Internal link generation method, data processing method, search method and computing device

Abstract

Provided in the embodiments of the present application are an internal link generation method, a data processing method, a search method and a computing device. The internal link generation method comprises: determining a plurality of matching pairs, each matching pair consisting of an object and key information having a matching relationship with the object, and the matching pairs being obtained from a key information library and an object library by means of identification; determining at least one page link relationship corresponding to a list page and a detail page; and according to the at least one page link relationship, performing association conversion on the plurality of matching pairs, so as to obtain two pages separately corresponding to the at least one page link relationship and key information used as an internal link between the two pages. The technical solution provided by the embodiments of the present application generates internal links on the basis of matching relationships between objects and key information, so as to ensure the correlation between pages corresponding to different page link relationships, helping to increase the weighting of websites and increase the inclusion ratio of websites, thereby ensuring the rationality of the internal links generated.

Description

Internal link generation method, data processing method, search method and computing device

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on September 4, 2023, with application number 202311137718.0 and application name “Internal link generation method, data processing method, search method and computing device”, the entire contents of which are incorporated by reference into this application. Technical Field

[0002] The embodiments of the present application relate to the field of network technology, and in particular to an internal link generation method, a data processing method, a search method, and a computing device. Background Art

[0003] Internal links refer to the internal links within a website, specifically the links between pages within the same domain. Internal links are typically placed within a website's internal pages as key information. Users can click on the key information to access other pages within the website, or search engines can crawl the pages pointed to by internal links.

[0004] In Search Engine Optimization (SEO), a reasonable internal link structure can enable search engines to quickly crawl and index the website, improve the website's weight, and increase user click-through rate.

[0005] Therefore, how to reasonably determine internal links becomes a technical problem that technicians in this field need to solve, especially in websites that provide objects such as goods and allow users to purchase, browse or other interactive behaviors with the objects. The page types are mainly divided into list pages and object details pages. The page link relationships involved usually include list pages linking to details pages, list pages linking to list pages, details pages linking to details pages, and details pages linking to object pages, etc. A reasonable internal link structure can recommend valuable list pages or details pages to search engines, improve page rankings, and thus bring more traffic to the website. Therefore, the above technical problems need to be solved urgently.

[0006] Summary of the Invention

[0007] The embodiments of the present application provide an internal link generation method, a data processing method, a search method and a computing device to solve the technical problem of unreasonable internal links in the prior art.

[0008] In a first aspect, an embodiment of the present application provides a method for generating an internal link, comprising:

[0009] Determine a plurality of matching pairs; the matching pairs are composed of objects and key information having a matching relationship with the objects; the matching pairs are identified and obtained from a key information library and an object library;

[0010] Determine at least one page link relationship corresponding to the list page and the detail page;

[0011] According to the at least one page link relationship, the multiple matching pairs are associated and converted to obtain two pages corresponding to the at least one page link relationship and key information between the two pages used as an internal link.

[0012] Optionally, the method further includes:

[0013] Clustering multiple key information in the key information database according to key information features to obtain multiple information groups;

[0014] Matching the objects in the object library with the plurality of information groups respectively according to object characteristics to determine at least one matched target information group;

[0015] The object is matched with key information in the at least one target information group respectively to determine at least one key information matching the object, and the object and the at least one key information form at least one matching pair.

[0016] Optionally, the method further includes:

[0017] The matching model is used to extract key information features corresponding to multiple key information in the key information database;

[0018] Extracting object features corresponding to a plurality of objects in an object library using the matching model;

[0019] The matching model is obtained by training based on multiple groups of sample matching pairs using a comparative learning method; the sample matching pairs include sample key information and sample objects.

[0020] Optionally, the at least one page link relationship includes a list page linking to a detail page;

[0021] The step of performing association conversion on the multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages includes:

[0022] Determining, from the multiple sets of matching pairs, multiple key information that have a matching relationship with the same object;

[0023] Determining first key information and second key information according to the matching degree between the plurality of key information and the object;

[0024] Based on the correspondence between key information and list pages, the first list page pointed to by the first key information and the first details page of the object are determined, and the second key information is used as an internal link in the first list page.

[0025] Optionally, determining the first key information and the second key information according to the matching degrees between the plurality of key information and the object includes:

[0026] According to the matching degrees between the plurality of key information and the object, first key information whose matching degree meets a first matching requirement is determined, and second key information whose matching degree meets a second matching requirement is determined.

[0027] Optionally, the at least one page link relationship includes a detail page link list page;

[0028] The step of performing association conversion on the multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages includes:

[0029] Searching for a target matching pair from the plurality of matching pairs according to the matching degree between the key information and the object;

[0030] Combined with the correspondence between key information and list pages, determine the target list page pointed to by the target key information in the target matching pair, the target details page of the target object, and use the target key information as an internal link in the target details page.

[0031] Optionally, the at least one page link relationship includes a details page linking to a details page;

[0032] The step of performing association conversion on the multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages includes:

[0033] determining a first object and a second object that satisfy a correlation requirement;

[0034] Searching for first key information matching the first object from the multiple matching pairs;

[0035] A first details page for the first object and a second details page for the second object are determined, and the first key information is used as an internal link in the second details page.

[0036] Optionally, the at least one page link relationship includes a list page link to a list page;

[0037] The step of performing association conversion on the multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages includes:

[0038] Searching for first key information and second key information that have a matching relationship with the same object from the multiple sets of matching pairs;

[0039] Based on the correspondence between key information and list pages, determine the first list page pointed to by the first key information, the second list page pointed to by the second key information, and use the first key information as an internal link in the second list page.

[0040] Optionally, searching for the first key information and the second key information that have a matching relationship with the same object from the multiple matching pairs includes:

[0041] Searching for multiple key information matching the same object from the multiple matching pairs;

[0042] According to the matching degrees between the plurality of key information and the object, the first key information whose matching degree meets the third matching requirement is determined, and the second key information whose matching degree meets the fourth matching requirement is determined.

[0043] Optionally, the method further includes:

[0044] Performing page configuration according to the associated conversion results corresponding to the at least one page link relationship to update the internal link structure of the website;

[0045] Alternatively, the associated conversion results corresponding to the at least one page link relationship are sent to the target person; the associated conversion results are used to generate page configuration information, and the page configuration information is used to update the internal link structure of the website.

[0046] Optionally, the method further includes:

[0047] Determining, based on search records generated within the website and / or outside the website, key information in the search records and a list page to which the key information points;

[0048] A corresponding relationship between the key information and the list page is established, and a key information library is constructed based on the key information.

[0049] Optionally, the matching model is pre-trained in the following manner:

[0050] Determine multiple sets of sample matching pairs corresponding to the current training batch;

[0051] Inputting the plurality of sample matching pairs into a matching model;

[0052] Any sample matching pair is taken as a positive sample, and the sample key information in any sample matching pair and the sample objects in at least one other sample matching pair are taken as negative samples, and the matching model is trained in a sampling contrast learning manner.

[0053] Optionally, taking any one sample matching pair as a positive sample, the sample key information in any one sample matching pair and the sample object in at least one other sample matching pair as negative samples, and training the matching model in a sampling contrast learning manner includes:

[0054] Extracting sample key information features corresponding to sample key information and sample object features corresponding to sample objects in multiple groups of sample matching pairs using the matching model;

[0055] Any sample matching pair is regarded as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one of the other sample matching pairs are regarded as negative samples;

[0056] Based on the sample information features and sample object features, calculate the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample;

[0057] The matching model is trained with the training objectives of maximizing the matching degree corresponding to the positive samples and minimizing the matching degree corresponding to the negative samples.

[0058] Optionally, the method further includes:

[0059] Based on the key information feature of at least one key information respectively included in the multiple information groups, respectively generate group features corresponding to the multiple information groups;

[0060] Using grouping features of the plurality of information groups as respective index identifiers;

[0061] The step of matching any object in the object library with the plurality of information groups according to the object characteristics and determining the matched target information group comprises:

[0062] According to the object characteristics, the target index identifier that matches any object is selected;

[0063] Determine the target information group corresponding to the target index identifier.

[0064] Optionally, the second key information for determining whether the matching degree satisfies the second matching requirement includes:

[0065] If there are at least two key information in addition to the first key information in the multiple key information, select the key information with the highest matching degree as the second key information;

[0066] The method further comprises:

[0067] When it is determined that the first list page corresponds to multiple internal links, and the multiple internal links point to the first detail page of the same object, at least one target internal link is screened and retained in descending order of matching degree with the object.

[0068] Optionally, searching for a target matching pair from the multiple groups of matching pairs according to a matching degree between key information and an object includes:

[0069] Determining, from the plurality of matching pairs, a plurality of key information having a matching relationship with any object;

[0070] According to the matching degrees between the plurality of key information and the object, it is determined that the key information with the highest matching degree and the object constitute a target matching pair.

[0071] Optionally, the second key information for determining whether the matching degree satisfies the fourth matching requirement includes:

[0072] If there are at least two key information in addition to the first key information in the multiple key information, select the key information with the highest matching degree as the second key information;

[0073] The method further comprises:

[0074] When it is determined that the first list page corresponds to multiple internal links, and the multiple internal links point to the same second list page, at least one target internal link is obtained by screening in order of the degree of matching with the respective corresponding objects from highest to lowest.

[0075] Optionally, determining the key information in the search record and the list page pointed to by the key information includes:

[0076] Determining key information in the search record and at least one candidate list page pointed to by the key information;

[0077] A list page corresponding to the key information is determined according to click information of the at least one candidate list page.

[0078] Optionally, identifying objects and key information having a matching relationship from the key information library and the object library to obtain multiple sets of matching pairs includes:

[0079] In response to the page update instruction, objects and key information having a matching relationship are identified from the key information library and the object library to obtain a plurality of matching pairs.

[0080] Optionally, the method further includes:

[0081] Counting the click rate corresponding to the website, and generating the page update instruction according to whether the click rate meets the first update condition;

[0082] or,

[0083] The crawling rate and / or collection rate corresponding to the website is counted, and the page update instruction is generated according to whether the crawling rate and / or collection rate meets the second update condition.

[0084] Optionally, the matching model is specifically used to extract key information features corresponding to multiple key information in the key information library, and for multiple objects in the object library, the matching model is used to extract respective object features from the relevant information of the multiple objects; the key information features and the object features are used to determine the objects and key information that have matching relationships in the key information library and the object library to obtain multiple groups of matching pairs.

[0085] In a second aspect, an embodiment of the present application provides a data processing method, including:

[0086] Obtain multiple sets of sample matching pairs;

[0087] The plurality of sample matching pairs are used as training data of the current training batch and input into the matching model;

[0088] Taking any sample matching pair as a positive sample, the sample key information in any sample matching pair and the sample object in at least one other sample matching pair as negative samples, and training the matching model in a sampling comparative learning method;

[0089] The matching model is used to identify objects and key information with matching relationships in the key information library and the object library to obtain multiple sets of matching pairs; the multiple sets of matching objects are used to perform association conversion to obtain any two pages corresponding to at least one page link relationship and the key information between the two pages used as internal links.

[0090] Optionally, taking any one sample matching pair as a positive sample, the sample key information in any one sample matching pair and the sample object in at least one other sample matching pair as negative samples, and training the matching model in a sampling contrast learning manner includes:

[0091] Extracting sample key information features corresponding to sample key information and sample object features corresponding to sample objects in multiple groups of sample matching pairs using the matching model;

[0092] Any sample matching pair is regarded as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one of the other sample matching pairs are regarded as negative samples;

[0093] Based on the sample information features and sample object features, calculate the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample;

[0094] The matching model is trained with the training objectives of maximizing the matching degree corresponding to the positive samples and minimizing the matching degree corresponding to the negative samples.

[0095] In a third aspect, an embodiment of the present application provides a search method, including:

[0096] Get search information;

[0097] Based on the search information, crawling a first page that matches the search information and a second page pointed to by an internal link in the first page; wherein the first page and the second page are list pages or detail pages; wherein the internal link determines by associating and converting multiple matching pairs according to at least one page link relationship; wherein the matching pair consists of an object and key information that has a matching relationship with the object;

[0098] Including the first page and the second page in the search results;

[0099] Determine a display result from the search results.

[0100] In a fourth aspect, an embodiment of the present application provides a computing device, including a processing component and a storage component;

[0101] The storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the internal link generation method described in the first aspect above, or the data processing method described in the second aspect above, or the search method described in the third aspect above.

[0102] In a fifth aspect, an embodiment of the present application provides a computer-readable storage medium storing a computer program. When the computer program is executed by a computer, it implements the internal link generation method described in the first aspect above, the data processing method described in the second aspect above, or the search method described in the third aspect above.

[0103] In an embodiment of the present application, by forming multiple groups of matching pairs consisting of objects with matching relationships and key information, and according to at least one page link relationship corresponding to the list page and the detail page, the multiple groups of matching pairs are associated and converted, so that any two pages that meet any page link relationship and the key information used as internal links between the two pages can be obtained. By combining the matching relationship between the objects and the key information to generate internal links, the correlation between pages corresponding to different page link relationships is guaranteed, which helps to improve the website weight and the website inclusion rate. Therefore, the embodiment of the present application improves the rationality of the generated internal links and ensures the rationality of the internal link placement.

[0104] These and other aspects of the present application will become more readily apparent from the description of the following embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0105] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, a brief introduction will be given below to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0106] FIG1 shows a flow chart of an embodiment of an internal link generation method provided by the present application;

[0107] FIG2a shows a schematic diagram of a page link relationship conversion in a practical application of an embodiment of the present application;

[0108] FIG2 b shows another schematic diagram of page link relationship conversion in an actual application of an embodiment of the present application;

[0109] FIG2c shows another schematic diagram of page link relationship conversion in an actual application of an embodiment of the present application;

[0110] FIG2 d shows another schematic diagram of page link relationship conversion in an actual application of an embodiment of the present application;

[0111] FIG3 shows a flow chart of an embodiment of a data processing method provided by the present application;

[0112] FIG4 shows a schematic diagram of scene interaction in an actual application of an embodiment of the present application;

[0113] Figures 5a and 5b are schematic diagrams of interface display in a practical application of the embodiment of the present application;

[0114] FIG6 shows a flowchart of an embodiment of a search method provided in an embodiment of the present application;

[0115] FIG7 shows a schematic structural diagram of an embodiment of an internal link generation device provided by the present application;

[0116] FIG8 shows a schematic structural diagram of an embodiment of a data processing device provided by the present application;

[0117] FIG9 shows a schematic structural diagram of an embodiment of a search device provided by the present application;

[0118] FIG10 shows a schematic structural diagram of an embodiment of a computing device provided by the present application. DETAILED DESCRIPTION

[0119] In order to enable those skilled in the art to better understand the solution of the present application, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application.

[0120] In some of the processes described in the specification and claims of this application and the above-mentioned figures, multiple operations that appear in a specific order are included, but it should be clearly understood that these operations may not be executed in the order in which they appear in this document or may be executed in parallel. The serial numbers of the operations, such as 101, 102, etc., are only used to distinguish between different operations, and the serial numbers themselves do not represent any order of execution. In addition, these processes may include more or fewer operations, and these operations may be executed in sequence or in parallel. It should be noted that the descriptions of "first", "second", etc. in this document are used to distinguish different messages, devices, modules, etc., and do not represent a sequential order, nor do they limit "first" and "second" to being different types.

[0121] The technical solution of the embodiment of the present application can be applied to the internal link construction scenario of the website, especially the website that provides objects such as goods and allows users to perform purchases, browsing or other interactive behaviors with the objects.

[0122] Websites typically include two types of pages that require internal linking: list pages and detail pages. List pages can be content aggregation pages containing multiple list elements. Internal links placed on list pages can be used as hyperlinks outside of the list elements within the list page. Detail pages are used to describe object information. Internal links placed on detail pages can be used as hyperlinks outside of the object information.

[0123] Since a reasonable internal link structure can recommend valuable list pages or detail pages to search engines, improve page rankings, and thus bring more traffic to the website, how to generate reasonable internal links and use them for internal link delivery has become a technical problem that needs to be solved.

[0124] In the process of implementing this application, the inventors discovered that the current page link relationships mainly include list page linking to detail page, list page linking to list page, detail page linking to detail page, and detail page linking to object page, etc., and the relevant key information between pages needs to be used as internal links for hyperlink connection. For example, when a list page links to a detail page, the list page needs to carry key information pointing to the detail page. If internal links are determined separately for different page link relationships, such as when a list page links to a detail page, it is necessary to determine the degree of matching between the list page and the key information, and the degree of matching between the key information and the detail page, etc., and a model is used for matching, the four types of page link relationships require the training of multiple models, resulting in low internal link generation efficiency and high cost.

[0125] In order to improve the rationality of internal links, ensure the efficiency of internal link generation, etc., the inventor has proposed the technical solution of the embodiment of the present application after a series of studies. In the embodiment of the present application, it is only necessary to determine the objects and key information with a matching relationship, so that the matching pairs of objects and key information can be converted into each other under the corresponding relationship between key information and list pages, so as to obtain two pages corresponding to different page link relationships, and key information between the two pages used as internal links. By generating internal links based on the matching relationship between objects and key information, the correlation between pages corresponding to different page link relationships is guaranteed, which helps to improve the weight of the website and the inclusion rate of the website. Therefore, the embodiment of the present application improves the rationality of the generated internal links, ensures the rationality of the internal link placement, and only needs to identify the matching relationship between objects and key information, thereby ensuring the efficiency of internal link generation and reducing the cost of internal link generation.

[0126] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0127] The following will clearly and completely describe the technical solutions in the embodiments of this application in conjunction with the accompanying drawings. Obviously, the described embodiments are only some of the embodiments of this application, not all of them. All other embodiments derived by those skilled in the art based on the embodiments of this application without inventive effort shall fall within the scope of protection of this application.

[0128] FIG1 is a flowchart of an embodiment of an internal link generation method provided in an embodiment of the present application. The technical solution of this embodiment can be executed by a server, which may include a server that provides various services.

[0129] It should be noted that the server 102 can be implemented as a distributed server cluster consisting of multiple servers, or as a single server. The server can also be a server of a distributed system, or a server combined with a blockchain. The server can also be a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms, or an intelligent cloud computing server or intelligent cloud host with artificial intelligence technology.

[0130] The internal link generation method of the embodiment shown in FIG1 may include the following steps:

[0131] 101: Determine multiple groups of matching objects.

[0132] Each matching pair consists of an object and key information that has a matching relationship with the object.

[0133] The multiple matching pairs are identified and obtained from a key information library and an object library.

[0134] Among them, the key information database stores a large amount of key information. In practical applications, the key information used as an internal link can specifically refer to keywords. Of course, this application does not exclude that the key information can also be in other data forms, such as pictures, videos, etc.

[0135] The key information in the key information library can be derived from search records generated within the website and / or search records generated externally to the website. The search records generated externally to the website can include search records generated by search engines or search records generated by external websites associated with the website. The present embodiment utilizes not only in-site data but also external data, thereby increasing the coverage of key information and making the key information library more comprehensive.

[0136] The object library stores information related to objects provided by the website, such as object images, object titles, object categories, and / or object click information such as CPV (Cost Per Click), etc., which is not limited in this application. In actual applications, the objects involved in the embodiments of this application can refer to commodities.

[0137] There may be a one-to-one correspondence between the key information in the key information and the list page, and each key information may be used to point to a corresponding list page.

[0138] The search records generated within and / or outside the website may include key information and search results corresponding to the key information, and the search results may include list pages provided by the website content. Therefore, optionally, based on the search records generated within and / or outside the website, the key information in the search records and the list pages pointed to by the key information are determined; a correspondence between the key information and the list pages is established, and a key information library is constructed based on the key information.

[0139] Since the search results for the same key information may include multiple list pages, optionally, determining the key information in the search record and the list page to which the key information points includes: determining the key information in the search record and at least one candidate list page to which the key information points; and determining the list page corresponding to the key information based on click information of the at least one candidate list page.

[0140] That is, at least one candidate list page can be sorted based on click information, such as the number of clicks, and the candidate list page whose click information meets the screening conditions can be selected as the list page corresponding to the key information to be saved in the key information library. The screening condition can be, for example, the page with the most clicks.

[0141] In an embodiment of the present application, the key information matched by each object can be identified from the key information library and the object library, so that a group of matching pairs can be formed by an object with a matching relationship and a key information, and then multiple groups of matching pairs can be obtained. Assuming that q (query) represents the key information and i (item) represents the object, each group of matching objects can be expressed as i2q (i to q). It should be noted that the i2q involved in this article, as well as the "2" in L2i, i2L, i2i and L2L that may be involved in the following text, is used to represent "to (pair)", therefore, L2i is also L to i; i2L is also i to L; i2i is also i to i; L2L is also L to L.

[0142] Among them, the objects and key information with a matching relationship can be determined based on the feature similarity of the object features and the key information features. The feature similarity represents the degree of matching between the object and the key information. For example, when the object features and the key information features are represented in vector form, the degree of matching can be determined based on the vector distance. Generally, the smaller the vector distance, the greater the degree of matching.

[0143] Of course, the model can also be used to identify objects with matching relationships and key information.

[0144] The matching pairs can be pre-generated and stored in a storage system, thereby determining multiple sets of matching objects from the storage system. The storage system can be a database or other network storage system. The specific method for obtaining the matching pairs can be described in detail in the embodiments below and will not be repeated here.

[0145] 102: Determine at least one page link relationship corresponding to the list page and the detail page.

[0146] In actual applications, the at least one page link relationship may include: a list page links to a detail page, a list page links to a list page, a detail page links to a detail page, and a detail page links to a list page.

[0147] The two pages involved in each page link relationship need to be connected through key information as internal links. For example, when a list page links to a detail page, it is necessary to determine the internal link placed in the list page to point to the detail page; when a list page links to a list page, it is necessary to determine the internal link placed in one list page to point to the other list page; when a detail page links to a detail page, it is necessary to determine the internal link placed in one detail page to point to the other detail page; when a detail page links to a list page, it is necessary to determine the internal link placed in the detail page to point to the list page.

[0148] 103: According to the at least one page link relationship, multiple matching pairs are associated and converted to obtain two pages corresponding to the at least one page link relationship and key information between the two pages used as an internal link.

[0149] In an embodiment of the present application, multiple groups of matching pairs can be associated through object i. There may be multiple key information q that have a matching relationship with the same object i, and each object i can correspond to a details page; based on the correspondence between the key information and the list page, the list page pointed to by each key information can be determined. Therefore, by performing association conversion on multiple groups of matching pairs, the two pages corresponding to at least one of the above-mentioned page link relationships and the key information used as an internal link between the two pages can be obtained.

[0150] The embodiment of the present application is based on matching pairs consisting of objects and key information in a matching relationship. By performing association conversion on multiple groups of matching pairs, any two pages corresponding to at least one of the above-mentioned page link relationships and the key information between the two pages used as internal links can be determined. The matching relationship between the object and the key information ensures the correlation between the pages corresponding to different page link relationships, which helps to improve the website weight and the website inclusion rate. Therefore, the rationality of the generated internal links is improved, and internal links can be generated by only determining the objects and key information in a matching relationship, thereby improving the efficiency of internal link generation.

[0151] For ease of understanding, the association conversion process for different page link relationships is introduced below.

[0152] In an optional method, at least one page link relationship may include a list page linking to a details page. Assuming that L is used to represent the list page, and each object i corresponds to a details page, the list page linking to the details page can be expressed as: L2i. Since it is necessary to determine the internal link q carried in the list page that links to the details page, the actual conversion result corresponding to L2i is: L2q2i.

[0153] The aforementioned associative conversion of multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages may include:

[0154] From multiple sets of matching pairs, determine multiple key information that have a matching relationship with the same object; determine the first key information and the second key information based on the matching degree between the multiple key information and the object; based on the correspondence between the key information and the list page, determine the first list page pointed to by the first key information and the first details page corresponding to the object, and use the second key information as an internal link in the first list page.

[0155] Among them, the embodiment of the present application can determine the degree of matching between a group of objects and key information that have a matching relationship.

[0156] Since the key information of this application corresponds to a list page, and the degree of matching between the object and the key information can be determined, for multiple matching pairs with the same object characteristics, the degree of matching can be combined to determine the first key information used to point to the list page and the second key information used to point to the details page in the multiple matching pairs.

[0157] For example, in the association conversion diagram shown in Figure 2a, for multiple matching pairs i2q with the same object characteristics, according to the matching degree between different key information and object i, it is assumed that the first key information q1 matching the same object i can be determined, and the list page corresponding to the first key information q1 is L1, and the list page corresponding to the first key information q1 is L1. Multiple matching pairs can be associated through object i to first obtain: L12i2q, and the second key information q2 can be determined through deduplication processing and other operations. Therefore, it can be further converted to obtain: L12q22i.

[0158] In some embodiments, determining the first key information and the second key information based on the matching degree between the plurality of key information and the object includes:

[0159] According to the matching degrees between the plurality of key information and the object, first key information whose matching degree meets the first matching requirement is determined, and second key information whose matching degree meets the second matching requirement is determined.

[0160] The first matching requirement may refer to the highest matching degree, for example.

[0161] Optionally, if there is one key information other than the first key information among the multiple key information, the key information may be used as the second key information. In addition, there may be two or more key information, and duplicate removal may be performed. Therefore, the second key information whose matching degree is determined to meet the second matching requirement may include:

[0162] When there are at least two key information in addition to the first key information in the plurality of key information, the key information with the highest matching degree is selected as the second key information.

[0163] In addition, according to the above internal link generation logic, the first list page may be determined to correspond to multiple internal links. If the multiple internal links are the same but point to the same object, deduplication processing is also required. Therefore, in some embodiments, the method may further include:

[0164] If it is determined that the first list page corresponds to multiple internal links, and that these multiple internal links point to the first detail page of the same object, at least one target internal link is selected and retained in descending order of matching degree with the object. This at least one target internal link is used as the final association conversion result for updating the internal link structure, so that only at least one target internal link from the multiple internal links is displayed on the first list page.

[0165] As another optional method, at least one page link relationship includes a details page linking to a list page. Assuming that L is used to represent a list page, and each object i can correspond to a details page, the details page linking to the list page can be expressed as: i2L. Since it is necessary to determine the internal link q carried in the details page that links to the list page, the actual corresponding association conversion result of i2L is: i2q2L.

[0166] Since the objects and key information that have a matching relationship have been determined, the multiple matching pairs are associated and converted according to the at least one page link relationship to obtain the key information used as internal links between the pages corresponding to the at least one page link relationship, which may include:

[0167] From multiple sets of matching pairs, find the target matching pair according to the matching degree between key information and objects;

[0168] Determine the target list page pointed to by the target key information in the target matching pair, the target details page corresponding to the target object, and use the target key information as an internal link in the target details page.

[0169] The target matching pair may refer to a matching degree greater than a specified threshold, for example.

[0170] As another optional method, searching for a target matching pair from the multiple groups of matching pairs according to the matching degree between the key information and the object can be: from the multiple groups of matching pairs, for any object, determining multiple key information that has a matching relationship with the object; and determining, based on the matching degree between the multiple key information and the object, that the key information with the highest matching degree constitutes a target matching pair with the object.

[0171] In the key transformation diagram shown in FIG2b, for multiple matching pairs i2q, duplicate removal can be performed according to the matching degree between the key information and the object to obtain the target matching pair i2q. t 2q t , since there is a corresponding relationship between key information and list pages, the target key information q can be determined based on this t The target list page L t , target key information q t That is, it can be used as an object i t The internal link in the target details page pointed to, so the corresponding association conversion result is i t 2q t 2L t .

[0172] As another optional method, at least one page link relationship includes a details page linking to a details page. Assume that L is used to represent a list page, and each object i can correspond to a details page. The details page linking to the details page can be expressed as: i2i. Since it is necessary to determine the internal link q carried in the details page that links to the details page, the actual conversion result corresponding to i2i is: i2q2i.

[0173] The aforementioned associative conversion of multiple matching pairs according to the at least one page link relationship to obtain key information used as internal links between pages corresponding to the at least one page link relationship may include:

[0174] determining a first object and a second object that satisfy a correlation requirement;

[0175] Searching for first key information that has a matching relationship with the first object from the multiple matching pairs;

[0176] A first details page corresponding to the first object and a second details page corresponding to the second object are determined, and the first key information is used as an internal link in the second details page.

[0177] The relevance requirements may include, for example, the same category and / or feature matching.

[0178] The first object may satisfy any one of the two objects with the correlation requirement in real time.

[0179] There are multiple key information that have a matching relationship with the first object, and the key information with the highest matching degree with the first object can be selected as the first key information.

[0180] For ease of understanding, in the association conversion diagram shown in Figure 2c, first determine the first object and the second object that meet the correlation requirements: i12i2. Assuming that the first object i1 matches the first key information q1, that is, i12q1, when the first object i1 and the second object i2 meet the correlation requirements, they are associated through object i1, and the association conversion result: i12q12i2 can be obtained.

[0181] As another optional method, at least one page link relationship includes a list page linking a list page. Assume that L is used to represent the list page, and each object i can correspond to a details page. The list page linking to the list page can be expressed as: L2L. Since it is necessary to determine the internal link q carried in the list page that links to the list page, the actual conversion result corresponding to L2L is: L2q2L.

[0182] The aforementioned associative conversion of multiple matching pairs according to the at least one page link relationship to obtain pages corresponding to the at least one page link relationship and key information used as internal links may include:

[0183] From multiple sets of matching pairs, search for the first key information and the second key information that have a matching relationship with the same object; based on the correspondence between the key information and the list page, determine that the first key information points to the first list page, the second key information points to the second list page, and use the first key information as an internal link in the second list page.

[0184] By using the same object, two related key information can be determined, and a link relationship can be established between the two list pages corresponding to the two key information.

[0185] Among the multiple key information matched with the same object, any one key information can be selected as the first key information, and any other key information can be selected as the second key information.

[0186] In addition, as another optional method, searching for the first key information and the second key information that have a matching relationship with the same object from multiple matching pairs may include:

[0187] From multiple sets of matching pairs, multiple key information that have a matching relationship with the same object are searched; according to the matching degrees between the multiple key information and the object, the first key information whose matching degree meets the third matching requirement and the second key information whose matching degree meets the fourth matching requirement are determined.

[0188] The third matching requirement may be, for example, a maximum matching degree. When the matching degree is represented by vector distance, it may be a minimum vector distance.

[0189] In some embodiments, determining the second key information whose matching degree meets the fourth matching requirement may include: when there are at least two key information in addition to the first key information in the multiple key information, selecting the key information with the highest matching degree as the second key information.

[0190] Through deduplication processing, the first key information and the second key information can be determined.

[0191] In addition, in some embodiments, the method may further include: when it is determined that the first list page corresponds to multiple internal links, and the multiple internal links point to the same second list page, screening and retaining at least one target internal link in order of the degree of matching with the respective corresponding objects from highest to lowest.

[0192] The at least one target internal link is used as the final association conversion result to update the internal link structure, so that only at least one target internal link among the multiple internal links can be placed in the first list page.

[0193] For example, in the association conversion diagram shown in Figure 2d, for multiple groups of matching pairs i2q with the same object characteristics, by associating through object i, the first key information q1 and the second key information q2 that have a matching relationship with the same object i can be determined. Assuming that the first list page corresponding to the first key information q1 is L1, multiple groups of matching pairs can be associated through object i to first obtain: L12i2q2. The second key information q2 can be determined through deduplication processing and other operations. The second list page corresponding to the second key information q2 is L2. Then, the two groups of matching pairs matching the same object can be associated and converted to obtain: L12q22L2.

[0194] In some embodiments, objects and key information that have a matching relationship can be identified based on key information features and object features to obtain multiple matching pairs. This can be done from the perspective of the object, by finding multiple key information that has a matching relationship with the object, thereby forming matching pairs between the object and the multiple key information, or from the perspective of the key information, by finding multiple objects that have a matching relationship with the key information, thereby forming matching pairs between the key information and the multiple objects.

[0195] Since the key information library and the object library contain a large amount of data, often in the billions, in order to improve matching efficiency, as an optional method, the above-mentioned method of determining objects and key information with matching relationships from the key information library and the object library based on the object characteristics and key information characteristics to obtain multiple sets of matching pairs may include:

[0196] Clustering multiple key information in the key information database according to key information characteristics to obtain multiple information groups;

[0197] According to the object characteristics, any object in the object library is matched with multiple information groups respectively to determine at least one matched target information group;

[0198] The objects are matched respectively with key information in at least one target information group to determine at least one key information matching the object, and the object and the at least one key information form at least one matching pair.

[0199] Specifically, the key information library is first grouped by clustering to obtain multiple information groups. Each object can then be matched with the information group, first determining at least one target information group to match, and then the object is matched with each key information contained in the at least one target information group. This can reduce the number of matches and the amount of computation. Alternatively, a graphics processing unit (GPU) can be used to cluster the multiple key information in the key information library to obtain multiple information groups, thereby further accelerating the matching operation and improving matching efficiency.

[0200] Among them, the matching of objects and information groups can be performed by combining the object characteristics and the grouping characteristics of the information grouping. Among them, the grouping characteristics can be calculated based on the various key information characteristics included in the information grouping. For example, it can be the average value or sum of the various key information characteristics, etc. This application does not limit this.

[0201] In some embodiments, the method may further include: generating group features corresponding to the plurality of information groups based on a key feature of at least one key information respectively included in the plurality of information groups; using the group features of the plurality of information groups as respective index identifiers;

[0202] For example, the average value or sum of key information features of at least one key information included in each information group may be used as the grouping feature.

[0203] The grouping characteristics of each information grouping can be used as the index identifier of each information grouping, so that any object in the object library is matched with multiple information groups according to the object characteristics. Determining the matched target information grouping may include: screening the target index identifier that matches any object according to the object characteristics; and determining the target information group corresponding to the target index identifier.

[0204] The target index identifier can be filtered according to the matching degree between the object characteristics and different index identifiers.

[0205] The above-mentioned object features or key information features can be represented in vector form, and the degree of matching between the object and the key information can be represented by vector distances such as Euclidean distance and Manhattan distance.

[0206] In addition, as another optional manner, the above-mentioned identifying objects and key information having matching relationships based on key information features and object features to obtain multiple sets of matching pairs may include:

[0207] Cluster multiple objects in the object library according to object features to obtain multiple object groups;

[0208] Matching the key information in the key information database with the plurality of object groups according to the key information characteristics, and determining at least one matched target object group;

[0209] The key information is matched with objects in at least one target object group to determine at least one object matching the key information, and the key information and the at least one object form at least one matching pair.

[0210] Specifically, the object library is first grouped using a clustering approach to obtain multiple object groups. Each key information item is then matched against the object group, and at least one target object group is determined to match. The key information item is then matched against each object contained in the at least one target object group. This reduces the number of matches and computational complexity. Alternatively, a graphics processing unit (GPU) can be used to cluster multiple objects in the object library to obtain multiple object groups, further accelerating the matching process and improving matching efficiency.

[0211] Among them, the key information is matched with the object, which can be performed by combining the object characteristics and the grouping characteristics of the object grouping. Among them, the grouping characteristics can be calculated based on the various object characteristics included in the object grouping. For example, it can be the average value or sum of the various object characteristics, etc. This application does not limit this.

[0212] In some embodiments, the method may further include: generating group features corresponding to the plurality of object groups based on the object features of at least one object respectively included in the plurality of object groups; using the group features of the plurality of object groups as respective index identifiers;

[0213] For example, the average value or sum of the object features of at least one object included in each object group may be used as the group feature.

[0214] The grouping characteristics of each object group can be used as the index identifier of each object group, so that any key information in the key information library is matched with multiple object groups according to the key information characteristics. Determining the matched target object group may include: screening the target index identifier that matches the key information according to the key information characteristics; and determining the target object group corresponding to the target index identifier.

[0215] The target index identifier can be filtered according to the matching degree between the key information characteristics and different index identifiers.

[0216] The above-mentioned object features or key information features can be represented in vector form, and the degree of matching between the object and the key information can be represented by vector distances such as Euclidean distance and Manhattan distance.

[0217] The object features or key information features may be extracted using a bag-of-words model or an N-gram (a Chinese language model). In order to improve feature accuracy and matching efficiency, in some embodiments, the method may further include: extracting key information features corresponding to multiple key information in the key information library using a matching model; extracting respective object features from the relevant information of multiple objects in the object library using the matching model;

[0218] Among them, the matching model is trained based on multiple groups of sample matching pairs using a contrastive learning method; the sample matching pairs include sample key information and sample objects.

[0219] For example, the matching model can be implemented using simcse (Simple Contrastive Learning of Sentence Embeddings, simple contrastive learning sentence vector model), etc. Of course, this application is not limited to this.

[0220] In some embodiments, the matching model can be trained in the following manner:

[0221] Determine multiple sets of sample matching pairs corresponding to the current training batch;

[0222] Input multiple sets of sample matching pairs into the matching model;

[0223] Any sample matching pair is taken as a positive sample, the sample key information in any sample matching pair and the sample objects in at least one other sample matching pair are taken as negative samples, and the matching model is trained using a sampling contrast learning method.

[0224] In some embodiments, any one sample matching pair is used as a positive sample, and the sample key information in any one sample matching pair and the sample object in at least one other sample matching pair are used as negative samples. Training the matching model using a sampling contrast learning method may include:

[0225] The matching model is used to extract the sample key information features corresponding to the sample key information in the multiple groups of sample matching pairs, and the sample object features corresponding to the sample objects;

[0226] Any sample matching pair is regarded as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one other sample matching pair are regarded as negative samples;

[0227] Based on the sample information features and sample object features, calculate the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample;

[0228] The matching model is trained to maximize the matching degree corresponding to the positive samples and minimize the matching degree corresponding to the negative samples.

[0229] The matching model can be trained using a self-supervised learning method, so that the matching model can extract key information features and object features respectively.

[0230] In addition, in some embodiments, identifying objects and key information with matching relationships from the key information library and the object library to obtain multiple sets of matching pairs can be achieved by using a recognition model to identify objects and key information with matching relationships from the key information library and the object library to obtain multiple sets of matching pairs.

[0231] That is, the relevant information and key information of any object can be input into the recognition model, and the recognition model can identify whether there is a matching relationship between the two.

[0232] The recognition model can be implemented using a neural network model, etc., and can be trained using a supervised training method, or can be trained using a self-supervised training method using a contrastive learning method, etc. When training using a contrastive learning method, the recognition model and the matching model can be trained in the same manner. The matching model can serve as a shared network architecture to extract corresponding features of key information and objects, respectively, while the recognition model can be used to calculate the degree of matching between the key information and the object.

[0233] In some embodiments, the method may further include:

[0234] Page configuration is performed according to the associated conversion results corresponding to at least one page link relationship to update the internal link structure of the website.

[0235] Among them, the associated conversion result corresponding to the page link relationship between the list page and the details page can include L12q22i, that is, the first list page L1 pointed to by the first key information, the first details page corresponding to the object i matched by the second key information, and the second key information q2 as an internal link in the first list page.

[0236] The associated conversion result corresponding to the page link relationship of the detail page link list page can include i t 2q t 2L t , that is, the target object i t Details page, target key information t Pointing to the target list page L t , and target key information q t As an internal link in the detail page;

[0237] The associated conversion result corresponding to the page link relationship of detail page linking to detail page may include i12q22i2, that is, the first detail page of the first object i1, the second detail page of the second object i2, and the second key information q2 matching the second object i2 as an internal link in the first detail page;

[0238] The associated conversion result corresponding to the page link relationship of the list page linking to the list page can include: L12q22L2, that is, the first list page pointed to by the first key information, the second list page pointed to by the second key information, and the second key information as an internal link in the first list page.

[0239] According to the above-mentioned association conversion results, page configuration can be performed, thereby updating the internal link structure of the website.

[0240] In some embodiments, the method may further include:

[0241] The associated conversion results corresponding to at least one page link relationship are sent to the target person; the associated conversion results are used to generate page configuration information, and the page configuration information is used to update the internal link structure of the website.

[0242] That is, the target personnel can configure the page according to the associated conversion results to obtain page configuration information, and the page configuration information can be configured in the website so that the website can update the internal link structure.

[0243] In some embodiments, identifying objects and key information having matching relationships from a key information library and an object library to obtain multiple sets of matching pairs includes:

[0244] In response to the page update instruction, objects and key information having a matching relationship are identified from the key information library and the object library to obtain a plurality of matching pairs.

[0245] As an optional method, the page update instruction can be generated periodically, so that the internal link structure can be updated regularly.

[0246] In addition, as another optional method, the click rate of the website may be counted, and a page update instruction may be generated based on the click rate satisfying the first update condition;

[0247] The click-through rate can represent the result of internal link optimization to a certain extent. If the click-through rate increases, it means that the internal link optimization is effective. The first update condition may, for example, mean that the click-through rate is lower than the click threshold.

[0248] The click-through rate can be calculated based on the number of clicks within the statistical period.

[0249] In addition, as another optional method, the crawling rate and / or collection rate corresponding to the website may be counted, and a page update instruction may be generated based on whether the crawling rate and / or collection rate meets the second update condition.

[0250] The crawl rate and / or index rate of the website to a certain extent indicates the results of internal link optimization. The second update condition may be, for example, that the crawl rate is lower than the crawl threshold and / or the index rate is lower than the index threshold.

[0251] The crawling rate can be calculated based on the number of crawls within the statistical period, and the inclusion rate can be calculated based on the number of inclusions within the statistical period, etc.

[0252] FIG3 is a flow chart of an embodiment of a data processing method provided by an embodiment of the present application. This embodiment introduces the technical solution of the embodiment of the present application from the perspective of model training. The technical solution of this embodiment can be executed by the server. The method may include the following steps:

[0253] 301: Obtain multiple sets of sample matching pairs.

[0254] The sample matching pair may include sample key information and sample objects.

[0255] It can be from historical record data, the sample key information can refer to historical search information, and the sample object can refer to the object corresponding to the detail page pointed to by the click behavior in the search results corresponding to the historical search information.

[0256] 302: Multiple groups of sample matching pairs are used as training data for the current training batch and input into the matching model.

[0257] Model training is performed in batches (bath), and each training batch corresponds to multiple groups of sample matching objects.

[0258] 303: Taking any sample matching pair as a positive sample, the sample key information in any sample matching pair and the sample object in at least one other sample matching pair as negative samples, and training the matching model using a sampling contrast learning method.

[0259] For example, assuming the first sample matching pair is a positive sample, the sample key information in the first sample matching pair can be combined with the sample objects in the remaining sample matching pairs except the first sample matching pair to form a negative sample. During the matching model training process, positive samples and negative samples can be constructed to achieve training.

[0260] Among them, the matching model is used to identify objects and key information with matching relationships in the key information library and the object library to obtain multiple sets of matching pairs; multiple sets of matching objects are used to perform association conversion to obtain any two pages corresponding to at least one page link relationship and the key information between the two pages used as internal links.

[0261] The matching model can be used to calculate the matching degree between the object and the key information, thereby determining whether there is a matching relationship between the object and the key information based on the matching degree.

[0262] In addition, as another optional method, the matching model can be specifically used to extract key information features corresponding to multiple key information in the key information library, and for multiple objects in the object library, use the matching model to extract respective object features from the relevant information of multiple objects; key information features and object features are used to determine objects and key information that have matching relationships in the key information library and the object library to obtain multiple groups of matching pairs.

[0263] In some embodiments, any one sample matching pair is used as a positive sample, and the sample key information in any one sample matching pair and the sample object in at least one other sample matching pair are used as negative samples. Training the matching model using a sampling contrast learning method may include:

[0264] The matching model is used to extract the sample key information features corresponding to the sample key information in the multiple groups of sample matching pairs, and the sample object features corresponding to the sample objects;

[0265] Any sample matching pair is regarded as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one of the other sample matching pairs are regarded as negative samples;

[0266] Based on the sample information features and sample object features, calculate the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample;

[0267] The matching model is trained with the goal of maximizing the matching degree corresponding to the positive samples and minimizing the matching degree corresponding to the negative samples.

[0268] The training goal is to maximize the matching degree corresponding to the positive samples and minimize the matching degree corresponding to the negative samples, which can be achieved by using a loss function.

[0269] In a practical application, the embodiment of the present application can be applied to an e-commerce website scenario. The e-commerce website can provide products for users to purchase. The product website mainly includes a list page and a product details page. The product details page is used to introduce product-related information, etc., so that users can understand the products, and can provide some product processing controls, such as add to cart, purchase, favorites, etc., for users to trigger corresponding operations. The list page can, for example, refer to a product aggregation page, which carries product prompt information that can be linked to different product detail pages. In the embodiment of the present application, the keywords pointing to the product details page can be used as product prompt information in the list page, and of course can also be used as hyperlinks after the product prompt information.

[0270] During the information exchange process between the client and the server, the list page and the detail page are sent from the server to the client and displayed to the user by the client. The user can operate the internal links in the list page or the detail page to jump to other pages.

[0271] The technical solution of the present application is introduced below using the e-commerce website scenario as an example. The keyword may specifically refer to a keyword. Referring to the interaction diagram shown in FIG4 , the technical solution of the embodiment of the present application can be executed by the server, and the server can be composed of a distributed server cluster. A flexible scheduling method is adopted. In the event of a failure in any server execution, the execution task can be transferred to another server for continued execution.

[0272] As shown in Figure 4, the model training operation can be performed by the first server 41: the sample products and the sample keywords that have a matching relationship with the sample products can constitute a sample matching pair 401, and each sample matching pair is input into the matching model as a positive sample. The matching model can construct a negative sample and, based on the positive samples and the negative samples, a comparative learning method is used to train to obtain a matching model 402.

[0273] The second server 42 can use the matching model obtained through training to extract keyword features 403 corresponding to multiple keywords in the keyword library. Then, based on the keyword features, the multiple keywords in the keyword library can be clustered to obtain multiple information groups. Based on the keyword features of at least one keyword respectively included in the multiple information groups, group features corresponding to the multiple information groups can be generated respectively, and the group features of the multiple information groups can be used as their respective index identifiers 404, wherein the index identifier corresponding to each information group can be saved 405, for example, it can be saved to a network storage system.

[0274] Afterwards, the third server 43 can use the matching model to extract the respective product features 406 from the relevant information of the multiple products in the product library using the matching model; based on the product features, any product in the product library can be matched with the index identifiers of the multiple information groups, and at least one target information group to be matched is determined, and then the products are matched with each keyword part in the at least one target information group, and can be sorted in order from high to low according to the degree of matching 407, and the top N keywords that match the object are selected 408, where N is an integer greater than or equal to 1. The top N keywords can be uploaded to the fourth server 44. The fourth server 44 can be implemented, for example, using odps (Open Data Processing Service, open data processing service), etc.

[0275] The fourth server 44 can combine the correspondence between keywords and list pages to perform association conversion 409 on multiple matching pairs of products and keywords, thereby obtaining two pages involved in different page link relationships and keywords 410 used as internal links between the two pages.

[0276] The fourth server 44 can send the association conversion results corresponding to different page link relationships to the configuration node 45. The target person can perform configuration operations in the configuration node 45 to generate page configuration information 411, which can be used to update the internal link structure of the e-commerce website.

[0277] Figure 5a shows a display diagram of an internal link 51 placed in a list page 50, where the internal link 51 can point to another list page or a product details page, etc.; Figure 5b shows a display diagram of an internal link 53 placed in a product details page 52, where the internal link 53 can point to a list page or another product details page.

[0278] In the embodiment of the present application, the four types of page link relationships involved in the e-commerce website can be attributed to the matching of goods and keywords, so that only the matching problem of goods and keywords needs to be solved, and the pages corresponding to the four types of page link relationships and the keywords between the pages can be obtained through association conversion. The internal links are generated according to the matching relationship between the objects and the keywords, which ensures the correlation between the pages corresponding to the different page link relationships, helps to improve the website weight, improve the website inclusion rate, and thus improve the rationality of the generated internal links. And the matching model can use the comparative learning method to complete the model training, billion-level index construction and billion-level product library model prediction, etc. The overall process runs for half a day to complete. On the basis of meeting the relevance requirements, the overall internal chain island rate has dropped significantly.

[0279] In addition, as shown in FIG6 , the embodiment of the present application further provides a search method, which can be performed by a search engine. The method may include the following steps:

[0280] 601: Get search information.

[0281] In actual application, the search information may refer to a search term input by a user.

[0282] 602: Based on the search information, crawl a first page that matches the search information and a second page pointed to by an internal link in the first page.

[0283] Among them, the first page and the second page are list pages or detail pages; wherein, the internal link determines the association conversion of multiple groups of matching pairs based on at least one page link relationship; the matching pair is composed of an object and key information that has a matching relationship with the object; the specific method of generating the internal link can be seen in the embodiment shown in Figure 1, and will not be repeated here.

[0284] When an internal link is placed on the first page, the second page pointed to by the internal link will also be crawled by the search engine.

[0285] 603: Include the first page and the second page in the search results.

[0286] 604: Determine the display result from the search results.

[0287] The captured first page and second page may be included in the search results.

[0288] The search results include multiple pages crawled from different websites, and the multiple pages include a first page and a second page from the same website.

[0289] Among them, the display results are determined from the search results, and the search engine can first sort the search results, that is, sort the first page and the second page in the search results. Among them, the search engine can first calculate the page weight of each page, and can sort the search results according to the page weight. Since the calculation of page weight is related to the correlation between pages, the more accurate the internal links placed in the first page, the higher the page weight score of the first page will be. Therefore, the sorting priority in the sorting process will be higher, thereby increasing the probability of the page being exposed and achieving SEO.

[0290] After the search engine sorts the search results, it can determine the display results based on the sorting results. The display results can be sent to the user end, and the user end displays the display results. The display results include page prompt information of one or more pages filtered according to the sorting priority.

[0291] FIG7 is a schematic diagram of the structure of an embodiment of an internal link generation device provided by an embodiment of the present application. The device may include:

[0292] Identification module 701, used to identify objects and key information that have matching relationships from the key information library and the object library to obtain multiple matching pairs;

[0293] A relationship determination module 702 is used to determine at least one page link relationship corresponding to a list page and a detail page;

[0294] The conversion module 703 is used to perform association conversion on the multiple matching pairs according to the at least one page link relationship to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages.

[0295] In some embodiments, the identification module is specifically used to cluster multiple key information in the key information library according to key information features to obtain multiple information groups; match any object in the object library with the multiple information groups according to object features to determine at least one matched target information group; match the object with the key information in the at least one target information group to determine at least one key information that matches the object, and the object and the at least one key information constitute at least one matching pair.

[0296] In some embodiments, the present invention further comprises:

[0297] The feature extraction module is used to extract key information features corresponding to multiple key information in the key information library using a matching model; wherein the matching model is trained using a comparative learning method based on multiple groups of sample matching pairs; the sample matching pairs include sample key information and sample objects.

[0298] In some embodiments, the apparatus may further include:

[0299] The model training module is used to determine multiple groups of sample matching pairs corresponding to the current training batch; input the multiple groups of sample matching pairs into the matching model; take any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample objects in at least one of the remaining sample matching pairs as negative samples, and train the matching model using a sampling comparison learning method.

[0300] In some embodiments, the model training module takes any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one remaining sample matching pair as negative samples, and the sampling comparative learning method to train the matching model includes: using the matching model to respectively extract the sample key information features corresponding to the sample key information in multiple groups of sample matching pairs, and the sample object features corresponding to the sample objects; taking any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one remaining sample matching pair as negative samples; based on the sample information features and the sample object features, calculating the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample; training the matching model with maximizing the matching degree corresponding to the positive sample and minimizing the matching degree corresponding to the negative sample as the training goal.

[0301] In some embodiments, the apparatus may further include:

[0302] An index creation module, configured to generate group features corresponding to the plurality of information groups based on key information features of at least one key information respectively included in the plurality of information groups; and use the group features of the plurality of information groups as respective index identifiers;

[0303] The identification module matches any object in the object library with the multiple information groups according to the object characteristics, and determines the matched target information group, including: screening the target index identifier that matches any object according to the object characteristics; and determining the target information group corresponding to the target index identifier.

[0304] In some embodiments, at least one page link relationship includes a list page linking to a detail page;

[0305] The above-mentioned conversion module is specifically used to determine multiple key information that have a matching relationship with the same object from multiple matching pairs with the same object characteristics; determine the first key information and the second key information based on the degree of matching between the multiple key information and the object; combine the correspondence between the key information and the list page to determine the first list page pointed to by the first key information and the first details page of the object, and use the second key information as an internal link in the first list page.

[0306] In some embodiments, the conversion module determines the first key information and the second key information based on the degree of matching between the multiple key information and the object, including: determining the first key information whose matching degree meets the first matching requirement and determining the second key information whose matching degree meets the second matching requirement based on the degree of matching between the multiple key information and the object.

[0307] In some embodiments, the conversion module determines the second key information whose matching degree meets the second matching requirement, including: when there are at least two key information in the multiple key information in addition to the first key information, selecting the key information with the highest matching degree as the second key information.

[0308] In some embodiments, at least one page link relationship includes a detail page linking to a list page;

[0309] The conversion module is specifically used to search for target matching pairs from the multiple groups of matching pairs according to the degree of matching between the key information and the object; based on the correspondence between the key information and the list page, determine the target list page pointed to by the target key information in the target matching pair, the target details page of the target object, and use the target key information as an internal link in the target details page.

[0310] In some embodiments, the conversion module searches for a target matching pair from the multiple groups of matching pairs according to the degree of matching between the key information and the object, including: determining, from the multiple groups of matching pairs, multiple key information that have a matching relationship with any object; and determining, based on the degree of matching between the multiple key information and the object, that the key information with the highest degree of matching constitutes a target matching pair with the object.

[0311] In some embodiments, at least one page link relationship includes a details page linking to a details page;

[0312] The conversion module is specifically used to determine a first object and a second object that meet the correlation requirements; search for the first key information matching the first object from the multiple matching pairs; determine the first details page of the first object and the second details page of the second object, and use the first key information as an internal link in the second details page.

[0313] In some embodiments, at least one page link relationship includes a list page linking a list page;

[0314] The conversion module specifically searches for the first key information and the second key information that have a matching relationship with the same object based on multiple groups of matching pairs with the same object characteristics; combines the correspondence between the key information and the list page to determine the first list page pointed to by the first key information and the second list page pointed to by the second key information, and uses the first key information as an internal link in the second list page.

[0315] In some embodiments, the conversion module searches for the first key information and the second key information that have a matching relationship with the same object based on multiple groups of matching pairs with the same object characteristics, including: searching for multiple key information that have a matching relationship with the same object from multiple groups of matching pairs with the same object characteristics; and determining the first key information whose matching degree meets the third matching requirement and determining the second key information whose matching degree meets the fourth matching requirement according to the matching degree between the multiple key information and the object respectively.

[0316] In some embodiments, the conversion module determines the second key information whose matching degree meets the fourth matching requirement, including: when there are at least two key information in the multiple key information in addition to the first key information, selecting the key information with the highest matching degree as the second key information.

[0317] In some embodiments, the apparatus may further include:

[0318] A page configuration module is used to configure the page according to the associated conversion results corresponding to the at least one page link relationship to update the internal link structure of the website; or, to send the associated conversion results corresponding to the at least one page link relationship to the target person; the associated conversion results are used to generate page configuration information, and the page configuration information is used to update the internal link structure of the website.

[0319] In some embodiments, the apparatus may further include:

[0320] The information library construction module is used to determine the key information in the search records and the list page pointed to by the key information based on the search records generated inside and / or outside the website; establish the corresponding relationship between the key information and the list page, and construct a key information library based on the key information.

[0321] In some embodiments, the information library construction model determines the key information in the search record and the list page pointed to by the key information, including: determining the key information in the search record and at least one candidate list page pointed to by the key information; and determining the list page corresponding to the key information based on the click information of the at least one candidate list page.

[0322] In some embodiments, the identification module is specifically configured to identify objects and key information having a matching relationship from the key information library and the object library in response to a page update instruction, so as to obtain a plurality of matching pairs.

[0323] In some embodiments, the apparatus may further include:

[0324] The instruction generation module is used to count the click rate corresponding to the website, and generate the page update instruction based on the click rate satisfying the first update condition; or to count the crawl rate and / or collection rate corresponding to the website, and generate the page update instruction based on the crawl rate and / or collection rate satisfying the second update condition.

[0325] The internal link generation device shown in FIG7 can execute the internal link generation method described in the embodiment shown in FIG1 , and its implementation principles and technical effects are not further described. The specific manner in which the various modules and units in the internal link generation device of the above embodiment perform operations has been described in detail in the embodiment of the method, and will not be further elaborated here.

[0326] FIG8 is a schematic structural diagram of an embodiment of a data processing device provided in an embodiment of the present application. The device may include:

[0327] The sample determination module 801 is used to obtain multiple groups of sample matching pairs;

[0328] A sample input module 802 is configured to input the plurality of sample matching pairs as training data of a current training batch into a matching model;

[0329] A training module 803 is configured to train the matching model using a sampling comparison learning method, using any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one other sample matching pair as negative samples;

[0330] The matching model is used to identify objects and key information with matching relationships in the key information library and the object library to obtain multiple sets of matching pairs; the multiple sets of matching objects are used to perform association conversion to obtain any two pages corresponding to at least one page link relationship and the key information between the two pages used as internal links.

[0331] In some embodiments, the matching model is specifically used to extract key information features corresponding to multiple key information in the key information library, and for multiple objects in the object library, the matching model is used to extract respective object features from the relevant information of the multiple objects; the key information features and the object features are used to determine the objects and key information that have matching relationships in the key information library and the object library to obtain multiple groups of matching pairs.

[0332] In some embodiments, the training module takes any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one remaining sample matching pair as negative samples, and the sampling comparative learning method to train the matching model includes: using the matching model to respectively extract the sample key information features corresponding to the sample key information in multiple groups of sample matching pairs, and the sample object features corresponding to the sample objects; taking any sample matching pair as a positive sample, and the sample key information in any sample matching pair and the sample object in at least one remaining sample matching pair as negative samples; calculating the matching degree corresponding to the positive sample and the matching degree corresponding to the negative sample based on the sample information features and the sample object features; training the matching model with maximizing the matching degree corresponding to the positive sample and minimizing the matching degree corresponding to the negative sample as the training goal.

[0333] The data processing device shown in FIG8 can execute the data processing method described in the embodiment shown in FIG3 , and its implementation principle and technical effects are not further described. The specific manner in which each module and unit in the internal link generation device of the above embodiment performs operations has been described in detail in the embodiment of the method, and will not be further elaborated here.

[0334] FIG9 is a schematic diagram of the structure of an embodiment of a search device provided in an embodiment of the present application. The device may include:

[0335] Acquisition module 901, used to acquire search information;

[0336] The crawling module 902 is configured to crawl, based on the search information, a first page that matches the search information and a second page pointed to by an internal link in the first page; wherein the first page and the second page are list pages or detail pages; wherein the internal link determines a plurality of matching pairs by performing an association conversion based on at least one page link relationship; wherein the matching pair is composed of an object and key information that has a matching relationship with the object;

[0337] The collecting module 903 is used to collect the first page and the second page into the search results.

[0338] The display module 904 is configured to determine a display result from the search results.

[0339] The search device shown in FIG9 can execute the search method described in the embodiment shown in FIG6 , and its implementation principle and technical effects are not described in detail here. The specific manner in which each module and unit performs operations in the internal link generation device of the above embodiment has been described in detail in the embodiment of the method, and will not be elaborated on here.

[0340] The embodiment of the present application further provides a computing device, as shown in FIG10 , which may include a storage component 1001 and a processing component 1002 ;

[0341] The storage component 1001 stores one or more computer instructions, wherein the one or more computer instructions are called and executed by the processing component to implement the internal link generation method shown in Figure 1 or the data processing method shown in Figure 3 or the search method shown in Figure 6.

[0342] Of course, a computing device may also include other components, such as input / output interfaces, display components, communication components, etc.

[0343] The input / output interface provides an interface between the processing component and peripheral interface modules, which may be output devices, input devices, etc. The communication component is configured to facilitate wired or wireless communication between the computing device and other devices.

[0344] The processing component may include one or more processors to execute computer instructions to complete all or part of the steps in the above method. Of course, the processing component may also be implemented as one or more application-specific integrated circuits (ASICs), digital signal processors (DSPs), digital signal processing devices (DSPDs), programmable logic devices (PLDs), field programmable gate arrays (FPGAs), controllers, microcontrollers, microprocessors, or other electronic components to perform the above method.

[0345] The storage component is configured to store various types of data to support operations at the terminal. The storage component can be implemented by any type of volatile or non-volatile memory device, or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk, or optical disk.

[0346] It should be noted that the computing device can be a physical device or an elastic computing host provided by a cloud computing platform, etc. It can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device.

[0347] The present application also provides a computer-readable storage medium storing a computer program. When executed by a computer, the computer program can implement the internal link generation method of the embodiment shown in FIG. 1 , the data processing method of the embodiment shown in FIG. 3 , or the search method shown in FIG. 6 . The computer-readable medium can be included in the electronic device described in the above embodiments, or it can exist independently and not be incorporated into the electronic device.

[0348] The present application also provides a computer program product, which includes a computer program carried on a computer-readable storage medium. When the computer program is executed by a computer, it can implement the internal link generation method of the embodiment shown in Figure 1 above, the data processing method of the embodiment shown in Figure 3 above, or the search method shown in Figure 6. In such an embodiment, the computer program can be downloaded and installed from a network and / or installed from a removable medium. When the computer program is executed by a processor, it performs various functions defined in the system of the present application.

[0349] It should be noted that the embodiments of the present application may involve the use of user data. In actual applications, user-specific personal data can be used in the scheme described herein within the scope permitted by applicable laws and regulations, subject to the requirements of applicable laws and regulations of the country where the user is located (for example, with the user's explicit consent, effective notification to the user, etc.).

[0350] Those skilled in the art will clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0351] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they may be located in one location or distributed across multiple network units. Some or all of the modules may be selected based on actual needs to achieve the objectives of the present embodiment. Persons of ordinary skill in the art will be able to understand and implement the present invention without inventive effort.

[0352] Through the description of the above embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus a necessary general hardware platform, or of course, by hardware. Based on this understanding, the essence of the above technical solution or the part that contributes to the existing technology can be embodied in the form of a software product. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or certain parts of the embodiments.

[0353] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for generating an internal link, characterized in that: include: determining a plurality of matching pairs; The matching pair is composed of an object and key information that has a matching relationship with the object; the matching pair is identified and obtained from a key information library and an object library; Determine at least one page link relationship corresponding to the list page and the detail page; According to the at least one page link relationship, the multiple groups of matching pairs are associated and converted to obtain two pages corresponding to the at least one page link relationship and key information between the two pages used as an internal link.

2. The method according to claim 1, characterized in that: Also includes: Clustering multiple key information in the key information database according to key information features to obtain multiple information groups; According to the object characteristics, the objects in the object library are matched with the multiple information groups respectively to determine at least one matched target information group; The object is matched with the key information in the at least one target information group respectively to determine at least one key information matching the object, and the object and the at least one key information form at least one matching pair.

3. The method according to claim 2, characterized in that Also includes: The matching model is used to extract key information features corresponding to multiple key information in the key information database; Extracting object features corresponding to a plurality of objects in an object library respectively using the matching model; The matching model is trained by contrastive learning based on multiple groups of sample matching pairs; the sample matching pairs include sample key information and sample objects.

4. The method according to claim 1, characterized in that The at least one page link relationship includes a list page linking to a detail page; According to the at least one page link relationship, the plurality of matching pairs are converted into associations to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages, including: Determining, from the multiple groups of matching pairs, multiple key information having a matching relationship with the same object; Determining first key information and second key information according to the matching degree between the plurality of key information and the object respectively; Based on the correspondence between the key information and the list page, the first list page pointed to by the first key information and the first detail page of the object are determined, and the second key information is used as an internal link in the first list page.

5. The method according to claim 4, characterized in that The determining the first key information and the second key information according to the matching degree between the plurality of key information and the object respectively comprises: According to the matching degree between the plurality of key information and the object, determining that the matching degree meets the first matching requirement The first key information and the second key information determining that the matching degree meets the second matching requirement.

6. The method according to claim 1, characterized in that The at least one page link relationship includes a detail page link list page; According to the at least one page link relationship, the plurality of matching pairs are converted into associations to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages, including: From the plurality of matching pairs, searching for a target matching pair according to the matching degree between the key information and the object; Combined with the correspondence between the key information and the list page, determine the target list page pointed to by the target key information in the target matching pair, the target details page of the target object, and use the target key information as an internal link in the target details page.

7. The method according to claim 1, characterized in that The at least one page link relationship includes a detail page linking a detail page; According to the at least one page link relationship, the plurality of matching pairs are converted into associations to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages, including: Determining a first object and a second object that meet a correlation requirement; Searching for first key information matching the first object from the multiple groups of matching pairs; A first details page of the first object and a second details page of the second object are determined, and the first key information is used as an internal link in the second details page.

8. The method according to claim 1, characterized in that The at least one page link relationship includes a list page linking a list page; According to the at least one page link relationship, the plurality of matching pairs are converted into associations to obtain two pages corresponding to the at least one page link relationship and key information used as an internal link between the two pages, including: Searching for first key information and second key information that have a matching relationship with the same object from the multiple sets of matching pairs; Based on the correspondence between key information and list pages, determine the first list page pointed to by the first key information, the second list page pointed to by the second key information, and use the first key information as an internal link in the second list page.

9. The method according to claim 8, characterized in that The searching, from the plurality of matching pairs, for the first key information and the second key information that have a matching relationship with the same object comprises: Searching for multiple key information that have a matching relationship with the same object from the multiple groups of matching pairs; According to the matching degrees between the plurality of key information and the object respectively, the first key information whose matching degree meets the third matching requirement is determined, and the second key information whose matching degree meets the fourth matching requirement is determined.

10. The method according to claim 1, characterized in that Also includes: Performing page configuration according to the associated conversion results corresponding to the at least one page link relationship, so as to update the internal link structure of the website; Alternatively, the associated conversion results corresponding to the at least one page link relationship are sent to the target person; The association conversion result is used to generate page configuration information, and the page configuration information is used to update the internal link structure of the website.

11. The method according to claim 1, characterized in that: Also includes: Determine key information in the search records and the list page pointed to by the key information according to the search records generated inside and / or outside the website; A corresponding relationship between the key information and the list page is established, and a key information library is constructed according to the key information.

12. A data processing method, characterized in that: include: Obtain multiple sets of sample matching pairs; Using the multiple groups of sample matching pairs as training data of the current training batch, and inputting them into the matching model; Taking any sample matching pair as a positive sample, the sample key information in any sample matching pair and the sample object in at least one other sample matching pair as negative samples, and training the matching model in a sampling contrast learning manner; The matching model is used to identify objects and key information that have matching relationships in the key information library and the object library to obtain multiple sets of matching pairs; The multiple groups of matching objects are used to perform association conversion to obtain any two pages corresponding to at least one page link relationship and key information between the two pages used as an internal link.

13. A search method, characterized in that: include: Get search information; Based on the search information, a first page matching the search information and a second page pointed to by an internal link in the first page are captured; wherein the first page and the second page are list pages or detail pages; wherein the internal link determines a plurality of matching pairs by association conversion according to at least one page link relationship; the matching pair is composed of an object and key information having a matching relationship with the object; Including the first page and the second page in the search results; A display result is determined from the search results.

14. A computing device, characterized in that including a processing component and a storage component; The storage component stores one or more computer instructions; the one or more computer instructions are used to be called and executed by the processing component to implement the internal link generation method as described in any one of claims 1 to 11 or the data processing method as described in claim 12 or the search method as described in claim 13.

15. A computer-readable storage medium, characterized in that: A computer program is stored, and when the computer program is executed by a computer, the internal link generation method according to any one of claims 1 to 11, the data processing method according to claim 12, or the search method according to claim 13 is implemented.

Images