Feature analysis method and device, electronic equipment and storage medium

Abstract

The application provides a feature analysis method and device, electronic equipment and storage medium, and belongs to the technical field of artificial intelligence, and comprises the following steps: obtaining meta information of each to-be-analyzed feature; the meta information comprises a feature name, an analyzer and a dependent parameter; merging each meta information with the same analyzer; establishing a feature dependency tree comprising each meta information; the feature name of a parent node in the feature dependency tree corresponds to the dependent parameter of a child node; based on the topological order of the feature dependency tree, the analyzer of each node is called to obtain the feature corresponding to each node. The application merges each feature meta information with the same analyzer, establishes a more simplified feature dependency tree, calls the analyzer of each node based on the topological order of the feature dependency tree, reduces the total calling times of the analyzer, and thus the time consumption of the feature acquisition stage can be reduced.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a feature parsing method, apparatus, electronic device, and storage medium. Background Technology

[0002] When using AI models for business recommendations, various business-related features need to be input into the model, which requires acquiring these features. Feature acquisition is achieved by calling a parser to parse Remote Procedure Call (RPC) requests. Current feature acquisition methods suffer from the problem of numerous parser calls, resulting in a long feature acquisition phase. Summary of the Invention

[0003] This invention provides a feature parsing method, apparatus, electronic device, and storage medium to solve the technical problem of long time consumption in the feature acquisition stage in the prior art.

[0004] This invention provides a feature parsing method, comprising: Obtain the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters; Merge the metadata that are identical in the parser; Establish a feature dependency tree containing each of the aforementioned meta-information; in the feature dependency tree, the feature name of the parent node corresponds to the dependency parameter of the child node; Based on the topological order of the feature dependency tree, the parser of each node is invoked to obtain the features corresponding to each node.

[0005] According to a feature parsing method provided by the present invention, the step of calling the parser of each node to obtain the features corresponding to each node includes: The parsers of each node at the same level are invoked in parallel to obtain the features corresponding to each node at the same level.

[0006] According to a feature parsing method provided by the present invention, the step of calling the parser of each node to obtain the features corresponding to each node includes: The parser of the parent node is invoked to obtain the features corresponding to the parent node; The dependency parameters of the child node are determined from the features corresponding to the parent node; The parser of the child node is invoked based on the dependency parameters of the child node to obtain the features corresponding to the child node.

[0007] According to a feature parsing method provided by the present invention, determining the dependency parameter of the child node from the features corresponding to the parent node includes: If the parent node has only one feature, then the feature corresponding to the parent node is directly used as the dependency parameter of the child node. If the parent node has multiple features, then the feature corresponding to the parent node is used as an index, and the feature with the same index is searched in the dependency parameters of the child node as the dependency parameter of the child node.

[0008] According to a feature parsing method provided by the present invention, before calling the parser of each node based on the topological order of the feature dependency tree to obtain the features corresponding to each node, the method further includes: Obtain context information; The context information is used as the dependency parameter of the root node of the feature dependency tree.

[0009] According to a feature parsing method provided by the present invention, the metadata further includes a cache identifier; After obtaining the features corresponding to each node by calling the parser of each node based on the topological order of the feature dependency tree, the process further includes: If the cache identifier indicates that the feature is cached, then the obtained feature will be stored in the cache.

[0010] The present invention also provides a feature parsing apparatus, comprising: The acquisition module is used to acquire the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters. The merging module is used to merge the metadata that are identical in the parser; A module is established to build a feature dependency tree containing each of the aforementioned meta-information; the feature name of the parent node in the feature dependency tree corresponds to the dependency parameter of the child node. The parsing module is used to call the parser of each node based on the topological order of the feature dependency tree to obtain the features corresponding to each node.

[0011] The present invention also provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the feature parsing method described above.

[0012] The present invention also provides a non-transitory computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the feature parsing method as described above.

[0013] The present invention also provides a computer program product, including a computer program that, when executed by a processor, implements the feature parsing method as described above.

[0014] The feature parsing method, apparatus, electronic device, and storage medium provided by this invention merge the identical feature metadata of the parser to establish a more simplified feature dependency tree. Based on the topological order of the feature dependency tree, the parser of each node is called, which reduces the total number of parser calls and thus reduces the time consumption of the feature acquisition stage. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in this invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 This is a schematic diagram of the feature dependency tree in the prior art.

[0017] Figure 2 This is a flowchart illustrating the feature parsing method provided by the present invention.

[0018] Figure 3 This is a schematic diagram of the feature dependency tree provided by the present invention.

[0019] Figure 4 This is a schematic diagram of the features and dependent parameters provided by the present invention.

[0020] Figure 5 This is a schematic diagram of the feature tree provided by the present invention.

[0021] Figure 6 This is a schematic diagram of the feature analysis device provided by the present invention.

[0022] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention. Detailed Implementation

[0023] To make the objectives, technical solutions, and advantages of this invention clearer, the technical solutions of this invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this invention. All other embodiments obtained by those skilled in the art based on the embodiments of this invention without creative effort are within the scope of protection of this invention.

[0024] The feature acquisition stage of a large model requires first constructing a feature dependency tree composed of feature meta-information, and then calling the parser of each meta-information according to the topological order of the feature dependency tree to obtain the required features.

[0025] The metadata of a feature can include the feature name, parser, dependency parameters, and cacheable flag. The naming convention for the feature name is "feature entity.feature name". The parser is the processor used to parse the feature, responsible for retrieving the feature from the RPC request; a single parser can parse multiple features. The dependency parameters are the parameters that the parser depends on; these can be context information or other features. The parser must obtain the dependency parameters before it can parse a feature. The cacheable flag indicates whether the feature is cached. If caching is enabled, the same feature can be retrieved directly from the cache without calling the parser the next time. The structure of the metadata can be: { "name": "feature entity.feature name", "parser": "xxx", "dependency": [ "xxx" "xxx" ], "cacheable": false / true } Here, `cacheable` being `false` means caching is disabled, and `cacheable` being `true` means caching is enabled.

[0026] An example of a feature dependency tree in existing technology is provided. Assume the features to be parsed are User.A, User.B, User.C, User.D, User.E, User.F, and Order.A, and the metadata of each feature is as follows: { "name": "User.A" "parser": "ParserUser" "dependency": [ "context: userID" ], "cacheable": false }; (Named Meta-information A) { "name": "User.B" "parser": "ParserUserIdentity", "dependency": [ "context: userID" ], "cacheable": false }; (Named Meta-information B) { "name": "User.C" "parser": "ParserUser" "dependency": [ "context: userID" ], "cacheable": false }; (named Meta-information C) { "name": "User.D", "parser": "ParserAccount" "dependency": [ "User.C" "User.B" ], "cacheable": false }; (named Meta-information D) { "name": "User.E" "parser": "ParserAccount" "dependency": [ "User.C" "User.B" ], "cacheable": false }; (named Meta-information E) { "name": "User.F" "parser": "ParserOrderCount", "dependency": [ "Order.A" "User.E" ], "cacheable": false }; (named Meta-information F) { "name": "Order.A" "parser": "ParserOrder" "dependency": [ "User.A" "User.D" ], "cacheable": false }; (named Meta-information O) Feature dependency trees established in existing technologies, such as Figure 1 As shown. The process of obtaining each feature is as follows: Get userID; Based on the userID, the ParserUser parser is used to obtain the feature User.C; based on the userID, the ParserUserIdentity parser is used to obtain the feature User.B; based on the features User.C and User.B, the ParserAccount parser is used to obtain the feature User.E. Based on the userID, the ParserUser parser is called to obtain the feature User.C; based on the userID, the ParserUserIdentity parser is called to obtain the feature User.B; based on the features User.C and User.B, the ParserAccount parser is called to obtain the feature User.D. Based on userID, the parser ParserUser is called to obtain feature User.A; based on feature User.D and feature User.A, the parser ParserOrder is called to obtain feature Order.A. Based on features Order.A and User.E, the parser ParserOrderCount is called to obtain feature User.F.

[0027] As can be seen, the above process repeatedly calls the parserParserUser 3 times, the parserParserUserIdentity 2 times, and the parserParserAccount 2 times, for a total of 9 calls to the parser.

[0028] The following is combined Figures 2-7 This invention describes the feature parsing method, apparatus, electronic device, and storage medium provided by the present invention.

[0029] Figure 2 This is a flowchart illustrating the feature parsing method provided by the present invention, as shown below. Figure 1 As shown, steps S1, S2, S3 and S4 are included but are not limited to.

[0030] Step S1: Obtain the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters.

[0031] The following text also uses Figure 1 The following explanation uses the various metadata as examples.

[0032] Step S2: Merge the metadata that are the same in the parser.

[0033] Figure 1 The parser for both metadata A and C is ParserUser, therefore the metadata obtained after merging metadata A and C is: { "name":["User.A","User.C"] "parser": "ParserUser" "dependency": [ "context: userID" ], "cacheable": false }; (Named Meta-information AC) The parser for both metadata D and E is ParserAccount, therefore, the metadata obtained after merging metadata D and E is: { "name":["User.D","User.E"] "parser": "ParserAccount" "dependency": [ "User.C" "User.B" ], "cacheable": false }; (named Meta-information DE) Therefore, the final metadata is obtained as AC, B, DE, F, and O.

[0034] Step S3: Establish a feature dependency tree containing various metadata; the feature name of the parent node in the feature dependency tree corresponds to the dependency parameter of the child node.

[0035] The feature dependency tree established in step S3 is as follows: Figure 3 As shown.

[0036] The feature name User.C of parent node AC corresponds to the dependency parameter User.C of child node DE, and the feature name User.B of parent node B corresponds to the dependency parameter User.B of child node DE.

[0037] The feature name User.D of parent node DE corresponds to the dependency parameter User.D of child node O, and the feature name User.A of parent node AC corresponds to the dependency parameter User.A of child node O.

[0038] The feature name Order.A of parent node O corresponds to the dependency parameter Order.A of child node F, and the feature name User.E of parent node DE corresponds to the dependency parameter User.E of child node F.

[0039] Step S4: Based on the topological order of the feature dependency tree, call the parser of each node to obtain the features corresponding to each node.

[0040] In one embodiment, the present invention invokes the parser of each node to obtain the features corresponding to each node, which may specifically include: Call the parser of the parent node to obtain the features corresponding to the parent node; Determine the dependency parameters of the child node from the features corresponding to the parent node; The parser of the child node is invoked based on the dependency parameters of the child node to obtain the features corresponding to the child node.

[0041] Specifically, based on Figure 3 Based on the topological order, the parser ParserUser of the parent node AC is called to obtain the features User.A and User.C corresponding to the parent node AC. The dependency parameter User.C of the child node DE is determined from the features User.A and User.C corresponding to the parent node AC. The parser ParserUserIdentity of the parent node B is called to obtain the feature User.B corresponding to the parent node B. The feature User.B corresponding to the parent node B is used as the dependency parameter User.B of the child node DE. Based on the dependency parameters User.B and User.C of the child node DE, the parser ParserAccount of the child node DE is called to obtain the features User.D and User.E corresponding to the child node DE.

[0042] Determine the dependency parameter User.A of child node O from the features User.A and User.C corresponding to parent node AC, and determine the dependency parameter User.D of child node O from the features User.D and User.E corresponding to parent node DE; call the parser ParserOrder of child node O based on the dependency parameters User.A and User.D of child node O to obtain the feature Order.A corresponding to child node O.

[0043] Determine the dependency parameter User.E of child node F from the features User.D and User.E corresponding to parent node DE, and use the feature Order.A corresponding to parent node O as the dependency parameter Order.A of child node F; call the parser ParserOrderCount of child node F based on the dependency parameters User.E and Order.A of child node F to obtain the feature User.F corresponding to child node F.

[0044] As can be seen, this invention only calls the parser ParserUser once, the parser ParserUserIdentity once, the parser ParserAccount once, the parser ParserOrder once, and the parser ParserOrderCount once, reducing the total number of parser calls from the original 9 to 5.

[0045] As can be seen from the above, the feature parsing method of the present invention merges the feature metadata that are the same in the parser, establishes a more simplified feature dependency tree, and calls the parser of each node based on the topological order of the feature dependency tree, thereby reducing the total number of parser calls and thus reducing the time consumption of the feature acquisition stage.

[0046] In one embodiment, step S4, which involves calling the parser of each node to obtain the features corresponding to each node, may specifically include: The parsers of each node at the same level are called in parallel to obtain the features corresponding to each node at the same level.

[0047] For example, Figure 3 If nodes AC and B are at the same level, the parser ParserUser and ParserUserIdentity can be called in parallel, thereby improving the efficiency of feature acquisition.

[0048] In one embodiment, the present invention determines the dependency parameters of child nodes from the features corresponding to the parent node, which may specifically include: If the parent node has only one feature, then the feature corresponding to the parent node is directly used as the dependency parameter of the child node. If a parent node has multiple features, then the feature corresponding to the parent node is used as an index, and the feature with the same index is searched in the dependency parameters of the child node and used as the dependency parameter of the child node.

[0049] Specifically, if the features corresponding to the parent node AC are User.A and User.C, then User.A and User.C are used as indexes to find User.C as a dependency parameter of the child node DE in the dependency parameters User.B and User.C.

[0050] If the parent node B corresponds to the feature User.B, then User.B is directly used as the dependency parameter of the child node DE.

[0051] If the parent node AC has the features User.A and User.C, then User.A and User.C are used as indices to find User.A as a dependency parameter of child node O in the dependency parameters User.A and User.D.

[0052] If the features corresponding to the parent node DE are User.D and User.E, then User.D and User.E are used as indexes to find User.D as a dependency parameter of child node O in the dependency parameters User.A and User.D of child node O.

[0053] If the parent node DE has the features User.D and User.E, then User.D and User.E are used as indexes to find User.E as a dependency parameter of child node F in the dependency parameters User.E and Order.A.

[0054] If the feature corresponding to the parent node O is Order.A, then Order.A is directly used as the dependency parameter of the child node F.

[0055] This allows us to determine the dependency parameters of child nodes from the features corresponding to the parent node using an index.

[0056] In one embodiment, prior to step S4, the feature parsing method of the present invention may further include: Obtain context information; Use context information as the dependency parameter of the root node of the feature dependency tree.

[0057] Considering Figure 3 The parser calls for the root nodes AC and B are based on the context information userID. Therefore, it is necessary to obtain the context information userID first and use it as a dependency parameter for the root nodes AC and B in order to implement the calls to the parser ParserUser and ParserUserIdentity.

[0058] In one embodiment, after step S4, the feature parsing method of the present invention may further include: If the cache flag indicates that caching is enabled for a feature, then the obtained feature will be stored in the cache.

[0059] If the node's cache flag is true, the obtained features can be stored in the cache, and the features can be retrieved directly from the cache next time, thereby improving the efficiency of feature retrieval.

[0060] After the parser for each node is invoked in this invention, the features and dependency parameters corresponding to each node are as follows: Figure 4 As shown, it is necessary to associate the features corresponding to each node to form a feature tree. Then, by traversing the nodes of the feature tree, a row of features can be obtained for input into the recommendation model.

[0061] like Figure 5 As shown, firstly, the context information userID is added to the Feature Tree. Then, features User.A, User.B, and User.C are added as child nodes of the userID node, thus establishing the association between userID and features User.A, User.B, and User.C. Features User.D and User.E are added as child nodes of the user.B and user.C nodes, thus establishing the association between features User.B, User.C and features User.D, User.E. Features Order.A is added as a child node of the user.A and user.D nodes, thus establishing the association between features Order.A and features User.D. Features User.F is added as a child node of the user.A and user.E nodes, thus establishing the association between features User.F and features Order.A, User.E.

[0062] This can reduce memory consumption during the feature association stage.

[0063] like Figure 6 As shown, the feature parsing apparatus provided by the present invention includes: The acquisition module is used to obtain the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters. The merging module is used to merge metadata that is identical in the parser; The module is used to build a feature dependency tree containing various metadata; the feature name of the parent node in the feature dependency tree corresponds to the dependency parameter of the child node. The parsing module is used to call the parser of each node based on the topological order of the feature dependency tree to obtain the features corresponding to each node.

[0064] The parsing module can also be used for: The parsers of each node at the same level are called in parallel to obtain the features corresponding to each node at the same level.

[0065] The parsing module can also be used for: Call the parser of the parent node to obtain the features corresponding to the parent node; Determine the dependency parameters of the child node from the features corresponding to the parent node; The parser of the child node is invoked based on the dependency parameters of the child node to obtain the features corresponding to the child node.

[0066] The parsing module can also be used for: If the parent node has only one feature, then the feature corresponding to the parent node is directly used as the dependency parameter of the child node. If a parent node has multiple features, then the feature corresponding to the parent node is used as an index, and the feature with the same index is searched in the dependency parameters of the child node and used as the dependency parameter of the child node.

[0067] The acquisition module can also be used to obtain context information; The module can also be used to use context information as a dependency parameter for the root node of the feature dependency tree.

[0068] The feature parsing device may also include a storage module for storing the obtained features into the cache if the cache identifier indicates that feature caching is enabled.

[0069] It should be noted that the feature parsing device provided by the present invention can execute the feature parsing method described in any of the above embodiments during specific operation, and this embodiment will not elaborate on this.

[0070] Figure 7 This is a schematic diagram of the structure of the electronic device provided by the present invention, such as... Figure 7 As shown, the electronic device may include a processor, a communications interface, memory, and a communication bus, wherein the processor, communications interface, and memory communicate with each other via the communication bus. The processor can invoke logical instructions in the memory to execute a feature parsing method. This method includes: acquiring metadata of each feature to be parsed; the metadata includes feature name, parser, and dependency parameters; merging metadata with the same parser; establishing a feature dependency tree containing all metadata; the feature name of the parent node in the feature dependency tree corresponds to the dependency parameters of its child nodes; and, based on the topological order of the feature dependency tree, invoking the parser of each node to obtain the feature corresponding to each node.

[0071] Furthermore, the logical instructions in the aforementioned memory can be implemented as software functional units and sold or used as independent products, and can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art, or a part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present invention. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0072] On the other hand, the present invention also provides a computer program product, the computer program product comprising a computer program stored on a non-transitory computer-readable storage medium, the computer program comprising program instructions, and when the program instructions are executed by a computer, the computer is able to execute the feature parsing method provided in the above embodiments, the method comprising: obtaining meta-information of each feature to be parsed; the meta-information including feature name, parser and dependency parameters; merging meta-information with the same parser; establishing a feature dependency tree containing each meta-information; the feature name of the parent node in the feature dependency tree corresponding to the dependency parameters of the child node; and, based on the topological order of the feature dependency tree, calling the parser of each node to obtain the feature corresponding to each node.

[0073] In another aspect, the present invention also provides a non-transitory computer-readable storage medium storing a computer program thereon. When executed by a processor, the computer program implements the feature parsing method provided in the above embodiments. The method includes: obtaining metadata of each feature to be parsed; the metadata includes feature name, parser, and dependency parameters; merging metadata with the same parser; establishing a feature dependency tree containing the metadata; the feature name of the parent node in the feature dependency tree corresponds to the dependency parameters of the child node; and, based on the topological order of the feature dependency tree, calling the parser of each node to obtain the feature corresponding to each node.

[0074] The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0075] Through the above description of the embodiments, those skilled in the art can clearly understand that each embodiment can be implemented by means of software plus necessary general-purpose hardware platforms, and of course, it can also be implemented by hardware. Based on this understanding, the above technical solutions, in essence or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments or some parts of the embodiments.

[0076] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some of the technical features; and these modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A feature parsing method, characterized in that, include: Obtain the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters; A parser is a processor used to parse features. It is responsible for parsing features from remote procedure call requests. A parser can parse multiple features. Merge the metadata that are identical in the parser; Establish a feature dependency tree containing all the aforementioned meta-information; The feature name of the parent node in the feature dependency tree corresponds to the dependency parameter of the child node; Based on the topological order of the feature dependency tree, the parser of each node is invoked to obtain the features corresponding to each node; The features corresponding to each node are associated to form a feature tree. Subsequently, traversing the nodes of the feature tree yields a row of features for inputting into the recommendation model.

2. The feature parsing method according to claim 1, characterized in that, The process of calling the parser of each node to obtain the features corresponding to each node includes: The parsers of each node at the same level are invoked in parallel to obtain the features corresponding to each node at the same level.

3. The feature parsing method according to claim 1, characterized in that, The process of calling the parser of each node to obtain the features corresponding to each node includes: The parser of the parent node is invoked to obtain the features corresponding to the parent node; The dependency parameters of the child node are determined from the features corresponding to the parent node; The parser of the child node is invoked based on the dependency parameters of the child node to obtain the features corresponding to the child node.

4. The feature parsing method according to claim 3, characterized in that, Determining the dependency parameter of the child node from the features corresponding to the parent node includes: If the parent node has only one feature, then the feature corresponding to the parent node is directly used as the dependency parameter of the child node. If the parent node has multiple features, then the feature corresponding to the parent node is used as an index, and the feature with the same index is searched in the dependency parameters of the child node as the dependency parameter of the child node.

5. The feature parsing method according to claim 1, characterized in that, Before invoking the parser of each node based on the topological order of the feature dependency tree to obtain the features corresponding to each node, the method further includes: Obtain context information; The context information is used as the dependency parameter of the root node of the feature dependency tree.

6. The feature parsing method according to claim 1, characterized in that, The metadata also includes a cache identifier; After obtaining the features corresponding to each node by calling the parser of each node based on the topological order of the feature dependency tree, the process further includes: If the cache identifier indicates that caching is enabled for the feature, then the obtained feature will be stored in the cache.

7. A feature parsing device, characterized in that, include: The acquisition module is used to acquire the metadata of each feature to be parsed; the metadata includes the feature name, parser, and dependency parameters. A parser is a processor used to parse features. It is responsible for parsing features from remote procedure call requests. A parser can parse multiple features. The merging module is used to merge the metadata that are identical in the parser; A module is established to build a feature dependency tree containing the various metadata. The feature name of the parent node in the feature dependency tree corresponds to the dependency parameter of the child node; The parsing module is used to call the parser of each node based on the topological order of the feature dependency tree to obtain the features corresponding to each node; The features corresponding to each node are associated to form a feature tree. Subsequently, traversing the nodes of the feature tree yields a row of features for inputting into the recommendation model.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the feature parsing method as described in any one of claims 1 to 6.

9. A non-transitory computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the feature parsing method as described in any one of claims 1 to 6.

10. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the feature parsing method as described in any one of claims 1 to 6.

Images