Data processing method and device, computer equipment, storage medium and program product

Abstract

The embodiment of the application provides a data processing method and device, computer equipment, a storage medium, and a program product. The method comprises the following steps: recalling content field data for representing the content field of media content recalled by historical query data matched with query intention information of target query data; publishing content field data for representing the content field of media content published by a publishing account of candidate media content of the target query data; performing matching processing on the recalled content field data and the publishing content field data of the publishing account of the candidate media content, to obtain the content field matching degree between the target query data and the publishing account of the candidate media content; and determining the recall result of the target query data in the candidate media content based on the content field matching degree. By using the embodiment of the application, media content published by a publishing account with a publishing content field matched with query data (i.e., a field-related account result of query data) can be recalled.

Description

Technical Field

[0001] This application relates to the field of computer technology, and more particularly to the field of intelligent search technology, specifically to a data processing method, a data processing device, a computer equipment, a computer-readable storage medium, and a computer program product. Background Technology

[0002] The essence of a search engine is to return media content (e.g., articles, videos, audio, images, etc.) that matches a given query (e.g., a search statement). Currently, search engines focus more on the matching degree between query data and media content, but neglect the matching degree between the query data and the publishing accounts of the media content. This can lead to mismatches between the content areas of the returned media content and the query data. For example, if a query seeks media content in the travel sector, the returned media content may indeed be in the travel sector, but the publishing account may primarily publish content in the food sector, while authoritative travel accounts might publish the media content the query seeks. Therefore, how to retrieve media content published by accounts whose content areas match the query data (i.e., how to retrieve results from accounts related to the query data's domain) has become a current research hotspot. Summary of the Invention

[0003] This application provides a data processing method, apparatus, computer equipment, storage medium, and program product that can recall media content published by publishing accounts whose published content domain matches the query data (i.e., recall domain-related account results of the query data).

[0004] On one hand, embodiments of this application provide a data processing method, which includes:

[0005] Retrieve candidate media content that matches the target query data;

[0006] Acquire recall content domain data that matches the query intent information of the target query data, and acquire publication content domain data of the publishing accounts of the candidate media content; wherein, the recall content domain data is used to characterize the content domain of the media content recalled by the historical query data that matches the query intent information; the publication content domain data is used to characterize the content domain of the media content published by the publishing accounts of the candidate media content;

[0007] The retrieved content domain data will be matched with the content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts.

[0008] Based on the degree of content domain matching, the recall results of the target query data are determined from the candidate media content.

[0009] Accordingly, embodiments of this application provide a data processing apparatus, which includes:

[0010] The acquisition unit is used to acquire candidate media content that matches the target query data;

[0011] The acquisition unit is also used to acquire recall content domain data that matches the query intent information of the target query data, and to acquire the publication content domain data of the publishing account of the candidate media content; wherein, the recall content domain data is used to characterize the content domain of the media content recalled by the historical query data that matches the query intent information; the publication content domain data is used to characterize the content domain of the media content published by the publishing account of the candidate media content;

[0012] The processing unit is used to match the recalled content domain data with the published content domain data of the candidate media content's publishing account to obtain the content domain matching degree between the target query data and the published account of the candidate media content.

[0013] The processing unit is also used to determine the recall results of the target query data in the candidate media content based on the degree of content domain matching.

[0014] In one implementation, the recall content domain data includes recall content type information, which includes: the viewing content type of the media content recalled by historical query data, and the viewing content ratio corresponding to each viewing content type; the publishing content domain data includes publishing content type information, which includes: the publishing content type and the publishing content ratio corresponding to each publishing content type.

[0015] The processing unit, used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts, and to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content, specifically performs the following steps:

[0016] The recall content type information is matched with the published content type information to determine the matching content type in the recall content type information and the published content type information;

[0017] Based on the proportion of viewed content and the proportion of published content corresponding to the matching content type, determine the degree of content domain matching between the target query data and the publishing accounts of the candidate media content.

[0018] In one implementation, the number of matched content types is multiple; the processing unit, when determining the content domain matching degree between the target query data and the publishing account of the candidate media content based on the viewing content ratio and publishing content ratio corresponding to the matched content type, specifically performs the following steps:

[0019] The matching degree information for each content type is calculated based on the proportion of viewed content and the proportion of published content for each matching content type.

[0020] Based on the matching degree information of each matching content type, the content domain matching degree between the target query data and the publishing account of the candidate media content is calculated; or, the maximum matching degree information among the matching degree information of each matching content type is determined as the domain matching degree between the target query data and the publishing account of the candidate media content.

[0021] In one implementation, the recall content domain data includes recall content representation information, which includes recall content representation terms; the publication content domain data includes publication content representation information, which includes publication content representation terms.

[0022] The processing unit, used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts, and to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content, specifically performs the following steps:

[0023] By performing similarity matching between the recall content representation information and the published content representation information, the content domain matching degree between the target query data and the publishing accounts of the candidate media content is obtained.

[0024] In one implementation, the recall content representation information also includes representation importance information for each recall content representation word;

[0025] The processing unit, used to perform similarity matching between the recall content representation information and the published content representation information to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content, specifically performs the following steps:

[0026] Determine the similarity information between each recall content characterization term in the recall content characterization information and the published content characterization information;

[0027] Based on the similarity information between each recall content characterization term and the published content characterization information, as well as the characterization importance information of each recall content characterization term, the content domain matching degree between the target query data and the publishing account of the candidate media content is determined.

[0028] In one implementation, the query intent information includes query type and query semantics; the historical query data includes: first historical query data matching the query type and second historical query data matching the query semantics; the recalled content domain data includes: recalled content type information of the viewed media content recalled by the first historical query data and recalled content representation information of the media content recalled by the second historical query data; and the published content domain data includes published content type information and published content representation information.

[0029] The processing unit, used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts, and to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content, specifically performs the following steps:

[0030] The type information of the recalled content is matched with the type information of the published content to obtain type matching information;

[0031] The recall content representation information is matched with the published content representation information to obtain representation matching information;

[0032] Based on type matching information and representation matching information, determine the degree of content domain matching between the target query data and the publishing accounts of candidate media content.

[0033] In one implementation, the query intent information is the query type of the target query data; the recall content domain data corresponding to the target query data includes recall content type information; the acquisition unit, when acquiring recall content domain data that matches the query intent information of the target query data, specifically performs the following steps:

[0034] Identify the query type of the target query data and determine that the query type of the target query data is the target query type;

[0035] The type of retrieved content corresponding to the target query type is determined as the type of retrieved content corresponding to the target query data.

[0036] In one implementation, the process of determining the type of retrieved content corresponding to the target query type includes:

[0037] Retrieve the viewed media content from the historical query data retrieved under the target query type;

[0038] The acquired media content is categorized to determine the types of content to be viewed;

[0039] Count the number of media contents under each content viewing type and determine the proportion of content viewed for each content viewing type;

[0040] The content type information for the target query type includes: the content type viewed and the proportion of content viewed for each content type.

[0041] In one implementation, the query intent information is the query semantics of the target query data; the recall content domain data corresponding to the target query data includes recall content representation information; the acquisition unit, when acquiring recall content domain data that matches the query intent information of the target query data, specifically performs the following steps:

[0042] Obtain reference historical query data that matches the query semantics of the target query data;

[0043] The recall content representation information corresponding to the historical query data will be used to determine the recall content representation information corresponding to the target query data.

[0044] In one implementation, the process of determining the recall content representation information corresponding to historical query data includes:

[0045] Retrieve media content retrieved by referencing historical query data;

[0046] Extract keywords from the acquired media content and determine the importance information of the keywords;

[0047] Keywords that meet the importance criteria are identified as recall content representation words;

[0048] The recall content representation information corresponding to the historical query data includes: recall content representation terms, and representation importance information corresponding to each recall content representation term.

[0049] In one implementation, when the processing unit determines the recall results of the target query data among candidate media content based on the content domain matching degree, it specifically performs the following:

[0050] The candidate media content is sorted from highest to lowest according to the degree of content domain matching, and the sorted candidate media content is used as the recall result of the target query data.

[0051] Alternatively, candidate media content with a content domain matching degree higher than the matching threshold can be used as the recall result of the target query data.

[0052] Accordingly, embodiments of this application provide a computer device, which includes:

[0053] A processor is a tool for implementing computer programs.

[0054] A computer-readable storage medium storing a computer program adapted to be loaded by a processor and executed by the above-described data processing method.

[0055] Accordingly, embodiments of this application provide a computer-readable storage medium storing a computer program, which, when read and executed by a processor of a computer device, causes the computer device to perform the aforementioned data processing method.

[0056] Accordingly, embodiments of this application provide a computer program product or computer program, which includes computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the data processing method described above.

[0057] In this embodiment, the recalled content domain data can be used to characterize the content domain of media content recalled by the relevant historical query data of the target query data, and the published content domain data can be used to characterize the content domain of media content published by the publishing accounts of the candidate media content of the target query data. By matching the recalled content domain data with the published content domain data, the degree of content domain matching between the target query data and the publishing accounts of the candidate media content can be determined. Therefore, based on the degree of content domain matching between the target query data and the publishing accounts of the candidate media content, media content published by publishing accounts whose published content domain matches the target query data can be recalled from the candidate media content as the recall result of the target query data, that is, the domain-related account results of the target query data can be recalled. Attached Figure Description

[0058] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0059] Figure 1 This is a schematic diagram illustrating an application scenario of an intelligent search technology provided in an embodiment of this application;

[0060] Figure 2 This is a schematic diagram of the architecture of a data processing system provided in an embodiment of this application;

[0061] Figure 3 This is a flowchart illustrating a data processing method provided in an embodiment of this application;

[0062] Figure 4 This is a schematic diagram illustrating the matching of query data with the publishing account of media content, provided in an embodiment of this application.

[0063] Figure 5 This is a flowchart illustrating another data processing method provided in an embodiment of this application;

[0064] Figure 6 This is a flowchart illustrating another data processing method provided in an embodiment of this application;

[0065] Figure 7 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application;

[0066] Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0067] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0068] This application relates to intelligent search technology. Intelligent search technology refers to a technology where, given query data (e.g., a search query), a search engine can return media content that matches the query data. Here, matching the query data with the media content means that the media content is semantically related to the query data, i.e., the similarity between the meaning expressed by the media content and the semantics of the query data is higher than a similarity threshold. The media content mentioned in this application may include, but is not limited to, any of the following: articles, videos, audio, images, etc. Figure 1 The diagram illustrates an intelligent search scenario where the query data 101 input by the query object is "strawberry cake style". The search engine returns a media content set 102 to the query object. For example, the media content set 102 includes media content titled "Complete Collection of Strawberry Cake Styles", media content titled "One Hundred Styles of Strawberry Cake", and so on.

[0069] By using intelligent search technology, candidate media content that matches the query data can be obtained. Based on this, this application proposes a data processing scheme. This data processing scheme determines the degree of content domain matching between the query data and the content publishing domain of the publishing accounts of the candidate media data by matching the query intent of the query data with the content publishing domain of the publishing accounts of the candidate media data. Thus, based on the degree of content domain matching, domain-related account results can be identified in the candidate media content as the recall results of the query data.

[0070] The following is combined Figure 2 This paper introduces a data processing system suitable for implementing the data processing scheme.

[0071] like Figure 2 As shown, the data processing system may include terminal device 201 and server 202. This application embodiment does not limit the number of terminal devices 201; the number of terminal devices 201 can be one or more. Terminal device 201 can be a smartphone, tablet, laptop, desktop computer, smart voice interaction device, smartwatch, vehicle terminal, smart home appliance, aircraft, etc., but is not limited to these. Server 202 can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. This application embodiment does not limit this. Terminal device 201 and server 202 can establish a direct communication connection via wired communication or an indirect communication connection via wireless communication. This application embodiment does not limit this.

[0072] Terminal device 201 runs a media content search application with media content search capabilities, and server 202 deploys a search engine for the media content search application. In this data processing system, the media content search process can be as follows: the query object inputs query data through the media content search application; the media content search application sends the query data to the search engine on server 202 through terminal device 201; after determining the domain-related account results for the query data, the search engine on server 202 feeds back the domain-related account results as the query data recall results to the media content search application; and the media content search application presents the query data recall results to the query object.

[0073] It is understood that the data processing system described in the embodiments of this application is for the purpose of more clearly illustrating the technical solutions of the embodiments of this application, and does not constitute a limitation on the technical solutions provided in the embodiments of this application. As those skilled in the art will know, with the evolution of system architecture and the emergence of new business scenarios, the technical solutions provided in the embodiments of this application are also applicable to similar technical problems.

[0074] The data processing scheme provided in the embodiments of this application will be described in more detail below with reference to the accompanying drawings.

[0075] This application provides a data processing method, which mainly describes the process of matching the query intent of query data with the publishing content domains of candidate media content publishing accounts from the perspective of media content type and / or media content characteristic terms, in order to determine the domain-related account results of the query data among the candidate media content. This data processing method can be executed by a computer device, which can be the aforementioned... Figure 2 Server 202 in the data processing system shown. (As...) Figure 3 As shown, the data processing method may include, but is not limited to, the following steps S301-S304:

[0076] S301, retrieve candidate media content that matches the target query data.

[0077] The database stores media content published by each publishing account among all registered publishing accounts in the media content search application. Upon receiving target query data input by a queryee through the application, the database retrieves candidate media content matching the target query data. The number of candidate media content can be multiple (i.e., two or more). Here, "matching the target query data with candidate media content" means that the candidate media content is semantically related to the target query data; that is, the similarity between the meaning expressed by the candidate media content and the semantics of the target query data is higher than a similarity threshold. More specifically, this could mean that the meaning expressed by the title of the candidate media content is semantically related to the target query data, or that the meaning expressed by the body text of the candidate media content is semantically related to the target query data.

[0078] S302, obtain recall content domain data that matches the query intent information of the target query data, and obtain the published content domain data of the publishing accounts of the candidate media content.

[0079] After obtaining candidate media content that matches the target query data, it is possible to obtain recall content domain data that matches the query intent information of the target query data; it is also possible to obtain the account identifier of the publishing account of the candidate media content, and thus, based on the account identifier of the publishing account of the candidate media content, obtain the publishing content domain data of the publishing account. Specifically, the query intent information of the target query data can be used to characterize the query intent of the target query data; the recall content domain data can be used to characterize the content domain of the media content recalled by the historical query data that matches the query intent information; and the publishing content domain data can be used to characterize the content domain of the media content published by the publishing account of the candidate media content.

[0080] S303, retrieve the content domain data and match it with the content domain data of the publishing accounts of the candidate media content to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content.

[0081] After obtaining the recall content domain data and the published content domain data, the recall content domain data can be matched with the published content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts. Specifically, the content domain matching degree refers to the degree of matching between the content domain of the media content published by the candidate media content's publishing account and the query intent of the target query data. A higher content domain matching degree indicates a higher degree of matching between the content domain of the media content published by the candidate media content's publishing account and the query intent of the target query data, making the candidate media content more suitable as a recall result for the target query data. Conversely, a lower content domain matching degree indicates a lower degree of matching between the content domain of the media content published by the candidate media content's publishing account and the query intent of the target query data, making the candidate media content less suitable as a recall result for the target query data.

[0082] The recall content domain data and the release content domain data can be matched from any of three perspectives: media content type, media content characteristic words, and a combination of media content types and media content characteristic words. These three perspectives are described in detail below:

[0083] (1) Matching perspective of media content type:

[0084] From the perspective of media content type matching, the query intent information of the target query data can refer to the query type of the target query data, that is, the content domain data of the retrieved content matching the query type of the target query data. The retrieved content domain data can be used to characterize the content domain of the media content retrieved by historical query data that matches the query type of the target query data. The query type of the target query data can be obtained by identifying the target query data using a pre-trained query data classification model. More specifically, the retrieved content domain data can include retrieved content type information, which can include: the viewed content type of the media content retrieved by historical query data that matches the query type of the target query data, and the viewing content ratio corresponding to each viewed content type. The viewing content ratio corresponding to any viewed content type refers to the ratio between the number of viewed media content belonging to that viewed content type and the total number of viewed media content. The retrieved content type information can exist in the form of a mapping list of <the query type of the target query data, <viewed content type 1: the viewing content ratio corresponding to viewed content type 1; viewed content type 2: the viewing content ratio corresponding to viewed content type 2; ...>>.

[0085] Similarly, from the perspective of media content type matching, the content domain data of any candidate media content's publishing account can include content type information. This content type information can include: the content types published by the candidate media content's publishing account, and the proportion of each content type. The proportion of any content type refers to the ratio between the number of media content belonging to that content type published by the candidate media content's publishing account and the total number of media content published by the candidate media content's publishing account. Content type information can exist in the form of a mapping list of <account identifier of the candidate media content's publishing account, <content type 1: proportion of content corresponding to content type 1; content type 2: proportion of content corresponding to content type 2; ...>>.

[0086] The matching process from the perspective of media content type matching can include: matching the recalled content type information with the published content type information to determine the matching content type in the recalled content type information and the published content type information; then, based on the viewing ratio and publishing ratio of the matching content type, determining the degree of content domain matching between the target query data and the publishing account of the candidate media content; the number of matching content types can be one or more.

[0087] When there is only one content type to match, the content domain matching degree between the target query data and the publishing account of the candidate media content can be calculated based on the viewing ratio and the publishing ratio of the matching content type. The content domain matching degree between the target query data and the publishing account of the candidate media content can be the result of a weighted sum of the viewing ratio and the publishing ratio of the single matching content type. For example, the formula for the weighted sum is "content domain matching degree = α × viewing ratio + β × publishing ratio", where α and β are hyperparameters that can be adjusted according to actual business needs, for example, α = 1, β = 1.

[0088] When there are multiple content types to match, the matching degree information for each content type can be calculated based on the viewing ratio and the publishing ratio of each content type. The matching degree information for any content type can be the result of a weighted sum of the viewing ratio and the publishing ratio of that content type. For example, the formula for the weighted sum is "matching degree information = α × viewing ratio + β × publishing ratio", where α and β are hyperparameters that can be adjusted according to actual business needs, such as α = 1 and β = 1. Then, based on the matching degree information of each content type, the content domain matching degree between the target query data and the publishing account of the candidate media content can be calculated. The content domain matching degree can be equal to the sum of the matching degree information of each content type; or, the maximum matching degree information among the matching degree information of each content type can be determined as the content domain matching degree.

[0089] Figure 4This example illustrates a matching perspective based on media content type. The target query data is for the "travel" category. Among the historical query data for "travel," 40% of the viewed media content belonged to the "travel guide" category, 35% to the "tourist attraction recommendation" category, and 25% to the "tourism group recommendation" category. Conversely, among the media content published by an account that matched the target query data, 50% belonged to the "travel guide" category, 30% to the "travel transportation recommendation" category, and the remaining 25% to the "tourist attraction recommendation" category. The proportion is 20%; then, the matching content types are "travel guide" and "tourist attraction recommendation". When α=1 and β=1, the matching degree information corresponding to the matching content type "travel guide" is 40%+50%=90%, and the matching degree information corresponding to the matching content type "tourist attraction recommendation" is 35%+20%=55%. Therefore, the content domain matching degree between the target query data and the publishing account of the candidate media content can be equal to the sum of the matching degree information of the two matching content types, which is 145%, or it can be equal to the maximum matching degree information of the two matching content types, which is 90%.

[0090] The content in (1) above measures the media content domain by media content type. On the one hand, it takes into account the viewing habits of the historical viewing objects in the media content recalled by the historical query data of the target query data. This can reflect the query intent of the target query data from the perspective of historical statistics. On the other hand, it takes into account the content type publishing situation of the publishing accounts of the candidate media content. Thus, the content domain matching degree between the target query data and the publishing accounts of the candidate media content can be calculated from the perspective of media content type matching.

[0091] (2) Matching angle of media content representation words:

[0092] From the perspective of matching media content representation terms, the query intent information of the target query data can refer to the query semantics of the target query data, that is, the recall content domain data matches the query semantics of the target query data. The recall content domain data can be used to represent the content domain of the media content recalled by historical query data that matches the query semantics of the target query data. More specifically, the recall content domain data can include recall content representation information, which can include: recall content representation terms (or recall content morphemes, i.e., recall content terms) in the media content recalled by historical query data that matches the query semantics of the target query data, and the representation importance information of each recall content representation term (here, representation importance information can refer to the IDF (Inverse Document Frequency) value of the recall content representation term). The recall content representation terms can be used to represent the meaning expressed by the recalled media content, and the representation importance information of any recall content representation term is used to represent the importance of the recall content representation term in the recalled media content.

[0093] Similarly, from the perspective of matching media content representation terms, the publication domain data of any candidate media content's publishing account can include publication content representation information. This publication content representation information can include: publication content representation terms (or publication content morphemes, i.e., publication content terms) of the media content published by the candidate media content's publishing account, and representation importance information for each publication content representation term (similarly, representation importance information here can refer to the IDF value of the publication content representation term). Publication content representation terms are used to represent the meaning expressed by the published media content, and the representation importance information of any publication content representation term can be used to represent the degree of importance of that publication content representation term in the published media content.

[0094] The matching process from the perspective of media content representation matching can include: performing similarity matching between the recalled content representation information and the published content representation information to obtain the content domain matching degree between the target query data and the publishing accounts of the candidate media content. The similarity matching can be performed using a similarity algorithm. This embodiment uses the BM (Best Matching)25 algorithm as an example. BM25 is commonly used to evaluate the similarity between a query and a document. Its main idea is: parsing the query into terms to generate a set of terms; then calculating the similarity score between each term and the document; finally, weighted summing of the similarity scores between each term and the document to obtain the similarity score between the query and the document.

[0095] Introducing BM25 into the similarity matching process of this application embodiment yields the following: either the recalled content representation information or the published content representation information can be used as a query, and the other as a document. For example, the recalled content representation information can be used as a query, and the published content representation information as a document. Thus, the similarity information between each recalled content representation term in the recalled content representation information and the published content representation information can be determined. Then, based on the similarity information between each recalled content representation term in the recalled content representation information and the published content representation information, as well as the representation importance information of each recalled content representation term, the content domain matching degree between the target query data and the publishing account of the candidate media content can be determined. More specifically, the representation importance information of each recalled content representation term can be used as a weight to perform a weighted summation of the similarity information between each recalled content representation term in the recalled content representation information and the published content representation information, thereby obtaining the content domain matching degree between the target query data and the publishing account of the candidate media content.

[0096] The content in (2) above measures the media content domain using media content representation words. On the one hand, it considers the representation words that can represent the meaning expressed by the recalled media content among the historical query data of the target query data. On the other hand, it considers the representation words that can represent the meaning expressed by the media content published by the publishing account of the candidate media content. Thus, the content domain matching degree between the target query data and the publishing account of the candidate media content can be calculated from the perspective of matching media content representation words. Compared with (1) above, the matching perspective of media content type in (1) above can be understood as a coarse-grained matching perspective, while the matching perspective of media content representation words in (2) above can be understood as a fine-grained matching perspective.

[0097] (3) Matching angles of media content type combinations and media content characteristic terms:

[0098] The matching angle of media content type combination and media content characteristic words is combined. In the matching angle of media content type combination and media content characteristic words, the query intent information of the target query data can include the query type and query semantics of the target query data. Historical query data can include: first historical query data that matches the query type of the target query data and second historical query data that matches the query semantics of the target query data. The recalled content domain data can include: the recalled content type information of the viewed media content in the media content recalled by the first historical query data (including the viewed content type and the viewing content ratio corresponding to each viewed content type), and the recalled content characteristic information of the media content recalled by the second historical query data (including the recalled content characteristic words and the characteristic importance information of each recalled content characteristic word). The published content domain data of any candidate media content's publishing account can include: the published content type information of the media content published by the candidate media's publishing account (including the published content type and the published content ratio corresponding to each published content type), and the published content characteristic word information (including the published content characteristic words and the characteristic importance information of each published content characteristic word).

[0099] The matching process from the perspective of matching media content type combination media content representation words can include: matching the recalled content type information with the published content type information to obtain type matching information. This process is the same as the matching process from the perspective of matching media content type in (1) above. For details, please refer to the matching process from the perspective of matching media content type in (1) above. It will not be repeated here. Also, matching the recalled content representation information with the published content representation information to obtain representation matching information. This process is the same as the matching process from the perspective of matching media content representation words in (2) above. For details, please refer to the matching process from the perspective of matching media content representation words in (2) above. It will not be repeated here. Then, based on the type matching information and the representation matching information, the content domain matching degree between the target query data and the publishing account of the candidate media content can be determined. More specifically, the content domain matching degree can be the result of weighted summation of type matching information and representation matching information, or it can be the product of type matching information and representation matching information. This application embodiment does not limit this.

[0100] The content in (3) above measures the media content domain by combining media content type and media content characteristic words. By combining coarse-grained and fine-grained matching perspectives, the content domain matching degree can more accurately represent the matching degree between the content domain of the media content published by the candidate media content publishing account and the query intent of the target query data.

[0101] It should be noted that the historical query data mentioned in the embodiments of this application refers to query data that matches the query intent of the target query data within a certain historical period. The historical period can be, for example, a week, half a month, a month, three months, half a year, etc.

[0102] S304, Based on the degree of content domain matching, determine the recall results of the target query data among the candidate media content.

[0103] After determining the content domain matching degree between the target query data and the publishing accounts of each candidate media content, the recall results of the target query data can be determined from the candidate media content based on the content domain matching degree. Specifically, the content domain matching degree can be used as a ranking factor to sort the candidate media content in descending order of content domain matching degree, and the sorted candidate media content can be used as the recall results of the target query data; alternatively, the content domain matching degree can be used as a filtering condition to select candidate media content with a content domain matching degree higher than a matching threshold as the recall results of the target query data.

[0104] In this embodiment, the degree of content domain matching between the query intent of the target query data and the publishing content domain of each candidate media content can be determined from different matching perspectives (such as the media content type matching perspective, media content characteristic word matching perspective, and media content type combination media content characteristic word matching perspective mentioned above). Thus, based on the degree of content domain matching, domain account-related results can be determined in each candidate media content as the recall results of the target query data, thereby improving the satisfaction of the query object with the recall results.

[0105] In the above Figure 3 Based on the illustrated embodiment, step S302 requires obtaining recall content domain data and publication content domain data. From the perspective of media content type matching, the recall content domain data may include recall content type information (including viewed content types and the proportion of viewed content corresponding to each viewed content type), and the publication content domain data may include publication content type information (including publication content types and the proportion of publication content for each publication content type). This application embodiment provides a data processing method, which mainly describes the process of obtaining recall content type information and publication content type information from the perspective of media content type matching. This data processing method can be executed by a computer device, which may be the aforementioned... Figure 2 Server 202 in the data processing system shown. (As...) Figure 5 As shown, the data processing method may include, but is not limited to, the following steps S501-S502:

[0106] S501, identify the query type of the target query data, and obtain the query type of the target query data as the target query type.

[0107] The query type of the target query data can be identified using a pre-trained query data classification model.

[0108] S502, determine the recall content type information corresponding to the target query type as the recall content type information corresponding to the target query data.

[0109] After determining the query type of the target query data as the target query type, the recall content type information corresponding to the target query type can be determined as the recall content type information corresponding to the target query data.

[0110] The following describes the process of determining the type of retrieved content corresponding to the target query type:

[0111] The server retrieves historical query data from the search engine over a period of time (e.g., one week, half a month, one month, three months, half a year, etc.), as well as the media content viewed from the media content retrieved by each historical query. For any historical query data, firstly, a pre-trained query data classification model can be used to identify the query type of the historical query data. Then, the viewed media content retrieved from the historical query data can be classified to obtain the viewed content type, generating a format of <historical query data, historical query data query type, viewed media content, viewed content type>. The viewed content type can be input by the media content creator, or it can be identified using a pre-trained media content classification model. Next, the query type of the historical query data can be aggregated as a key, summarizing the viewed content types of the viewed media content under the same query type and calculating the proportion of query content corresponding to each viewed content type. For each query type, retrieved content type information in the format of <query type, <viewed content type 1: the proportion of viewed content corresponding to viewed content type 1; viewed content type 2: the proportion of viewed content corresponding to viewed content type 2; ...>> is generated.

[0112] Specifically, to accelerate the matching speed of subsequent content type recall information, content types with a low viewing ratio in the recall content type information can be discarded. Specifically, a viewing ratio threshold can be set, discarding content types with a viewing ratio less than the threshold and retaining those with a viewing ratio greater than or equal to the threshold. Alternatively, the content types can be sorted in descending order of viewing ratio, retaining the top N content types and discarding those after the Nth position, where N is a positive integer. After obtaining the recall content type information corresponding to each query type, this information is stored for subsequent matching.

[0113] In other words, the process of determining the recall content type information corresponding to the target query type can include: obtaining the viewed media content from the historical query data recalled under the target query type, classifying the obtained media content to obtain the viewed content type; then, counting the number of media content under each viewed content type to determine the viewed content ratio corresponding to each viewed content type. The viewed content ratio under any viewed content type refers to the ratio between the number of viewed media content belonging to that viewed content type and the total number of viewed media content. Among them, the recall content type information corresponding to the target query type includes: the viewed content type (including the viewed content type before discarding the viewed content type with a lower viewed content ratio, or the viewed content type after discarding the viewed content type with a lower viewed content ratio) and the viewed content ratio corresponding to each viewed content type.

[0114] Steps S501-S502 above mainly describe the process of determining the recall content type information that matches the query type of the target query data. The process of determining the publishing content type information of the publishing account of the candidate media content is similar, and can be found in the following content:

[0115] The database stores media content published by each of all publishing accounts registered in the media content search application. For any given publishing account, the published media content can be categorized to obtain the content type, and the proportion of each content type can be calculated. This generates and stores the publishing account's content type information in the following format: <Account ID of Publishing Account, <Content Type 1: Proportion of Content Type 1; Content Type 2: Proportion of Content Type 2; ...>>. The content type can be input by the media content creator, or it can be identified using a pre-trained media content classification model.

[0116] Specifically, to expedite the matching of subsequent content type information, content types with a low proportion can be discarded. Specifically, a content proportion threshold can be set, discarding content types with a proportion less than the threshold and retaining those with a proportion greater than or equal to the threshold. Alternatively, content types can be sorted in descending order of proportion, retaining the first M content types and discarding those after the Mth position, where M is a positive integer. After obtaining the content type information corresponding to a publishing account, this information is stored for subsequent matching.

[0117] In other words, the process of determining the content type information of the candidate media content publishing accounts can include: obtaining the media content published by the candidate media content publishing accounts, classifying the obtained media content to obtain the content type; then, counting the number of media content under each content type, determining the proportion of content published under each content type, where the proportion of content published under any content type refers to the ratio between the number of media content belonging to that content type published by the candidate media content publishing accounts and the total number of media content published by the candidate media content publishing accounts; wherein, the content type information of the candidate media content publishing accounts includes: the content type (including the content type before discarding the content type with the lower proportion of content published, or the content type after discarding the content type with the lower proportion of content published) and the proportion of content published for each major content type.

[0118] In this embodiment, the classification system for categorizing media content retrieved from historical query data that matches the query type of the target query data is the same as the classification system for media content published by the publishing account. This helps to solve the problem of inconsistent classification systems when matching media content types between the target query data and the publishing accounts of candidate media content, and also helps to solve the problem of overlapping content types when matching media content types between the target query data and the publishing accounts of candidate media content.

[0119] In the above Figure 3 Based on the illustrated embodiment, step S302 requires acquiring recall content domain data and publication content domain data. From the perspective of matching media content representation terms, the recall content domain data can include recall content representation information (including recall content representation terms and the representation importance information corresponding to each recall content representation term), and the publication content domain data can include publication content representation information (including publication content representation terms and the representation importance information for each publication content representation term). This application embodiment provides a data processing method, which mainly describes the process of acquiring recall content representation information and publication content representation information from the perspective of matching media content representation terms. This data processing method can be executed by a computer device, which can be the aforementioned... Figure 2 Server 202 in the data processing system shown. (As...) Figure 6 As shown, the data processing method may include, but is not limited to, the following steps S601-S602:

[0120] S601, Obtain reference historical query data that matches the query semantics of the target query data.

[0121] Matching the query semantics of the target query data with the referenced historical query data can mean that: the query semantics of the target query data are the same as the query semantics of the referenced historical query data, or that the similarity between the query semantics of the target query data and the query semantics of the referenced historical query data is higher than the similarity threshold.

[0122] S602, referencing the recall content representation information corresponding to the historical query data, determines the recall content representation information corresponding to the target query data.

[0123] After identifying the reference historical query data that matches the query semantics of the target query data, the recall content representation information corresponding to the reference historical query data can be used as the recall content representation information corresponding to the target query data. The process of determining the recall content representation information corresponding to the reference historical query data is described below:

[0124] The server retrieves historical query data from the search engine over a specific period (e.g., one week, two weeks, one month, three months, six months, etc.), along with the media content retrieved for each historical query. Here, the media content retrieved for any given historical query can be all the media content retrieved from that historical query, or it can be the media content ranked higher among all the retrieved media content. This is because the retrieved media content is typically ordered, and the earlier-ranked media content matches the historical query data better than the later-ranked media content. Therefore, media content with a higher match to the historical query data can be selected for further processing. For any given historical query:

[0125] First, a pre-trained keyword extraction model can be used to extract keywords from each media content retrieved from historical query data. Each extracted media content can be represented as a keyword list in the form of <historical query data, content identifier of media content, keyword 1, keyword 2, keyword 3…>. Keywords are words that express the central meaning of the media content and are commonly used in computer systems to index paper content features, information retrieval, and system aggregation for reader review. Keyword extraction is a branch of text mining and is the foundation of text mining research such as text retrieval, document comparison, summary generation, document classification, and clustering. Different keyword extraction algorithms exist depending on the scoring strategy. For example, TF-IDF (Term Frequency–Inverse Document Frequency), a keyword extraction algorithm based on statistical features, and LDA (Latent Dirichlet Allocation), a keyword extraction algorithm based on a subject model, etc. Any keyword extraction model can be used to extract keywords from media content.

[0126] Secondly, the media content recalled from historical query data can be collected as a set. Each piece of media content is added to this set as a list of its keywords. Then, the importance information of each keyword can be calculated. Keywords whose importance information meets the importance criteria are used as recall content representation terms. These recall content representation terms and their corresponding importance information together constitute the recall content representation information corresponding to the historical query data. The keyword representation importance information TF-IDF is composed of the first representation importance information TF and the second representation importance information IDF. The calculation process for the keyword importance representation information can be found in Formulas 1-3 below:

[0127]

[0128]

[0129] Importance information TF-IDF = TF × IDF Formula 3

[0130] It should be noted that the representation importance information in the recall content representation information can specifically adopt the second representation importance information IDF; the representation importance information meets the importance condition, specifically, it can mean: setting a representation importance threshold, determining keywords with representation importance information TF-IDF greater than the representation importance threshold as recall content representation words, or, sorting keywords in descending order of representation importance information TF-IDF, and taking the top X keywords as recall content representation words, where X is a positive integer.

[0131] In other words, the process of determining the recall content representation information corresponding to the historical query data may include: obtaining the media content recalled by the historical query data; extracting keywords from the obtained media content and determining the representation importance information of the keywords; and determining the keywords whose representation importance information meets the importance conditions as recall content representation words; wherein, the recall content representation information corresponding to the historical query data includes: recall content representation words, and representation importance information (i.e., second representation importance information IDF) corresponding to each recall content representation word.

[0132] Steps S601-S602 above mainly describe the process of determining the recall content representation information that matches the query semantics of the target query data. The process of determining the publishing content representation information of the publishing account of the candidate media content is similar, and can be found in the following content:

[0133] The database stores media content published by each of all publishing accounts registered in the media content search application. For any given publishing account: First, a pre-trained keyword extraction model can be used to extract keywords for each piece of media content published by that account. Each extracted piece of media content can be represented as a keyword list in the form of <account identifier, content identifier, keyword 1, keyword 2, keyword 3…>. Second, the media content published by the publishing account can be treated as a set. Any piece of media content can be added to the set as a list of its keywords. Then, the importance information of each keyword can be calculated. Keywords whose importance information meets the importance criteria are used as the content representation terms. The content representation terms and their corresponding importance information together constitute the content representation information for the publishing account.

[0134] It should be noted that the representation importance information in the content representation information can specifically adopt the second representation importance information IDF; the representation importance information meets the importance condition, which can specifically mean: setting a representation importance threshold, determining keywords with representation importance information TF-IDF greater than the representation importance threshold as the representation words of the content, or sorting keywords in descending order of representation importance information TF-IDF, and taking the top Y keywords as the representation words of the content, where Y is a positive integer.

[0135] In other words, the process of determining the content representation information of the publishing accounts of candidate media content may include: obtaining the media content published by the publishing accounts of candidate media content; extracting keywords from the obtained media content and determining the representation importance information of the keywords; identifying keywords whose representation importance information meets the importance conditions as the representation words of the published content; wherein, the content representation information of the publishing accounts of candidate media content includes: the representation words of the published content, and the representation importance information (i.e., the second representation importance information IDF) corresponding to each representation word of the published content.

[0136] In this embodiment of the application, the recall content representation information of the media content recalled by historical query data over a certain period of time, and the publication content representation information of the media content published by the registered publishing account are stored. During matching, the corresponding recall content representation information and publication content representation information can be directly retrieved from the storage space for matching, which is very convenient and fast.

[0137] It should be further noted that the embodiments of this application involve processes such as classifying and identifying query data using a query data classification model, classifying and identifying media content using a media content classification model, and performing semantic understanding on query data. All of these processes involve artificial intelligence technologies such as natural language processing and machine learning. Among them:

[0138] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0139] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0140] Natural Language Processing (NLP) is an important field within computer science and artificial intelligence. It studies the theories and methods for enabling effective communication between humans and computers using natural language. NLP is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language people use in daily life—and thus it has a close relationship with linguistic research. NLP techniques typically include text processing, semantic understanding, machine translation, question answering, and knowledge graphs.

[0141] Machine learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning.

[0142] The methods of the embodiments of this application have been described in detail above. In order to facilitate better implementation of the above solutions of the embodiments of this application, the apparatus of the embodiments of this application is provided below.

[0143] Please see Figure 7 , Figure 7 This is a schematic diagram of the structure of a data processing device provided in an embodiment of this application. The data processing device can be installed in the computer equipment provided in the embodiment of this application. The computer equipment can be the server 202 mentioned in the above method embodiment. Figure 7 The data processing apparatus shown may be a computer program (including program code) running on a computer device, which can be used to execute... Figure 3 , Figure 5 or Figure 6 Some or all of the steps in the method embodiments shown. Please refer to [link / reference]. Figure 7The data processing apparatus may include the following units:

[0144] The acquisition unit 701 is used to acquire candidate media content that matches the target query data;

[0145] The acquisition unit 701 is also used to acquire recall content domain data that matches the query intent information of the target query data, and to acquire the publication content domain data of the publishing account of the candidate media content; wherein, the recall content domain data is used to characterize the content domain of the media content recalled by the historical query data that matches the query intent information; the publication content domain data is used to characterize the content domain of the media content published by the publishing account of the candidate media content;

[0146] The processing unit 702 is used to match the recalled content domain data with the published content domain data of the candidate media content's publishing account to obtain the content domain matching degree between the target query data and the candidate media content's publishing account.

[0147] The processing unit 702 is also used to determine the recall results of the target query data in the candidate media content based on the content domain matching degree.

[0148] In one implementation, the recall content domain data includes recall content type information, which includes: the viewing content type of the media content recalled by historical query data, and the viewing content ratio corresponding to each viewing content type; the publishing content domain data includes publishing content type information, which includes: the publishing content type and the publishing content ratio corresponding to each publishing content type.

[0149] Processing unit 702 is used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts. Specifically, it performs the following steps:

[0150] The recall content type information is matched with the published content type information to determine the matching content type in the recall content type information and the published content type information;

[0151] Based on the proportion of viewed content and the proportion of published content corresponding to the matching content type, determine the degree of content domain matching between the target query data and the publishing accounts of the candidate media content.

[0152] In one implementation, the number of matched content types is multiple; the processing unit 702, when determining the content domain matching degree between the target query data and the publishing account of the candidate media content based on the viewing content ratio and publishing content ratio corresponding to the matched content type, specifically performs the following steps:

[0153] The matching degree information for each content type is calculated based on the proportion of viewed content and the proportion of published content for each matching content type.

[0154] Based on the matching degree information of each matching content type, the content domain matching degree between the target query data and the publishing account of the candidate media content is calculated; or, the maximum matching degree information among the matching degree information of each matching content type is determined as the domain matching degree between the target query data and the publishing account of the candidate media content.

[0155] In one implementation, the recall content domain data includes recall content representation information, which includes recall content representation terms; the publication content domain data includes publication content representation information, which includes publication content representation terms.

[0156] Processing unit 702 is used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts. Specifically, it performs the following steps:

[0157] By performing similarity matching between the recall content representation information and the published content representation information, the content domain matching degree between the target query data and the publishing accounts of the candidate media content is obtained.

[0158] In one implementation, the recall content representation information also includes representation importance information for each recall content representation word;

[0159] Processing unit 702 is used to perform similarity matching between the recall content representation information and the published content representation information to obtain the content domain matching degree between the target query data and the publishing account of the candidate media content. Specifically, it is used to perform the following steps:

[0160] Determine the similarity information between each recall content characterization term in the recall content characterization information and the published content characterization information;

[0161] Based on the similarity information between each recall content characterization term and the published content characterization information, as well as the characterization importance information of each recall content characterization term, the content domain matching degree between the target query data and the publishing account of the candidate media content is determined.

[0162] In one implementation, the query intent information includes query type and query semantics; the historical query data includes: first historical query data matching the query type and second historical query data matching the query semantics; the recalled content domain data includes: recalled content type information of the viewed media content recalled by the first historical query data and recalled content representation information of the media content recalled by the second historical query data; and the published content domain data includes published content type information and published content representation information.

[0163] Processing unit 702 is used to match the recalled content domain data with the published content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts. Specifically, it performs the following steps:

[0164] The type information of the recalled content is matched with the type information of the published content to obtain type matching information;

[0165] The recall content representation information is matched with the published content representation information to obtain representation matching information;

[0166] Based on type matching information and representation matching information, determine the degree of content domain matching between the target query data and the publishing accounts of candidate media content.

[0167] In one implementation, the query intent information is the query type of the target query data; the recall content domain data corresponding to the target query data includes recall content type information; the acquisition unit 701, when acquiring recall content domain data that matches the query intent information of the target query data, specifically performs the following steps:

[0168] Identify the query type of the target query data and determine that the query type of the target query data is the target query type;

[0169] The type of retrieved content corresponding to the target query type is determined as the type of retrieved content corresponding to the target query data.

[0170] In one implementation, the process of determining the type of retrieved content corresponding to the target query type includes:

[0171] Retrieve the viewed media content from the historical query data retrieved under the target query type;

[0172] The acquired media content is categorized to determine the types of content to be viewed;

[0173] Count the number of media contents under each content viewing type and determine the proportion of content viewed for each content viewing type;

[0174] The content type information for the target query type includes: the content type viewed and the proportion of content viewed for each content type.

[0175] In one implementation, the query intent information is the query semantics of the target query data; the recall content domain data corresponding to the target query data includes recall content representation information; the acquisition unit 701, when acquiring recall content domain data that matches the query intent information of the target query data, specifically performs the following steps:

[0176] Obtain reference historical query data that matches the query semantics of the target query data;

[0177] The recall content representation information corresponding to the historical query data will be used to determine the recall content representation information corresponding to the target query data.

[0178] In one implementation, the process of determining the recall content representation information corresponding to historical query data includes:

[0179] Retrieve media content retrieved by referencing historical query data;

[0180] Extract keywords from the acquired media content and determine the importance information of the keywords;

[0181] Keywords that meet the importance criteria are identified as recall content representation words;

[0182] The recall content representation information corresponding to the historical query data includes: recall content representation terms, and representation importance information corresponding to each recall content representation term.

[0183] In one implementation, when processing unit 702 determines the recall result of the target query data in candidate media content based on the content domain matching degree, it specifically performs the following:

[0184] The candidate media content is sorted from highest to lowest according to the degree of content domain matching, and the sorted candidate media content is used as the recall result of the target query data.

[0185] Alternatively, candidate media content with a content domain matching degree higher than the matching threshold can be used as the recall result of the target query data.

[0186] According to another embodiment of this application, Figure 7The data processing apparatus shown can be constructed by combining each unit individually or entirely into one or more other units, or one or more of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. The above-mentioned units are based on logical function division. In practical applications, the function of one unit can be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the data processing apparatus may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0187] According to another embodiment of this application, the following can be achieved by running on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM), a device capable of performing operations such as... Figure 3 , Figure 5 or Figure 6 The computer program (including program code) for each step involved in some or all of the methods shown, to construct such... Figure 7 The data processing apparatus shown herein, and the data processing method for implementing the embodiments of this application, are described. A computer program may be recorded on, for example, a computer-readable storage medium, loaded onto the aforementioned computing device via the computer-readable storage medium, and executed therein.

[0188] In this embodiment, the recalled content domain data can be used to characterize the content domain of media content recalled by the relevant historical query data of the target query data, and the published content domain data can be used to characterize the content domain of media content published by the publishing accounts of the candidate media content of the target query data. By matching the recalled content domain data with the published content domain data, the degree of content domain matching between the target query data and the publishing accounts of the candidate media content can be determined. Therefore, based on the degree of content domain matching between the target query data and the publishing accounts of the candidate media content, media content published by publishing accounts whose published content domain matches the target query data can be recalled from the candidate media content as the recall result of the target query data, that is, the domain-related account results of the target query data can be recalled.

[0189] Based on the above methods and apparatus embodiments, this application provides a computer device, which may be the aforementioned server 202. Please refer to... Figure 8 , Figure 8 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Figure 8The computer device shown includes at least a processor 801, an input interface 802, an output interface 803, and a computer-readable storage medium 804. The processor 801, input interface 802, output interface 803, and computer-readable storage medium 804 can be connected via a bus or other means.

[0190] Computer-readable storage medium 804 can be stored in the memory of a computer device. Computer-readable storage medium 804 is used to store computer programs, including computer instructions. Processor 801 is used to execute the program instructions stored in computer-readable storage medium 804. Processor 801 (or CPU (Central Processing Unit)) is the computing and control core of the computer device. It is suitable for implementing one or more computer instructions, specifically for loading and executing one or more computer instructions to achieve corresponding method flows or corresponding functions.

[0191] This application also provides a computer-readable storage medium (Memory), which is a memory device in a computer device used to store programs and data. It is understood that the computer-readable storage medium here can include both built-in storage media in the computer device and extended storage media supported by the computer device. The computer-readable storage medium provides storage space that stores the operating system of the computer device. Furthermore, the storage space also stores one or more computer instructions suitable for loading and execution by a processor. These computer instructions can be one or more computer programs (including program code). It should be noted that the computer-readable storage medium here can be high-speed RAM or non-volatile memory, such as at least one disk storage device; optionally, it can also be at least one computer-readable storage medium located remotely from the aforementioned processor.

[0192] In some embodiments, the processor 801 may load and execute one or more computer instructions stored in the computer-readable storage medium 804 to implement the aforementioned related... Figure 3 , Figure 5 or Figure 6 The corresponding steps of the data processing method shown. In a specific implementation, the computer instructions in the computer-readable storage medium 804 are loaded by the processor 801 and executed as follows:

[0193] Retrieve candidate media content that matches the target query data;

[0194] Acquire recall content domain data that matches the query intent information of the target query data, and acquire publication content domain data of the publishing accounts of the candidate media content; wherein, the recall content domain data is used to characterize the content domain of the media content recalled by the historical query data that matches the query intent information; the publication content domain data is used to characterize the content domain of the media content published by the publishing accounts of the candidate media content;

[0195] The retrieved content domain data will be matched with the content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts.

[0196] Based on the degree of content domain matching, the recall results of the target query data are determined from the candidate media content.

[0197] In one implementation, the recall content domain data includes recall content type information, which includes: the viewing content type of the media content recalled by historical query data, and the viewing content ratio corresponding to each viewing content type; the publishing content domain data includes publishing content type information, which includes: the publishing content type and the publishing content ratio corresponding to each publishing content type.

[0198] When the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to retrieve content domain data and match it with the content domain data of the publishing accounts of candidate media content to obtain the content domain matching degree between the target query data and the publishing accounts of candidate media content, the following steps are specifically performed:

[0199] The recall content type information is matched with the published content type information to determine the matching content type in the recall content type information and the published content type information;

[0200] Based on the proportion of viewed content and the proportion of published content corresponding to the matching content type, determine the degree of content domain matching between the target query data and the publishing accounts of the candidate media content.

[0201] In one implementation, the number of matched content types is multiple; when the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to determine the content domain matching degree between the target query data and the publishing account of the candidate media content according to the viewing content ratio and publishing content ratio corresponding to the matched content type, the instructions are specifically used to perform the following steps:

[0202] The matching degree information for each content type is calculated based on the proportion of viewed content and the proportion of published content for each matching content type.

[0203] Based on the matching degree information of each matching content type, the content domain matching degree between the target query data and the publishing account of the candidate media content is calculated; or, the maximum matching degree information among the matching degree information of each matching content type is determined as the domain matching degree between the target query data and the publishing account of the candidate media content.

[0204] In one implementation, the recall content domain data includes recall content representation information, which includes recall content representation terms; the publication content domain data includes publication content representation information, which includes publication content representation terms.

[0205] When the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to retrieve content domain data and match it with the content domain data of the publishing accounts of candidate media content to obtain the content domain matching degree between the target query data and the publishing accounts of candidate media content, the following steps are specifically performed:

[0206] By performing similarity matching between the recall content representation information and the published content representation information, the content domain matching degree between the target query data and the publishing accounts of the candidate media content is obtained.

[0207] In one implementation, the recall content representation information also includes representation importance information for each recall content representation word;

[0208] When the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to perform similarity matching between the recall content representation information and the published content representation information to obtain the content domain matching degree between the target query data and the publishing account of the candidate media content, the instructions are specifically used to perform the following steps:

[0209] Determine the similarity information between each recall content characterization term in the recall content characterization information and the published content characterization information;

[0210] Based on the similarity information between each recall content characterization term and the published content characterization information, as well as the characterization importance information of each recall content characterization term, the content domain matching degree between the target query data and the publishing account of the candidate media content is determined.

[0211] In one implementation, the query intent information includes query type and query semantics; the historical query data includes: first historical query data matching the query type and second historical query data matching the query semantics; the recalled content domain data includes: recalled content type information of the viewed media content recalled by the first historical query data and recalled content representation information of the media content recalled by the second historical query data; and the published content domain data includes published content type information and published content representation information.

[0212] When the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to retrieve content domain data and match it with the content domain data of the publishing accounts of candidate media content to obtain the content domain matching degree between the target query data and the publishing accounts of candidate media content, the following steps are specifically performed:

[0213] The type information of the recalled content is matched with the type information of the published content to obtain type matching information;

[0214] The recall content representation information is matched with the published content representation information to obtain representation matching information;

[0215] Based on type matching information and representation matching information, determine the degree of content domain matching between the target query data and the publishing accounts of candidate media content.

[0216] In one implementation, the query intent information is the query type of the target query data; the recall content domain data corresponding to the target query data includes recall content type information; when the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to obtain the recall content domain data matching the query intent information of the target query data, they are specifically used to perform the following steps:

[0217] Identify the query type of the target query data and determine that the query type of the target query data is the target query type;

[0218] The type of retrieved content corresponding to the target query type is determined as the type of retrieved content corresponding to the target query data.

[0219] In one implementation, the process of determining the type of retrieved content corresponding to the target query type includes:

[0220] Retrieve the viewed media content from the historical query data retrieved under the target query type;

[0221] The acquired media content is categorized to determine the types of content to be viewed;

[0222] Count the number of media contents under each content viewing type and determine the proportion of content viewed for each content viewing type;

[0223] The content type information for the target query type includes: the content type viewed and the proportion of content viewed for each content type.

[0224] In one implementation, the query intent information is the query semantics of the target query data; the recall content domain data corresponding to the target query data includes recall content representation information; when the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to obtain the recall content domain data that matches the query intent information of the target query data, they are specifically used to perform the following steps:

[0225] Obtain reference historical query data that matches the query semantics of the target query data;

[0226] The recall content representation information corresponding to the historical query data will be used to determine the recall content representation information corresponding to the target query data.

[0227] In one implementation, the process of determining the recall content representation information corresponding to historical query data includes:

[0228] Retrieve media content retrieved by referencing historical query data;

[0229] Extract keywords from the acquired media content and determine the importance information of the keywords;

[0230] Keywords that meet the importance criteria are identified as recall content representation words;

[0231] The recall content representation information corresponding to the historical query data includes: recall content representation terms, and representation importance information corresponding to each recall content representation term.

[0232] In one implementation, when the computer instructions in the computer-readable storage medium 804 are loaded and executed by the processor 801 to determine the recall results of the target query data among candidate media content based on the degree of content domain matching, the instructions specifically perform the following:

[0233] The candidate media content is sorted from highest to lowest according to the degree of content domain matching, and the sorted candidate media content is used as the recall result of the target query data.

[0234] Alternatively, candidate media content with a content domain matching degree higher than the matching threshold can be used as the recall result of the target query data.

[0235] In this embodiment, the recalled content domain data can be used to characterize the content domain of media content recalled by the relevant historical query data of the target query data, and the published content domain data can be used to characterize the content domain of media content published by the publishing accounts of the candidate media content of the target query data. By matching the recalled content domain data with the published content domain data, the degree of content domain matching between the target query data and the publishing accounts of the candidate media content can be determined. Therefore, based on the degree of content domain matching between the target query data and the publishing accounts of the candidate media content, media content published by publishing accounts whose published content domain matches the target query data can be recalled from the candidate media content as the recall result of the target query data, that is, the domain-related account results of the target query data can be recalled.

[0236] According to one aspect of this application, a computer program product or computer program is provided, comprising computer instructions stored in a computer-readable storage medium. A processor of a computer device reads the computer instructions from the computer-readable storage medium and executes the computer instructions, causing the computer device to perform the data processing methods provided in the various alternative embodiments described above.

[0237] The above description is merely a specific embodiment of this application, but the scope of protection of this application is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the scope of the technology disclosed in this application should be included within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A data processing method, characterized in that, The method includes: Retrieve candidate media content that matches the target query data; Acquire recall content domain data that matches the query intent information of the target query data, and acquire publication content domain data of the publishing accounts of the candidate media content; wherein, when the query intent information refers to the query type of the target query data, the recall content domain data includes the recall content type information, and the publication content domain data includes the publication content type information; the recall content type information includes: the viewing content type of the media content recalled from historical query data that matches the query type of the target query data, and the viewing content ratio corresponding to each viewing content type; the publication content type information includes: the publication content type of the media content published by the publishing accounts of the candidate media content, and the publication content ratio corresponding to each publication content type; The process of matching the recalled content domain data with the published content domain data of the candidate media content's publishing accounts to obtain the content domain matching degree between the target query data and the candidate media content's publishing accounts includes: matching the recalled content type information with the published content type information to determine the matching content types in the recalled content type information and the published content type information; determining the content domain matching degree between the target query data and the candidate media content's publishing accounts based on the viewing content ratio and publishing content ratio corresponding to the matching content type; the content domain matching degree is the result of a weighted sum of the viewing content ratio and publishing content ratio of the matching content type; Based on the content domain matching degree, the recall result of the target query data is determined in the candidate media content.

2. The method as described in claim 1, characterized in that, The number of matched content types is multiple; determining the content domain matching degree between the target query data and the publishing account of the candidate media content based on the viewing content ratio and publishing content ratio corresponding to the matched content type includes: The matching degree information for each content type is calculated based on the proportion of viewed content and the proportion of published content for each matching content type. Based on the matching degree information of each matching content type, the content domain matching degree between the target query data and the publishing account of the candidate media content is calculated; or, the maximum matching degree information among the matching degree information of each matching content type is determined as the content domain matching degree between the target query data and the publishing account of the candidate media content.

3. The method as described in claim 1, characterized in that, When the query intent information refers to the query semantics of the target query data, the recalled content domain data includes recalled content representation information, the recalled content representation information includes recalled content representation words, the published content domain data includes published content representation information, and the published content representation information includes published content representation words; The step of matching the recalled content domain data with the published content domain data of the candidate media content's publishing account to obtain the content domain matching degree between the target query data and the published account of the candidate media content includes: The similarity matching between the recalled content representation information and the published content representation information is performed to obtain the content domain matching degree between the target query data and the publishing account of the candidate media content.

4. The method as described in claim 3, characterized in that, The recall content representation information also includes the representation importance information of each recall content representation word; The step of performing similarity matching between the recalled content representation information and the published content representation information to obtain the content domain matching degree between the target query data and the publishing account of the candidate media content includes: Determine the similarity information between each recall content characterization word in the recall content characterization information and the published content characterization information; Based on the similarity information between each recall content characterization term in the recall content characterization information and the publication content characterization information, as well as the characterization importance information of each recall content characterization term, the content domain matching degree between the target query data and the publishing account of the candidate media content is determined.

5. The method as described in claim 1, characterized in that, When the query intent information refers to the query type and query semantics of the target query data, the recalled content domain data includes: the recalled content type information of the media content viewed in the media content recalled by the first historical query data that matches the query type of the target query data, and the recalled content representation information of the media content recalled by the second historical query data that matches the query semantics of the target query data. The published content domain data includes published content type information and published content representation information. The step of matching the recalled content domain data with the published content domain data of the candidate media content's publishing account to obtain the content domain matching degree between the target query data and the published account of the candidate media content includes: The recalled content type information is matched with the published content type information to obtain type matching information; The recall content representation information is matched with the published content representation information to obtain representation matching information; Based on the type matching information and the representation matching information, the degree of content domain matching between the target query data and the publishing account of the candidate media content is determined.

6. The method as described in claim 1, characterized in that, When the query intent information is the query type of the target query data, the recall content domain data corresponding to the target query data includes recall content type information; The acquisition of recall content domain data that matches the query intent information of the target query data includes: Identify the query type of the target query data and determine that the query type of the target query data is the target query type; The recall content type information corresponding to the target query type is determined as the recall content type information corresponding to the target query data.

7. The method as described in claim 6, characterized in that, The process of determining the type of retrieved content corresponding to the target query type includes: Retrieve the viewed media content from the historical query data retrieved under the target query type; The acquired media content is categorized to determine the types of content to be viewed; Count the number of media contents under each content viewing type and determine the proportion of content viewed for each content viewing type; The content type information corresponding to the target query type includes: the content type to be viewed and the proportion of content to be viewed for each content type.

8. The method as described in claim 1, characterized in that, When the query intent information is the query semantics of the target query data, the recall content domain data corresponding to the target query data includes recall content representation information; The acquisition of recall content domain data that matches the query intent information of the target query data includes: Obtain reference historical query data that matches the query semantics of the target query data; The recall content representation information corresponding to the reference historical query data is determined as the recall content representation information corresponding to the target query data.

9. The method as described in claim 8, characterized in that, The process of determining the recall content representation information corresponding to the reference historical query data includes: Obtain the media content recalled by the reference historical query data; Extract keywords from the acquired media content and determine the keyword importance information; Keywords that meet the importance criteria are identified as recall content representation words; The recall content representation information corresponding to the reference historical query data includes: the recall content representation words, and the representation importance information corresponding to each recall content representation word.

10. The method as described in claim 1, characterized in that, The step of determining the recall result of the target query data in the candidate media content based on the content domain matching degree includes: The candidate media content is sorted from highest to lowest according to the degree of matching in the content domain, and the sorted candidate media content is used as the recall result of the target query data; or, Candidate media content with a content domain matching degree higher than the matching threshold is used as the recall result of the target query data.

11. A data processing apparatus, characterized in that, The data processing device includes: The acquisition unit is used to acquire candidate media content that matches the target query data; The acquisition unit is further configured to acquire recall content domain data that matches the query intent information of the target query data, and acquire the publication content domain data of the publishing account of the candidate media content; wherein, when the query intent information refers to the query type of the target query data, the recall content domain data includes the recall content type information, and the publication content domain data includes the publication content type information; the recall content type information includes: the viewing content type of the media content recalled from the historical query data that matches the query type of the target query data, and the viewing content ratio corresponding to each viewing content type; the publication content type information includes: the publication content type of the media content published by the publishing account of the candidate media content, and the publication content ratio corresponding to each publication content type; The processing unit is configured to match the recalled content domain data with the published content domain data of the candidate media content's publishing account to obtain the content domain matching degree between the target query data and the candidate media content's publishing account. This includes: matching the recalled content type information with the published content type information to determine the matching content type in the recalled content type information and the published content type information; determining the content domain matching degree between the target query data and the candidate media content's publishing account based on the viewing content ratio and publishing content ratio corresponding to the matching content type; the content domain matching degree is the result of a weighted sum of the viewing content ratio and publishing content ratio of the matching content type. The processing unit is further configured to determine the recall result of the target query data in the candidate media content based on the content domain matching degree.

12. A computer device, characterized in that, The computer device includes: A processor is a tool for implementing computer programs. A computer-readable storage medium storing a computer program adapted to be loaded by the processor and executed as described in any one of claims 1 to 10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program adapted to be loaded by a processor and executed as described in any one of claims 1 to 10.

14. A computer program product, characterized in that, The computer program product includes computer instructions that, when executed by a processor, implement the data processing method as described in any one of claims 1 to 10.

Images