way, the amount of data calculation can be greatly reduced and the efficiency of determining the classifier set can be improved.

[0104] For details, see Table 5, which shows examples of several classifier sets:

[0105] Table 5

[0106]

[0107] It can be seen from Table 5 that classifiers of different levels expressing different meanings can appear in the same classifier set. For example, please refer to the classifier set shown in sequence number 2 in Table 5 for details. The second-level classifier "education and training / class" of the "industry" category and the first-level classifier "student" of the "occupation" category are in the same classifier set.

[0108] Comparing the sequence number 1 and sequence number 2 shown in Table 5, it can be seen that, except for the difference between the classifier "student / high school student" and the classifier "student", the other classifiers included in the two classifier sets are the same. By constructing the two classifier sets of sequence number 1 and sequence number 2, it is possible to achieve simultaneous attention to: a classifier set with high abstractness and a classifier set with high concreteness.

[0109] For example, suppose you get three pieces of text data as shown in the following table:

[0110] Table 6

[0111]

[0112] By extracting classifiers from the three text data shown in Table 6, multiple classifiers corresponding to each text data can be obtained (the process of obtaining multiple classifiers corresponding to the text data will be described in detail below), as shown in Table 7:

[0113] Table 7

[0114]

[0115] For the sequence numbers 1 and 2 shown in Table 5, it can be seen that the category word set corresponding to the sequence number 1 in Table 5, "disasters / epidemics; environmental pollution / noise; education and training / classes; students / high school students; buildings / decoration", can only hit the text data corresponding to the sequence number 1 in Table 7; while the category word set corresponding to the sequence number 2 in Table 5, "disasters / epidemics; environmental pollution / noise; education and training / classes; students; buildings / decoration", can hit the text data corresponding to the sequence number 1 in Table 7 and the text data corresponding to the sequence number 2 at the same time. Therefore, when the category word set corresponding to the sequence number 1 in Table 5 cannot identify abnormal data, the category word set corresponding to the sequence number 2 in Table 5 may identify abnormal data.

[0116] Optionally, in a specific implementation, the step of "screening multiple combinations of classifiers to be screened" in step 1122 may specifically include the following steps A1 to A3:

[0117] A1. Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed.

[0118] The category phrase to be processed is a phrase composed of category words to be processed that are manually selected.

[0119] The classifiers to be processed are manually selected and used to implement the screening of the classifier combinations to be screened. The classifiers to be processed are used to implement the calculation of a certain parameter of the classifier combinations to be screened; for example, the number of hits of the classifier combinations to be screened with respect to multiple classifier phrases to be processed can be calculated, and the confidence of the classifier combinations to be screened with respect to multiple classifier phrases to be processed can also be calculated.

[0120] A2. For each of the to-be-screened category word combinations, calculate the number of hits of the to-be-screened category word combination with respect to the multiple to-be-processed category word groups.

[0121] For each of the multiple classifier combinations to be screened, the hit count of the classifier combination to be screened with respect to the multiple classifier phrases to be processed acquired in A1 can be calculated. That is, each classifier combination to be screened can calculate its own hit count in step A2.

[0122] If the category words to be processed included in a category phrase to be processed are all included in a category word combination to be screened, it is determined that the category phrase to be processed hits the above category word combination to be screened, so that the number of hits of the above category word combination to be screened can be increased by one.

[0123] A3. For the plurality of to-be-screened category word combinations, eliminate the to-be-screened category word combinations whose hit times are higher than a first threshold, and eliminate the to-be-screened category word combinations whose hit times are lower than a second threshold, to obtain at least one preliminary screening category word combination.

[0124] It should be understood that the second threshold is lower than the first threshold, and the specific values ​​of the first threshold and the second threshold should not be understood as limiting the present application.

[0125] The number of hits is compared with the first threshold and the second threshold respectively, and the category word combinations to be screened whose hit times are not higher than the first threshold and not lower than the second threshold are retained, and these category word combinations to be screened are recorded as preliminary screening category word combinations.

[0126] In the above implementation, the category word combinations to be screened can be screened according to the hit counts of the category word combinations to be screened. If the hit count is higher than the first threshold, it means that the combination of the category word combinations to be screened corresponding to the hit count is too conventional and has less analytical value. For example, for the category word combination: "traffic accident + road", since traffic accidents usually occur on roads, the analytical value is relatively small; if the hit count is lower than the second threshold, it means that the combination of the category word combinations to be screened corresponding to the hit count is more accidental and also has less analytical value. By eliminating category word combinations with less analytical value, category word combinations with high analytical value can be allocated to more computing resources, which is beneficial to improving the accuracy of screening of abnormal category word sets.

[0127] Optionally, in another specific implementation, the step of "screening multiple combinations of classifiers to be screened" in step 1122 may specifically include the following steps B1 to B4:

[0128] B1. Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed.

[0129] The category phrase to be processed is a phrase composed of category words to be processed that are manually selected.

[0130] The classifiers to be processed are manually selected and used to implement the screening of the classifier combinations to be screened. The classifiers to be processed are used to implement the calculation of a certain parameter of the classifier combinations to be screened; for example, the number of hits of the classifier combinations to be screened with respect to multiple classifier phrases to be processed can be calculated, and the confidence of the classifier combinations to be screened with respect to multiple classifier phrases to be processed can also be calculated.

[0131] B2. For each of the to-be-screened category word combinations, calculate the confidence of the to-be-screened category word combination with respect to the plurality of to-be-processed category word groups.

[0132] For each of the multiple classifier combinations to be screened, the confidence of the classifier combination to be screened with respect to the multiple classifier phrases to be processed acquired in step B1 can be calculated. That is, each classifier combination to be screened can calculate its own confidence through step B2.

[0133] B3. Sort the multiple to-be-screened category word combinations in descending order of confidence to obtain a sorted sequence of to-be-screened category word combinations.

[0134] B4. intercepting the first n to-be-screened category word combinations of the to-be-screened category word combination sequence to obtain at least one preliminary screening category word combination; wherein n is a positive integer.

[0135] After calculating the confidence corresponding to each of the to-be-screened classifier combinations, the multiple to-be-screened classifier combinations can be sorted in descending order of confidence to obtain a to-be-screened classifier combination sequence. Then, the first n to-be-screened classifier combinations of the to-be-screened classifier combination sequence are intercepted, and the first n to-be-screened classifier combinations are used as the initial screening classifier combinations.

[0136] In the above implementation, the classifier combinations to be screened may be screened according to the confidences corresponding to the classifier combinations to be screened, thereby obtaining classifier combinations to be screened with high confidences.

[0137] It should be understood that the basis for screening multiple combinations of classifiers to be screened may include not only the number of hits and confidence, but also other methods. For example, multiple combinations of classifiers to be screened may be screened based on expert experience. The specific screening method for the combinations of classifiers to be screened should not be understood as a limitation to the present application.

[0138] Optionally, in a specific implementation, step 112 may also be implemented in the following specific manner:

[0139] For multiple classifiers, first combine all the classifier combinations to be screened. Each classifier combination to be screened is composed of at least two classifiers from the multiple classifiers. Optionally, there may not be different levels of classifiers expressing the same meaning between the at least two classifiers. For example, for the two classifiers "students" and "high school students" that express the same meaning but at different levels, they do not appear as two classifiers in the same classifier combination to be screened.

[0140] All the combinations of classifiers to be screened are screened to obtain a screened classifier set.

[0141] Screening out a set of category words from all the category word combinations to be screened can be achieved based on the number of hits corresponding to each category word combination to be screened, for details, please refer to the above steps A1 to A3; it can also be achieved based on the confidence level corresponding to each category word combination to be screened, for details, please refer to the above steps B1 to B4.

[0142] 120. For multiple historical time periods, obtain the number of hits of the text data corresponding to each historical time period to each of the classifier sets.

[0143] The historical time period is a historical time period with a fixed time length, for example, one day can be used as a historical time period, or one week can be used as a historical time period. The specific time length of the historical time period should not be understood as a limitation to the present application.

[0144] The text data corresponding to the historical time period refers to multiple text data generated in the historical time period. For example, for the convenience service industry, it can refer to the text data received from citizens complaining about livelihood issues, and multiple pieces of this text data may be generated in each historical time period; for the medical service industry, it can refer to the medical record text data received from patients when they see a doctor, and multiple pieces of this text data may also be generated in each historical time period. For the sake of ease of description, the text data of the convenience service industry is still used as an example for description below.

[0145] Optionally, in a specific implementation, step 120 includes the following steps 121 to 123:

[0146] 121. For each piece of text data among the multiple pieces of text data corresponding to the historical time period, extract multiple keywords from the text data.

[0147] The process of extracting keywords from text data can be specifically implemented through a trained Natural Language Processing (NLP) model.

[0148] Let’s take the text data of citizens’ complaints about people’s livelihood issues as an example: “Children in the first and second year of high school have been taking online classes for the entire semester during the intensive teaching period of difficult and key high school knowledge. Now that the epidemic situation has improved, the renovation project has been carried out vigorously, and the high-intensity noise has made them wonder how to take online classes...”

[0149] After processing the above text data with the NLP model, we can obtain the following keywords: children in the first and second year of high school, online classes, epidemic, decoration, and noise.

[0150] 122. Obtain a classifier corresponding to each of the keywords, thereby obtaining multiple classifiers corresponding to each piece of text data.

[0151] Optionally, in a specific implementation, "obtaining the category words corresponding to each of the keywords" can be implemented in the following manner: performing natural language processing on the keywords to obtain the category words corresponding to the keywords.

[0152] For each keyword, natural language processing can be performed through the trained NLP model to obtain the category word corresponding to the keyword.

[0153] Continue with the above example:

[0154] For the keyword "children in the first and second year of high school", after natural language processing by the NLP model, we can get the corresponding category words: high school students and the parent level of "high school students" "students";

[0155] For the keyword "online class", after natural language processing by the NLP model, we can get the corresponding category words: class and the parent level of "class" "education and training";

[0156] For the keyword "epidemic", after natural language processing by the NLP model, we can get the corresponding category words: epidemic and the parent level of "epidemic" "disaster accident";

[0157] For the keyword "decoration", after natural language processing by the NLP model, we can get the corresponding category word: decoration and the parent level of "decoration" "architecture";

[0158] For the keyword: noise, after natural language processing by the NLP model, we can get the corresponding category word: noise and the parent level of "noise" "environmental pollution".

[0159] In the above implementation, by performing natural language processing on the keywords to obtain the classifiers corresponding to the keywords, the amount of computation consumed in obtaining the classifiers can be reduced, time can be saved, and efficiency can be improved.

[0160] Optionally, in another specific implementation, step 122 may be implemented by: calculating the similarity between the keyword and multiple category words in the category word library; obtaining the category word with the greatest similarity, the category word with the greatest similarity being the category word corresponding to the keyword.

[0161] Continue with the above example:

[0162] Suppose the category word library includes X category words, where X is a positive integer.

[0163] For the keyword: children in the first and second year of high school, the similarity between the keyword and the X category words in the category word library can be calculated one by one, so as to obtain the category word with the greatest similarity. For example, let the category word "high school student" be the category word with the greatest similarity to the keyword "children in the first and second year of high school", so the category word "high school student" can be determined as the category word corresponding to the keyword "children in the first and second year of high school", and the parent level "student" of the category word "high school student" can be obtained at the same time.

[0164] For the keyword: online class, the similarity between the keyword and the X classifiers in the classifier database can be calculated one by one, so as to obtain the classifier with the greatest similarity. For example, let the classifier "class" be the classifier with the greatest similarity to the keyword "online class", so the classifier "class" can be determined as the classifier corresponding to the keyword "online class", and the parent level "education and training" of the classifier "class" can be obtained at the same time.

[0165] For the keyword: epidemic, the similarity between the keyword and the X category words in the category word library can be calculated one by one to obtain the category word with the greatest similarity. For example, let's assume that the category word "epidemic" is the category word with the greatest similarity to the keyword "epidemic", so the category word "epidemic" can be determined as the category word corresponding to the keyword "epidemic", and the parent level "disaster accident" of the category word "epidemic" can be obtained at the same time.

[0166] For the keyword: decoration, the similarity between the keyword and the X category words in the category word library can be calculated one by one, so as to obtain the category word with the greatest similarity. For example, let's assume that the category word "decoration" is the category word with the greatest similarity to the keyword "decoration", so the category word "decoration" can be determined as the category word corresponding to the keyword "decoration", and the parent level "architecture" of the category word "decoration" can be obtained at the same time.

[0167] For the keyword: noise, the similarity between the keyword and the X classifiers in the classifier database can be calculated one by one, so as to obtain the classifier with the greatest similarity. For example, let the classifier "noise" be the classifier with the greatest similarity to the keyword "noise", so the classifier "noise" can be determined as the classifier corresponding to the keyword "noise", and the parent level "environmental pollution" of the classifier "noise" can be obtained at the same time.

[0168] Therefore, for the text data: "Children in the first and second year of high school have been taking online classes for the entire semester during the period of intensive teaching of key and difficult high school knowledge. Now that the epidemic situation has improved, the renovation project has been carried out vigorously, and there is high-intensity noise. How to take online classes...", through step 122, multiple category words corresponding to the text data can be obtained: students / high school students, education and training / classes, disasters and accidents / epidemics, construction / decoration, environmental pollution / noise.

[0169] It should be understood that the method provided in the embodiment of the present application can identify abnormal data for more specific category words in addition to identifying abnormal data for more general category word combinations (such as the above-mentioned combination of high school students, classes, epidemics, decoration, and noise). For example, abnormal data can also be identified for the category word combination of XX company + labor-capital dispute.

[0170] 123. If there is a piece of text data and multiple classifiers corresponding to the multiple classifiers are all included in a classifier set, it is determined that the text data hits the classifier set, and the hit count of the classifier set is increased by one.

[0171] Suppose that the multiple category words corresponding to the text data are: Z1, Z2, Z3, Z4; the category words included in the category word set are: Y1, Z1, Z2, Z3, Z4, Z5, then the category words corresponding to the text data are all included in the category word set, that is, the category words corresponding to the text data are a subset of the category word set. Therefore, it can be determined that the text data hits the category word set.

[0172] It should be understood that if a classifier in the classifier set is a higher level of a classifier in the text data, it can also be determined that the classifier set includes the classifier in the text data. For example, if the classifier in the classifier set is "student", and the text data includes the classifier "senior high school student", since "student" is a higher level of "senior high school student", it can still be determined that the classifier set includes the classifier "senior high school student" in the text data.

[0173] Optionally, in a specific implementation, if a classifier in the classifier set is a lower level of a classifier in the text data, it can also be determined that the classifier set does not include the classifier in the text data. For example, if the classifier in the classifier set is "high school student", and the text data includes the classifier "student", since "high school student" is a higher level of "student", it can be determined that the classifier set does not include the classifier "student" in the text data.

[0174] By executing steps 121 to 123 for the multiple text data included in each historical time period and each classifier set, the number of hits of the text data corresponding to each historical time period with respect to each classifier set can be obtained.

[0175] 130. If the hit difference between the number of hits of the first category word set by the text data corresponding to the first historical time period and the number of hits of the first category word set by the text data corresponding to the second historical time period exceeds a preset threshold, the first category word set is determined to be an abnormal category word set.

[0176] Among them, the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first category word set is any category word set among the multiple category word sets.

[0177] The second historical time period may be a historical time period that can be compared with the first historical time period, such as the first r historical time periods or the last s historical time periods adjacent to the first historical time period, where r and s are both positive integers. For example, let the historical time period be one day, and the first historical time period be June 22, 2022. Then the second historical time period may include June 20, 2022, June 21, 2022, June 23, 2022, June 24, 2022, June 25, 2022, or June 26, 2022, so that the first historical time period can be compared with any day from June 20, 2022 to June 26, 2022, except June 22, 2022.

[0178] The second historical time period may also be a historical time period that can be compared with the first historical time period, such as a historical time period separated from the first historical time period by a preset number of days, and the specific value of the preset number of days should not be understood as a limitation to the present application. For example, let the historical time period be one day, and if the preset number of days is 7 days, then the second historical time period may be a day separated from the first historical time period by one week; if the preset number of days is 30 days, then the second historical time period may be a day separated from the first historical time period by one month.

[0179] In addition to the above-mentioned month-on-month and year-on-year comparisons of the number of hits for the same category word set in text data in different historical time periods, the changes in the number of hits can also be intuitively reflected by outputting control charts for staff to view intuitively.

[0180] If the hit difference between the number of hits of the text data corresponding to the first historical time period to the first category word set and the number of hits of the text data corresponding to the second historical time period to the first category word set exceeds a preset threshold, it means that the text data corresponding to the first historical time period has undergone a more obvious change compared with the text data corresponding to the second historical time period: either the category word combination that did not appear frequently before suddenly appears in large numbers, or the category word combination that appeared frequently before suddenly decreases greatly; in either of the above situations, it can be considered that the category word combination that has undergone obvious changes is abnormal, that is, the first category word set containing the category word combination is an abnormal category word set.

[0181] For each classifier set in the multiple classifier sets, the above step 130 may be performed, so that all abnormal classifier sets in the multiple classifier sets may be obtained.

[0182] 140. Display at least one abnormal category word set.

[0183] In the above implementation, at least one abnormal classifier set can be displayed to draw the staff's attention to the abnormal classifier set and the multiple classifiers included in the abnormal classifier set, thereby facilitating the staff to obtain abnormal data in a targeted manner based on the multiple classifiers, saving the staff's time and improving processing efficiency.

[0184] Optionally, in a specific implementation, the multiple abnormal classifier sets obtained in step 130 may be compared in pairs. If there is an abnormal classifier set (which may be set as abnormal classifier set A) which is a subset of another abnormal classifier set (which may be set as abnormal classifier set B), then only the abnormal classifier set with a higher degree of specificity as a subset (i.e., abnormal classifier set A) may be retained.

[0185] Optionally, in a specific implementation, step 140 may specifically include the following steps 141 to 143:

[0186] 141. For each abnormal classifier set in the plurality of abnormal classifier sets, calculate an evaluation score of the abnormal classifier set based on popularity, specificity, complexity, and abnormality of the abnormal classifier set.

[0187] The heat represents the number of text data hit by the abnormal category word set within a preset historical time period; the more text data there are, the greater the heat value is, and the heat can be represented by rh.

[0188] The specificity represents the number of levels of the classifiers included in the abnormal classifier set; the more levels there are, the greater the specificity value is, and the specificity can be represented by rs. For example, let's assume that the specificity value corresponding to the abnormal classifier set with only one level is 0.3, the specificity value corresponding to the abnormal classifier set with two levels is 0.5, and the specificity value corresponding to the abnormal classifier set with three levels is 0.7... Then the specific value of the specificity can be determined according to the number of levels of the abnormal classifier set.

[0189] The complexity represents the number of types of classifiers included in the abnormal classifier set; the greater the number of types, the greater the complexity value, and the complexity can be represented by rc.

[0190] The abnormality degree represents the ratio of the abnormal text data hit by the abnormal classifier set to the total text data hit, and the abnormality degree can be represented by ra. The above ratio can be the ratio of the data volume of the abnormal text data to the data volume of the total text data, or the ratio of the number of historical time periods where the abnormal text data is located to the number of historical time periods where the total text data is located.

[0191] Optionally, the popularity, specificity, complexity and abnormality can all be normalized to avoid the impact of numerical differences on the calculation of evaluation scores.

[0192] The evaluation score is used to reflect the degree to which the abnormal category word set deserves attention. The higher the evaluation score, the higher the degree to which the abnormal category word set corresponding to the evaluation score deserves attention.

[0193] Optionally, the evaluation score can be calculated according to the formula: score=ahrh+asrs+acrc+aara. Among them, ah is the score weight of heat rh, as is the score weight of specificity rs, ac is the score weight of complexity rc, and aa is the score weight of abnormality ra. The values ​​of the above score weights: ah, as, ac, aa can be assigned by the staff according to their work experience.

[0194] 142. According to the evaluation scores, sort the abnormal category word sets corresponding to the evaluation scores in descending order to obtain an abnormal category word set sequence.

[0195] 143. Display the abnormal category word set sequence.

[0196] In the above implementation, after calculating the evaluation score of each abnormal classifier set, the multiple abnormal classifier sets can be sorted in descending order of the evaluation score, and the sorted abnormal classifier set sequence can be displayed. Since the evaluation score reflects the degree of attention of the corresponding abnormal classifier set, displaying the abnormal classifier set sequence can facilitate the staff to intuitively understand the degree of attention of the multiple abnormal classifier set sequences, so that the staff can focus on the abnormal classifier set that is more worthy of attention, and further improve the processing efficiency of the staff.

[0197] In the data identification method provided by the embodiment of the present application, in the data identification method provided by the embodiment of the present application, multiple classifier sets can be obtained; for each historical time period in multiple historical time periods, the number of hits of multiple text data corresponding to the historical time period to each classifier set can be obtained. If there is a historical time period in multiple historical time periods, the hit difference between the number of hits of the text data corresponding to the historical time period to a certain classifier set and the number of hits of the text data corresponding to other historical time periods to the above classifier set exceeds the preset threshold, then the above classifier set can be determined to be an abnormal classifier set. At least one abnormal classifier set is displayed so that the staff can identify the corresponding abnormal data based on the abnormal classifier set shown. Compared with the prior art that manually analyzes and processes a large amount of text data, the embodiment of the present application can screen and display the abnormal classifier set. After checking the abnormal classifier set, the staff can summarize the corresponding abnormal data more quickly and conveniently, and analyze it, which reduces the consumption time and reduces the consumption of human resources. And compared with the prior art, the method provided by the embodiment of the present application has a wider coverage and can better cope with the flexible and changeable needs of business scenarios. The method for sorting the abnormal category word set proposed in this application can help staff to find valuable abnormal data more easily.

[0198] In this application, the consumption of human resources can be reduced.

[0199] In order to better implement the above method, the embodiment of the present application also provides a data identification device, which can be integrated in an electronic device, and the electronic device can be a terminal, a server, etc. Among them, the terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, or a personal computer (PC) and other devices; the server can be a single server or a server cluster composed of multiple servers. For example, Figure 3 As shown, the data identification device may include:

[0200] A set acquisition unit 301 is used to acquire a plurality of classifier sets, wherein each of the classifier sets includes at least two classifiers;

[0201] A hit count acquisition unit 302 is used to acquire, for multiple historical time periods, the hit count of the text data corresponding to each historical time period for each of the classifier sets;

[0202] The abnormality determination unit 303 is used to determine that the first classifier set is an abnormal classifier set if the hit difference between the hit count of the text data corresponding to the first historical time period and the hit count of the text data corresponding to the second historical time period and the first classifier set exceeds a preset threshold value; wherein the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the multiple classifier sets;

[0203] The set display unit 304 is used to display at least one abnormal category word set.

[0204] In some embodiments, the set acquisition unit 301 includes:

[0205] A classifier acquisition subunit is used to acquire multiple classifiers;

[0206] The set determination subunit is used to determine the plurality of classifier sets based on the plurality of classifiers.

[0207] In some embodiments, the set determining subunit includes:

[0208] The to-be-screened category sub-subunit is used to determine a to-be-screened category word combination consisting of any two category words among a plurality of category words;

[0209] The primary screening combination sub-subunit is used to screen multiple classifier combinations to be screened to obtain at least one primary screening classifier combination;

[0210] The classifier adding sub-subunit is used to add any remaining classifier to each of the preliminary screening classifier combinations with remaining classifiers to obtain a new classifier combination to be screened, wherein the remaining classifier is a classifier among the multiple classifiers except the classifiers included in the preliminary screening classifier combination;

[0211] The step jump sub-subunit is used to jump to the step: screen multiple classifier word combinations to be screened to obtain at least one preliminary screening classifier word combination until there are no remaining classifier words; wherein, when there are no remaining classifier words, the corresponding multiple preliminary screening classifier word combinations are the multiple classifier word sets.

[0212] In some embodiments, the primary screening combination sub-subunit is specifically used for:

[0213] Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed;

[0214] For each of the to-be-screened category word combinations, calculating the number of hits of the to-be-screened category word combination with respect to the plurality of to-be-processed category word groups;

[0215] For the plurality of to-be-screened category word combinations, the to-be-screened category word combinations with hit times higher than a first threshold are eliminated, and the to-be-screened category word combinations with hit times lower than a second threshold are eliminated, to obtain at least one preliminary screening category word combination.

[0216] In some embodiments, the primary screening combination sub-subunit is specifically used for:

[0217] Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed;

[0218] For each of the to-be-screened classifier word combinations, calculating the confidence of the to-be-screened classifier word combination with respect to the plurality of to-be-processed classifier word groups;

[0219] Sorting the plurality of to-be-screened classifier word combinations in descending order of confidence to obtain a sorted sequence of to-be-screened classifier word combinations;

[0220] The first n to-be-screened classifier word combinations of the to-be-screened classifier word combination sequence are intercepted to obtain at least one preliminary screening classifier word combination; wherein n is a positive integer.

[0221] In some embodiments, the hit count acquisition unit 302 includes:

[0222] A keyword extraction subunit, configured to extract a plurality of keywords from each of the plurality of text data corresponding to the historical time period;

[0223] A classifier acquisition subunit is used to acquire a classifier corresponding to each of the keywords, thereby obtaining a plurality of classifiers corresponding to each piece of text data;

[0224] The hit determination subunit is used to determine whether the text data hits the classifier set when multiple classifiers corresponding to a text data are all included in a classifier set, and increase the hit count of the classifier set by one.

[0225] In some embodiments, the classifier acquisition subunit is specifically used to perform natural language processing on the keyword to obtain the classifier corresponding to the keyword.

[0226] In some embodiments, the classifier acquisition subunit is specifically used to calculate the similarity between the keyword and a plurality of classifiers in the classifier library;

[0227] The classifier with the greatest similarity is obtained, where the classifier with the greatest similarity is the classifier corresponding to the keyword.

[0228] In some embodiments, the collection display unit 304 includes:

[0229] A score calculation subunit is used to calculate, for each abnormal classifier set in the plurality of abnormal classifier sets, an evaluation score of the abnormal classifier set based on the popularity, specificity, complexity and abnormality of the abnormal classifier set;

[0230] A sorting subunit, used to sort the abnormal classifier set corresponding to the evaluation score in descending order according to the evaluation score, to obtain an abnormal classifier set sequence;

[0231] The sequence display subunit is used to display the abnormal category word set sequence.

[0232] In specific implementation, the above units can be implemented as independent entities, or can be arbitrarily combined to be implemented as the same or several entities. The specific implementation of the above units can refer to the previous method embodiments, which will not be repeated here.

[0233] In the data identification method provided in the embodiment of the present application, multiple classifier sets can be obtained; for each historical time period in multiple historical time periods, the number of hits of multiple text data corresponding to the historical time period to each classifier set can be obtained. If there is a historical time period in multiple historical time periods, and the difference between the number of hits of the text data corresponding to the historical time period to a certain classifier set and the number of hits of the text data corresponding to other historical time periods to the above classifier set exceeds a preset threshold, then the above classifier set can be determined to be an abnormal classifier set. At least one abnormal classifier set is displayed so that the staff can identify the corresponding abnormal data based on the abnormal classifier set shown. Compared with the prior art that manually analyzes and processes a large amount of text data, the embodiment of the present application can screen and display abnormal classifier sets. After viewing the abnormal classifier set, the staff can summarize the corresponding abnormal data more quickly and conveniently, and analyze it, which reduces the consumption time and reduces the consumption of human resources.

[0234] In this application, the consumption of human resources can be reduced.

[0235] The embodiment of the present application also provides an electronic device, which can be a terminal, a server, etc. The terminal can be a mobile phone, a tablet computer, a smart Bluetooth device, a laptop, a personal computer, etc. The server can be a single server or a server cluster composed of multiple servers, etc.

[0236] In some embodiments, the data identification device may also be integrated into multiple electronic devices. For example, the data identification device may be integrated into multiple servers, and the data identification method of the present application may be implemented by multiple servers.

[0237] In this embodiment, the electronic device of this embodiment is an electronic device as an example for detailed description, for example, Figure 4 As shown, it shows a schematic diagram of the structure of the electronic device involved in the embodiment of the present application, specifically:

[0238] The electronic device may include components such as a processor 401 with one or more processing cores, a memory 402 with one or more computer-readable storage media, a power supply 403, an input module 404, and a communication module 405. Those skilled in the art will appreciate that Figure 4 The electronic device structure shown in the figure does not constitute a limitation on the electronic device, and may include more or fewer components than shown in the figure, or combine certain components, or arrange the components differently.

[0239] The processor 401 is the control center of the electronic device. It uses various interfaces and lines to connect various parts of the entire electronic device. By running or executing software programs and / or modules stored in the memory 402 and calling data stored in the memory 402, it performs various functions of the electronic device and processes data, thereby monitoring the electronic device as a whole. In some embodiments, the processor 401 may include one or more processing cores; in some embodiments, the processor 401 may integrate an application processor and a modem processor, wherein the application processor mainly processes the operating system, user interface, and application programs, and the modem processor mainly processes wireless communications. It is understandable that the above-mentioned modem processor may not be integrated into the processor 401.

[0240] The memory 402 can be used to store software programs and modules. The processor 401 executes various functional applications and data processing by running the software programs and modules stored in the memory 402. The memory 402 may mainly include a program storage area and a data storage area, wherein the program storage area may store an operating system, an application required for at least one function (such as a sound playback function, an image playback function, etc.), etc.; the data storage area may store data created according to the use of the electronic device, etc. In addition, the memory 402 may include a high-speed random access memory, and may also include a non-volatile memory, such as at least one disk storage device, a flash memory device, or other volatile solid-state storage devices. Accordingly, the memory 402 may also include a memory controller to provide the processor 401 with access to the memory 402.

[0241] The electronic device also includes a power supply 403 for supplying power to various components. In some embodiments, the power supply 403 can be logically connected to the processor 401 through a power management system, so as to manage charging, discharging, and power consumption through the power management system. The power supply 403 can also include any components such as one or more DC or AC power supplies, recharging systems, power failure detection circuits, power converters or inverters, and power status indicators.

[0242] The electronic device may further include an input module 404, which may be used to receive input digital or character information and generate keyboard, mouse, joystick, optical or trackball signal input related to user settings and function control.

[0243] The electronic device may further include a communication module 405. In some embodiments, the communication module 405 may include a wireless module. The electronic device may perform short-range wireless transmission through the wireless module of the communication module 405, thereby providing the user with wireless broadband Internet access. For example, the communication module 405 may be used to help the user send and receive emails, browse web pages, and access streaming media.

[0244] Although not shown, the electronic device may further include a display unit, etc., which will not be described in detail herein. Specifically in this embodiment, the processor 401 in the electronic device will load the executable files corresponding to the processes of one or more application programs into the memory 402 according to the following instructions, and the processor 401 will run the application programs stored in the memory 402, thereby realizing various functions, as follows:

[0245] Acquire multiple classifier sets, wherein each of the classifier sets includes at least two classifiers; for multiple historical time periods, acquire the number of hits of the text data corresponding to each historical time period to each of the classifier sets; if the hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold, determine that the first classifier set is an abnormal classifier set; wherein the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the multiple classifier sets; display at least one abnormal classifier set so that the staff can identify the corresponding abnormal data based on the at least one abnormal classifier set.

[0246] The specific implementation of the above operations can be found in the previous embodiments, which will not be described in detail here.

[0247] A person of ordinary skill in the art will appreciate that all or part of the steps in the various methods of the above embodiments may be completed by instructions, or by controlling related hardware through instructions. The instructions may be stored in a computer-readable storage medium and loaded and executed by a processor.

[0248] To this end, an embodiment of the present application provides a computer-readable storage medium, in which a plurality of instructions are stored, and the instructions can be loaded by a processor to execute the steps in any data identification method provided in the embodiment of the present application. For example, the instructions can execute the following steps:

[0249] Acquire multiple classifier sets, wherein each of the classifier sets includes at least two classifiers; for multiple historical time periods, acquire the number of hits of the text data corresponding to each historical time period to each of the classifier sets; if the hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold, determine that the first classifier set is an abnormal classifier set; wherein the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the multiple classifier sets; display at least one abnormal classifier set so that the staff can identify the corresponding abnormal data based on the at least one abnormal classifier set.

[0250] The storage medium may include: a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, etc.

[0251] According to one aspect of the present application, a computer program product or a computer program is provided, the computer program product or the computer program including computer instructions, the computer instructions being stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes the methods provided in the various optional implementations provided in the above embodiments.

[0252] Since the instructions stored in the storage medium can execute the steps in any data identification method provided in the embodiments of the present application, the beneficial effects that can be achieved by any data identification method provided in the embodiments of the present application can be achieved. Please refer to the previous embodiments for details and will not be repeated here.

[0253] The above is a detailed introduction to a data identification method, device, electronic device and computer-readable storage medium provided in the embodiments of the present application. Specific examples are used herein to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for technical personnel in this field, according to the idea of ​​the present application, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A data identification method, characterized in that: The method comprises: Acquire multiple classifier sets, wherein each of the classifier sets includes at least two classifiers; the acquiring multiple classifier sets includes: Get multiple category words; Determine a classifier combination to be screened consisting of any two classifiers among the plurality of classifiers; Acquire a plurality of to-be-processed category phrase groups, wherein each of the to-be-processed category phrase groups includes at least two to-be-processed category words, and the to-be-processed category words are used to implement the screening of the to-be-screened category word combination; For each of the to-be-screened category word combinations, calculating the number of hits of the to-be-screened category word combination with respect to the plurality of to-be-processed category word groups; For the plurality of to-be-screened classifier word combinations, the to-be-screened classifier word combinations whose hit times are higher than a first threshold are eliminated, and the to-be-screened classifier word combinations whose hit times are lower than a second threshold are eliminated, to obtain at least one preliminary screening classifier word combination; Determining whether there are any remaining classifiers, wherein the remaining classifiers are classifiers among the plurality of classifiers other than the classifiers included in the initial screening classifier combination; If yes, for each of the preliminary screening classifier combinations with remaining classifiers, add any remaining classifier to obtain a new classifier combination to be screened, and jump to step: screen multiple classifier combinations to be screened to obtain at least one preliminary screening classifier combination; If not, record at least one combination of preliminary screening classifiers as a classifier set; For multiple historical time periods, obtaining the number of hits of the text data corresponding to each historical time period to each of the classifier sets; If the hit difference between the hit count of the first set of classifiers for the text data corresponding to the first historical time period and the hit count of the first set of classifiers for the text data corresponding to the second historical time period exceeds a preset threshold, the first set of classifiers is determined to be an abnormal set of classifiers; wherein the first historical time period is any one of the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first set of classifiers is any one of the multiple sets of classifiers; Display at least one set of abnormal category words.

2. The method according to claim 1, characterized in that The step of screening a plurality of to-be-screened classifier combinations to obtain at least one preliminary screening classifier combination includes: Acquire a plurality of category phrases to be processed, wherein each of the category phrases to be processed includes at least two category words to be processed; For each of the to-be-screened classifier word combinations, calculating the confidence of the to-be-screened classifier word combination with respect to the plurality of to-be-processed classifier word groups; Sorting the plurality of to-be-screened classifier word combinations in descending order of confidence to obtain a sorted sequence of to-be-screened classifier word combinations; The first n to-be-screened classifier word combinations of the to-be-screened classifier word combination sequence are intercepted to obtain at least one preliminary screening classifier word combination; wherein n is a positive integer.

3. The method according to claim 1, characterized in that The obtaining of the number of hits of the text data corresponding to each historical time period to each of the classifier sets includes: For each piece of text data in the plurality of pieces of text data corresponding to the historical time period, extracting a plurality of keywords from the text data; Obtaining a classifier corresponding to each of the keywords, thereby obtaining a plurality of classifiers corresponding to each piece of text data; If there is a text data corresponding to multiple classifiers that are all included in a classifier set, it is determined that the text data hits the classifier set, and the hit count of the classifier set is increased by one.

4. The method according to claim 3, characterized in that The obtaining of the classifier corresponding to each keyword includes: Natural language processing is performed on the keywords to obtain category words corresponding to the keywords.

5. A data identification device, characterized in that: The device comprises: A set acquisition unit is used to acquire multiple classifier sets, wherein each of the classifier sets includes at least two classifiers; the acquisition of multiple classifier sets includes: acquiring multiple classifiers; determining a classifier combination to be screened consisting of any two classifiers from the multiple classifiers; acquiring multiple classifier phrases to be processed, wherein each of the classifier phrases to be processed includes at least two classifiers to be processed, and the classifiers to be processed are used to implement the screening of the classifier combination to be screened; for each of the classifier combinations to be screened, calculating the number of hits of the classifier combination to be screened with respect to the multiple classifier phrases to be processed; for the multiple classifier combinations to be screened, eliminating the hits The classifier combinations to be screened with hit times higher than a first threshold value, and the classifier combinations to be screened with hit times lower than a second threshold value are eliminated to obtain at least one preliminary screening classifier combination; judging whether there are remaining classifiers, wherein the remaining classifiers are classifiers among the multiple classifiers other than the classifiers included in the preliminary screening classifier combination; if so, for each of the preliminary screening classifier combinations with remaining classifiers, any remaining classifier is added to obtain a new classifier combination to be screened, and jumping to the step: screening the multiple classifier combinations to be screened to obtain at least one preliminary screening classifier combination; if not, recording the at least one preliminary screening classifier combination as a classifier set; A hit count acquisition unit, used for acquiring, for multiple historical time periods, the hit count of the text data corresponding to each historical time period for each of the classifier sets; an abnormality determination unit, configured to determine that the first classifier set is an abnormal classifier set if a hit difference between the number of hits of the text data corresponding to the first historical time period to the first classifier set and the number of hits of the text data corresponding to the second historical time period to the first classifier set exceeds a preset threshold; wherein the first historical time period is any historical time period among the multiple historical time periods, the second historical time period is at least one historical time period that satisfies a preset time relationship with the first historical time period, and the first classifier set is any classifier set among the multiple classifier sets; The set display unit is used to display at least one abnormal category word set.

6. An electronic device, characterized in that: It comprises a processor and a memory, wherein the memory stores a plurality of instructions; the processor loads instructions from the memory to execute the steps in the data identification method according to any one of claims 1 to 4.

7. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a plurality of instructions, and the instructions are suitable for being loaded by a processor to execute the steps in the data identification method according to any one of claims 1 to 4.

Images