Information Recommendation Method, Apparatus, Electronic Device and Storage Medium

Abstract

The present application relates to an information recommendation method, apparatus, electronic device and storage medium. The method includes: obtaining information to be recommended, where the information to be recommended is information to be pushed to a target client; extracting named entities corresponding to a preset entity type by performing named entity recognition on the information to be recommended; extracting a plurality of target entities associated with the named entities by performing named entity recognition on user data on the target client; and performing information recommendation on the information to be recommended according to the degree of association between the plurality of target entities and the named entities. This method uses the degree of association between the target client and the information to be recommended as the judgment criterion, realizing automatic information delivery. Compared with the related art where information delivery depends on manual subjective judgment, this method does not require manual judgment, solving the problem in the related art that inaccurate advertisement delivery is caused by manual advertisement delivery.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and in particular, to an information recommendation method, apparatus, electronic device, and storage medium. Background Art

[0002] Currently, when placing advertisements on media advertising platforms, it is often necessary for advertising optimizers to perform manual operations based on their understanding of the promoted products and past experience. New advertisements are tested using different advertising creative combinations to achieve advertising targeting exploration, screen out preferred creatives, and reach a stable advertising placement state. The above advertising placement method relies on the subjective judgment of advertising optimizers, which places high requirements on the advertising placement level of advertising optimizers. Once the subjective judgment of the advertising optimizer deviates, it will lead to deviations in advertising placement. Summary of the Invention

[0003] This application provides an information recommendation method, apparatus, electronic device, and storage medium to solve the problem of inaccurate advertising placement caused by manual advertising placement in related technologies.

[0004] In a first aspect, this application provides an information recommendation method, including: obtaining information to be recommended, where the information to be recommended is information to be pushed to a target client; extracting named entities corresponding to a preset entity type by performing named entity recognition on the information to be recommended; extracting a plurality of target entities associated with the named entities by performing named entity recognition on user data on the target client; and performing information recommendation on the information to be recommended according to the association degree between the plurality of target entities and the named entities.

[0005] In a second aspect, this application provides an information recommendation apparatus, including: an obtaining unit, configured to obtain information to be recommended, where the information to be recommended is information to be pushed to a target client; a first extraction unit, configured to extract named entities corresponding to a preset entity type by performing named entity recognition on the information to be recommended; a second extraction unit, configured to extract a plurality of target entities associated with the named entities by performing named entity recognition on user data on the target client; and a recommendation unit, configured to perform information recommendation on the information to be recommended according to the association degree between the plurality of target entities and the named entities.

[0006] In a third aspect, an electronic device is provided, including a processor, a communication interface, a memory, and a communication bus, where the processor, the communication interface, and the memory communicate with each other through the communication bus;

[0007] The memory is used to store a computer program;

[0008] A processor, when executing a program stored in a memory, implements the steps of the information recommendation method described in any embodiment of the first aspect.

[0009] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a processor, the steps of the information recommendation method described in any embodiment of the first aspect are implemented.

[0010] The technical solution of the present application can be applied to the technical field of knowledge graph based on graph construction. The above technical solutions provided by the embodiments of the present application have the following advantages compared with the prior art:

[0011] In the method provided by the embodiment of the present application, first, obtain the information to be recommended to be pushed to the target client, then perform named entity recognition on the information to be recommended, extract named entities corresponding to the preset entity type, and then perform named entity recognition on the user data on the target client, extract multiple target entities associated with the named entities, and finally perform information recommendation according to the degree of association between the multiple target entities and the named entities. This method uses the degree of association between the target client and the information to be recommended as the judgment criterion, realizes automatic information delivery. Compared with the related art that relies on manual subjective judgment for information delivery, this method does not need to rely on manual judgment, so there is no problem that the manual subjective judgment is biased and the delivery is inaccurate, thus solving the problem that the related art needs manual advertising delivery and the advertising delivery is inaccurate. Description of the Drawings

[0012] The drawings here are incorporated into the specification and constitute a part of this specification, showing embodiments consistent with the present invention and used together with the specification to explain the principles of the present invention.

[0013] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, for those of ordinary skill in the art, without creative efforts, other drawings can also be obtained based on these drawings.

[0014] Figure 1 It is a schematic flowchart of an information recommendation method provided by an embodiment of the present application;

[0015] Figure 2 It is a schematic composition diagram of an advertisement;

[0016] Figure 3 It is a schematic diagram of knowledge fusion provided by an embodiment of the present application;

[0017] Figure 4 It is a schematic diagram of named entity recognition of the information to be recommended provided by an embodiment of the present application;

[0018] Figure 5 A flowchart showing the process of an advertising placement strategy provided by an embodiment of the present application;

[0019] Figure 6 A schematic structural diagram of an information recommendation device provided by an embodiment of the present application;

[0020] Figure 7 A schematic structural diagram of an electronic device provided by an embodiment of the present application. Detailed implementation manners

[0021] To make the objectives, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application. Apparently, the described embodiments are some but not all of the embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present application without creative efforts shall fall within the protection scope of the present application.

[0022] According to one aspect of the embodiments of the present application, an information recommendation method is provided. Optionally, in this embodiment, the above information recommendation method may be applied to a hardware environment composed of a terminal and a server. The server is connected to the terminal through a network and can be used to provide services for the terminal or a client installed on the terminal. A database may be set up on the server or independently of the server for providing data storage services for the server.

[0023] The above network may include but is not limited to at least one of the following: a wired network, a wireless network. The above wired network may include but is not limited to at least one of the following: a wide area network, a metropolitan area network, a local area network. The above wireless network may include but is not limited to at least one of the following: WIFI (Wireless Fidelity), Bluetooth. The terminal is not limited to a PC, a mobile phone, a tablet computer, etc.

[0024] The information recommendation method of the embodiments of the present application may be executed by the server, or may be executed by the terminal, or may be jointly executed by the server and the terminal. Among them, when the terminal executes the information recommendation method of the embodiments of the present application, it may also be executed by a client installed thereon.

[0025] Taking the execution of the information recommendation method in this embodiment by the server as an example, Figure 1 A flowchart showing the process of an information recommendation method provided by an embodiment of the present application. As Figure 1 shown, the method includes the following steps:

[0026] Step S201: Obtain the information to be recommended, where the information to be recommended is the information to be pushed to the target client.

[0027] In this embodiment, the information to be recommended may be an advertisement to be placed, and the target client may be an Internet media platform.

[0028] Step S202: Through named entity recognition of the information to be recommended, extract the named entities corresponding to the preset entity types.

[0029] In this embodiment, named entity recognition of the information to be recommended can be performed using natural language processing methods or neural network methods, etc. Figure 2 It is a schematic diagram of the composition of the advertisement, as Figure 2 shown. The advertisement includes parts such as advertisement attributes and advertisement elements. The advertisement elements include endorsed stars, brands, and product categories, etc. The above preset entity types may be advertisement elements.

[0030] For example, the advertisement to be placed is a product of XX series of brand A endorsed by star B, theme: AAAAA, style: BBBB. Named entity recognition is performed on the above advertisement to be placed, and the obtained named entities are brand A and star B.

[0031] Step S203: Through named entity recognition of the user data on the target client, extract multiple target entities associated with the named entities.

[0032] In this embodiment, named entity recognition of the user data on the target client can be performed using the same method as named entity recognition of the information to be recommended. The source of the above user data can be the public data of the media platform or the data collected by a third-party data platform.

[0033] The above user data includes data such as user usage, active periods, user age distribution, user gender distribution, hot issues, hot topics, hot stars, programs, variety shows, etc.

[0034] Step S204: Perform information recommendation on the information to be recommended according to the degree of association between the multiple target entities and the named entities.

[0035] In this embodiment, according to the degree of association between the multiple target entities and the named entities, the information to be recommended is pushed to the target client.

[0036] In one embodiment, the information recommendation based on the association degree between the multiple target entities and the named entity includes: determining a weight value between each target entity and each named entity according to a pre-constructed knowledge graph, where the knowledge graph is constructed based on the named entity, and each weight value is used to represent the association degree between the target entity and each named entity; obtaining a target weight through weighted operation on the weight values corresponding to all target entities; and recommending the information to be recommended to the target client when the target weight is greater than or equal to a first predetermined value, where the information to be recommended will not be recommended to the target client when the target weight is less than the first predetermined value.

[0037] In this embodiment, according to the pre-constructed knowledge graph, the weight value corresponding to each target entity for representing the association degree with the named entity is obtained, and then weighted operation is performed to obtain the target weight. When the target weight is greater than or equal to the first predetermined value, it indicates that the association degree between the target client and the information to be recommended is relatively high, so the information to be recommended can be recommended to the target client. Otherwise, the information to be recommended will not be recommended to the target client. The knowledge graph can connect seemingly unrelated data together and enable the relationships between various entities to be calculated, so that more accurate weight values can be obtained, and further improve the accuracy of advertising placement.

[0038] Specifically, the weighted operation can be an addition operation, adding up the weight values corresponding to all target entities to obtain the target weight. When there are multiple named entities, the weighted operation can also have different weights for different named entities. For example, the weight of the brand is 0.5 and the weight of the spokesperson is 0.2.

[0039] The first predetermined value can be 0.5. If the weight value is greater than or equal to 0.5, it is considered that the correlation is relatively high, and the information to be recommended is pushed to the target client.

[0040] Figure 3 This is a schematic diagram of knowledge fusion provided by the embodiment of the present application, as Figure 3As shown, in one embodiment, before determining the weight value between each target entity and each named entity according to the pre-constructed knowledge graph, the above method further includes: searching for associated information associated with the named entity from a preset database; performing knowledge extraction on the associated information to obtain a plurality of associated entities and an association weight for characterizing the degree of association between the associated entities; determining a first similarity between each associated entity and the named entity according to the first word feature vector corresponding to each associated entity and the second word feature vector corresponding to the named entity; selecting a target similarity with the largest numerical value from the first similarities, and in the case where the target similarity is greater than or equal to a second predetermined value, performing knowledge fusion on the associated entity and the named entity corresponding to the target similarity to obtain a fusion result; constructing the knowledge graph according to the plurality of associated entities, the association weights between the associated entities, and the fusion result.

[0041] In this embodiment, search for associated information associated with the named entity from a preset database. The preset database can be the database of a media platform or the database of a third-party data platform. Perform knowledge extraction on the associated information to obtain associated entities and association weights. The associated entities are named entities extracted from the associated information, and then perform knowledge fusion according to the similarity between the associated entities and the named entity to obtain a knowledge graph. The knowledge graph in this embodiment is obtained by targeted supplementation according to the named entity, so the coverage rate of the constructed knowledge graph is higher. In this way, when determining the weight value according to the knowledge graph subsequently, a more accurate weight value can be obtained, thereby making the placement more accurate.

[0042] The above second predetermined value can be set according to the actual situation. When the second predetermined value is 100%, it means that the associated entity and the named entity are exactly the same.

[0043] The first word feature vector of the above associated entity and the second word feature vector of the above named entity can be calculated by means of a neural network or obtained by means of natural language processing.

[0044] The above knowledge extraction includes the following subtasks:

[0045] (1) Named entity recognition

[0046] Detection: Beijing is a busy city. [Beijing]: Entity

[0047] Classification: Beijing is a busy city. [Beijing]: Place name

[0048] (2) Term extraction: Discover related terms consisting of multiple words from the corpus.

[0049] (3) Relationship extraction

[0050] Xiaoming and Xiaohong are good friends.

[0051] The result of relationship extraction is: [Xiaoming]<friend>[Xiaohong]

[0052] (4) Event extraction: For example, extract information such as trigger words, time, and location when an event occurs from a news report.

[0053] (5) Coreference resolution: Figure out the referents of pronouns ("he", "she", "it") in a sentence.

[0054] In order to further improve the efficiency of placement, in one embodiment, determining the weight value between each target entity and each named entity according to the pre-constructed knowledge graph includes: determining the second similarity between the associated entity and each target entity according to the first attribute value of the attribute corresponding to the associated entity and the second attribute value of the attribute corresponding to each target entity; searching for the first associated entity whose second similarity with each target entity is greater than or equal to a third predetermined value from the multiple associated entities; determining the association weight between the first associated entity and each named entity as the weight between the target entity corresponding to the first associated entity and each named entity.

[0055] In this embodiment, according to the attribute value of the associated entity and the attribute value of the target entity, search for the associated entity corresponding to the target entity, and determine the weight value between the target entity and the named entity according to the weight value between the associated entity and the named entity in the knowledge graph. Using the knowledge graph, the associated entity corresponding to the target entity can be quickly found, so as to more quickly obtain the weight value between the target entity and the named entity.

[0056] In one embodiment, obtaining multiple associated entities and the association weights used to characterize the association degree between the associated entities by performing knowledge extraction on the above-mentioned association information includes: determining the first text corpus in the above-mentioned association information; performing word segmentation processing on the first text corpus to obtain a first word segmentation sequence; using a first entity recognition model to obtain the entity attribute identifiers corresponding to each word in the first word segmentation sequence, where the first entity recognition model is used to obtain the entity attribute identifiers corresponding to each word through a feedforward neural network according to the word segmentation features corresponding to each word in the word segmentation sequence, and the entity attribute identifiers are used to indicate whether each word in the word segmentation sequence belongs to a named entity; determining the associated entities in the above-mentioned association information according to the entity attribute identifiers corresponding to each word in the first word segmentation sequence; determining the association weights between the multiple associated entities according to the punctuation marks in the sentences where the multiple associated entities are located.

[0057] In this embodiment, the associated information needs to be converted into the first text corpus before word segmentation processing can be performed. When the above-mentioned associated information is voice information, a speech recognition model can be used to convert the voice information into text corpus. Perform word segmentation processing on the above-mentioned first text corpus to obtain a first word segmentation sequence. For example, perform word segmentation processing on "Xiaoming graduated from University A with a master's degree", and the obtained word segmentation sequence is "Xiaoming / master / graduated from / A / university". Use the first entity model to determine whether the above-mentioned word segmentation is a named entity according to the characteristics of each word segmentation, so as to obtain associated entities. Finally, according to the punctuation marks in the sentence where the associated entity is located, the association weight between the associated entities can be accurately analyzed. In this way, the weight value between the subsequently obtained associated entity and the named entity is also more accurate, so as to obtain a more accurate placement strategy.

[0058] In one embodiment, recommending the to-be-recommended information to the target client includes: obtaining the user login location information and online time information in the user data; according to the user login location information and the online time information, pushing the to-be-recommended information to the target client at a preset time and a preset login location.

[0059] In this embodiment, obtain the user's login location and online time information. From the user's online time, it can be seen at which time period the user likes to use the target client, so as to obtain the user's active time. The active time of users in different regions can be obtained. Then, according to the user's active time and login location, recommendations are made in different time windows in different regions, so as to achieve more accurate placement.

[0060] In one embodiment, when the to-be-recommended information is an advertisement, the relevant sales data of the advertisement product can also be analyzed, such as product preferences in different regions, product preference data for different ages, genders, and then obtain the product preference information of users in different regions for the product. Then, different product advertisements are placed for users in different regions.

[0061] Figure 4 Schematic diagram of named entity recognition of the to-be-recommended information provided by the embodiments of the present application, such as Figure 4As shown, in one embodiment, the above-mentioned extraction of named entities corresponding to the preset entity type by performing named entity recognition on the information to be recommended includes: determining the second text corpus in the information to be recommended; performing word segmentation on the text corpus to obtain a second word segmentation sequence; using a second entity recognition model to obtain the entity attribute identifiers corresponding to each word in the second word segmentation sequence, where the second entity recognition model is used to obtain the entity attribute identifiers corresponding to each word through a feedforward neural network according to the word segmentation features corresponding to each word in the word segmentation sequence, and the entity attribute identifiers are used to indicate whether each word in the word segmentation sequence belongs to a named entity; annotating the words corresponding to the preset entity type according to the entity attribute identifiers corresponding to each word in the second word segmentation sequence; and extracting the words in the second word segmentation sequence after annotation to obtain the named entities.

[0062] In order to obtain more accurate named entities and thus more accurate placement results, in this embodiment, a natural language processing method is used to perform named entity recognition on the information to be recommended. First, a speech recognition model can be used to convert the information to be recommended into a second text corpus, and then a second entity recognition model is used to obtain the named entities. The second entity recognition model and the first entity recognition model can be the same model or different models. Then, the words are annotated according to the preset entity type, and the annotated words are extracted to obtain the named entities.

[0063] The technical solution of the present application will be further described in detail below in conjunction with specific embodiments. Figure 5 It is a schematic flowchart of an advertising placement strategy provided by an embodiment of the present application, as Figure 5 shown. The above-mentioned advertising placement strategy includes three modules, namely the NLP (Natural Language Processing) module, the knowledge graph module, and the media big data module.

[0064] The NLP module includes: using an NLP algorithm to perform named entity recognition on the advertisement to be placed, extracting the advertisement elements of the advertisement to be placed, such as the brand and the spokesperson, etc., to obtain the named entities of the advertisement to be placed.

[0065] The knowledge graph module includes the following steps:

[0066] Step 1, searching for associated data related to the advertisement elements from the public data of the third-party data management platform.

[0067] Step 2, performing knowledge extraction on the above-mentioned associated data to extract multiple associated entities of the associated data and the association weights used to characterize the association degree between each associated entity.

[0068] Step 3, according to the similarity between each associated entity and the named entity, when the similarity is greater than or equal to the second predetermined value, fusing the associated entity and the named entity.

[0069] Step 4: Construct a knowledge graph based on the above multiple associated entities, the association weights between each associated entity, and the above fusion result.

[0070] The media big data module includes the following steps:

[0071] Step 1: Collect historical user data of the media user side.

[0072] Step 2: Analyze the above historical user data to obtain data such as user usage volume, active time periods, user age distribution, user gender distribution, hot issues, hot topics, hot stars, programs, variety shows, etc. in different regions, and then perform named entity recognition on the above data to obtain multiple target entities.

[0073] Step 3: According to the above knowledge graph, determine the weight value between each target entity and each named entity, perform a superposition operation on the weight values corresponding to all target entities to obtain a target weight value. When the target weight value is greater than 0.5, select the above media user side as the media platform to be pushed.

[0074] Analyze the relevant sales data of advertising products, such as product preferences in different regions, and analyze data on product preferences for different ages, genders, and optimize the matching with the user data of the above media user side to obtain a placement strategy (for example, if platform A is selected as the media platform to be pushed and it is found through analyzing the relevant sales data of advertising products that users in different cities have different taste preferences for products, then different advertising materials will be used on platform A, and since users in different cities have different active times, targeted placements will be carried out in different time windows for different regions).

[0075] Figure 6 It is a schematic structural diagram of an information recommendation device provided by an embodiment of the present application. As Figure 6 shown, the device includes:

[0076] An acquisition unit 10, configured to acquire information to be recommended, where the information to be recommended is information to be pushed to a target client;

[0077] A first extraction unit 20, configured to extract named entities corresponding to a preset entity type by performing named entity recognition on the information to be recommended;

[0078] A second extraction unit 30, configured to extract multiple target entities associated with the named entities by performing named entity recognition on the user data on the target client;

[0079] A recommendation unit 40, configured to perform information recommendation on the information to be recommended according to the association degree between the multiple target entities and the named entities.

[0080] In one embodiment, the above-mentioned recommendation unit includes an input module, a calculation module, and a recommendation module. Among them, the input module is used to determine the weight value between each target entity and each named entity according to a pre-constructed knowledge graph. The knowledge graph is constructed based on the named entities, and each weight value is used to represent the degree of association between the target entity and each named entity. The calculation module is used to obtain a target weight through weighted calculation of the weight values corresponding to all target entities. The recommendation module is used to recommend the information to be recommended to the target client when the target weight is greater than or equal to a first predetermined value. When the target weight is less than the first predetermined value, the information to be recommended will not be recommended to the target client.

[0081] In one embodiment, the above-mentioned device further includes a search unit, a third extraction unit, a determination unit, a fusion unit, and a construction unit. Among them, the search unit is used to search for associated information associated with the named entity from a preset database before determining the weight value between each target entity and each named entity according to a pre-constructed knowledge graph. The third extraction unit is used to obtain a plurality of associated entities and the association weights representing the degree of association between the associated entities through knowledge extraction of the associated information. The determination unit is used to determine the first similarity between each associated entity and the named entity according to the first word feature vector corresponding to each associated entity and the second word feature vector corresponding to the named entity. The fusion unit is used to select the target similarity with the largest value from the first similarities. When the target similarity is greater than or equal to a second predetermined value, the associated entity and the named entity corresponding to the target similarity are fused to obtain a fusion result. The construction unit is used to construct the knowledge graph according to the plurality of associated entities, the association weights between the associated entities, and the fusion result.

[0082] In one embodiment, the input module includes a first determination module, a search module, and a second determination module. Among them, the first determination module is used to determine the second similarity between the associated entity and each target entity according to the first attribute value of the attribute corresponding to the associated entity and the second attribute value of the attribute corresponding to each target entity. The search module is used to search for the first associated entity with a second similarity greater than or equal to a third predetermined value between each target entity and the plurality of associated entities. The second determination module is used to determine the association weight between the first associated entity and each named entity as the weight between the target entity corresponding to the first associated entity and each named entity.

[0083] In one embodiment, the third extraction unit includes a third determination module, a first processing module, a first acquisition module, a fourth determination module, and a fifth determination module. Among them, the third determination module is used to determine the first text corpus in the association information; the first processing module is used to perform word segmentation processing on the first text corpus to obtain a first word segmentation sequence; the first acquisition module is used to use a first entity recognition model to obtain entity attribute identifiers corresponding to each word segment in the first word segmentation sequence. The first entity recognition model is used to obtain entity attribute identifiers corresponding to each word segment through a feedforward neural network according to the word segment features corresponding to each word segment in the word segmentation sequence. The entity attribute identifier is used to indicate whether each word segment in the word segmentation sequence belongs to a named entity; the fourth determination module is used to determine the associated entities in the association information according to the entity attribute identifiers corresponding to each word segment in the first word segmentation sequence; the fifth determination module is used to determine the association weights between the multiple associated entities according to the punctuation marks in the sentences where the multiple associated entities are located.

[0084] In one embodiment, the recommendation module includes an acquisition sub-module and a push sub-module. Among them, the acquisition sub-module is used to acquire the user login location information and online time information in the user data; the push sub-module is used to push the information to be recommended to the target client at a preset time and a preset login location according to the user login location information and the online time information.

[0085] In one embodiment, the first extraction unit includes a sixth determination module, a second processing module, a second acquisition module, a labeling module, and an extraction module. Among them, the sixth determination module is used to determine the second text corpus in the information to be recommended; the second processing module is used to perform word segmentation processing on the text corpus to obtain a second word segmentation sequence; the second acquisition module is used to use a second entity recognition model to obtain entity attribute identifiers corresponding to each word segment in the second word segmentation sequence. The second entity recognition model is used to obtain entity attribute identifiers corresponding to each word segment through a feedforward neural network according to the word segment features corresponding to each word segment in the word segmentation sequence. The entity attribute identifier is used to indicate whether each word segment in the word segmentation sequence belongs to a named entity; the labeling module is used to label the word segments of the preset entity type according to the entity attribute identifiers corresponding to each word segment in the second word segmentation sequence; the extraction module is used to extract the word segments in the second word segmentation sequence after labeling to obtain the named entities.

[0086] As Figure 7 shown, an embodiment of the present application provides an electronic device, including a processor 111, a communication interface 112, a memory 113, and a communication bus 114. Among them, the processor 111, the communication interface 112, and the memory 113 complete mutual communication through the communication bus 114.

[0087] A memory 113 for storing a computer program;

[0088] In an embodiment of the present application, when the processor 111 executes the program stored on the memory 113, it implements the control method for information recommendation provided by any one of the foregoing method embodiments.

[0089] The embodiment of the present application also provides a computer-readable storage medium, on which a computer program is stored. When the computer program is executed by a processor, the steps of the information recommendation method provided by any one of the foregoing method embodiments are implemented.

[0090] It should be noted that in this article, relational terms such as "first" and "second" are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "including" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device. Without further limitation, an element defined by the statement "comprising one..." does not exclude the presence of another identical element in the process, method, article or device comprising the element.

[0091] The above are only specific embodiments of the present invention, enabling those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be obvious to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but rather to the broadest scope consistent with the principles and novel features claimed herein.

Claims

1. An information recommendation method, characterized in that, Including: Obtain information to be recommended, where the information to be recommended is information to be pushed to a target client; By performing named entity recognition on the information to be recommended, extract named entities corresponding to a preset entity type, including: determining a second text corpus in the information to be recommended; performing word segmentation processing on the text corpus to obtain a second word segmentation sequence; using a second entity recognition model to obtain entity attribute identifiers corresponding to each word segmentation in the second word segmentation sequence, where the second entity recognition model is used to obtain entity attribute identifiers corresponding to each word segmentation through a feedforward neural network according to the word segmentation features corresponding to each word segmentation in the word segmentation sequence, and the entity attribute identifier is used to indicate whether each word segmentation in the word segmentation sequence belongs to a named entity; according to the entity attribute identifiers corresponding to each word segmentation in the second word segmentation sequence, label the word segmentations of the preset entity type; extract the word segmentations in the labeled second word segmentation sequence to obtain the named entities; By performing named entity recognition on user data on the target client, extract multiple target entities associated with the named entity; Perform information recommendation on the information to be recommended according to the association degree between the multiple target entities and the named entity, including: according to a pre-constructed knowledge graph, determine the weight value between each target entity and each named entity, where the knowledge graph is constructed according to the named entity, and each weight value is used to characterize the association degree between the target entity and each named entity; perform weighted operation on the weight values corresponding to all target entities to obtain a target weight; in the case where the target weight is greater than or equal to a first predetermined value, recommend the information to be recommended to the target client, and in the case where the target weight is less than the first predetermined value, the information to be recommended will not be recommended to the target client.

2. The method according to claim 1, wherein Before determining the weight value between each target entity and each named entity according to the pre-constructed knowledge graph, the method further includes: Search for associated information associated with the named entity from a preset database; Through knowledge extraction of the associated information, obtain multiple associated entities and an association weight for characterizing the association degree between each associated entity; According to the first word feature vector corresponding to each associated entity and the second word feature vector corresponding to the named entity, determine the first similarity between each associated entity and the named entity; Select the target similarity with the largest value from the first similarities, and in the case where the target similarity is greater than or equal to a second predetermined value, perform knowledge fusion on the associated entity and the named entity corresponding to the target similarity to obtain a fusion result; Construct the knowledge graph according to the multiple associated entities, the association weights between each associated entity, and the fusion result.

3. The method according to claim 2, wherein The determining the weight value between each target entity and each named entity according to the pre-constructed knowledge graph includes: According to the first attribute value of the attribute corresponding to the associated entity and the second attribute value of the same attribute corresponding to each target entity, determine the second similarity between the associated entity and each target entity; Find a first associated entity from the multiple associated entities, where the second similarity between each first associated entity and each target entity is greater than or equal to a third predetermined value; Determine the association weight between the first associated entity and each named entity as the weight value between the target entity corresponding to the first associated entity and each named entity.

4. The method according to claim 2, wherein The obtaining of multiple associated entities and the association weights representing the degree of association between the various associated entities through knowledge extraction of the association information includes: Determine a first text corpus in the association information; Perform word segmentation processing on the first text corpus to obtain a first word segmentation sequence; Use a first entity recognition model to obtain the entity attribute identifier corresponding to each word segment in the first word segmentation sequence, where the first entity recognition model is used to obtain the entity attribute identifier corresponding to each word segment through a feedforward neural network according to the word segment features corresponding to the word segments in the word segmentation sequence, and the entity attribute identifier is used to indicate whether each word segment in the word segmentation sequence belongs to a named entity; Determine the associated entities in the association information according to the entity attribute identifiers corresponding to the word segments in the first word segmentation sequence; Determine the association weights between the multiple associated entities according to the punctuation marks in the sentences where the multiple associated entities are located.

5. The method according to claim 1, characterized in that, The recommending the information to be recommended to the target client includes: Obtain the user login location information and online time information in the user data; Push the information to be recommended to the target client at a preset time and a preset login location according to the user login location information and the online time information.

6. An information recommendation device, characterized in that, including: An obtaining unit, configured to obtain information to be recommended, where the information to be recommended is information to be pushed to a target client; A first extraction unit, configured to extract named entities corresponding to a preset entity type by performing named entity recognition on the information to be recommended, including: determining a second text corpus in the information to be recommended; performing word segmentation processing on the text corpus to obtain a second word segmentation sequence; using a second entity recognition model to obtain the entity attribute identifier corresponding to each word segment in the second word segmentation sequence, the second entity recognition model is used to obtain the entity attribute identifier corresponding to each word segment through a feedforward neural network according to the word segment features corresponding to the word segments in the word segmentation sequence, and the entity attribute identifier is used to indicate whether each word segment in the word segmentation sequence belongs to a named entity; annotating the word segments of the preset entity type according to the entity attribute identifiers corresponding to the word segments in the second word segmentation sequence; extracting the word segments in the second word segmentation sequence after annotation to obtain the named entities; A second extraction unit, configured to extract multiple target entities associated with the named entities by performing named entity recognition on the user data on the target client; A recommendation unit for recommending the information to be recommended according to the degree of association between the multiple target entities and the named entity, including: determining a weight value between each target entity and each named entity according to a pre-constructed knowledge graph, where the knowledge graph is constructed based on the named entity, and each weight value is used to characterize the degree of association between the target entity and each named entity; obtaining a target weight through weighted calculation of the weight values corresponding to all target entities; recommending the information to be recommended to the target client when the target weight is greater than or equal to a first predetermined value, and not recommending the information to be recommended to the target client when the target weight is less than the first predetermined value.

7. An electronic device, characterized in that, It includes a processor, a communication interface, a memory, and a communication bus. Among them, the processor, the communication interface, and the memory complete mutual communication through the communication bus; The memory is used for storing computer programs; The processor is used to implement the steps of the information recommendation method according to any one of claims 1-5 when executing the programs stored on the memory.

8. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the information recommendation method according to any one of claims 1-5.

Images