Advertisement search method, model training method and device

Abstract

The present disclosure proposes an advertisement search method, a model training method and a device, which relate to the field of artificial intelligence technology, specifically the technical fields of deep learning, natural language processing and big models, and can be applied to scenarios such as intelligent search. The method comprises: obtaining user search terms; inputting the user search terms into a big model, and determining advertisers associated with candidate advertisements based on the user search terms through the big model; and determining target advertisements to be searched from the candidate advertisements based on the advertisers. Thus, the present disclosure can determine advertisers associated with candidate advertisements through the big model, and determine target advertisements to be searched under the restrictions of the advertisers, thereby improving the efficiency of advertisement search, ensuring the accuracy of advertisement search results, and enhancing user experience.

Description

Technical Field

[0001] The present disclosure relates to the field of artificial intelligence technology, specifically to the technical fields of deep learning, natural language processing and large models, and in particular to an advertisement search method, a model training method, an apparatus, an electronic device, a storage medium and a computer program product. Background Art

[0002] At present, with the continuous development of artificial intelligence technology, large language models have the advantages of good generalization and have been widely used in information extraction, text credibility assessment, machine translation and other fields. However, the advertising search method in the related art has the problem of low accuracy and efficiency in the advertising search process. Summary of the invention

[0003] The present disclosure proposes an advertisement search method, a model training method, a device, an electronic device, a storage medium and a computer program product.

[0004] According to the first aspect of the present disclosure, an advertisement search method is proposed, comprising: obtaining user search terms; inputting the user search terms into a large model, and determining advertisers associated with candidate advertisements based on the user search terms through the large model; and determining target advertisements to be searched from the candidate advertisements based on the advertisers.

[0005] According to a second aspect of the present disclosure, a model training method is proposed, comprising: obtaining training samples, wherein the training samples include sample advertising information and sample advertisers; and training a large model according to the training samples.

[0006] According to a third aspect of the present disclosure, an advertisement search device is proposed, comprising: an acquisition module for acquiring user search terms; a first determination module for inputting the user search terms into a large model, and determining advertisers associated with candidate advertisements based on the user search terms through the large model; and a second determination module for determining target advertisements to be searched from the candidate advertisements based on the advertisers.

[0007] According to a fourth aspect of the present disclosure, a model training device is proposed, including a first acquisition module for acquiring training samples, wherein the training samples are for sample advertising information and sample advertisers; and a training module for training a large model based on the training samples.

[0008] According to the fifth aspect of the present disclosure, an electronic device is proposed, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the advertisement search method proposed in the first aspect above, or execute the model training method proposed in the second aspect above.

[0009] According to the sixth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions is proposed, wherein the computer instructions are used to enable the computer to execute the advertisement search method proposed in the first aspect, or to execute the model training method proposed in the second aspect.

[0010] According to a seventh aspect of the present disclosure, a computer program product is proposed, including a computer program, which, when executed by a processor, implements the advertisement search method proposed in the first aspect, or implements the model training method proposed in the second aspect.

[0011] It should be understood that the content described in this section is not intended to identify the key or important features of the embodiments of the present disclosure, nor is it intended to limit the scope of the present disclosure. Other features of the present disclosure will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] The accompanying drawings are used to better understand the present solution and do not constitute a limitation of the present disclosure.

[0013] Figure 1 A flowchart of an advertisement search method according to an embodiment of the present disclosure;

[0014] Figure 2 A flowchart of an advertisement search method according to another embodiment of the present disclosure;

[0015] Figure 3 A flowchart of an advertisement search method according to another embodiment of the present disclosure;

[0016] Figure 4 A flowchart of a model training method according to another embodiment of the present disclosure;

[0017] Figure 5 A flowchart of a model training method according to another embodiment of the present disclosure;

[0018] Figure 6 A flowchart of a model training method according to another embodiment of the present disclosure;

[0019] Figure 7 A flowchart of a model training method according to another embodiment of the present disclosure;

[0020] Figure 8 It is a structural schematic diagram of a model training device according to an embodiment of the present disclosure;

[0021] Fig. 9 A schematic block diagram of an electronic device according to an embodiment of the present disclosure. DETAILED DESCRIPTION

[0022] The following is a description of exemplary embodiments of the present disclosure in conjunction with the accompanying drawings, including various details of the embodiments of the present disclosure to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present disclosure. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0023] Artificial Intelligence (AI) is a technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. At present, AI technology has the advantages of high automation, high precision and low cost, and has been widely used.

[0024] Deep Learning (DL) is a new research direction in the field of Machine Learning (ML). It is a science that studies the inherent laws and representation levels of sample data, enabling machines to have analytical learning capabilities like humans and to recognize data such as text, images, and sounds. It is widely used in speech and image recognition.

[0025] Natural Language Processing (NLP) is an important field in the fields of computer science and artificial intelligence. It studies various theories and methods that can achieve effective communication between humans and computers using natural language. Natural Language Processing is a science that integrates linguistics, computer science, and mathematics.

[0026] A large model refers to a machine learning model with a large parameter scale and complexity. It requires a large amount of computing resources and storage space to train and store, and often requires distributed computing and special hardware acceleration technology. Large models have stronger generalization and expression capabilities.

[0027] Figure 1 FIG. 1 is a flow chart of an advertisement search method according to an embodiment of the present disclosure. Figure 1 As shown, the method includes:

[0028] S101, obtaining user search terms.

[0029] It should be noted that the execution subject of the advertisement search method of the embodiment of the present disclosure may be a hardware device with data information processing capabilities and / or the necessary software required to drive the hardware device to work. Optionally, the execution subject may include a workstation, a server, a computer, a user terminal and other intelligent devices. Among them, the user terminal includes but is not limited to a mobile phone, a computer, an intelligent voice interaction device, an intelligent home appliance, a vehicle-mounted terminal, etc.

[0030] It should be noted that the user search term (query) is the search information entered by the user on the terminal device, for example: the search information entered by the user on the mobile phone, the search information entered by the user on the computer, etc.

[0031] It should be noted that after the user inputs search information (user search term) on the terminal device, the user search term can be directly obtained.

[0032] It should be noted that the present disclosure does not limit the form of the user search term. For example, the user search term can be a word, a sentence, a text paragraph, etc.

[0033] For example, the user search term may be "headphones"; the user search term may also be "registration time for exam X".

[0034] S102, inputting the user search words into the big model, and determining the advertisers associated with the candidate advertisements based on the user search words through the big model.

[0035] Among them, advertisers refer to legal persons, other economic organizations or individuals who design, produce and publish advertisements by themselves or entrust others to promote goods or provide services. Advertisers are the initiators of advertising activities, merchants who sell or promote their own products and services online, and providers of affiliate marketing advertisements. Any merchant who promotes and sells its products or services can be an advertiser.

[0036] It should be noted that the big model is pre-trained using a deep learning algorithm. When the user search terms are input into the big model, the big model can output advertisers associated with the candidate advertisements based on the user search terms.

[0037] Optionally, a large base model may be selected and model training may be performed on the selected large base model.

[0038] For example, a supervised fine-tuning (SFT) strategy can be used to train large models.

[0039] It should be noted that there are often multiple advertisers associated with candidate advertisements.

[0040] S103: Determine a target advertisement to be searched from the candidate advertisements based on the advertiser.

[0041] In the embodiment of the present disclosure, after the advertiser is acquired, the target advertisement to be searched can be determined from the candidate advertisements based on the advertiser.

[0042] Optionally, the advertiser's business information may be obtained, and a target advertisement may be determined from candidate advertisements based on the user search terms and the business information.

[0043] Optionally, the similarity between the user search term and the business information may be obtained, and the target advertisement may be determined from the candidate advertisements based on the similarity.

[0044] The advertisement search method proposed in the present disclosure obtains user search terms, inputs the user search terms into a big model, determines the advertisers associated with the candidate advertisements based on the user search terms through the big model, and determines the target advertisements to be searched from the candidate advertisements based on the advertisers. Thus, the present disclosure can determine the advertisers associated with the candidate advertisements through the big model, and determine the target advertisements to be searched under the restrictions of the advertisers, thereby improving the efficiency of advertisement search, ensuring the accuracy of advertisement search results, and enhancing user experience.

[0045] Figure 2 It is a flowchart of an advertisement search method according to the second embodiment of the present disclosure.

[0046] like Figure 2 As shown, in Figure 1 Based on the illustrated embodiment, the pattern generating method of the embodiment of the present disclosure may specifically include the following steps:

[0047] S201, obtaining user search terms.

[0048] For the relevant contents of step S201, please refer to the above embodiment, which will not be described again here.

[0049] S202, inputting the user search terms into the big model, and determining the advertisers associated with the candidate advertisements based on the user search terms through the big model.

[0050] In the disclosed embodiment, the user search term is identified as a first category user search term through a large model, and based on the first category user search term, a first advertiser is determined, wherein the first advertiser is a real advertiser.

[0051] Among them, the first advertiser is a real advertiser, that is, a real advertiser.

[0052] It should be noted that the resource value (commercial value) of the first type of user search terms is often higher than that of the second type of user search terms, namely the “high business” sample search terms.

[0053] In an embodiment of the present disclosure, a user search term is identified as a second category user search term through a large model, and based on the second category user search term, a second advertiser is determined and indication information is generated, wherein the second advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category user search term.

[0054] The second advertiser is a virtual advertiser, that is, an advertiser that does not really exist and is a virtual advertiser constructed in advance.

[0055] It should be noted that the resource value (commercial value) of the second category of sample search terms is often low, that is, the “low-commercial” sample search terms.

[0056] Further, after identifying that the user search term is a second-category user search term, indication information may be sent to the source client of the second-category user search term.

[0057] The step S103 of “determining the target advertisement to be searched from the candidate advertisements based on the advertiser” in the above embodiment may specifically include the following steps S203 and S203 .

[0058] S203, obtaining the advertiser's business information.

[0059] It should be noted that the business information corresponding to different advertisers may be the same or different, and there may be one or more business information of an advertiser.

[0060] For example, the business information of advertiser 1 is education and finance; the business information of advertiser 2 is education and medical care, etc.

[0061] Optionally, after acquiring the advertiser's business information, an advertiser business information database may be constructed.

[0062] S204, determining a target advertisement from candidate advertisements based on the user search term and the business information.

[0063] The step S204 of the above embodiment, “determining a target advertisement from candidate advertisements based on the user search term and the business information”, may specifically include the following steps S205 and S206 .

[0064] S205, obtaining the similarity between the user search term and the business information.

[0065] In the disclosed embodiment, after acquiring the advertiser's business information database, the similarity between the user's search term and the business information may be acquired.

[0066] It should be noted that the present disclosure does not limit the specific method for obtaining the similarity between the user search term and the business information, and it can be selected according to the actual situation.

[0067] Optionally, the similarity between the user search term and the business information may be obtained by using an approximate nearest neighbor search algorithm (ANN for short).

[0068] S206: Determine a target advertisement from the candidate advertisements based on the similarity.

[0069] In the embodiment of the present disclosure, after the similarity between the user search term and the business information is obtained, a target advertisement can be determined from the candidate advertisements according to the similarity between the user search term and the business information.

[0070] Optionally, if the similarity between the user search term and the business information is greater than a similarity threshold, the candidate advertisement corresponding to the business information is determined as the target advertisement.

[0071] It should be noted that the present disclosure does not limit the setting of the similarity threshold, and it can be set according to actual conditions. For example, the similarity threshold can be set to 80%; for another example, the similarity threshold can be set to 85%.

[0072] For example, for a similarity threshold of 80%, if the similarity between the user search term and the business information is greater than 80%, the candidate advertisement corresponding to the part of the business information is determined as the target advertisement.

[0073] Optionally, multiple candidate advertisements may be sorted in descending order of similarity, and the top N candidate advertisements may be determined as target advertisements, where N is a positive integer.

[0074] It should be noted that the present disclosure does not limit the number of N, and it can be set according to actual conditions. For example, N can be set to 500; for another example, N can be set to 200.

[0075] For example, when N is 500, multiple candidate advertisements may be sorted in descending order of similarity, and the top 500 candidate advertisements may be determined as target advertisements according to the sorting result.

[0076] In summary, the advertisement search method proposed in the present invention obtains user search terms, inputs the user search terms into a big model, determines the advertiser associated with the candidate advertisement based on the user search terms through the big model, obtains the business information of the advertiser, obtains the similarity between the user search terms and the business information, and determines the target advertisement from the candidate advertisements based on the similarity. Through the big model, the understanding of the user search terms is improved, and then the accuracy of the search results is improved. Through the big model, advertisers are generated based on the user search terms, which ensures the diversity of advertisements, thereby improving the efficiency of advertisement search and realizing end-to-end advertisement targeting.

[0077] Figure 3FIG. 1 is a flow chart of a model training method according to an embodiment of the present disclosure. Figure 3 As shown, the method includes:

[0078] S301, obtaining training samples, wherein the training samples include sample advertising information and sample advertisers.

[0079] It should be noted that the execution subject of the model training method of the embodiment of the present disclosure may be a hardware device with data information processing capabilities and / or the necessary software required to drive the hardware device to work. Optionally, the execution subject may include a workstation, a server, a computer, a user terminal and other intelligent devices. Among them, the user terminal includes but is not limited to a mobile phone, a computer, an intelligent voice interaction device, a smart home appliance, a car terminal, etc.

[0080] It should be noted that the present application does not limit the specific method for obtaining training samples, which can be selected according to actual conditions.

[0081] Optionally, sample advertisement information may be obtained, and the sample advertisement information and sample advertisers may be associated to obtain training samples.

[0082] In the disclosed embodiment, the sample advertisement information may be sample advertisement keywords and / or sample advertisement landing page information.

[0083] For example, the sample advertisement keyword a1 and the sample advertisement landing page information b1 may be associated with the sample advertiser A to obtain training sample 1.

[0084] In the disclosed embodiment, the sample advertising information may also be sample user search terms. The sample user search terms may be classified into first-category sample user search terms and second-category sample user search terms. The corresponding sample advertisers may be determined based on different types of user search terms.

[0085] Optionally, for the first type of sample user search terms, the corresponding sample advertisers are real advertisers, and the first type of sample user search terms and the real advertisers may be associated to obtain training samples.

[0086] For example, the user search term query1 can be associated with sample advertiser B in the first category to obtain training sample 2.

[0087] Optionally, for the second category sample user search terms, the corresponding sample advertiser is a virtual advertiser, and the second category sample user search terms are associated with the virtual advertiser and indication information to obtain training samples, wherein the indication information is used to indicate that there is no target advertisement to be searched for the second category sample user search terms.

[0088] For example, the user search term query2 may be associated with the sample advertiser A and the instruction information in the second category to obtain training sample 3.

[0089] S302: Train the large model based on the training samples.

[0090] In the embodiment of the present disclosure, after the training samples are obtained, the large model can be trained according to the training samples.

[0091] It should be noted that the type of large model can be selected according to actual conditions, and no excessive restrictions are made here.

[0092] It should be noted that for the training method of large models, the supervised fine-tuning SFT strategy can be used to train the large model.

[0093] Optionally, based on the multiple training samples obtained, sample advertising information in the training samples can be input into the big model, and the big model can output predicted advertisers. The big model can be trained based on the predicted advertisers and the sample advertisers in the training samples.

[0094] For example, the loss function of the large model can be obtained based on the predicted advertisers and sample advertisers, the model parameters of the large model can be updated based on the loss function, and the next training sample can be returned to continue training the large model with the adjusted model parameters until the model training end conditions are met.

[0095] It should be noted that the present disclosure does not limit the setting of the model training end condition, and the model training end condition can be set according to actual conditions.

[0096] Optionally, the model training termination condition can be set as the loss function value being less than a preset loss threshold; optionally, the model training termination condition can also be set as the number of adjustments to the model parameters of the large model reaching a preset number threshold.

[0097] The model training method proposed in the present invention obtains training samples, wherein the training samples include sample advertising information and sample advertisers, and trains the big model based on the training samples. Thus, the big model is trained by obtaining a variety of training samples, which improves the comprehensiveness and accuracy of the training samples, improves the accuracy and reliability of the big model training process, and lays a foundation for advertising search processing based on the big model.

[0098] Figure 4 FIG. 1 is a flow chart of a model training method according to another embodiment of the present disclosure. Figure 4 As shown, the method includes:

[0099] The step S301 “obtaining training samples” in the above embodiment may specifically include the following steps S401 and S402 .

[0100] S401, obtaining sample advertisement keywords and / or sample advertisement landing page information as first sample advertisement information.

[0101] It should be noted that the sample advertiser can set sample advertisement keywords and / or sample advertisement landing page information according to different advertisements, wherein the sample advertisement keywords and sample advertisement landing page can be used to reflect the advertisement content.

[0102] Among them, the advertising landing page information refers to the web page information that is displayed to the user after the user clicks on the advertisement or searches using a search engine.

[0103] For example, the sample advertising keyword may be the registration time for the X exam, the sample advertising landing page information may be the 2023 X exam preparation tutoring center, and the first sample advertiser may be {gMASK1}.

[0104] S402: Associating first sample advertisement information with first sample advertisers to generate a first training sample.

[0105] In the embodiment disclosed in the present application, after the first sample advertisement information and the first sample advertiser are acquired, the first sample advertisement information and the first sample advertiser may be associated to generate a first training sample.

[0106] In the above embodiment, step S302 “training the large model according to the training samples” may specifically include the following steps S403 and S404.

[0107] S403: input the first sample advertising information into the big model, and determine the first predicted advertiser based on the first sample advertising information through the big model.

[0108] For example, for the first sample advertising information (the sample advertising keywords are the registration time for the X exam and the sample advertising landing page information is the 2023 X exam preparation and tutoring center), by inputting the above first sample advertising information into the big model, the first predicted advertiser can be output as "Y Education and Training Co., Ltd."

[0109] S404: Training the large model based on the first predicted advertiser and the first sample advertiser.

[0110] It should be noted that for the training method of large models, the supervised fine-tuning SFT strategy can be used to train the large model.

[0111] For example, based on the first predicted advertiser and the first sample advertiser, the loss function of the large model can be obtained, the model parameters of the large model can be updated based on the loss function, and the next training sample can be returned to continue training the large model with adjusted model parameters until the model training end conditions are met.

[0112] The model training method proposed in the present invention obtains sample advertising keywords and / or sample advertising landing page information as first sample advertising information, associates the first sample advertising information with the first sample advertiser, generates a first training sample, inputs the first sample advertising information into the big model, determines the first predicted advertiser based on the first sample advertising information through the big model, trains the big model based on the first predicted advertiser and the first sample advertiser, and trains the big model by using the obtained sample advertising keywords and / or sample advertising landing page information as sample advertising information, so that the big model can learn the mapping relationship between advertising content and advertisers, laying a foundation for advertising search processing based on the big model.

[0113] Figure 5 FIG. 1 is a flow chart of a model training method according to another embodiment of the present disclosure. Figure 5 As shown, the method includes:

[0114] The step S301 “obtaining training samples” in the above embodiment may specifically include the following steps S501 and S502 .

[0115] S501, obtaining the first type of sample user search terms as the second sample advertising information.

[0116] In the embodiment of the present disclosure, there is no limitation on the specific method for obtaining the search terms of the first type of sample users, and the method may be selected according to actual conditions.

[0117] In the disclosed embodiment, the user's click search behavior log may be obtained, and the first type of sample user search terms may be determined based on the click search behavior log.

[0118] Optionally, the user's historical click search behavior can be obtained to generate a user's click search behavior log, and feature extraction can be performed on the click search behavior log to determine the first type of sample user search terms, that is, the second sample advertising information.

[0119] It should be noted that the present disclosure does not limit the specific method of determining the search terms of the first type of sample users based on the click search behavior logs, and it can be selected according to actual conditions.

[0120] Optionally, a preset feature extraction model may be used to extract the first type of sample user search terms from the click search behavior log to obtain the second sample advertising information.

[0121] It should be noted that the resource value (commercial value) of the first type of sample search terms is often higher than that of the second type of sample search terms, namely the “high business” sample search terms.

[0122] S502: Associating the second sample advertisement information with the second sample advertiser to generate a second training sample, wherein the second sample advertiser is a real advertiser.

[0123] In the disclosed embodiment of the present application, after the second sample advertisement information and the second sample advertiser are acquired, the second sample advertisement information and the second sample advertiser may be associated to generate a second training sample.

[0124] Among them, the second sample advertiser is a real advertiser, that is, an actual advertiser.

[0125] In the above embodiment, step S502 “training the large model according to the training samples” may specifically include the following steps S503 and S504.

[0126] S503: Input the second sample advertising information into the big model, and determine the second predicted advertiser based on the second sample advertising information through the big model.

[0127] For example, for the second sample advertising information (the first type of sample user search term query1) being "X exam registration time and registration conditions", by inputting the first type of sample user search term query1 into the big model, the second predicted advertiser can be output as "Z Education and Training Co., Ltd."

[0128] S504: Training the large model based on the second predicted advertiser and the second sample advertiser.

[0129] It should be noted that for the training method of large models, the supervised fine-tuning SFT strategy can be used to train the large model.

[0130] For example, based on the second predicted advertiser and the second sample advertiser, the loss function of the large model can be obtained, the model parameters of the large model can be updated based on the loss function, and the next training sample can be returned to continue training the large model with adjusted model parameters until the model training end conditions are met.

[0131] The model training method proposed in the present invention obtains the first category sample user search terms as the second sample advertising information, associates the second sample advertising information with the second sample advertiser, generates a second training sample, wherein the second sample advertiser is a real advertiser, inputs the second sample advertising information into the big model, determines the second predicted advertiser based on the second sample advertising information through the big model, trains the big model based on the second predicted advertiser and the second sample advertiser, and trains the big model by using the first category sample user search terms obtained as the sample advertising information, so that the big model can identify the type of user search terms and learn the mapping relationship between the first category user search terms and advertisers, thereby laying a foundation for advertising search processing based on the big model.

[0132] Figure 6 FIG. 1 is a flow chart of a model training method according to another embodiment of the present disclosure. Figure 6 As shown, the method includes:

[0133] The step S301 “obtaining training samples” in the above embodiment may specifically include the following steps S601 and S602 .

[0134] S601, obtaining the second type of sample user search terms as third sample advertising information.

[0135] It should be noted that in order to improve the recall accuracy of the large model, the second category of sample user search terms are added as the third sample advertising information, where the resource value (commercial value) of the second category of sample search terms is often low, that is, the "low-commerce" sample user search terms.

[0136] For example, the third sample advertising information (the second type of sample user search term query2) is "What does Jiefu mean?"

[0137] S602, associating the third sample advertising information with the third sample advertiser and the indication information to generate a third training sample, wherein the third sample advertiser is a virtual advertiser, and the indication information is used to indicate that the target advertisement to be searched does not exist for the search term of the second type of sample user.

[0138] In the embodiment disclosed in the present application, after the third sample advertisement information, the third sample advertiser, and the instruction information are acquired, the third sample advertisement information, the third sample advertiser, and the instruction information may be associated to generate a third training sample.

[0139] Among them, the third sample advertiser is a virtual advertiser, that is, an advertiser that does not really exist and is a virtual advertiser constructed in advance.

[0140] In the embodiment of the present disclosure, the indication information may indicate that there is no target advertisement to be searched for the search term of the second type of sample users.

[0141] In the above embodiment, step S302 “training the large model according to the training samples” may specifically include the following steps S603 and S604.

[0142] S603: Input the third sample advertising information into the big model, and determine a third predicted advertiser based on the third sample advertising information through the big model.

[0143] For example, for the third sample advertising information (second-category sample user search term query2) being "What does Jiefu mean", by inputting the second-category sample user search term query2 into the big model, the third predicted advertiser can be output as "M".

[0144] S604: Training the large model based on the third predicted advertiser, the third sample advertiser and the instruction information.

[0145] It should be noted that for the training method of large models, the supervised fine-tuning SFT strategy can be used to train the large model.

[0146] The model training method proposed in the present invention obtains the second category sample user search terms as the third sample advertising information, associates the third sample advertising information with the third sample advertiser and the indication information, and generates a third training sample, wherein the third sample advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category sample user search terms, inputs the third sample advertising information into the big model, determines the third predicted advertiser based on the third sample advertising information through the big model, and trains the big model based on the third predicted advertiser, the third sample advertiser and the indication information. By training the big model with the obtained second category sample user search terms as the sample advertising information, the big model can be enabled to identify the type of user search terms, and the big model can learn the mapping relationship between the second category user search terms, the virtual advertiser and the indication information, thereby laying a foundation for advertising search processing based on the big model.

[0147] To sum up, the model training method proposed in this disclosure, by training the large model with multiple training samples obtained, improves the comprehensiveness and accuracy of the training samples, improves the accuracy and reliability of the large model training process, and lays the foundation for advertising search processing based on the large model.

[0148] In the technical solution of the present disclosure, the collection, storage, use, processing, transmission, provision and disclosure of user personal information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0149] According to an embodiment of the present disclosure, the present disclosure also provides an advertisement search device for implementing the above-mentioned advertisement search method.

[0150] Figure 7 is a block diagram of an advertisement search device according to an embodiment of the present disclosure.

[0151] like Figure 7 As shown, the advertisement search device 700 includes: an acquisition module 701 , a first determination module 702 and a second determination module 703 .

[0152] The acquisition module 701 is used to acquire the user's search terms;

[0153] A first determination module 702, configured to input the user search term into a large model, and determine an advertiser associated with a candidate advertisement based on the user search term through the large model;

[0154] The second determination module 703 is used to determine the target advertisement to be searched from the candidate advertisements based on the advertiser.

[0155] In one embodiment of the present disclosure, the second determination module 703 is used to: obtain the business information of the advertiser; and determine the target advertisement from the candidate advertisements based on the user search term and the business information.

[0156] In one embodiment of the present disclosure, the second determination module 703 is used to: obtain the similarity between the user search term and the business information; and determine the target advertisement from the candidate advertisements based on the similarity.

[0157] In one embodiment of the present disclosure, the second determination module 703 is used to: if the similarity between the user search term and the business information is greater than the similarity threshold, determine the candidate advertisement corresponding to the business information as the target advertisement.

[0158] In one embodiment of the present disclosure, the second determination module 703 is used to: sort the multiple candidate advertisements in descending order according to the similarity, and determine the top N candidate advertisements as the target advertisements, where N is a positive integer.

[0159] In one embodiment of the present disclosure, the first determination module 702 is used to: identify the user search term as a first category of user search terms through the large model, and determine a first advertiser based on the first category of user search terms, wherein the first advertiser is a real advertiser.

[0160] In one embodiment of the present disclosure, the first determination module 702 is used to: identify the user search term as a second category of user search term through the large model, determine a second advertiser based on the second category of user search term and generate indication information, wherein the second advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category of user search term.

[0161] In one embodiment of the present disclosure, the device 700 is further used to: send the indication information to the source client of the user search term.

[0162] The advertisement search device proposed in the present disclosure obtains user search terms, inputs the user search terms into a big model, determines the advertisers associated with the candidate advertisements based on the user search terms through the big model, and determines the target advertisements to be searched from the candidate advertisements based on the advertisers. Thus, the present disclosure can determine the advertisers associated with the candidate advertisements through the big model, and determine the target advertisements to be searched under the restrictions of the advertisers, thereby improving the efficiency of advertisement search, ensuring the accuracy of advertisement search results, and enhancing user experience.

[0163] According to an embodiment of the present disclosure, the present disclosure also provides a model training device for implementing the above-mentioned model training method.

[0164] Figure 8 It is a block diagram of a model training device according to an embodiment of the present disclosure.

[0165] like Figure 8 As shown, the model training device 800 includes: a first acquisition module 801 and a training module 802.

[0166] A first acquisition module 801 is used to acquire training samples, wherein the training samples are used for sample advertising information and sample advertisers;

[0167] The training module 802 is used to train the large model according to the training samples.

[0168] In one embodiment of the present disclosure, the first acquisition module 801 is used to: acquire sample advertisement keywords and / or sample advertisement landing page information as first sample advertisement information; associate the first sample advertisement information with the first sample advertiser to generate a first training sample.

[0169] In one embodiment of the present disclosure, the training module 802 is used to: input the first sample advertising information into the big model, determine the first predicted advertiser based on the first sample advertising information through the big model; and train the big model based on the first predicted advertiser and the first sample advertiser.

[0170] In one embodiment of the present disclosure, the first acquisition module 801 is used to: acquire a first type of sample user search terms as second sample advertising information; associate the second sample advertising information with a second sample advertiser to generate a second training sample, wherein the second sample advertiser is a real advertiser.

[0171] In one embodiment of the present disclosure, the first acquisition module 801 is used to: acquire the user's click search behavior log; and determine the first type of sample user search terms according to the click search behavior log.

[0172] In one embodiment of the present disclosure, the training module 802 is used to: input the second sample advertising information into the big model, determine the second predicted advertiser based on the second sample advertising information through the big model; and train the big model based on the second predicted advertiser and the second sample advertiser.

[0173] In one embodiment of the present disclosure, the first acquisition module 801 is used to: obtain the second category sample user search terms as third sample advertising information; associate the third sample advertising information with the third sample advertiser and indication information to generate a third training sample, wherein the third sample advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category sample user search terms.

[0174] In one embodiment of the present disclosure, the training module 802 is further used to: input the third sample advertising information into the big model, determine the third predicted advertiser based on the third sample advertising information through the big model; and train the big model based on the third predicted advertiser, the third sample advertiser and the indication information.

[0175] The model training device proposed in the present invention obtains training samples, wherein the training samples include sample advertising information and sample advertisers, and trains the big model according to the training samples. Thus, the big model is trained by obtaining a variety of training samples, which improves the comprehensiveness and accuracy of the training samples, improves the accuracy and reliability of the big model training process, and lays a foundation for advertising search processing based on the big model.

[0176] According to an embodiment of the present disclosure, the present disclosure also proposes an electronic device, a readable storage medium, and a computer program product.

[0177] Fig. 9A schematic block diagram of an example electronic device 900 that can be used to implement an embodiment of the present disclosure is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present disclosure described and / or required herein.

[0178] like Fig. 9 As shown, the device 900 includes a computing unit 901, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 902 or a computer program loaded from a storage unit 908 into a random access memory (RAM) 903. In the RAM 903, various programs and data required for the operation of the device 900 can also be stored. The computing unit 901, the ROM 902, and the RAM 903 are connected to each other via a bus 904. An input / output (I / O) interface 905 is also connected to the bus 904.

[0179] A number of components in the device 900 are connected to the I / O interface 905, including: an input unit 906, such as a keyboard, a mouse, etc.; an output unit 907, such as various types of displays, speakers, etc.; a storage unit 908, such as a disk, an optical disk, etc.; and a communication unit 909, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 909 allows the device 900 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0180] The computing unit 901 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 901 performs the various methods and processes described above, such as an advertisement search method and a model training method. For example, in some embodiments, the advertisement search method and the model training method may be implemented as a computer software program, which is tangibly contained in a machine-readable medium, such as a storage unit 908. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 900 via the ROM 902 and / or the communication unit 909. When the computer program is loaded into the RAM 903 and executed by the computing unit 901, one or more steps of the advertisement search method and the model training method described above may be executed. Alternatively, in other embodiments, the computing unit 901 may be configured to execute the advertisement search method and the model training method in any other appropriate manner (for example, by means of firmware).

[0181] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0182] The program code for implementing the method of the present disclosure may be written in any combination of one or more programming languages. These program codes may be presented to a processor or controller of a general-purpose computer, a special-purpose computer, or other programmable data processing device, so that the program code, when executed by the processor or controller, implements the functions / operations specified in the flow chart and / or block diagram. The program code may be executed entirely on the machine, partially on the machine, partially on the machine and partially on a remote machine as a stand-alone software package, or entirely on a remote machine or server.

[0183] In the context of the present disclosure, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0184] To propose interactions with a user account, the systems and techniques described herein may be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user account; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user account can propose input to the computer. Other types of devices may also be used to propose interactions with the user account; for example, feedback proposed to the user account may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user account may be received in any form (including acoustic input, voice input, or tactile input).

[0185] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user account computer with a graphical user account interface or a web browser through which a user account can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0186] A computer system may include a client and a server. The client and the server are generally remote from each other and usually interact through a communication network. The relationship of client and server is generated by computer programs running on respective computers and having a client-server relationship with each other. The server may be a cloud server, a server of a distributed system, or a server combined with a blockchain.

[0187] According to an embodiment of the present disclosure, the present disclosure also provides a computer program product, including a computer program, wherein when the computer program is executed by a processor, the steps of the advertisement search method and the model training method described in the above embodiments of the present disclosure are implemented.

[0188] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps recorded in this disclosure can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solutions disclosed in this disclosure can be achieved, and this document does not limit this.

[0189] The above specific implementations do not constitute a limitation on the protection scope of the present disclosure. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present disclosure shall be included in the protection scope of the present disclosure.

Claims

1. An advertisement search method, comprising: Get user search terms; Inputting the user search term into a large model, and determining an advertiser associated with a candidate advertisement based on the user search term through the large model; Based on the advertiser, determining a target advertisement to be searched from the candidate advertisements; The step of determining the advertiser associated with the candidate advertisement based on the user search term by using the large model includes: The user search term is identified as a second category user search term through the large model, and based on the second category user search term, a second advertiser is determined and indication information is generated, wherein the second advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category user search term.

2. The search method according to claim 1, wherein: The step of determining a target advertisement to be searched from the candidate advertisements based on the advertiser includes: Obtaining business information of the advertiser; The target advertisement is determined from the candidate advertisements based on the user search term and the business information.

3. The search method according to claim 2, wherein: The step of determining the target advertisement from the candidate advertisements based on the user search term and the business information includes: Obtaining the similarity between the user search term and the business information; The target advertisement is determined from the candidate advertisements according to the similarity.

4. The search method according to claim 3, wherein: The step of determining the target advertisement from the candidate advertisements according to the similarity comprises: If the similarity between the user search term and the business information is greater than a similarity threshold, the candidate advertisement corresponding to the business information is determined as the target advertisement.

5. The search method according to claim 3, wherein: The step of determining the target advertisement from the candidate advertisements according to the similarity comprises: The plurality of candidate advertisements are sorted in descending order according to the similarity, and the top N candidate advertisements are determined as the target advertisements, wherein N is a positive integer.

6. The search method according to any one of claims 1 to 5, wherein: The determining, by using the large model and based on the user search term, an advertiser associated with the candidate advertisement includes: The user search term is identified as a first category of user search terms through the large model, and a first advertiser is determined based on the first category of user search terms, wherein the first advertiser is a real advertiser.

7. The search method according to claim 1, wherein: The method further comprises: The indication information is sent to a source client of the user search term.

8. A model training method, comprising: Acquire a training sample, wherein the training sample includes sample advertising information and sample advertisers; Training the large model according to the training samples; The step of obtaining training samples includes: Obtaining the search terms of the second type of sample users as the third sample advertising information; The third sample advertising information is associated with a third sample advertiser and indication information to generate a third training sample, wherein the third sample advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the search term of the second type of sample user.

9. The training method according to claim 8, wherein: The obtaining of training samples comprises: Obtaining sample advertisement keywords and / or sample advertisement landing page information as first sample advertisement information; The first sample advertisement information and the first sample advertiser are associated to generate a first training sample.

10. The training method according to claim 9, wherein: The step of training the large model according to the training samples includes: inputting the first sample advertising information into the large model, and determining a first predicted advertiser based on the first sample advertising information through the large model; The large model is trained based on the first predicted advertiser and the first sample advertiser.

11. The training method according to claim 8, wherein: The obtaining of training samples comprises: Obtaining the search terms of the first type of sample users as the second sample advertising information; The second sample advertising information is associated with a second sample advertiser to generate a second training sample, wherein the second sample advertiser is a real advertiser.

12. The training method according to claim 11, wherein: The process of obtaining the search terms of the first type of sample users includes: Obtain the user's click and search behavior log; The first category of sample user search terms is determined according to the click search behavior log.

13. The training method according to claim 11, wherein: The step of training the large model according to the training samples includes: inputting the second sample advertising information into the large model, and determining a second predicted advertiser based on the second sample advertising information through the large model; The large model is trained based on the second predicted advertiser and the second sample advertiser.

14. The training method according to claim 8, wherein: The step of training the large model according to the training samples includes: inputting the third sample advertising information into the large model, and determining a third predicted advertiser based on the third sample advertising information through the large model; The large model is trained based on the third predicted advertiser, the third sample advertiser and the indication information.

15. An advertisement search device, comprising: The acquisition module is used to obtain the user's search terms; A first determination module, configured to input the user search term into a large model, and determine an advertiser associated with a candidate advertisement based on the user search term through the large model; A second determination module, configured to determine a target advertisement to be searched from the candidate advertisements based on the advertiser; Wherein, the first determining module is used to: The user search term is identified as a second category user search term through the large model, and based on the second category user search term, a second advertiser is determined and indication information is generated, wherein the second advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the second category user search term.

16. The search device according to claim 15, wherein: The second determining module is used to: Obtaining business information of the advertiser; The target advertisement is determined from the candidate advertisements based on the user search term and the business information.

17. The search device according to claim 16, wherein: The second determining module is used to: Obtaining the similarity between the user search term and the business information; The target advertisement is determined from the candidate advertisements according to the similarity.

18. The search device according to claim 17, wherein: The second determining module is used to: If the similarity between the user search term and the business information is greater than a similarity threshold, the candidate advertisement corresponding to the business information is determined as the target advertisement.

19. The search device according to claim 17, wherein: The second determining module is used to: The plurality of candidate advertisements are sorted in descending order according to the similarity, and the top N candidate advertisements are determined as the target advertisements, wherein N is a positive integer.

20. The search device according to any one of claims 15 to 19, wherein: The first determining module is used to: The user search term is identified as a first category of user search terms through the large model, and a first advertiser is determined based on the first category of user search terms, wherein the first advertiser is a real advertiser.

21. The search device according to claim 15, wherein: The device is also used for: The indication information is sent to a source client of the user search term.

22. A model training device, used for: The first acquisition module is used to acquire training samples, wherein: The training samples are used for sample advertising information and sample advertisers; A training module, used for training the large model according to the training samples; Wherein, the first acquisition module is used to: Obtaining the search terms of the second type of sample users as the third sample advertising information; The third sample advertising information is associated with a third sample advertiser and indication information to generate a third training sample, wherein the third sample advertiser is a virtual advertiser, and the indication information is used to indicate that there is no target advertisement to be searched for the search term of the second type of sample user.

23. The training device according to claim 22, wherein: The first acquisition module is used to: Obtaining sample advertisement keywords and / or sample advertisement landing page information as first sample advertisement information; The first sample advertisement information and the first sample advertiser are associated to generate a first training sample.

24. The training device according to claim 23, wherein: The training module is used to: inputting the first sample advertising information into the large model, and determining a first predicted advertiser based on the first sample advertising information through the large model; The large model is trained based on the first predicted advertiser and the first sample advertiser.

25. The training device according to claim 22, wherein: The first acquisition module is used to: Obtaining the search terms of the first type of sample users as the second sample advertising information; The second sample advertising information is associated with a second sample advertiser to generate a second training sample, wherein the second sample advertiser is a real advertiser.

26. The training device according to claim 25, wherein: The first acquisition module is used to: Obtain the user's click and search behavior log; The first category of sample user search terms is determined according to the click search behavior log.

27. The training device according to claim 26, wherein: The training module is used to: inputting the second sample advertising information into the large model, and determining a second predicted advertiser based on the second sample advertising information through the large model; The large model is trained based on the second predicted advertiser and the second sample advertiser.

28. The training device according to claim 22, wherein: The training module is also used for: inputting the third sample advertising information into the large model, and determining a third predicted advertiser based on the third sample advertising information through the large model; The large model is trained based on the third predicted advertiser, the third sample advertiser and the indication information.

29. An electronic device, characterized in that: including a processor and a memory; The processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, so as to implement the method according to any one of claims 1 to 7 or claims 8 to 14.

30. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 7 or claims 8 to 14 is implemented.

31. A computer program product comprising a computer program, which, when executed by a processor, implements the method according to any one of claims 1-7 or claims 8-14.

Images