A data processing system for obtaining app tags

Abstract

The application relates to a data processing system for obtaining APP labels, and the system comprises a database, a processor and a memory storing a computer program, wherein the database comprises a first database and a second database, the first database comprises an initial APP list and an initial text list, the second database comprises a first sample APP set, a second sample APP set and a third sample APP set, and when the computer program is executed by the processor, the following steps are implemented: obtaining a target word vector list according to the first database, inputting the target word vector into a first neural network model to obtain a first intermediate label list, and marking according to the label quantity corresponding to the first intermediate label, so as to obtain the label corresponding to the APP. The application provides a new method for obtaining APP labels, sample labels are obtained through model training, and the sample labels are processed, so that the accuracy of the obtained APP labels is relatively high.

Description

Technical Field

[0001] This invention relates to the field of APP processing technology, and in particular to a data processing system for acquiring APP tags. Background Technology

[0002] With the popularization and development of smart technology, more and more application software (APP) is appearing on device application platforms. The continuous innovation and development of technology has led to an increasing number of application software types and a growing number of applications within each type. How to clearly classify and process the application software on device application platforms has become a hot research issue. By obtaining APP tags and accurately classifying APPs, more precise services can be provided to users.

[0003] Currently, existing technologies for obtaining APP classification involve: retrieving corresponding description information based on the APP's ID, inputting this description information into a model to generate vectors, retrieving a preset installation sequence based on the APP's ID to generate vectors, and inputting all generated vectors into a classification model to output the APP type. The problems with this method of obtaining APP type are as follows:

[0004] On the one hand, due to the limitations of the method, the simplistic APP tags result in low accuracy in APP classification;

[0005] On the other hand, due to the sheer number of apps, classifying only a specific portion of them cannot fully and automatically cover all apps, resulting in lower accuracy of the obtained app tags. Summary of the Invention

[0006] To address the aforementioned technical problems, the present invention adopts the following technical solution: a data processing system for acquiring APP tags, the system comprising: a first database, a processor, and a memory storing computer programs, wherein the first database includes: an initial APP list B = {B1, ..., B...} i , ..., B m The initial text list C corresponding to B is C = {C1, ..., C...} i , ..., C m}, B i Let C be the name of the i-th initial app. i For B i The corresponding initial APP text, i = 1...m, where m is the number of initial APPs, implements the following steps when the computer program is executed by the processor:

[0007] S100, according to B i and C i , obtain B i The corresponding target word vector list D i={B 0 i C 0 i1 , ..., C 0 ix , ..., C 0 ip(i)}, B 0 i For B i The corresponding first word vector, C 0 ix C i The corresponding x-th second word vector, x = 1...p(i), p(i) is C i The number of corresponding second word vectors.

[0008] S200, D i Input into the preset first neural network model to obtain B i The corresponding first intermediate tag list D 0 i .

[0009] S300, when D 0 i When =Null, B i The corresponding initial APP is used as the first APP.

[0010] S400, when D 0 i ≠Null and K 0 i ≥K 0 At that time, B i The corresponding initial APP is used as the first APP, where K 0 i D 0 i The corresponding number of first intermediate tags, K 0 This is a preset threshold for the number of tags.

[0011] S500, when D 0 i ≠Null and K 0 i <K 0 At that time, B i The corresponding initial app is used as the second app.

[0012] S600, perform a first marking process on the first APP to obtain the target tag corresponding to the first APP.

[0013] S700, perform a second marking process on the second APP to obtain the target tag corresponding to the second APP.

[0014] Compared with the prior art, the present invention has significant advantages. Through the above technical solution, the data processing system for obtaining APP tags provided by the present invention achieves considerable technological progress and practicality, and has broad industrial application value. It has at least the following advantages:

[0015] This invention provides a data processing system for obtaining APP tags. The system includes a first database, a processor, and a memory storing a computer program. The first database includes an initial APP list and an initial text list corresponding to the initial APP list. When the computer program is executed by the processor, the following steps are implemented: obtaining a list of target word vectors corresponding to the initial APPs based on the initial APP list and the initial text list corresponding to the initial APPs; inputting the target word vectors into a preset first neural network model to obtain a first intermediate label list corresponding to the initial APPs; processing the first intermediate labels to obtain a first APP and a second APP; and performing labeling processing on the first APP and the second APP respectively to obtain the target labels corresponding to the first APP and the target labels corresponding to the second APP, thereby obtaining the APP tags. It can be seen that, on the one hand, this invention obtains sample labels through model training and processes the sample labels, making the APP tags more complex and resulting in higher accuracy of the obtained APP tags; on the other hand, it is not limited to classifying specific APPs, but by using different processing methods for APPs, it can comprehensively and automatically cover all APPs, resulting in higher accuracy of the obtained APP tags. Attached Figure Description

[0016] Figure 1 This is a flowchart of the execution computer program of a data processing system for obtaining APP tags, provided as an embodiment of the present invention. Detailed Implementation

[0017] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0018] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or server that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or devices.

[0019] This invention provides a data processing system for acquiring APP tags. The system includes: a first database, a processor, and a memory storing computer programs, wherein the first database includes: an initial APP list B = {B1, ..., B...} i , ..., B m The initial text list C corresponding to B is C = {C1, ..., C...} i , ..., C m}, B i Let C be the name of the i-th initial app. i For B i The corresponding initial APP text, i = 1...m, where m is the number of initial APPs, when the computer program is executed by the processor, implements the following steps, such as... Figure 1 As shown:

[0020] S100, according to B i and C i , obtain B i The corresponding target word vector list D i ={B 0 i C 0 i1 , ..., C 0 ix , ..., C 0 ip(i)}, B 0 i For B i The corresponding first word vector, C 0 ix C i The corresponding x-th second word vector, x = 1...p(i), p(i) is C i The number of corresponding second word vectors.

[0021] Specifically, the initial text is the text used to describe the initial APP.

[0022] Specifically, the system also includes a second database.

[0023] Furthermore, the second database includes a first sample APP set H = {H1, ..., H}. j H n}, H j ={H j1 H jy H jq(j)}, H jy Let y be the information of the first sample APP corresponding to the j-th application platform, j = 1...n, where n is the number of APP application platforms, and y = 1...q(j), where q(j) is the number of first sample APPs corresponding to the j-th APP application platform. The information of the first sample APP includes the name of the first sample APP and the APP text of the first sample APP.

[0024] Specifically, the following steps are included before S100:

[0025] S1, iterate through H and obtain H from H. jy Corresponding key APP list U jy ={U 1 jy , ..., U e jy , ..., U f jy}, where U e jy For H jy The corresponding e-th key APP information, e = 1...f, where f is the number of key APPs, wherein the key APP information includes the key APP name and the APP text of the key APP, wherein each U e jy Key APP names and H jy The first sample app had the same name.

[0026] S2, according to U jy , get U jy The corresponding list of text character counts SL jy ={SL 1 jy , ...,SL e jy , ...,SL f jy}, SL e jy For U e jy The number of characters in the corresponding APP text.

[0027] S3. Traverse SL jy And when SL e jy is the SL jy with the largest number of characters in the APP text, insert U e jy into value B.

[0028] As described above, by obtaining the text with the largest number of literal characters corresponding to the key APP, the information of the obtained APP is more accurate, effectively determining the APP name and the corresponding APP text, and thus avoiding different texts describing the same APP due to different APP application platforms.

[0029] In a specific embodiment, before 100, the following steps are further included:

[0030] S101. Perform word segmentation on C i to obtain the first initial word list corresponding to C i Among them, those skilled in the art know that any word segmentation method in the prior art can be adopted according to actual needs, which will not be elaborated here.

[0031] S103. Delete the preset words in the preset word list from the first initial word list corresponding to C i to obtain the second initial word list corresponding to C<了 i Among them, those skilled in the art know that the preset word list is set according to actual needs. For example, the preset words are "de", "děi", etc., which will not be elaborated here.

[0032] S105. Input any second initial word in the second initial word list corresponding to C i into the word embedding model to obtain the second word vector.

[0033] Furthermore, the vector dimensions of B i and C i are the same. Those skilled in the art know that any word embedding model is determined according to actual needs to obtain the word vector corresponding to the initial APP name and the word vector corresponding to any word in the initial APP text, which will not be elaborated here.

[0034] Specifically, the second database further includes a second sample APP set, where each second sample APP information in the second sample APP set includes: the second sample APP name, the APP text of the second sample APP name, and the APP tag list corresponding to the second sample APP name.

[0035] S200. Input D i into the preset first neural network model to obtain B iThe corresponding first intermediate tag list D 0 i .

[0036] Specifically, the first neural network model is obtained in S200 through the following steps:

[0037] S201, Based on the second sample APP set, obtain the first sample word vector corresponding to the second sample APP set.

[0038] Specifically, the method for obtaining the first sample word vector is the same as the method for obtaining the target word vector.

[0039] As mentioned above, by using the same method to obtain word vectors, it can be ensured that the representation of the first sample word vector and the target word vector are consistent, providing effective data for subsequent APP classification and resulting in higher accuracy of the obtained APP tags.

[0040] S203, the first sample word vector set corresponding to the second sample APP set is used as a training dataset and input into a preset first initial neural network model to obtain the sample label set corresponding to the second sample APP set. The sample label set includes a list of first sample labels corresponding to the names of the second sample APPs. Those skilled in the art know the process of training a neural network model, so it will not be described in detail here.

[0041] S205, obtain the first neural network model based on the first sample label list corresponding to the second sample APP name and the APP label list corresponding to the second sample APP name.

[0042] The above describes how a neural network model is trained using the APP name and APP text to obtain the first neural network model, which can accurately determine the APP tag and thus accurately identify the APP type without text description.

[0043] S300, when D 0 i When =Null, B i The corresponding initial APP is used as the first APP.

[0044] S400, when D 0 i ≠Null and K 0 i ≥K 0 At that time, B i The corresponding initial APP is used as the first APP, where K 0 i D 0 i The corresponding number of first intermediate tags, K 0This is a preset threshold for the number of tags.

[0045] Specifically, the preset tag quantity threshold K 0 The value range of K is 1 to 3, where, as those skilled in the art know, K is determined according to actual needs. 0 The selection of [the appropriate element] will not be elaborated upon here.

[0046] As mentioned above, for apps with an excessive number of tags obtained from the first neural network model, the app tags are further determined to obtain more accurate tags, thereby accurately identifying the app type without text description.

[0047] S500, when D 0 i ≠Null and K 0 i <K 0 At that time, B i The corresponding initial app is used as the second app.

[0048] The above method can accurately determine the initial app tag of an app by using the initial app name and initial app text, and then accurately obtain the app type without text description based on the initial app tag.

[0049] S600, perform a first marking process on the first APP to obtain the target tag corresponding to the first APP.

[0050] Specifically, the first tagging process involves tagging the APP by a tagger.

[0051] S700, perform a second marking process on the second APP to obtain the target tag corresponding to the second APP.

[0052] Specifically, the S700 also includes the following steps:

[0053] S701, Based on the first intermediate tag list corresponding to any second APP, obtain the second intermediate tag list G = {G1, ..., G...} corresponding to the second APP. r , ..., G s}, G r Let r be the r-th intermediate tag corresponding to the second APP, where r = 1...s, and s is the number of intermediate tags corresponding to the second APP.

[0054] S703, Get the preset APP tag list G 0 ={G 0 1, ..., G 0 t , ..., G 0g}, G 0 t Let t be the t-th preset APP tag, t = 1...g, where g is the number of preset APP tags, and the preset APP tags are the preset tags corresponding to the APPs that represent virtual reality.

[0055] S705, when G r ≠G 0 t When the first intermediate tag corresponding to the second APP is used as the target tag corresponding to the second APP.

[0056] S707, when G r =G 0 t At that time, the target word vector list corresponding to the second APP is input into the preset second neural network model to obtain the third intermediate label list corresponding to the second APP.

[0057] Specifically, the second database also includes a third set of sample apps.

[0058] Furthermore, the information for each third sample APP in the third sample APP set includes: the name of the third sample APP, the APP text of the third sample APP name, and a list of APP tags corresponding to the third sample APP name, wherein the third sample APP is an APP representing virtual reality.

[0059] Specifically, the second neural network model is obtained in S707 through the following steps:

[0060] S7071, Based on the third sample APP set, obtain the second sample word vector corresponding to the third sample APP set.

[0061] Specifically, the method for obtaining the second sample word vector is the same as the method for obtaining the target word vector.

[0062] As mentioned above, by using the same method to obtain word vectors, it can be ensured that the representation of the second sample word vectors and the target word vectors are consistent, providing effective data for subsequent APP classification and resulting in higher accuracy of the obtained APP tags.

[0063] S7073, the second sample word vectors corresponding to the third sample APP set are used as training datasets and input into a preset second initial neural network model to obtain the second sample label set corresponding to the third sample APP set. The second sample label set includes a list of second sample labels corresponding to the names of the third sample APPs. Those skilled in the art know the process of training a neural network model, so it will not be described in detail here.

[0064] S7075, based on the second sample label list corresponding to the third sample APP name and the APP label list corresponding to the third sample APP name, obtain the second neural network model.

[0065] As mentioned above, the second neural network model obtained through model training can process data in parallel on a large scale and can autonomously learn to process different information, resulting in higher accuracy of the obtained APP tags.

[0066] S709, take the third intermediate tag in the third intermediate tag list corresponding to the second APP as the target tag corresponding to the second APP.

[0067] As mentioned above, for apps representing virtual reality, labels need to be obtained through separate neural network models. This avoids the inability to accurately determine the labels of such apps due to the use of the same neural network model, which would lead to the inability to accurately obtain the app type without text description. By performing further labeling processing on virtual reality apps, more accurate labels can be generated for virtual reality apps, thereby accurately obtaining the app type without text description.

[0068] Specifically, in the system, when the computer program is executed by the processor, the following steps are also performed:

[0069] S800, based on the target tags corresponding to the first APP and the second APP, obtain the target tag set D' = {D'1, ..., D''} corresponding to B. i , ..., D' m},D' i For B i The corresponding list of target tags.

[0070] The above method accurately obtains the app's tags through the app name and app text, enabling the app to be categorized based on the app's tags, thus giving most apps on the app application platform tags.

[0071] S900, based on D' i Obtain the final tag set D” = {D”1, ..., D”} corresponding to B. i , ..., D” m}, D” i For B i The corresponding final tag list.

[0072] Specifically, the S900 also includes the following steps:

[0073] S910, obtain the target APP in B, and the tagger performs tagging processing on the target APP.

[0074] Specifically, when DN > N0 and WP > N1, the APP is selected as the target APP, where DN is the number of APP installations, the number of installations is the number of devices that have the APP installed, WP is the APP retention rate, N0 is a preset installation threshold, and N1 is a preset retention rate threshold.

[0075] Specifically, WP meets the following conditions:

[0076] US M US1 represents the number of users who logged into the app at least once from the first day to the Mth day, and US1 represents the number of users who logged into the app for the first time on the first day.

[0077] Specifically, the preset installation threshold N0 ranges from 1×10-1. 8 ~2×10 8 As those skilled in the art will know, N0 is selected based on actual needs, and will not be elaborated further here.

[0078] Specifically, the preset retention rate threshold N1 ranges from 30% to 40%. As those skilled in the art know, N1 is selected according to actual needs, and will not be elaborated here.

[0079] Furthermore, when the text corresponding to the target app contains tags that are not included in the target tag, the tagger needs to perform tagging processing.

[0080] As mentioned above, manually labeling the target app by labelers ensures the accuracy of the app information and the comprehensiveness of the app tags, thus resulting in a high degree of accuracy in the obtained app tags.

[0081] Specifically, the system also includes a third database, and further, the third database includes a non-sample APP set, wherein the non-sample APP set is the set of APPs in the original APP set other than the sample APP set.

[0082] Specifically, when the computer program is executed by the processor, the following steps are performed:

[0083] S10, Obtain the first non-sample APP list FA = {FA1, ..., FA2} from the non-sample APP set. α , ..., FA λ The second non-sample APP list FB = {FB1, ..., FB2} β ..., FB γ}, FA α For the α-th non-sample APP information, α = 1...λ, where λ is the number of non-sample APPs. FB βLet β be the information of the second non-sample APP, where β = 1...γ, and γ is the number of second non-sample APPs.

[0084] Specifically, the first non-sample APP information includes the name of the first non-sample APP but does not include the APP text and pkg name of the first non-sample APP.

[0085] Specifically, the second non-sample APP information includes the pkg name of the second non-sample APP but does not include the APP text of the second non-sample APP or the name of the second non-sample APP.

[0086] S20, according to FA α and B i , obtain FA α The corresponding final tag list;

[0087] Specifically, S20 also includes the following steps:

[0088] S210, Based on FA and B, obtain the first similarity list F = {F1, ..., F2} corresponding to FA. α , ..., F λ}, F α ={F 1 α , ..., F i α , ..., F m α}, F i α For FA α and B i The first similarity between them.

[0089] Specifically, the following steps are included before S210:

[0090] S2101, Obtain FA α The corresponding first intermediate string.

[0091] S2103, Traverse FA α The corresponding first intermediate string and FA α Remove non-Chinese characters from the corresponding first intermediate string and obtain FA. α The corresponding second intermediate string.

[0092] S2105, according to FA α The corresponding second intermediate string, obtain FA. α The corresponding first intermediate word vector.

[0093] S2107, according to FA α The corresponding first intermediate word vector and B0 i , obtain F i α Those skilled in the art can determine the method for obtaining the similarity between vectors based on actual needs, such as cosine similarity between vectors, which will not be elaborated here.

[0094] The above steps involve removing non-Chinese words from the app to avoid affecting the accuracy of the similarity between app names. This allows for the accurate identification of such app tags, enabling the accurate acquisition of app types without text descriptions.

[0095] S230, when F i α The highest first similarity and F i α ≥F 0 At 1 o'clock, D' i As FA α The corresponding final tag list, where F 0 1 represents the preset first similarity threshold.

[0096] Specifically, the preset first similarity threshold F 0 The value of 1 ranges from 0.8 to 0.9. Those skilled in the art can set the first similarity threshold according to actual needs, which will not be elaborated here.

[0097] S250, when F i α The highest first similarity and F i α <F 0 At 1 o'clock, according to B i and C i , obtain FA α The corresponding final tag list.

[0098] Specifically, S250 also includes the following steps:

[0099] S2501, obtain C i The corresponding second initial word list.

[0100] S2503, according to C i The corresponding second initial word list, obtain C i The priority YX of any keyword in the corresponding second initial word list v i ,in, Where, η v i C i The vth second initial word in the corresponding second initial word list is in C i The number of times it appears in λv i C i The total number of times the v-th second initial word in the corresponding second initial word list appears in B and C, μ v i For in D' i In the corresponding target label B i The quantity.

[0101] S2507, when YX v i ≥YX 0 At that time, YX v i The corresponding second initial word is D' i The keywords, among which, YX 0 The priority threshold for keywords is known to those skilled in the art. The threshold for keyword priority is set according to actual needs, and will not be elaborated here.

[0102] S2509, according to D' i Keywords and FA α The similarity between them is used to obtain the FA. α The corresponding final tag list; can be understood as: D' i Keywords transformed into D' i The keyword vector, based on D' i Keyword vectors and FA α Obtain D' from the corresponding first intermediate word vector. i Keywords and FA α The similarity between them, when D' i Keywords and FA α When the similarity between them is not less than the first similarity threshold, based on D' i Construct FA by finding all target labels corresponding to the keywords in D'. α The corresponding final tag list.

[0103] The above approach avoids the inability to obtain tags for apps without text descriptions when the app name cannot be matched, thus improving the accuracy and scope of obtaining tags for apps without text descriptions.

[0104] S30, according to FB β and Q i Get FB β The corresponding final tag list.

[0105] Specifically, S30 also includes the following steps:

[0106] S310, Based on FB and Q, obtain the second similarity list F' = {F'1, ..., F'} corresponding to FB.β , ..., F' γ},F' β ={F' 1 β , ..., F' i β , ..., F' m β}, F' i β For FB β and Q i The second similarity between them.

[0107] Specifically, the method for obtaining the second similarity list is the same as the method for obtaining the first similarity list.

[0108] S330, when F' i β The second highest similarity and F' i β ≥F 0 At 2 o'clock, D' i As FB β The corresponding final tag list, where F 0 2 is the preset second similarity threshold.

[0109] Specifically, the preset second similarity threshold F 0 The value of 2 ranges from 0.8 to 0.9. Those skilled in the art can set a second similarity threshold according to actual needs, which will not be elaborated here.

[0110] S350, when F i α The second highest similarity and F i α <F 0 At 2 o'clock, no FB was generated. β The corresponding final tag list.

[0111] In another specific embodiment, the non-sample APP set further includes a third non-sample APP list, wherein the third non-sample APP information includes the name of the third non-sample APP and the pkg name of the third non-sample APP, but does not include the APP text. When the computer program is executed by the processor, the following steps are implemented:

[0112] S'1, Based on the name of the third non-sample APP and B, determine the final tag list corresponding to the third non-sample APP. The method for determining the final tag list corresponding to the third non-sample APP based on the name of the third non-sample APP and B is the same as that in S40, and will not be described again here.

[0113] S'2, when the name of the third non-sample APP is the same as D' i When the similarity between keywords is less than the first similarity threshold, the final tag list corresponding to the third non-sample APP is determined based on the pkg name and Q of the third non-sample APP. The method for determining the final tag list corresponding to the third non-sample APP based on the pkg name and Q of the third non-sample APP is the same as that in S50, and will not be described again here.

[0114] As mentioned above, by using different processing methods for different apps, it is possible to fully and automatically cover all apps, resulting in a high degree of accuracy in obtaining the app types.

[0115] S40, according to the FA α The corresponding final tag list and the FB β The corresponding final tag list, obtain FA α Corresponding APP types and FB β The corresponding APP type.

[0116] The above describes how, by obtaining tags for apps with text descriptions, apps without text descriptions can be identified. Based on these tags, apps on the app application platform can be accurately categorized, allowing more apps to be classified into more accurate types and making it easier for users to find them.

[0117] This invention provides a data processing system for obtaining app tags. The system includes a first database, a processor, and a memory storing a computer program. The first database includes an initial app list and an initial text list corresponding to the initial app list. When the computer program is executed by the processor, the following steps are implemented: obtaining a list of target word vectors corresponding to the initial apps based on the initial app list and the initial text list corresponding to the initial apps; inputting the target word vectors into a preset first neural network model to obtain a first intermediate label list corresponding to the initial apps; processing the first intermediate labels to obtain a first app and a second app; performing labeling processing on the first app and the second app respectively to obtain the target labels corresponding to the first app and the second app, thereby obtaining the app tags. This invention provides a data processing system for obtaining app tags. On the one hand, by obtaining sample labels through model training and processing the sample labels, the app tags are made more complex, resulting in higher accuracy of the obtained app tags. On the other hand, it is not limited to classifying specific parts of apps; by using different processing methods for apps, it can comprehensively and automatically cover all apps, resulting in higher accuracy of the obtained app tags.

[0118] While specific embodiments of the invention have been described in detail by way of example, those skilled in the art should understand that the examples are for illustrative purposes only and not intended to limit the scope of the invention. It should also be understood that various modifications can be made to the embodiments without departing from the scope and spirit of the invention. The scope of the invention is defined by the appended claims.

Claims

1. A data processing system for obtaining an APP tag, characterized by The system includes: a first database, a second database, a processor, and a memory storing computer programs, wherein the first database includes: an initial APP list B = {B1, ..., B...} i , ..., B m The initial text list C = {C1, ..., C} corresponding to B is also given. i , ..., C m }, B i Let C be the name of the i-th initial app. i For B i The corresponding initial APP text, i=1……m, where m is the number of initial APPs, when the computer program is executed by the processor, implements the following steps: S100, according to B i and C i , obtain B i The corresponding target word vector list D i ={B 0 i C 0 i1 , ..., C 0 ix , ..., C 0 ip（i） }, B 0 i For B i The corresponding first word vector, C 0 ix C i The corresponding x-th second word vector, x=1……p(i), p(i) is C i The number of corresponding second word vectors; S200, the D i input into a preset first neural network model, obtaining B i The corresponding first intermediate label list D 0 i ; S300, when D 0 i =Null, B i The corresponding initial APP is taken as the first APP; S400, when D 0 i ≠Null and K 0 i ≥K 0 At that time, B i The corresponding initial APP is used as the first APP, where K 0 i D 0 i The corresponding number of first intermediate tags, K 0 This is a preset threshold for the number of tags; S500, when D 0 i ≠ Null and K 0 i < K 0 , B i corresponding to the initial APP as the second APP; S600, perform a first marking process on the first APP to obtain the target tag corresponding to the first APP; S700, perform a second tagging process on the second APP to obtain the target tag corresponding to the second APP; S700 also includes the following steps: S701, based on the first intermediate tag list corresponding to any second APP, obtain the second intermediate tag list G={G1, ..., G...} corresponding to the second APP. r , ..., G s }, G r Let r be the r-th intermediate tag corresponding to the second APP, where r = 1...s, and s is the number of intermediate tags corresponding to the second APP; S703, Get the preset APP tag list G 0 ={G 0 1, ..., G 0 t , ..., G 0 g }, G 0 t Let t be the t-th preset APP tag, t=1……g, where g is the number of preset APP tags, and the preset APP tags are the preset tags corresponding to the APPs that represent virtual reality; S705, when G r ≠G 0 t the first intermediate label corresponding to the second APP as the target label corresponding to the second APP; S707, when G r =G 0 t the target word vector list corresponding to the second APP is input into a preset second neural network model to obtain a third intermediate label list corresponding to the second APP. S709, take the third intermediate tag in the third intermediate tag list corresponding to the second APP as the target tag corresponding to the second APP; The second database further includes a third sample app set, wherein each third sample app in the third sample app set includes: the name of the third sample app, the app text of the third sample app name, and a list of app tags corresponding to the third sample app name. The third sample app is an app representing virtual reality. When the computer program is executed by the processor, the second neural network model is obtained in S707 through the following steps: S7071, Based on the third sample APP set, obtain the second sample word vector corresponding to the third sample APP set; S7073, the second sample word vectors corresponding to the third sample APP set are used as training datasets and input into a preset second initial neural network model to obtain the second sample label set corresponding to the third sample APP set. The second sample label set includes a list of second sample labels corresponding to the names of the third sample APPs. S7075, based on the second sample label list corresponding to the third sample APP name and the APP label list corresponding to the third sample APP name, obtain the second neural network model.

2. The data processing system for acquiring APP tags according to claim 1, wherein, The initial text is the text used to describe the initial APP.

3. The data processing system for acquiring APP tags according to claim 1, characterized in that, The second database includes a first sample APP set H = {H1, ..., H2}. j H n }, H j ={H j1 H jy H jq（j） }, H jy The information for the y-th first sample APP corresponding to the j-th application platform is given, where j = 1...n, n is the number of APP application platforms, and y = 1...q(j), where q(j) is the number of first sample APPs corresponding to the j-th APP application platform. The first sample APP information includes the first sample APP name and the APP text of the first sample APP. When the computer program is executed by the processor, the following steps are performed before S100: S1, iterate through H and obtain H from H. jy Corresponding key APP list U jy ={U 1 jy , ..., U e jy , ..., U f jy }, where U e jy For H jy The corresponding e-th key APP information, e=1……f, where f is the number of key APPs, wherein the key APP information includes the key APP name and the APP text of the key APP, wherein each U e jy Key APP names and H jy The first sample app had the same name; S2, according to U jy , get U jy The corresponding list of text character counts SL jy ={SL 1 jy , ...,SL e jy , ...,SL f jy }, SL e jy For U e jy The corresponding number of characters in the APP text; S3, traversing SL jy And when SL e jy For SL jy When determining the maximum number of characters in the text of an app, U e jy Insert value B.

4. The data processing system for acquiring APP tags according to claim 1, characterized in that, The following steps are included before S100: S101, to C i word segmentation processing, obtaining C i corresponding first initial word list; S103, from C i The corresponding first initial word list deletes the preset word in the preset word list, and obtains C i The corresponding second initial word list; S105, C i Any second initial word in the corresponding second initial word list is input into the word embedding model to obtain a second word vector.

5. The data processing system for acquiring APP tags according to claim 3, characterized in that, The second database also includes a second sample app set, wherein the information of each second sample app in the second sample app set includes: the name of the second sample app, the app text of the second sample app name, and a list of app tags corresponding to the second sample app name. When the computer program is executed by the processor, the first neural network model is obtained in S200 through the following steps: S201, Based on the second sample APP set, obtain the first sample word vector corresponding to the second sample APP set; S203, the first sample word vector set corresponding to the second sample APP set is used as a training dataset and input into a preset first initial neural network model to obtain the sample label set corresponding to the second sample APP set, wherein the sample label set includes a list of first sample labels corresponding to the second sample APP name; S205, obtain the first neural network model based on the first sample label list corresponding to the second sample APP name and the APP label list corresponding to the second sample APP name.

6. The data processing system for acquiring APP tags according to claim 1, wherein, The preset tag quantity threshold K 0 The value range is 1 to 3.

7. The data processing system for acquiring APP tags according to claim 1, wherein, The first tagging process involves tagging the APP by a tagger.

8. The data processing system for acquiring APP tags according to claim 1, wherein, In the system, when the computer program is executed by the processor, the following steps are also performed: S800, based on the target tags corresponding to the first APP and the second APP, obtain the target tag set D'={D'1, ..., D''} corresponding to B. i , ..., D' m },D' i For B i The corresponding list of target tags; S900, based on D' i , obtaining the final label set D'' corresponding to B i ,..., D'' m}, D'' i is the final label list corresponding to B i .

Images