Search word recommendation method and device, computer equipment and storage medium

Abstract

The invention relates to a search term recommendation method and device, computer equipment and a storage medium. The method comprises the steps that a recommendation word list is provided for existing different videos through a large language model, recommendation words output by the large language model are natural and smooth in semantics and accurate and appropriate in expression, and search words are accurately recommended for a video client in combination with the access popularity of all the recommendation words in the recommendation word list. Compared with the existing method for generating the search terms based on short video tags and user search behaviors, the method can solve the problem that the search terms recommended by the existing video software are stiff.

Description

Technical Field

[0001] The present application relates to the field of computer technology, and in particular to a search term recommendation method, apparatus, computer device and storage medium. Background Art

[0002] In order to encourage users to search when watching short videos and improve their experience in video apps, search terms related to watching short videos are usually recommended to users when searching for short videos, so that users can view more content of interest.

[0003] However, the current production logic of short video recommendation search terms is mainly based on short video tags and user search behavior. The search terms produced according to this production method are relatively rigid, that is, the search terms are not natural, smooth, accurate or appropriate enough. Summary of the invention

[0004] The present application provides a search term recommendation method, apparatus, computer device and storage medium to solve the problem that the search terms recommended by existing video software are awkward.

[0005] In a first aspect, the present application provides a search term recommendation method, the method comprising:

[0006] When a search request sent by a video client is obtained, a target video corresponding to the search request is determined, wherein the target video is a video played by the video client before sending the search request;

[0007] Querying a target recommendation word list corresponding to the target video in the cloud storage, wherein each recommendation word in the recommendation word list is an output result of the large language model;

[0008] When the target recommendation word list is found in the cloud storage, each recommendation word in the target recommendation word list is sorted according to the access popularity within a preset time period to obtain a sorted list;

[0009] The recommended words in the sorted list that are before the preset position are output as target search words to the video client for display.

[0010] In a second aspect, the present application provides a search term recommendation device, the device comprising:

[0011] A determination module, configured to determine a target video corresponding to a search request sent by a video client upon obtaining the search request, wherein the target video is a video played by the video client before sending the search request;

[0012] A query module, used for querying a target recommendation word list corresponding to the target video in the cloud storage, wherein each recommendation word in the recommendation word list is an output result of the large language model;

[0013] A sorting module, configured to sort each recommended word in the target recommended word list according to the access popularity within a preset time period to obtain a sorted list when the target recommended word list is found in the cloud storage;

[0014] The recommendation module is used to output the recommendation words located before the preset position in the sorting list as target search words to the video client for display.

[0015] In a third aspect, the present application provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned search term recommendation method when executing the computer program.

[0016] In a fourth aspect, the present application also provides a computer storage medium storing computer executable instructions, wherein the computer executable instructions are used to execute the above-mentioned search term recommendation method.

[0017] The above-mentioned technical scheme provided by the embodiment of the present application has the following advantages over the prior art: the method provided by the embodiment of the present application, when obtaining a search request sent by a video client, determines the target video corresponding to the search request, wherein the target video is the video played by the video client before sending the search request; queries the cloud storage for a target recommendation word list corresponding to the target video, wherein each recommendation word in the recommendation word list is an output result of a large language model; when the target recommendation word list is queried in the cloud storage, sorts each recommendation word in the target recommendation word list according to the access popularity within a preset time period to obtain a sorted list; and outputs the recommendation word located before the preset position in the sorted list as the target search word to the video client for display.

[0018] Based on the above method, a large language model is used to provide a list of recommended words for existing different videos. The recommended words output by the large language model are semantically natural and fluent, and the expressions are accurate and appropriate. The access popularity of each recommended word in the recommended word list is combined to accurately recommend search words to the video client. Compared with the existing search words generated based on short video tags and user search behavior, this can solve the problem of stiff search words recommended by existing video software. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the invention and, together with the description, serve to explain the principles of the invention.

[0020] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0021] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0022] Figure 1 An application environment diagram of a search term recommendation method provided in an embodiment of the present application;

[0023] Figure 2 A flowchart of a search term recommendation method provided in an embodiment of the present application;

[0024] Figure 3 A structural block diagram of a search term recommendation device provided in an embodiment of the present application;

[0025] Figure 4 A schematic diagram of the internal structure of a computer device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0026] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0027] The disclosure below provides many different embodiments or examples to implement different structures of the present invention. In order to simplify the disclosure of the present invention, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present invention. In addition, the present invention can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0028] Figure 1 FIG. 1 is an application environment diagram of a search term recommendation method in an embodiment. Figure 1, the search term recommendation method is applied to a search term recommendation system. The search term recommendation system includes a terminal 110 and a server 120. The terminal 110 and the server 120 are connected via a network. The terminal 110 includes a video client, and the video client supports playing long videos and / or short videos. The terminal 110 can be a desktop terminal or a mobile terminal, and the mobile terminal can be at least one of a mobile phone, a tablet computer, a laptop computer, etc. The server 120 can be implemented as an independent server or a server cluster composed of multiple servers.

[0029] In one embodiment, Figure 2 FIG. 1 is a flow chart of a search term recommendation method in an embodiment, referring to FIG. Figure 2 , provides a search term recommendation method. This embodiment mainly applies this method to the above Figure 1 Taking the server 120 in the example as an example, the search term recommendation method specifically includes the following steps:

[0030] Step S210, when a search request sent by a video client is obtained, a target video corresponding to the search request is determined, wherein the target video is a video played by the video client before sending the search request.

[0031] Specifically, the search request is a request triggered by a user initiating a search behavior when playing a target video on a video client, and is used to request a search behavior on the video client. The search request is most likely used to search for other videos related to the target video, and the target video can be a long video or a short video.

[0032] Step S220, querying the cloud storage for a target recommendation word list corresponding to the target video, wherein each recommendation word in the recommendation word list is an output result of the large language model.

[0033] Specifically, the cloud storage is used to store recommendation word lists corresponding to multiple different videos, that is, to store the correspondence between video identifiers and recommendation word lists. Different videos correspond to different video identifiers (IDs), and each video identifier is used to indicate a unique video, which can be a long video or a short video. The recommendation word list includes at least one recommendation word, which is output by the large language model for the corresponding video.

[0034] Step S230, when the target recommendation word list is found in the cloud storage, each recommendation word in the target recommendation word list is sorted according to the access popularity within a preset time period to obtain a sorted list.

[0035] Specifically, each recommended word in the target recommended word list retrieved from the cloud storage is sorted in sequence according to the access popularity within a preset time period. The preset time period can be a pre-defined time period, or a time period of a preset length before the current moment. In this embodiment, the preset time period is a time period of a preset length before the current moment, that is, the preset time period is the time period closest to the current moment. Based on the access popularity within the most recent period, each recommended word in the target recommended word list is sorted to form a sorted list. The higher the recommended word is ranked in the sorted list, the higher its access popularity.

[0036] Step S240: outputting the recommended words before the preset ranking in the sorting list as target search words to the video client for display.

[0037] Specifically, the preset position is recorded as N, that is, the first N recommended words in the sorted list are output as target search words to the video client for display, so as to provide users with search words related to the target video, and use the large language model to provide a recommendation word list for existing different videos. The recommendation words output by the large language model are semantically natural and fluent, and the expression is accurate and appropriate. The access popularity of each recommended word in the recommendation word list is combined to accurately recommend search words to the video client. Compared with the existing search words generated based on short video tags and user search behavior, this can solve the problem of stiff search words recommended by existing video software, and the recommendation words provided by the large language model for each video can also improve the video coverage of the recommended words.

[0038] In one embodiment, when a search request sent by a video client is obtained, a target video corresponding to the search request is determined, wherein the target video is a video played by the video client before sending the search request. The method further includes:

[0039] Generate feature files corresponding to each sample video according to the key features of each sample video in the database;

[0040] Constructing prompt information corresponding to each of the sample videos according to the feature files corresponding to each of the sample videos;

[0041] Outputting the recommendation words corresponding to each of the sample videos based on the prompt information corresponding to each of the sample videos using the large language model;

[0042] The plurality of recommendation words corresponding to the sample video are combined into a recommendation word list and then saved in the cloud storage.

[0043] Specifically, the sample video is a long video or short video already in the cloud storage. The key features include the title, content description, and tags of the sample video. The key features of the sample video are integrated to generate a corresponding feature file, that is, the feature file includes all the key features of the sample video. For example, the feature file of the sample video includes {'title':'Who is more responsible for the collision between the two cars due to the mutual high beam lights?', 'content description':'The collision between the two cars due to the mutual high beam lights', 'tags': ['Zhejiang', 'turn off the lights', 'car accident', 'information']}. Based on the feature file, the prompt information corresponding to the sample video is constructed. The prompt information is used to input into the large language model. The large language model outputs the recommendation words corresponding to the sample video based on the prompt information. The recommendation words output by the large language model are words that have been optimized for text, thereby avoiding the problem of stiff recommendation words.

[0044] After the recommendation words output by the large language model for each sample video are organized into a recommendation word list, it is saved in the cloud storage so as to provide relevant search words accurately for the user's search behavior while watching the video.

[0045] In one embodiment, the step of assembling a plurality of recommendation words corresponding to the sample video into a recommendation word list and then saving the list to the cloud storage comprises:

[0046] The recommendation words corresponding to the sample video and conforming to a preset data format are combined into a recommendation word list and then saved in the cloud storage.

[0047] Specifically, after the large language model outputs the corresponding recommendation words for the sample video, it is necessary to perform format judgment on the recommendation words, that is, to judge whether the recommendation words conform to the preset data format. The preset data format is the standard phrase format. That is, to judge whether the recommendation words have semantic errors, unnaturalness, or fluency. Only the recommendation words that conform to the preset data format are used to form a recommendation word list and then saved to the cloud storage, that is, unqualified recommendation words are filtered out, thereby improving the recommendation quality of the recommendation words.

[0048] In one embodiment, constructing prompt information corresponding to each sample video according to the feature file corresponding to each sample video includes:

[0049] Generate corresponding output tasks according to the corresponding feature files of each sample video;

[0050] Prompt information corresponding to each of the sample videos is constructed according to the example information and the output tasks corresponding to each of the sample videos.

[0051] Specifically, the output task is a task to be executed related to the sample video. The output task includes the feature file of the sample video and the output requirements. For example, the output task is: "Please give the recommended words based on the following information and output them in the form of an array. {'Title':'Funny video: Hold back your laughter, how many can you hold back? 23 funny moments of human cubs', 'Content description':'Funny moments of human cubs', 'Labels':['Funny']}".

[0052] The sample information includes the sample content of at least one sample video. The sample content includes the sample task, the feature file of the sample video, and the output result. For example, the sample task is: "You are a search term recommendation expert. Please provide high-quality search terms to guide users to search based on the video information I provide." The sample task is used to inform the large language model of the task content that needs to be learned. For example, the sample content includes {'title':'Who is more responsible for the collision between the two cars caused by the high beam lights?', 'content description':'The collision between the two cars caused by the high beam lights', 'labels': ['Zhejiang', 'turn off the lights at high beams', 'car accident', 'information']}, and the output is ['high beam usage specifications', 'safety precautions for driving at night']. The sample content is used to let the large language model learn to provide recommendation words based on video information.

[0053] The example information and the output task corresponding to the sample video are used to construct the prompt information corresponding to the sample video. The prompt information includes the video information of the current sample video and the example content that indicates the learning of the large language model, thereby improving the recommendation accuracy and quality of the recommendation words output by the large language model for the sample video.

[0054] In one embodiment, after searching the cloud storage for a target recommendation word list corresponding to the target video, the method further includes:

[0055] When the target recommendation word list corresponding to the target video is not found in the cloud storage, the target search word is refused to be output to the video client.

[0056] Specifically, if the target recommendation word list corresponding to the target video is not found in the cloud storage, no search word recommendation will be made for the target video, that is, the search word will be refused to be output to the video client. In this scenario, the user is required to enter the search word in the search box by himself, and then a mapping relationship between the video identifier of the target video and the search word can be established in the cloud storage based on the search word entered by the user for the target video. In this way, the mapping relationship between the video identifier and the recommendation word in the cloud storage can be updated, so that the search word manually entered by the user can be used as a recommendation word for other users to search while watching the target video.

[0057] In one embodiment, when the target recommendation word list is found in the cloud storage, each recommendation word in the target recommendation word list is sorted according to the access popularity within a preset time period to obtain a sorted list, including:

[0058] When the target recommendation word list is found in the cloud storage, the access popularity of each recommendation word in the target recommendation word list is determined according to the search volume and click rate within a preset time period;

[0059] The recommended words in the target recommended word list are sorted according to the access popularity to obtain a sorted list.

[0060] Specifically, the access heat of each recommended word in the target recommended word list is determined according to the search volume and click-through rate within a preset time period. The access heat can be the sum of the search volume and click-through rate of the recommended word within the preset time period, the weighted sum of the search volume and click-through rate of the recommended word within the preset time period, or the product of the search volume and click-through rate of the recommended word within the preset time period. The recommended words are sorted according to the access heat to obtain a sorted list. The sorted list can be used to obtain the recommended words that are related to the target video and have high heat within the preset time period. The access heat of the recommended words can reflect the understanding of the target video by most users and the search direction of related information, further improving the recommendation accuracy of the target search words.

[0061] In one embodiment, determining the access popularity of each recommended word in the target recommended word list according to the search volume and click rate within a preset time period includes:

[0062] When the search volume of the current recommended word in the target recommended word list within a preset time period is greater than or equal to the preset search volume, the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word, wherein the current recommended word is any one of the recommended words in the target recommended word list; or,

[0063] When the search volume of the current recommended word in the target recommended word list within the preset time period is less than the preset search volume, the product of the ratio between the search volume of the current recommended word within the preset time period and the preset search volume and the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word.

[0064] Specifically, the preset search volume can be any positive integer, and can be customized according to the actual application scenario. In this embodiment, the preset search volume is set to 100. If the search volume of the current recommended word in the preset time period is greater than or equal to the preset search volume, that is, the search volume is greater than or equal to 100, the click rate of the current recommended word in the preset time period is determined as the access heat, that is, access heat = click rate; if the search volume of the current recommended word in the preset time period is less than the preset search volume, that is, the search volume is less than 100, the access heat at this time is search volume / 100*click rate. In this way, the access heat is determined by combining the search volume and click rate, thereby improving the recommendation accuracy of the recommended word.

[0065] Figure 2 FIG. 1 is a flowchart of a search term recommendation method in one embodiment. It should be understood that although Figure 2 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this document, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Moreover, Figure 2 At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.

[0066] In one embodiment, Figure 3 As shown, a search term recommendation device is provided, comprising:

[0067] The determination module 310 is used to determine the target video corresponding to the search request when the search request sent by the video client is obtained, wherein the target video is the video played by the video client before sending the search request;

[0068] A query module 320, configured to query a target recommendation word list corresponding to the target video in the cloud storage, wherein each recommendation word in the recommendation word list is an output result of the large language model;

[0069] A sorting module 330 is used to sort the target recommendation word list according to the access popularity within a preset time period when the target recommendation word list is found in the cloud storage to obtain a sorted list;

[0070] The recommendation module 340 is used to output the recommendation words that are located before the preset position in the sorting list as target search words to the video client for display.

[0071] In one embodiment, the determining module 310 is further configured to:

[0072] Generate feature files corresponding to each sample video according to the key features of each sample video in the database;

[0073] Constructing prompt information corresponding to each of the sample videos according to the feature files corresponding to each of the sample videos;

[0074] Outputting the recommendation words corresponding to each of the sample videos based on the prompt information corresponding to each of the sample videos using the large language model;

[0075] The plurality of recommendation words corresponding to the sample video are combined into a recommendation word list and then saved in the cloud storage.

[0076] In one embodiment, the determining module 310 is further configured to:

[0077] The recommendation words corresponding to the sample video and conforming to a preset data format are combined into a recommendation word list and then saved in the cloud storage.

[0078] In one embodiment, the determining module 310 is further configured to:

[0079] Generate corresponding output tasks according to the feature files corresponding to each of the sample videos; and construct prompt information corresponding to each of the sample videos according to the example information and the output tasks corresponding to each of the sample videos.

[0080] In one embodiment, the query module 320 is further configured to:

[0081] When the target recommendation word list corresponding to the target video is not found in the cloud storage, the target search word is refused to be output to the video client.

[0082] In one embodiment, the sorting module 330 is further configured to:

[0083] When the target recommendation word list is found in the cloud storage, the access popularity of each recommendation word in the target recommendation word list is determined according to the search volume and click rate within a preset time period;

[0084] The recommended words in the target recommended word list are sorted according to the access popularity to obtain a sorted list.

[0085] In one embodiment, the sorting module 330 is further configured to:

[0086] When the search volume of the current recommended word in the target recommended word list within a preset time period is greater than or equal to the preset search volume, the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word, wherein the current recommended word is any one of the recommended words in the target recommended word list; or,

[0087] When the search volume of the current recommended word in the target recommended word list within the preset time period is less than the preset search volume, the product of the ratio between the search volume of the current recommended word within the preset time period and the preset search volume and the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word.

[0088] like Figure 4 As shown, an embodiment of the present application provides a computer device, including a processor 711, a communication interface 712, a memory 713 and a communication bus 714, wherein the processor 711, the communication interface 712, and the memory 713 communicate with each other through the communication bus 714;

[0089] Memory 713, used for storing computer programs;

[0090] The processor 711 is used to implement the search term recommendation method provided by any one of the aforementioned method embodiments when executing the program stored in the memory 713.

[0091] Those skilled in the art will understand that Figure 4 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0092] In one embodiment, the search term recommendation device provided by the present application can be implemented in the form of a computer program. Figure 4 The memory of the computer device may store various program modules constituting the search term recommendation device, for example, Figure 3 The determination module 310, query module 320, sorting module 330 and recommendation module 340 are shown. The computer program composed of various program modules enables the processor to execute the search word recommendation method of each embodiment of the present application described in this specification.

[0093] Figure 4 The computer device shown can be Figure 3The determination module 310 in the search term recommendation device shown determines the target video corresponding to the search request when the search request sent by the video client is obtained, wherein the target video is the video played by the video client before sending the search request. The computer device can query the cloud storage for a target recommendation word list corresponding to the target video through the query module 320, wherein each recommendation word in the recommendation word list is the output result of the large language model. The computer device can query the cloud storage for the target recommendation word list through the sorting module 330, and sort the each recommendation word in the target recommendation word list according to the access popularity within a preset time period to obtain a sorted list. The computer device can output the recommendation word that is located before the preset position in the sorted list as the target search word to the video client for display through the recommendation module 340.

[0094] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the search term recommendation method provided by any of the aforementioned method embodiments is implemented.

[0095] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0096] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution is essentially or the part that contributes to the relevant technology can be embodied in the form of a software product, and the computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server 120, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0097] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative can be used.

[0098] The foregoing is merely a specific embodiment of the present invention, which enables those skilled in the art to understand or implement the present invention. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features claimed herein.

Claims

1. A search term recommendation method, characterized in that: The method comprises: When a search request sent by a video client is obtained, a target video corresponding to the search request is determined, wherein the target video is a video played by the video client before sending the search request; Querying a target recommendation word list corresponding to the target video in the cloud storage, wherein each recommendation word in the recommendation word list is an output result of the large language model; When the target recommendation word list is found in the cloud storage, each recommendation word in the target recommendation word list is sorted according to the access popularity within a preset time period to obtain a sorted list; The recommended words in the sorted list that are before the preset position are output as target search words to the video client for display.

2. The method according to claim 1, characterized in that When a search request sent by a video client is obtained, a target video corresponding to the search request is determined, wherein the target video is a video played by the video client before sending the search request. The method further includes: Generate feature files corresponding to each sample video according to the key features of each sample video in the database; Constructing prompt information corresponding to each of the sample videos according to the feature files corresponding to each of the sample videos; Outputting the recommendation words corresponding to each of the sample videos based on the prompt information corresponding to each of the sample videos using the large language model; The plurality of recommendation words corresponding to the sample video are combined into a recommendation word list and then saved in the cloud storage.

3. The method according to claim 2, characterized in that The step of forming a recommendation word list from the plurality of recommendation words corresponding to the sample video and then saving the list to the cloud storage comprises: The recommendation words corresponding to the sample video and conforming to a preset data format are combined into a recommendation word list and then saved in the cloud storage.

4. The method according to claim 2, characterized in that: The step of constructing prompt information corresponding to each of the sample videos according to the feature files corresponding to each of the sample videos includes: Generate corresponding output tasks according to the corresponding feature files of each sample video; Prompt information corresponding to each of the sample videos is constructed according to the example information and the output tasks corresponding to each of the sample videos.

5. The method according to claim 1, characterized in that After searching the cloud storage for a target recommendation word list corresponding to the target video, the method further includes: When the target recommendation word list corresponding to the target video is not found in the cloud storage, the target search word is refused to be output to the video client.

6. The method according to claim 1, characterized in that When the target recommendation word list is found in the cloud storage, each recommendation word in the target recommendation word list is sorted according to the access popularity within a preset time period to obtain a sorted list, including: When the target recommendation word list is found in the cloud storage, the access popularity of each recommendation word in the target recommendation word list is determined according to the search volume and click rate within a preset time period; The recommended words in the target recommended word list are sorted according to the access popularity to obtain a sorted list.

7. The method according to claim 6, characterized in that Determining the access popularity of each recommended word in the target recommended word list according to the search volume and click rate within a preset time period includes: When the search volume of the current recommended word in the target recommended word list within a preset time period is greater than or equal to the preset search volume, the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word, wherein the current recommended word is any one of the recommended words in the target recommended word list; or, When the search volume of the current recommended word in the target recommended word list within the preset time period is less than the preset search volume, the product of the ratio between the search volume of the current recommended word within the preset time period and the preset search volume and the click rate of the current recommended word within the preset time period is determined as the access popularity of the current recommended word.

8. A search term recommendation device, characterized in that: The device comprises: A determination module, configured to determine a target video corresponding to a search request sent by a video client upon obtaining the search request, wherein the target video is a video played by the video client before sending the search request; A query module, used for querying a target recommendation word list corresponding to the target video in the cloud storage, wherein each recommendation word in the recommendation word list is an output result of the large language model; A sorting module, configured to sort each recommended word in the target recommended word list according to the access popularity within a preset time period to obtain a sorted list when the target recommended word list is found in the cloud storage; The recommendation module is used to output the recommendation words located before the preset position in the sorting list as target search words to the video client for display.

9. A computer device comprising a memory, a processor and a computer program stored in the memory and executable on the processor, characterized in that: When the processor executes the computer program, the method according to any one of claims 1 to 7 is implemented.

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

Images