Resource putting word mining method and device, electronic equipment and storage medium

Abstract

The application discloses a resource putting word mining method and device, electronic equipment and a storage medium, relates to the technical field of computers, and particularly relates to the field of artificial intelligence such as deep learning and large models. The specific implementation scheme is as follows: a candidate historical search word is acquired, and the candidate historical search word is clustered to obtain a candidate search word set; a plurality of search word samples are extracted from the candidate search word set; a large model is used to perform intent determination on the search word samples, and an intent determination result of the search word samples is determined; wherein the intent determination result is used to indicate whether the search word samples meet a resource putting intent of a target product; the candidate search word set is filtered according to the intent determination results of the plurality of search word samples, and a retained target search word set is obtained; and a target putting word is determined from the target search word set.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to the field of artificial intelligence such as deep learning and large models, specifically to a resource delivery word mining method, device, electronic device and storage medium. Background Technology

[0002] With the development of internet technology, various online resources have achieved rapid growth by leveraging massive traffic. Taking advertising as an example, compared to traditional brand advertising that focuses on building brand image, current online advertising emphasizes driving substantial user conversion through precise keyword targeting and other methods. Summary of the Invention

[0003] This application provides a method, apparatus, electronic device, and storage medium for resource placement keyword mining. The specific solution is as follows:

[0004] According to one aspect of this application, a method for mining keywords for resource placement is provided, comprising: Obtain candidate historical search terms and cluster them to obtain a set of candidate search terms; Extract multiple search term samples from the candidate search term set; A large model is used to determine the intent of search term samples and determine the intent determination result of the search term samples; the intent determination result is used to indicate whether the search term samples are consistent with the resource allocation intent of the target product; Based on the intent determination results of multiple search term samples, the candidate search term set is filtered to obtain the retained target search term set; Target keywords are determined from the target search term set.

[0005] According to another aspect of this application, a resource placement term mining apparatus is provided, comprising: The acquisition module is used to retrieve candidate historical search terms; The clustering module is used to cluster candidate historical search terms to obtain a set of candidate search terms; The extraction module is used to extract multiple search term samples from the candidate search term set; The intent determination module is used to perform intent determination on search term samples using a large model, and to determine the intent determination result of the search term samples; the intent determination result is used to indicate whether the search term samples are consistent with the resource allocation intent of the target product; The filtering module is used to filter the candidate search term set based on the intent determination results of multiple search term samples to obtain the retained target search term set. The determination module is used to identify target keywords from the target search term set.

[0006] According to another aspect of this application, an electronic device is provided, comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores instructions that can be executed by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method described in the above embodiments.

[0007] According to another aspect of this application, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to cause the computer to perform the method described in the above embodiments.

[0008] According to another aspect of this application, a computer program product is provided, including a computer program that, when executed by a processor, implements the steps of the method described in the above embodiments.

[0009] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of this application, nor is it intended to limit the scope of this application. Other features of this application will become readily apparent from the following description. Attached Figure Description

[0010] The accompanying drawings are provided for a better understanding of this solution and do not constitute a limitation of this application. Wherein: Figure 1 A flowchart illustrating a resource delivery keyword mining method provided in an embodiment of this application; Figure 2 A flowchart illustrating a resource delivery keyword mining method provided in another embodiment of this application; Figure 3 A flowchart illustrating a resource delivery keyword mining method provided in another embodiment of this application; Figure 4 A flowchart illustrating a resource delivery keyword mining method provided in another embodiment of this application; Figure 5 This is a schematic diagram illustrating a resource placement keyword mining process provided in an embodiment of this application; Figure 6 This is a schematic diagram of an intent determination process provided in an embodiment of this application; Figure 7 A schematic diagram illustrating the evaluation process for keyword placement levels in resource placement, provided as an embodiment of this application; Figure 8 A schematic diagram of the structure of a resource placement word mining device provided in an embodiment of this application; Figure 9 This is a block diagram of an electronic device used to implement the resource placement term mining method of the embodiments of this application. Detailed Implementation

[0011] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of this application, including various details to aid understanding. These should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of this application. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0012] It should be noted that the acquisition, storage, use, and processing of data in this application comply with the relevant provisions of national laws and regulations and do not violate public order and good morals.

[0013] The resource placement word mining method, apparatus, electronic device, and storage medium of this application are described below with reference to the accompanying drawings.

[0014] Figure 1 This is a flowchart illustrating a resource-targeting keyword mining method provided in an embodiment of this application. In this embodiment, "resource" in "resource-targeting keyword" can be advertising or other resources with promotional functions, and is not limited thereto.

[0015] The resource delivery term mining method of this application embodiment can be executed by the resource delivery term mining device of this application embodiment, which can be configured in an electronic device.

[0016] Among them, electronic devices can be any device with computing capabilities, such as personal computers, mobile terminals, servers, etc. Mobile terminals can be hardware devices with various operating systems, touch screens and / or displays, such as in-vehicle devices, mobile phones, tablets, personal digital assistants, wearable devices, etc.

[0017] like Figure 1 As shown, the keyword mining method for resource placement includes: Step 101: Obtain candidate historical search terms and cluster them to obtain a set of candidate search terms.

[0018] In this application, candidate historical search terms may be obtained from the search logs of the search engine or obtained by other means, and this application does not limit this.

[0019] In this application, there can be multiple candidate historical search terms. A clustering algorithm can be used to cluster the candidate historical search terms to obtain one or more candidate search term sets (i.e., clusters). For example, a candidate search term set may include one or more candidate historical search terms.

[0020] For example, the clustering algorithm used may be K-means clustering algorithm, K-medoids clustering algorithm, or other clustering algorithm, and this application does not limit it.

[0021] Step 102: Extract multiple search term samples from the candidate search term set.

[0022] In this application, for each candidate search term set, multiple search term samples can be extracted from the candidate search term set.

[0023] For example, multiple candidate historical search terms can be randomly selected from the candidate search term set, and each candidate historical search term can be used as a search term sample. Alternatively, multiple candidate historical search terms can be selected from the candidate search term set as search term samples according to certain rules. There are no restrictions on this.

[0024] Step 103: Using a large model, perform intent determination on the search term samples to determine the intent determination result of the search term samples.

[0025] In this application, the intent determination result is used to indicate whether the search term sample matches the resource delivery intent of the target product. For example, the target product can be any product promoted through resource channels.

[0026] Taking internet advertising as an example, resource placement intent can be understood as advertising placement intent. The intent determination result can be used to indicate whether the search term sample matches the advertising placement intent of the target product. The target product can be any product promoted through advertising channels.

[0027] For example, the intent determination prompt template can be filled in based on search term samples to generate intent determination prompt information. This intent determination prompt information can then be input into a large model for intent determination, and the intent determination result output by the large model can be obtained.

[0028] For example, the intent determination prompt template may include, but is not limited to, intent determination rules, output requirements, intent determination examples, etc.

[0029] For example, intent determination rules can include that search term samples that match the allowed delivery rules are consistent with the resource delivery intent, otherwise they are not consistent with the resource delivery intent.

[0030] Step 104: Based on the intent determination results of multiple search term samples, filter the candidate search term set to obtain the retained target search term set.

[0031] In this application, the candidate search term set can be filtered based on the intent determination results of multiple search term samples, and the retained candidate search term set can be used as the target search term set.

[0032] For example, for any candidate search term set, the number of search term samples that match the resource allocation intent of the target product can be determined based on the intent determination results of multiple search term samples corresponding to the candidate search term set. Based on this number, it can be determined whether to filter out the candidate search term set, that is, whether to filter out all candidate historical search terms in the candidate search term set in batches.

[0033] Step 105: Determine the target keywords from the target search term set.

[0034] In this application, target keywords corresponding to a target product can be determined from candidate historical search terms in a target search term set. These target keywords can be applied to resources (such as advertisements) used to promote the target product. For example, target keywords can also be described as search terms to be targeted; this application does not limit the name of the target keywords.

[0035] For example, all candidate historical search terms in the target search term set can be identified as target terms.

[0036] For example, it is also possible to determine whether candidate historical search terms in the target search term set match the campaign tone, and then use candidate historical search terms in the target search term set that match the campaign tone as target campaign terms.

[0037] Here, "candidate historical search terms matching the campaign tone" can be understood as the user intent and content preferences represented by the candidate historical search terms matching the product positioning, content ecosystem, and target user group served by the current campaign system.

[0038] The resource placement keyword mining method of this application embodiment can be applied to the mining of advertising placement keywords, in Internet advertising placement and traffic distribution scenarios, or other scenarios involving placement keyword mining, and is not limited thereto.

[0039] Taking internet advertising and traffic distribution scenarios as an example, a large model can be used to determine the intent of search term samples, which can be understood as the determination of advertising intent. Based on the determination of advertising intent, the candidate search term set is filtered to obtain the set of retained target search terms. Candidate historical search terms that match the product tone in the target search term set can be used as target advertising terms for advertising.

[0040] The phrase "candidate historical search terms match the product's tone" can be understood as the candidate historical search terms matching the product's core attributes, target audience, usage scenarios, or positioning.

[0041] In this embodiment, candidate historical search terms are clustered to form a set of candidate search terms with similar semantics. Search term samples are extracted from these samples, and a large model is used to determine the intent of the search term samples to identify whether they match the resource allocation intent of the target product. Based on the highly accurate intent determination results of a small number of samples, the entire set of candidate search terms is efficiently filtered. This not only improves the screening efficiency and accuracy of the candidate search term set, thereby improving the mining efficiency and accuracy of the target terms, but also reduces computational overhead.

[0042] Figure 2 This is a flowchart illustrating a resource delivery term mining method provided in another embodiment of this application.

[0043] like Figure 2 As shown, the keyword mining method for resource placement includes: Step 201: Obtain candidate historical search terms and cluster them to obtain a set of candidate search terms.

[0044] Step 202: Extract multiple search term samples from the candidate search term set.

[0045] In this application, steps 201-202 can be implemented in any of the embodiments of this application, so they will not be described in detail here.

[0046] Step 203: Using a large model, perform intent determination on the search term samples to determine the intent determination result of the search term samples.

[0047] In this application, step 203 can be implemented in any of the embodiments of this application, so it will not be described in detail here.

[0048] In some embodiments, a large model can be used to determine the intent of search term samples according to judgment rules, thereby obtaining the intent determination result of the search term samples.

[0049] For example, the judgment rule can refer to the rule used to determine the intent of the search term sample. The judgment rule can include, but is not limited to, one or more of the first rule, the second rule, the target business positioning, etc.

[0050] The first rule can be a rule that refuses to deliver search term samples. For example, the first rule may include search terms involving harmful information, sensitive topics, or unrelated to the resource domain. For instance, the first rule may also be called a refusal rule or an undeliverable rule, etc., and this application does not limit the name of the first rule.

[0051] The second rule can be a rule that allows the delivery of search term samples. For example, the second rule can include positive matching of search needs that are fully covered or can be solved by the resource based on the resource domain or included capabilities. For example, the second rule can also be called an allowable delivery rule or a deliverable rule, etc., and this application does not limit the name of the second rule.

[0052] For example, a large model can be used to determine the semantic similarity between the search term sample and the first rule. If the semantic similarity is greater than the first threshold, it means that the search term sample is rejected and the intent determination result is determined to be inconsistent with the resource delivery intent of the target product. If the semantic similarity is less than the first threshold, it means that the search term sample is not rejected. In order to improve the accuracy of delivery, the first matching degree between the search term sample and the second rule is further determined to determine whether the search term sample meets the delivery rules and is allowed to be delivered.

[0053] If the first matching degree is greater than the second threshold, it means the search term sample meets the placement rules and is allowed to be placed. The intent determination result is then determined to be consistent with the resource placement intent of the target product. If the first matching degree is less than the second threshold, it means the search term sample does not meet the placement rules, and then the second matching degree between the search term sample and the target business positioning is determined. For example, the target business positioning could be the core business positioning of the target product.

[0054] If the second matching degree is greater than the third threshold, it means that the search term sample matches the target business positioning, and the intent determination result is consistent with the resource allocation intent of the target product. If the second matching degree is less than the third threshold, it means that the search term sample does not match the target business positioning, and the intent determination result is inconsistent with the resource allocation intent of the target product.

[0055] For example, intent determination based on the semantic similarity between the search term sample and the first rule can be described as red line determination; intent determination based on the matching degree between the search term sample and the second rule can be described as green line determination; and intent determination based on the matching degree between the search term sample and the target business positioning can be described as business positioning alignment determination. That is, a large model can be used to perform one-dimensional or multi-dimensional intent determination on the search term sample.

[0056] It should be noted that when the semantic similarity equals the first threshold, it can be determined that the intent judgment result does not conform to the resource allocation intent of the target product. Alternatively, it can continue to make a judgment based on the first matching degree between the search term sample and the second rule. The specific settings can be set according to actual needs and are not limited thereto. When the first matching degree equals the second threshold, it can be determined that the intent judgment result conforms to the resource allocation intent of the target product. Alternatively, it can continue to make a judgment based on the second matching degree between the search term sample and the target business positioning. The specific settings can be set according to actual needs and are not limited thereto. When the second matching degree equals the third threshold, it can be determined that the intent judgment result conforms to the resource allocation intent of the target product, or it can be determined that the intent judgment result does not conform to the resource allocation intent of the target product. The specific settings can be set according to actual needs and are not limited thereto.

[0057] In this embodiment, the semantic similarity between the search term sample and the first rule is first compared with the first threshold to determine whether the search term sample conforms to the first rule. If it does not conform, the search term sample does not conform to the resource allocation intention of the target product. Otherwise, the matching degree between the search term sample and the second rule is used to determine whether the search term sample conforms to the second rule. If the matching degree is greater than the second threshold, the search term sample is determined to conform to the resource allocation intention of the target product. By adopting this cascaded determination logic, not only can the accuracy and robustness of the intent determination be significantly improved, but the risk of misjudgment by a single positive or negative rule can also be avoided, realizing a refined and structured judgment of the resource allocation intent of search terms.

[0058] Furthermore, if the matching degree between the search term sample and the second rule is less than the second threshold, it is possible to further determine whether the search term sample is aligned with the target business positioning based on the matching degree between the search term sample and the target business positioning. If it is aligned with the target business positioning, the search term is determined to meet the resource placement intention; otherwise, the search term is determined not to meet the resource placement intention. Thus, a target business positioning matching mechanism is further introduced on the basis of two-level rule discrimination. This can effectively alleviate the problem of high-potential long-tail keywords or emerging expressions being mistakenly rejected due to rigid thresholds, enhance the tolerance for marginal but highly adaptable search terms, and improve the breadth of search term coverage and adaptability to the business ecosystem while ensuring the accuracy of placement. This results in a more robust and intelligent resource placement intention recognition effect.

[0059] Step 204: Determine the delivery category of the candidate search term set based on the intent determination results of multiple search term samples.

[0060] In this application, the targeting category can be used to indicate whether candidate historical search terms in the candidate search term set are targetingable. For example, the targeting category may include targetingable, non-targetable, etc.

[0061] In some embodiments, for each candidate search term set, the number of search term samples whose intent determination results are consistent with the resource placement intent of the target product can be determined based on the intent determination results of multiple search term samples corresponding to the candidate search term set, and the placement category of the candidate search term set can be determined based on the number of such samples.

[0062] For example, for each candidate search term set, the number of search term samples that match the resource allocation intent of the target product can be determined based on the intent determination results of multiple search term samples. The ratio of the number of search term samples matching the resource allocation intent to the total number of search term samples can then be determined. Based on this ratio, the allocation category for the candidate search term set can be determined. In other words, the proportion of search term samples matching the resource allocation intent of the target product among the extracted search term samples can be determined based on the intent determination results of multiple search term samples. Based on this proportion, the allocation category for the candidate search term set can be determined.

[0063] For example, the delivery category can be determined based on the relationship between the ratio and a set threshold. For example, if the ratio is greater than the set threshold, the delivery category can be determined to be deliverable; if the ratio is less than or equal to the set threshold, the delivery category can be determined to be undeliverable.

[0064] For example, if a threshold of 0.6 is set, and M search term samples are extracted from a set of candidate search terms, if N of the M search term samples match the resource allocation intent of the target product, the allocation category of the set of candidate search terms can be determined based on whether the ratio of N to M is greater than 0.6.

[0065] Therefore, by statistically analyzing the number of samples that match the resource allocation intent of the target product among multiple search term samples extracted from the candidate search term set, and calculating their proportion to the total number of samples, the alignment between the entire set and the allocation target can be quantified through sampling inference. This ratio serves as a set-level intent purity indicator, providing an interpretable and comparable classification basis for the candidate search term set. It avoids judging each search term in the candidate search term set one by one, reducing computational overhead. At the same time, statistical representativeness ensures the reliability of classification decisions, enabling rapid stratification and efficient screening of large-scale search term sets while maintaining accuracy.

[0066] Step 205: In response to the target category being the delivery category, the candidate search term set is determined as the target search term set.

[0067] In this application, the target category can be eligible for delivery, and correspondingly, the non-target category can be non-deliverable.

[0068] For example, if the target category is the target category, it means that the set of candidate search terms is targetable. In this case, the set of candidate search terms can be retained and determined as the target search term set.

[0069] For example, if the target category is not the target category, the candidate search term set can be filtered out, thereby filtering out candidate historical search terms in the candidate search term set in batches.

[0070] Step 206: Determine the target keywords from the target search term set.

[0071] In this application, step 206 can be implemented in any of the embodiments of this application, so it will not be described in detail here.

[0072] In this embodiment, the entire candidate search term set is uniformly classified based on the intent determination results of multiple search term samples. If the overall classification of the set is the target category, the candidate search term set is retained; if the overall classification of the set is a non-target category, the entire set is filtered. This fully utilizes the prior characteristic of semantic homogeneity within the cluster, reducing redundant computation caused by processing each search term individually. It not only achieves efficient removal of low-relevance search term clusters but also ensures the overall quality of the target search term set in terms of semantic consistency and business adaptability, significantly improving the accuracy of keyword mining and system throughput efficiency.

[0073] Figure 3 This is a flowchart illustrating a resource delivery term mining method provided in another embodiment of this application.

[0074] like Figure 3 As shown, the keyword mining method for resource placement includes: Step 301: Using a large model, generate target keywords based on seed keywords.

[0075] In this application, seed keywords can refer to core keywords that serve as the semantic starting point.

[0076] In some embodiments, a large model can be used to semantically expand the seed keywords to obtain candidate keywords, and the candidate keywords can be quality verified. If a candidate keyword passes the quality verification, it is determined as the target keyword.

[0077] For example, quality verification may include one or more of the following: original word deduplication verification, compliance verification, semantic similarity verification, etc.

[0078] For example, original word deduplication verification (or rootless verification) is used to verify whether candidate keywords are compound words containing seed keywords. If they are compound words, those containing seed keywords are removed to ensure incremental value. If compound words are removed or no compound words are found, the original word deduplication verification can be considered successful.

[0079] For example, compliance verification verifies whether candidate keywords meet business requirements. If they meet the business requirements, the candidate keywords pass the compliance verification; otherwise, they fail.

[0080] For example, semantic similarity verification is used to check whether the semantic similarity between candidate keywords and seed keywords is greater than a similarity threshold. If it is greater than the similarity threshold, the semantic similarity verification passes; if it is less than or equal to the similarity threshold, the semantic similarity verification fails.

[0081] For example, the following approach can be used: using a large model, semantic expansion of seed keywords can be performed to obtain candidate keywords. Based on the seed keywords, semantic expansion prompt templates can be filled to generate semantic expansion prompt information. This prompt information can then be input into the large model for processing to generate candidate keywords.

[0082] For example, this semantic expansion prompt is used to prompt a large model to perform a semantic expansion task to generate new keywords.

[0083] For example, semantic expansion prompts can be generated by filling in the semantic expansion prompt template based on keywords, keyword generation requirements, and vertical category names. Keyword generation requirements can describe the requirements for keyword generation, such as "we do not want to obtain keywords through simple word root concatenation, but only through deep semantic synonym or near-synonymous expansion (e.g., generalizing 'report' to 'judgment')"; the vertical category name can be the vertical field to which the keyword to be generated belongs.

[0084] Therefore, by performing deep semantic expansion on seed keywords using a large model to generate semantically related but diverse candidate keywords, the limitations of expansion based on literal matching or rule templates can be effectively overcome, significantly improving the breadth of keyword coverage and linguistic diversity. Furthermore, by introducing quality verification including original word deduplication verification, compliance verification, and semantic similarity verification, not only can low-quality generated results that are duplicated, violate regulations, or deviate semantically be eliminated, but the retained candidate keywords can also be ensured to maintain their core meaning in relation to the seed keywords. Figure 1 It combines business availability and campaign security by identifying candidate keywords that have passed quality verification as target keywords, thus ensuring both semantic richness and accuracy and reliability.

[0085] Step 302: Obtain candidate historical search terms related to the target keyword from the search log.

[0086] In this application, candidate historical search terms related to the target keyword can be found in the search logs of the search engine based on the target keyword.

[0087] Step 303: Cluster the candidate historical search terms to obtain a set of candidate search terms.

[0088] In this application, step 303 can be implemented in any of the embodiments of this application, so it will not be described in detail here.

[0089] Step 304: Extract multiple search term samples from the candidate search term set.

[0090] In this application, step 304 can be implemented in any of the embodiments of this application, so it will not be described in detail here.

[0091] In some embodiments, for each candidate search term set, multiple search term samples can be extracted from the candidate search term set based on the distance between each candidate historical search term in the candidate search term set and the cluster center of the set.

[0092] For example, a first preset number of candidate historical search terms that are closest to the cluster center of the candidate search term set and a second preset number of candidate historical search terms that are furthest apart from the cluster center of the candidate search term set can be extracted from the candidate search term set, and the first preset number of candidate historical search terms and the second preset number of candidate historical search terms can be used as search term samples.

[0093] Among them, the candidate historical search term with the shortest distance from the cluster center can reflect the relevance to the theme of the candidate search term set, while the candidate historical search term with the largest distance from the cluster center can reflect the boundary or diversity of the candidate search term set.

[0094] For example, for each set of candidate search terms, the words closest to and farthest from the cluster center can be extracted. For instance, the 10 search terms closest to the cluster center can be extracted as representatives of the search terms within the cluster to reflect their topic relevance. At the same time, the 10 search terms farthest from the cluster center can be extracted to represent the boundaries or diversity of the cluster.

[0095] It should be noted that the first preset quantity and the second preset quantity can be the same or different, and this application does not limit this.

[0096] Therefore, extracting the candidate historical search terms closest to the cluster center and the candidate historical search terms furthest away as search term samples can not only reflect the theme of the search term set, but also preserve high-value intentions at the margins and enrich the diversity of search term samples.

[0097] Step 305: Using a large model, perform intent determination on the search term samples to determine the intent determination result of the search term samples.

[0098] Step 306: Based on the intent determination results of multiple search term samples, filter the candidate search term set to obtain the retained target search term set.

[0099] Step 307: Determine the target keywords from the target search term set.

[0100] In this application, steps 305-307 can be implemented in any of the embodiments of this application, so they will not be described in detail here.

[0101] In this embodiment, the large model generates semantically rich and expressive target keywords based on seed keywords, which can effectively overcome the limitations of manual setting or rule expansion and cover a wider range of user language habits and potential intentions. On this basis, candidate historical search terms related to these generated terms are traced back from search logs and extracted. This not only combines the generation capability of the large model with real user behavior data, but also ensures that the mined terms are both semantically reasonable and have a practical search basis and traffic potential, which can significantly improve the relevance, practicality and business adaptability of the target term mining.

[0102] Figure 4 This is a flowchart illustrating a resource delivery term mining method provided in another embodiment of this application.

[0103] like Figure 4 As shown, the keyword mining method for resource placement includes: Step 401: Obtain candidate historical search terms and cluster them to obtain a set of candidate search terms.

[0104] Step 402: Extract multiple search term samples from the candidate search term set.

[0105] Step 403: Using a large model, perform intent determination on the search term samples to determine the intent determination result of the search term samples.

[0106] Step 404: Based on the intent determination results of multiple search term samples, filter the candidate search term set to obtain the retained target search term set.

[0107] Step 405: Determine the target keywords from the target search term set.

[0108] In this application, steps 404-405 can be implemented in any of the embodiments of this application, so they will not be described in detail here.

[0109] Step 406: Evaluate the target keywords and obtain the target keyword evaluation results.

[0110] To improve the accuracy of ad targeting, this application allows for a rating assessment of each target keyword to classify them and obtain the rating assessment results. This rating assessment can also be understood as an ad targeting rating assessment.

[0111] For example, the rating results can be used to indicate the targeting level of a keyword. For instance, there could be five targeting levels: Level 1, Level 2, ..., Level 5, with the levels decreasing sequentially.

[0112] For example, the higher the placement level of a keyword, the higher its conversion rate and the higher its value. In advertising scenarios, conversion rate can refer to whether a user registers or purchases a membership through the advertising page.

[0113] It should be noted that the number and level of the deployment levels can be determined according to actual needs, and this application does not impose any restrictions on this.

[0114] For example, in this embodiment, the target keywords for ranking evaluation can be all candidate historical search terms in the target search term set, or a selected portion of candidate historical search terms.

[0115] In some embodiments, a large model can be used to evaluate the target keywords and obtain the evaluation results of the target keywords.

[0116] For example, a large model can be used to identify the intent of the target keywords, obtain the search intent, and then perform a quality assessment on the target keywords based on the search intent to obtain the quality assessment results. Finally, the grade assessment results can be determined based on the quality assessment results.

[0117] For example, quality assessment may include, but is limited to, one or more of the following: matching determination based on historical conversion keyword database, business relevance identification based on external entity-specific keywords, intent granularity classification based on keyword structure parsing, etc.

[0118] For example, matching based on a historical conversion keyword database is used to query the database to determine if there are any conversion keywords that match the target keyword. For example, if a conversion keyword matches the target keyword, the higher the conversion rate of the matched keyword, the higher the placement level of the target keyword.

[0119] For example, business relevance identification based on external entity-specific keywords is used to determine whether a target keyword is an external entity-specific keyword. For example, if a target keyword is an external entity-specific keyword, its placement priority is relatively low.

[0120] For example, intent granularity classification based on keyword structure parsing is used to perform structural parsing of target keywords to determine the granularity of the search intent for the target keywords. For example, the intent granularity of the search intent can be determined based on the semantics of the target keywords.

[0121] For example, the larger the intent granularity, the higher the placement level of the target keywords. For example, intent granularity can include core broad keywords, core secondary keywords, and long-tail keywords, with intent granularity decreasing in size in that order.

[0122] For example, core keywords can refer to general keywords that cover a certain topic area without involving specific sub-categories. These keywords usually represent the broadest user needs, often corresponding to large search volumes but relatively low conversion rates. For example, "ai ppt", "free ppt", and "ppt template" reflect users' basic needs for finding PPT creation tools or templates, but without specific content limitations, and are applicable to all scenarios that require PPT templates.

[0123] For example, secondary core keywords fall between core keywords and long-tail keywords. Secondary core keywords refer to keywords that target specific needs or application scenarios. For instance, the secondary core keyword "third grade, first semester, Chinese language PPT" refers to a specific educational stage and subject content, but multiple PPT files are still needed to fully cover the course requirements, making it more targeted than core keywords.

[0124] For example, long-tail keywords can refer to specific, clearly targeted keywords. These keywords reflect users' personalized needs and usually only require one or a few specific resources to satisfy them. For instance, the long-tail keyword "university physics power definition PPT" has a higher conversion rate because the user's needs are more precise, it has higher content specificity and lower search frequency.

[0125] For example, each of the above quality assessment methods can correspond to a tool. Large models can call the corresponding tool to perform the corresponding quality assessment and obtain the quality assessment results returned by the tool.

[0126] For example, by calling a historical category matching tool, a matching judgment based on the historical conversion keyword library can be performed on the target keywords, and the quality assessment result returned by the tool can be obtained. The quality assessment result includes the matching result. If the matching result shows that there is a conversion keyword that matches the target keywords, the quality assessment result can also include the conversion rate of that conversion keyword.

[0127] For example, by calling a keyword structure parsing tool, you can perform intent granularity classification on the target keywords based on keyword structure parsing and obtain the quality assessment results returned by the tool. These assessment results can include the intent granularity corresponding to the target keywords.

[0128] For example, by calling a dedicated keyword checking tool, business relevance identification based on external entity dedicated keywords can be performed on the target keywords, and the quality assessment results returned by the tool can be obtained. The assessment results can include the query results of whether the target keywords are external entity dedicated keywords.

[0129] For example, the ranking assessment result can be determined based on the quality assessment results in the following way: the initial ranking of the target keywords can be adjusted based on the quality assessment results, and the adjusted ranking is used as the ranking assessment result of the target keywords. For example, different keywords can have the same initial ranking.

[0130] In this embodiment, the intent of the target keywords is identified through a large model, and the quality of the target keywords is evaluated in combination with the search intent. The effectiveness of the search keywords is measured from the perspectives of semantic understanding and business adaptability. The resulting rating evaluation results not only reflect the comprehensive level of the search keywords in terms of user intent clarity, conversion potential and platform ecosystem fit, but also provide a quantifiable and graded basis for subsequent advertising decisions, thereby improving the accuracy of keyword targeting.

[0131] Step 407: Target keywords are delivered based on the rating assessment results.

[0132] In this application, the placement information, such as exposure position and placement order, of the target keywords can be determined based on their placement level in the evaluation results, and the target keywords can be placed accordingly. For example, high-level keywords have better exposure positions, while low-level keywords are placed in poorer exposure positions.

[0133] Taking internet advertising as an example, the target keywords of different advertising levels (i.e., value levels) may have different advertising information such as placement and frequency. For example, the higher the advertising level, the better the placement.

[0134] In this embodiment, after determining the target keywords, a ranking assessment is performed on the target keywords to obtain the ranking assessment results indicating the ranking of the target keywords. Based on this, a differentiated targeting strategy is implemented, so that high-ranking keywords receive higher priority exposure or resource allocation, while low-ranking keywords are allocated lower priority exposure or resource allocation. This achieves a shift from "undifferentiated targeting" to "tiered and precise targeting," which can not only improve the matching accuracy between the targeted content and the user's real needs, but also optimize the utilization efficiency of the platform's traffic resources.

[0135] To facilitate understanding of the resource placement term mining method in this application's embodiments, the following is combined with... Figures 5-7 Please provide an explanation. Figure 5This is a schematic diagram illustrating a resource placement keyword mining process provided in an embodiment of this application. Figure 6 This is a schematic diagram of an intent determination process provided in an embodiment of this application. Figure 7 This is a schematic diagram illustrating the evaluation process for keyword placement levels in resource placement, provided as an embodiment of this application.

[0136] like Figure 5 As shown, the process of keyword mining for resource placement includes three stages: Phase 1: Seed Keyword Generalization Based on the vertical category name, core description, etc., a large model is used to generalize and perform quality verification on seed keywords, outputting an independent expanded vocabulary that includes target keywords as a benchmark for subsequent mining. Quality verification may include original word deduplication verification, compliance verification, and semantic similarity verification, as detailed in the above embodiments.

[0137] Phase Two: Cluster Sampling Filtering Based on the target keywords in the thesaurus, candidate historical search terms are obtained from the search logs and clustered to obtain a candidate search term set. Search term samples are extracted from each candidate search term set, and then a large model is used for sampling arbitration. Specifically, the large model is used to determine intent. Based on the intent determination result, the proportion of search term samples that match the target intent is determined. If the proportion is greater than 60%, the candidate search term set is retained (i.e., the search terms in the set are retained in batches). If the proportion is less than or equal to 60%, the search terms in the candidate search term set are filtered in batches (i.e., the search terms in the set are filtered in batches).

[0138] The specific process of intent determination using a large model can be described as follows: Figure 6 As shown, multiple search term samples are extracted from the candidate search term set, and a large model is used to perform three-dimensional judgment on each individual search term sample. Specifically, the search term sample is first judged by the red line: whether the semantic similarity between the search term sample and the non-targeting rule is greater than or equal to 80%. If so, the search term sample is determined to be inconsistent with the resource targeting intention of the target product. If not, the green line judgment is performed: whether the proportion of entries matching the search term sample with the targetable rule is greater than or equal to 50%. If it is greater than or equal to 50%, the search term sample is determined to be consistent with the resource targeting intention. Otherwise, the positioning alignment judgment is performed: whether the search term sample matches the core business positioning. If so, the search term sample is determined to be consistent with the resource targeting intention; otherwise, the search term sample is determined to be inconsistent with the resource targeting intention.

[0139] Next, based on the intent determination results of multiple search term samples in a candidate search term set, the proportion of search term samples that meet the resource delivery intent in the multiple search term samples in the candidate search term set is calculated, and it is determined whether the proportion is greater than 60%. If it is greater than 60%, all candidate historical search terms in the candidate search term set are retained; otherwise, all candidate historical search terms in the candidate search term set are filtered.

[0140] Phase Three: Level Assessment The candidate historical search terms in the candidate search term set retained in Phase 2 are evaluated for their ranking.

[0141] The specific process of the rating assessment is as follows: Figure 7 As shown, based on a single candidate historical search term in the candidate search term set, combined with a prompt template (such as including role definition + task description), the prompt information is input into the large model for processing. The large model performs inference (such as identifying search intent for candidate historical search terms). Based on the inference result, it determines whether external information is needed. If external information is needed, a tool call instruction is generated, which can call historical classification matching tools, proprietary word checking tools, keyword structure parsing tools, etc., to obtain the tool return results and inject them into the context.

[0142] The large model, combined with the results returned by the tools, further determines whether external information is needed. If it is determined that no external information is needed, the large model performs multiple level assessments on the search terms based on the results returned by the tools, and judges whether the current inference result (i.e., the level assessment result) of the large model is consistent with the historical inference results (the previous level assessment results). If they are consistent, the targeting level is output. If they are inconsistent, the targeting level determined based on the results returned by each tool is weighted according to the priority and confidence of each tool to obtain the final targeting level and output it.

[0143] To achieve the above embodiments, this application also proposes a resource placement word mining device. Figure 8 This is a schematic diagram of the structure of a resource placement word mining device provided in an embodiment of this application.

[0144] like Figure 8 As shown, the resource placement keyword mining device 800 includes: Module 810 is used to obtain candidate historical search terms; Clustering module 820 is used to cluster candidate historical search terms to obtain a set of candidate search terms; The extraction module 830 is used to extract multiple search term samples from the candidate search term set; The intent determination module 840 is used to perform intent determination on search term samples using a large model, and to determine the intent determination result of the search term samples; wherein, the intent determination result is used to indicate whether the search term samples are consistent with the resource allocation intent of the target product; The filtering module 850 is used to filter the candidate search term set based on the intent determination results of multiple search term samples to obtain the retained target search term set. The determination module 860 is used to determine the target keywords from the target search term set.

[0145] Optionally, the filter module 850 is used for: Based on the intent determination results of multiple search term samples, the targeting category of the candidate search term set is determined; whereby the targeting category is used to indicate whether the candidate historical search terms in the candidate search term set can be targeted. In response to the target category being the target category, the set of candidate search terms is determined as the set of target search terms; If the target category is not the target category, filter out the set of candidate search terms.

[0146] Optionally, the filter module 850 is used for: Based on the intent determination results of multiple search term samples, determine the ratio of the number of search term samples that match the resource delivery intent to the number of multiple search term samples; The distribution category is determined based on the ratio.

[0147] Optionally, the intent determination module 840 is used for: A large model is used to determine the semantic similarity between the search term sample and the first rule; wherein, the first rule is the rule that refuses to serve the search term sample. If the semantic similarity is greater than the first threshold, the intent determination result is determined to be inconsistent with the resource delivery intent. In response to a semantic similarity less than a first threshold, a first matching degree is determined between the search term sample and the second rule; wherein, the second rule is a rule that allows the delivery of the search term sample; If the first matching degree is greater than the second threshold, the intent determination result is determined to be consistent with the resource delivery intent.

[0148] Optionally, the intent determination module 840 is used for: In response to the first matching degree being less than the second threshold, a second matching degree between the search term sample and the target business positioning is determined; If the second matching degree is greater than the third threshold, the intent determination result is determined to be consistent with the resource delivery intent. If the second matching degree is less than the third threshold, the intent determination result is determined to be inconsistent with the resource delivery intent.

[0149] Optionally, module 810 is used for: A large model is used to generate target keywords based on seed keywords; Retrieve candidate historical search terms related to the target keyword from the search log.

[0150] Optionally, module 810 is used for: Using a large model, semantic expansion of seed keywords is performed to obtain candidate keywords; Perform quality verification on candidate keywords; the quality verification includes at least one of the following: original word deduplication verification, compliance verification, and semantic similarity verification; In response to the candidate keywords passing quality verification, the candidate keywords are determined as target keywords.

[0151] Optionally, the extraction module 830 is used for: Extract a first preset number of candidate historical search terms that are closest to the cluster centers of the candidate search term set and a second preset number of candidate historical search terms that are furthest from the cluster centers of the candidate search term set. Based on a first preset number of candidate historical search terms and a second preset number of candidate historical search terms, multiple search term samples are determined.

[0152] Optionally, the device may further include: The evaluation module is used to evaluate the target keywords and obtain the evaluation results; the evaluation results are used to indicate the target keyword's placement level. The ad delivery module is used to deliver ad to target keywords based on the rating assessment results.

[0153] Optionally, the evaluation module is used for: A large model is used to identify the intent of the target keywords and obtain the search intent. Based on search intent, the target keywords are evaluated for quality, and the results are obtained. The quality evaluation includes at least one of the following: matching determination based on historical conversion keyword database, business relevance identification based on external entity-specific keywords, and intent granularity classification based on keyword structure parsing. Based on the quality assessment results, the grade assessment results are determined.

[0154] It should be noted that the explanation of the above-mentioned resource placement term mining method embodiment also applies to the resource placement term mining device of this embodiment, so it will not be repeated here.

[0155] In this embodiment, candidate historical search terms are clustered to form a set of candidate search terms with similar semantics. Search term samples are extracted from these samples, and a large model is used to determine the intent of the search term samples to identify whether they match the resource allocation intent of the target product. Based on the highly accurate intent determination results of a small number of samples, the entire set of candidate search terms is efficiently filtered. This not only improves the screening efficiency and accuracy of the candidate search term set, thereby improving the mining efficiency and accuracy of the target terms, but also reduces computational overhead.

[0156] According to embodiments of this application, this application also provides an electronic device, a readable storage medium, and a computer program product.

[0157] Figure 9 A schematic block diagram of an example electronic device 900 that can be used to implement embodiments of this application 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 may also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the application described and / or claimed herein.

[0158] like Figure 9 As shown, device 900 includes a computing unit 901, which can perform various appropriate actions and processes based on a computer program stored in ROM (Read-Only Memory) 902 or a computer program loaded from storage unit 908 into RAM (Random Access Memory) 903. RAM 903 can also store various programs and data required for the operation of device 900. The computing unit 901, ROM 902, and RAM 903 are interconnected via bus 904. I / O (Input / Output) interface 905 is also connected to bus 904.

[0159] Multiple components in device 900 are connected to I / O interface 905, including: input unit 906, such as keyboard, mouse, etc.; output unit 907, such as various types of monitors, speakers, etc.; storage unit 908, such as disk, optical disk, etc.; and communication unit 909, such as network card, modem, wireless transceiver, etc. Communication unit 909 allows device 900 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0160] The computing unit 901 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of the computing unit 901 include, but are not limited to, CPUs (Central Processing Units), GPUs (Graphics Processing Units), various special-purpose AI (Artificial Intelligence) computing chips, various computing units running machine learning model algorithms, DSPs (Digital Signal Processors), and any suitable processor, controller, microcontroller, etc. The computing unit 901 performs the various methods and processes described above, such as the resource-targeted word mining method. For example, in some embodiments, the resource-targeted word mining method can be implemented as a computer software program tangibly contained in a machine-readable medium, such as storage unit 908. In some embodiments, part or all of the computer program can be loaded and / or installed on device 900 via ROM 902 and / or communication unit 909. When the computer program is loaded into RAM 903 and executed by the computing unit 901, one or more steps of the resource-targeted word mining method described above can be performed. Alternatively, in other embodiments, the computing unit 901 may be configured to perform the resource delivery term mining method by any other suitable means (e.g., by means of firmware).

[0161] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, FPGAs (Field Programmable Gate Arrays), ASICs (Application-Specific Integrated Circuits), ASSPs (Application-Specific Standard Products), SOCs (System-on-Chips), CPLDs (Complex Programmable Logic Devices), computer hardware, firmware, software, and / or combinations thereof. These various implementations may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0162] The program code used to implement the methods of this application may be written in any combination of one or more programming languages. This program code may be provided to a processor or controller of a general-purpose computer, special-purpose computer, or other programmable data processing device, such that when executed by the processor or controller, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The program code may be executed entirely on a machine, partially on a machine, as a standalone software package partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0163] In the context of this application, a machine-readable medium can be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, apparatus, or device. A machine-readable medium can be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium can be, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination of the foregoing. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, RAM, ROM, EPROM (Electrically Programmable Read-Only Memory) or flash memory, optical fiber, CD-ROM (Compact Disc Read-Only Memory), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0164] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device for displaying information to the user (e.g., a CRT (Cathode-Ray Tube) or LCD (Liquid Crystal Display) monitor); and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the computer. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0165] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include LANs (Local Area Networks), WANs (Wide Area Networks), the Internet, and blockchain networks.

[0166] Computer systems can include clients and servers. Clients and servers are generally geographically separated and typically interact via communication networks. The client-server relationship is established by computer programs running on the respective computers and having a client-server relationship with each other. A server can be a cloud server, also known as a cloud computing server or cloud host, a hosting product within the cloud computing service ecosystem, addressing the shortcomings of traditional physical hosts and VPS (Virtual Private Server) services, such as high management difficulty and weak business scalability. Servers can also be servers for distributed systems or servers incorporating blockchain technology.

[0167] According to an embodiment of this application, this application also provides a computer program product that, when the instruction processor in the computer program product is executed, performs the resource placement term mining method proposed in the above embodiments of this application.

[0168] It should be understood that the various forms of processes shown above can be used to rearrange, add, or delete steps. For example, the steps described in this application can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution disclosed in this application can be achieved, and this is not limited herein.

[0169] The specific embodiments described above do not constitute a limitation on the scope of protection of this application. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this application should be included within the scope of protection of this application.

Claims

1. A method for mining keywords for resource placement, comprising: Obtain candidate historical search terms and cluster them to obtain a candidate search term set; Multiple search term samples are extracted from the candidate search term set; A large model is used to determine the intent of the search term samples, and the intent determination result of the search term samples is determined; wherein, the intent determination result is used to indicate whether the search term samples are consistent with the resource allocation intent of the target product; Based on the intent determination results of multiple search term samples, the candidate search term set is filtered to obtain the retained target search term set; The target keywords are determined from the set of target search terms.

2. The method as described in claim 1, wherein, The step of filtering the candidate search term set based on the intent determination results of multiple search term samples to obtain a retained target search term set includes: Based on the intent determination results of multiple search term samples, the targeting category of the candidate search term set is determined; wherein, the targeting category is used to indicate whether the candidate historical search terms in the candidate search term set can be targeted; In response to the fact that the delivery category is the target category, the set of candidate search terms is determined as the set of target search terms; In response to the fact that the delivery category is not the target category, the set of candidate search terms is filtered out.

3. The method as described in claim 2, wherein, The step of determining the delivery category of the candidate search term set based on the intent determination results of multiple search term samples includes: Based on the intent determination results of multiple search term samples, determine the ratio of the number of search term samples that match the resource delivery intent to the number of multiple search term samples; The delivery category is determined based on the ratio.

4. The method of claim 1, wherein, The process of using a large model to determine the intent of the search term samples and determining the intent determination result of the search term samples includes: Using the large model, the semantic similarity between the search term sample and the first rule is determined; wherein, the first rule is a rule that refuses to serve the search term sample. In response to the semantic similarity being greater than a first threshold, the intent determination result is determined to be inconsistent with the resource delivery intent; In response to the semantic similarity being less than the first threshold, a first matching degree between the search term sample and the second rule is determined; wherein, the second rule is a rule that allows the delivery of the search term sample; In response to the first matching degree being greater than the second threshold, the intent determination result is determined to be consistent with the resource delivery intent.

5. The method of claim 4, wherein, The method further includes: In response to the first matching degree being less than the second threshold, a second matching degree between the search term sample and the target business location is determined; In response to the second matching degree being greater than the third threshold, the intent determination result is determined to be consistent with the resource delivery intent; In response to the second matching degree being less than the third threshold, the intent determination result is determined to be inconsistent with the resource delivery intent.

6. The method of claim 1, wherein, The process of obtaining candidate historical search terms includes: A large model is used to generate target keywords based on seed keywords; Obtain the candidate historical search terms related to the target keyword from the search log.

7. The method of claim 6, wherein, The method employs a large model to generate target keywords based on seed keywords, including: Using the aforementioned large model, the seed keywords are semantically expanded to obtain candidate keywords; The candidate keywords are subjected to quality verification; wherein the quality verification includes at least one of the following: original word deduplication verification, compliance verification, and semantic similarity verification; In response to the candidate keywords passing quality verification, the candidate keywords are determined as the target keywords.

8. The method of claim 1, wherein, The step of extracting multiple search term samples from the candidate search term set includes: Extract a first preset number of candidate historical search terms that are closest to the cluster centers of the candidate search term set and a second preset number of candidate historical search terms that are furthest from the cluster centers of the candidate search term set; Based on the first preset number of candidate historical search terms and the second preset number of candidate historical search terms, a plurality of search term samples are determined.

9. The method according to any one of claims 1-8, the method further comprising: The target keywords are evaluated for their ranking, and the ranking evaluation result is obtained; wherein the ranking evaluation result is used to indicate the ranking of the target keywords. Based on the evaluation results, the target keywords are then targeted.

10. The method of claim 9, wherein, The step of evaluating the target keywords and obtaining the evaluation results includes: A large model is used to identify the intent of the target keywords and obtain the search intent. Based on the search intent, the target keywords are evaluated for quality, and the quality evaluation results are obtained; wherein, the quality evaluation includes at least one of the following: matching determination based on historical conversion keyword library, business relevance identification based on external entity-specific keywords, and intent granularity classification based on keyword structure parsing; Based on the quality assessment results, the grade assessment results are determined.

11. A resource placement term mining device, comprising: The acquisition module is used to retrieve candidate historical search terms; The clustering module is used to cluster the candidate historical search terms to obtain a set of candidate search terms; An extraction module is used to extract multiple search term samples from the candidate search term set; The intent determination module is used to perform intent determination on the search term samples using a large model, and to determine the intent determination result of the search term samples; wherein, the intent determination result is used to indicate whether the search term samples are consistent with the resource allocation intent of the target product; The filtering module is used to filter the candidate search term set based on the intent determination results of multiple search term samples to obtain the retained target search term set. The determination module is used to determine the target keywords from the target search term set.

12. An electronic device, comprising: At least one processor; as well as A memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, which, when executed by the at least one processor, enables the at least one processor to perform the method of any one of claims 1-10.

13. A non-transitory computer-readable storage medium storing computer instructions, wherein, The computer instructions are used to cause the computer to perform the method according to any one of claims 1-10.

14. A computer program product comprising a computer program that, when executed by a processor, implements the steps of the method according to any one of claims 1-10.

Images