Object recommendation method and device, electronic equipment and storage medium
By extracting search phrases and object introduction from the recommendation text to sort the object library, the time-consuming recommendation problem caused by the large number of objects is solved, and fast and efficient object recommendation is achieved, improving the user experience.
Patent Information
- Application Number
- CN202410858666.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-28
- Publication Date
- 2024-11-08
AI Technical Summary
In the prior art, due to the large number of objects, the recommended object takes a long time and is inefficient, which affects the user experience.
By extracting search phrases from the recommended text and initially sorting and filtering the object library based on the search phrases and object introduction, we obtain an object sequence that meets the recommended text, and then further determine the object collection of objects to be recommended based on the object introduction, so as to achieve fast and efficient Recommended objects.
It realizes the rapid identification of objects to be recommended, saves computing resources, improves user experience, and improves the efficiency of object recommendations.
Smart Images

Figure CN118916539A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of computer technology, and in particular to an object recommendation method, device, electronic device and storage medium. Background Art
[0002] With the development of video software, in order to improve user experience and enable users to quickly and easily determine the TV series they want to watch, general video software has set up a TV series list. The list generally corresponds to a type of TV series. By screening films that match the type and adding the film to the list, users can quickly and easily find films of interest.
[0003] In the prior art, when creating a film and television drama list, it is necessary to find a film that meets the film and television drama list from an existing film library and recommend the film to be added to the film and television drama list.
[0004] However, in actual applications, the number of movies in a movie library is often in the order of thousands to tens of thousands. The method of determining the movies to be recommended one by one will take days to complete, and it is impossible to screen out suitable movies in time, which results in a long time spent on screening movies that meet the movie list. In this regard, if the movies in the movie library are regarded as an object, then when recommending objects to be added to the movie list from the movie library, due to the large number of objects, it will take a long time to recommend objects, which is inefficient and affects the user experience. Summary of the invention
[0005] The present application provides an object recommendation method, device, electronic device and storage medium to solve the technical problem in the prior art that due to the large number of objects, object recommendation takes a long time and is inefficient, which affects the user experience.
[0006] In a first aspect, the present application provides an object recommendation method, the method comprising:
[0007] Obtaining a preset recommendation text, and extracting a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended;
[0008] According to the search phrase, multiple objects in a preset object library are sorted to obtain an object sequence;
[0009] Obtaining an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object;
[0010] According to the object introduction and the recommendation text of each object, an object set of objects to be recommended is determined from the object sequence to recommend objects in the object set.
[0011] As a possible implementation, the extracting the search phrase from the recommended text includes:
[0012] Writing the recommended text into a preset first prompt template to obtain a first prompt text;
[0013] The first prompt text is input into a preset first language model to obtain a search phrase extracted from the recommended text.
[0014] As a possible implementation, the step of sorting a plurality of objects in a preset object library according to the search phrase to obtain an object sequence includes:
[0015] For each object in the object library, input the object identifier of the object and the search phrase into a preset search engine, and obtain the first N search results obtained by the search engine;
[0016] Determine the similarity between the object and the recommended text according to the first N search results;
[0017] The multiple objects in the object library are arranged based on the similarity to obtain an object sequence.
[0018] As a possible implementation, determining the similarity between the object and the recommended text according to the first N search results includes:
[0019] Deleting search results that do not contain the object identifier in the search results, and using the remaining search results as target search results;
[0020] Encoding the recommended text into a recommended text vector of a preset dimension, and encoding each of the target search results into a target search result vector of a preset dimension;
[0021] For each of the target search results, determining the cosine similarity between the target search result vector corresponding to the target search result and the recommended text vector;
[0022] The maximum cosine similarity is used as the similarity between the object and the recommended text.
[0023] As a possible implementation manner, the arranging the multiple objects in the object library based on the similarity to obtain an object sequence includes:
[0024] Determine the publishing time and attention level of each object, where the attention level is used to represent the degree to which the object is paid attention to by multiple users;
[0025] Arranging the objects in the object library in descending order of the similarity to obtain an initial object sequence;
[0026] According to the release time and the attention degree, the order of objects in the initial object sequence is adjusted to obtain the object sequence.
[0027] As a possible implementation, adjusting the order of objects in the initial object sequence according to the publishing time and the popularity to obtain the object sequence includes:
[0028] Determining a maximum similarity and a minimum similarity in the initial object sequence;
[0029] Calculating a similarity difference between the maximum similarity and the minimum similarity, and dividing the similarity difference by a preset number to obtain an interval length of each similarity interval in the preset number of similarity intervals;
[0030] Dividing the similarity in the initial object sequence into intervals according to the interval lengths to obtain the preset number of similarity intervals;
[0031] According to the similarity of each object in the initial object sequence, the objects in the initial object sequence are divided according to the similarity interval to obtain a preset number of similarity interval sequences;
[0032] For each of the similarity interval sequences, a first score is given to the objects in the similarity interval according to the release time of each object in the similarity interval sequence to obtain a first score; and a second score is given to the objects in the similarity interval according to the attention of each object in the similarity interval sequence to obtain a second score;
[0033] Performing a preset calculation on the first score and the second score of each object to obtain a target score for each object;
[0034] According to the target score, the objects in the similarity interval sequence are sorted to obtain an object sequence.
[0035] As a possible implementation, determining an object set of objects to be recommended from the object sequence according to the object introduction and the recommendation text of each object includes:
[0036] For each of the objects in the object sequence, writing the object introduction and the recommended text of the object into a preset second prompt template to obtain a second prompt text;
[0037] Inputting the second prompt text into a preset second language model, so that the second language model determines whether the content of the object profile description matches the content of the recommendation text description according to the second prompt text, and outputs a matching result;
[0038] In the case where the matching result indicates that the content of the object profile description of the object matches the content of the recommendation text description, the object is determined as an object to be recommended;
[0039] The object to be recommended is classified into a preset object set of objects to be recommended.
[0040] As a possible implementation manner, the recommending the object in the object set includes:
[0041] Obtaining a detailed description text of each object in the object set, wherein the detailed description text is a text describing the content of the object, and the number of words in the detailed description text is more than that of the object introduction;
[0042] Encoding the detailed description text into a description text vector of a preset dimension, and encoding the recommended text into a recommended text vector of a preset dimension;
[0043] Determining the text cosine similarity between the description text vector and the recommendation text vector;
[0044] An object whose text cosine similarity is greater than a preset similarity threshold is determined as a target object, and the target object is recommended.
[0045] In a second aspect, the present application provides an object recommendation device, the device comprising:
[0046] A first acquisition module is used to acquire a preset recommendation text and extract a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended;
[0047] A sorting module, used to sort multiple objects in a preset object library according to the search phrase to obtain an object sequence;
[0048] A second acquisition module is used to acquire an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object;
[0049] The recommendation module is used to determine an object set of objects to be recommended from the object sequence according to the object introduction and the recommendation text of each object, so as to recommend objects in the object set.
[0050] In a third aspect, the present application provides an electronic device, comprising: a processor and a memory, wherein the processor is used to execute an object recommendation program stored in the memory to implement the object recommendation method described in any one of the first aspects.
[0051] In a fourth aspect, the present application provides a storage medium storing one or more programs, wherein the one or more programs can be executed by one or more processors to implement the object recommendation method described in any one of the first aspects.
[0052] The technical solution provided by the embodiment of the present application obtains a preset recommendation text, the above recommendation text is used to describe the object characteristics of the object to be recommended, extracts a search phrase from the recommendation text, the above search phrase is used as a keyword to search for the object to be recommended, sorts multiple objects in the preset object library according to the above search phrase, obtains an object sequence, obtains an object profile of each object in the above object sequence, the above object profile is a text describing the object content, and determines the object set of the object to be recommended from the object sequence according to the object profile of each object and the above recommendation text, so as to recommend the object in the object set. This technical solution extracts the search phrase from the recommendation text, and based on this, preliminarily sorts and screens the objects in the preset object library from the two dimensions of the search phrase and the object profile, obtains an object sequence that conforms to the recommendation text, and then further determines the object set of the object to be recommended according to the object profile, so that the object to be recommended can be quickly determined, and fast and efficient object recommendation is achieved, saving computing resources and improving user experience. BRIEF DESCRIPTION OF THE DRAWINGS
[0053] 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.
[0054] 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.
[0055] 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.
[0056] Figure 1 A flowchart of an embodiment of an object recommendation method provided in an embodiment of the present application;
[0057] Figure 2 A flowchart of another method for recommending an object provided in an embodiment of the present application;
[0058] Figure 3 A flowchart of another method for recommending an object provided in an embodiment of the present application;
[0059] Figure 4 A flowchart of another method for recommending an object provided in an embodiment of the present application;
[0060] Figure 5 A schematic diagram of a system flow of an object recommendation system provided in an embodiment of the present application;
[0061] Figure 6 A block diagram of an embodiment of an object recommendation device provided in an embodiment of the present application;
[0062] Figure 7 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION
[0063] 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.
[0064] 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.
[0065] In order to solve the technical problem in the prior art that due to the large number of objects, object recommendation is time-consuming and inefficient, which affects the user experience, the present application provides an object recommendation method, device, electronic device and storage medium, which can extract search phrases from the recommendation text, and based on this, preliminarily sort and screen the objects in the preset object library from the two dimensions of search phrases and object profiles to obtain an object sequence that conforms to the recommendation text, and then further determine the object set of the object to be recommended according to the object profile, so as to quickly determine the object to be recommended, realize fast and efficient determination of object recommendation, save computing resources, and improve user experience.
[0066] The object recommendation method provided by the present application is further explained below with reference to specific embodiments in conjunction with the accompanying drawings. The embodiments do not constitute a limitation on the embodiments of the present invention.
[0067] join Figure 1, is a flow chart of an embodiment of an object recommendation method provided in an embodiment of the present application. Figure 1 As shown, the process may include the following steps:
[0068] Step 101: Obtain preset recommendation text, where the recommendation text is used to describe the object features of the object to be recommended.
[0069] The above-mentioned recommendation text refers to a preset text used to describe the object characteristics of the object to be recommended. The object to be recommended here may include an object to be recommended but not yet determined, and may also include an object that has been determined to be recommended.
[0070] Furthermore, the recommendation text may be used as a requirement text for determining the object to be recommended, that is, an object whose object features are consistent with the object features described in the recommendation text is selected as the recommended object according to the recommendation text.
[0071] For example, when recommending movies to a playlist in a video player software, each movie in the movie library can be used as an object, and the object to be recommended is selected from multiple objects in the movie library to recommend the object to the playlist. Based on this, since the object is recommended to the playlist, the object features of the objects stored in the playlist can be obtained first. Furthermore, each playlist can have at least one description text for describing the object features. Therefore, the execution subject of the embodiment of the present application can obtain the description text of the playlist and use it as the recommendation text for describing the object features of the object to be recommended.
[0072] In the embodiment of the present application, when determining to recommend an object, the execution subject of the embodiment of the present application may first obtain a preset recommendation text, so as to subsequently determine the object to be recommended based on the recommendation text.
[0073] As an exemplary implementation, the user can create a set of objects to be recommended through the visual interface, and set corresponding recommendation text for the set of objects to be recommended. It can be understood that the set of objects to be recommended at this time is a newly created empty set.
[0074] Based on this, the execution subject of the embodiment of the present application can obtain the recommended text set by the above user through the above visualization interface.
[0075] As another exemplary implementation, the execution subject of the embodiment of the present application may have a database to be recommended, and the database to be recommended includes multiple preset sets of objects to be recommended, and each set of objects to be recommended may have at least one recommendation text.
[0076] Based on this, when the execution subject of the embodiment of the present application recommends objects for any set of objects to be recommended in the database to be recommended, the recommendation text corresponding to the set of objects to be recommended can be directly obtained from the database.
[0077] Step 102: extract a search phrase from the recommendation text. The search phrase is used as a keyword to search for the recommended object.
[0078] Step 103: sort multiple objects in the preset object library according to the search phrase to obtain an object sequence.
[0079] The following is a unified description of step 102 and step 103:
[0080] The above search phrase is used as a keyword to search for the recommended object. For example, the search phrase is input into a preset search engine, and the object corresponding to the feature described by the above search phrase can be searched from the search engine.
[0081] The above-mentioned preset object library refers to a database that stores multiple objects. The objects stored in the database can be objects of the same type or objects of different types, and the embodiment of the present application does not impose any restrictions on this.
[0082] The above-mentioned object may be a TV series, a movie, a novel, or news, etc., and the embodiments of the present application do not impose any limitation on this.
[0083] In the prior art, if you want to obtain objects that match the recommended text from the above-mentioned preset object library, you generally use a large language model to determine whether the content described in the detailed introduction text of each object matches the content described in the recommended text, thereby determining the object that matches the recommended text. For example, when recommending movies to a playlist in a video player software, and taking the movies in the movie library as objects to determine the objects recommended to the playlist, in the prior art, you can determine whether the content described in the plot introduction text of each movie matches the content described in the recommended text, thereby determining whether the movie is recommended to the playlist.
[0084] However, in actual applications, the above-mentioned preset object library generally stores a large number of objects (such as all the movies in the video playback software), and the number is in the thousands to tens of thousands. If a large language model is relied upon to match and analyze the detailed introduction text and the recommended text of each object in the preset object library, this is undoubtedly a huge workload and time-consuming, resulting in the inability to timely determine the objects that meet the recommended text from the preset object library, affecting the user experience.
[0085] In this regard, the execution subject of the embodiment of the present application may first pre-sort the objects in the above-mentioned preset object library according to the above-mentioned recommendation text to obtain an object sequence. Based on this, when recommending objects from the preset database according to the recommendation text, the objects to be recommended may be directly selected from the above-mentioned object sequence. Since the object sequence is a sequence pre-sorted according to the recommendation text, when determining the objects that meet the object characteristics described in the recommendation text for the object sequence, it is faster, more accurate and more efficient than determining the objects that meet the object characteristics described in the recommendation text for the unsorted objects in the preset object library.
[0086] As an exemplary implementation, the execution subject of the embodiment of the present application may use a preset first language model to extract a search phrase from the recommended text, and use the search phrase to sort multiple objects in a preset object library to obtain an object sequence.
[0087] As for how to use the preset first language model to extract the search phrase from the recommended text, and how to sort multiple objects in the preset object library according to the search phrase to obtain the object sequence, it can be respectively described below. Figure 2 and Figure 3 The process shown is explained and will not be described in detail here.
[0088] Step 104: Obtain an object introduction of each object in the object sequence. The object introduction is a text describing the content of the object.
[0089] Step 105: According to the object profile of each object and the above-mentioned recommendation text, determine an object set of objects to be recommended from the object sequence to recommend objects in the object set.
[0090] The following is a unified description of step 104 and step 105:
[0091] The object introduction is a text describing the object content. Further, the number of words in the object introduction may be less than a preset value.
[0092] The above-mentioned object set refers to a set of determined objects to be recommended.
[0093] In an embodiment of the present application, in order to further improve the efficiency of object recommendation and quickly determine the objects to be recommended, the execution subject of the embodiment of the present application may again filter the object sequence according to the object profile of each object in the object sequence to obtain an object set.
[0094] As an optional implementation method, each object in the preset object library may have at least one object profile. Based on this, the execution subject of the embodiment of the present application may directly obtain the object profile of each object from the preset object library.
[0095] As an optional implementation method, when the execution subject of the embodiment of the present application determines the object set of objects to be recommended from the object sequence, it can use a preset large language model (for the sake of ease of description, referred to as the second large language model below) to judge the object profile and recommended text of each object in the above object sequence, and determine whether the object features described in the object profile of each object match the object features described in the recommended text, so as to determine the objects to be recommended whose object profiles match the recommended text from the above object sequence, and classify the objects to be recommended into the preset object set.
[0096] As for how to use the preset second largest language model to determine the object set in the above object sequence that matches the object features described in the object introduction with the object features described in the recommendation text, it can be seen in the following text. Figure 4 The process shown is explained and will not be described in detail here.
[0097] It can be seen from the above-described steps that the above-mentioned object set is a set of objects to be recommended after sorting and filtering the objects included in the preset object library. Compared with the objects in the unsorted and unfiltered preset object library, it can more quickly determine the objects to be recommended that match the recommended text.
[0098] Based on this, the execution subject of the embodiment of the present application may recommend objects in the above-mentioned object set when recommending objects.
[0099] As an exemplary implementation, the execution subject of the embodiment of the present application may obtain the detailed description text of each object in the above-mentioned object set, and recommend the objects in the above-mentioned object set according to the detailed description text of each object. The detailed description text refers to the text describing the content of the object, and the number of words in the detailed description text is more than the object introduction, that is, the detailed description text of the object describes the above-mentioned object in more detail. For example, the object in the object set may be a film or TV series, and the detailed description text is the plot description of the film or TV series.
[0100] As an optional implementation method, the execution subject of the embodiment of the present application may, for each object, encode the detailed description text of the object into a description text vector of a preset dimension, and encode the above-mentioned recommendation text into a recommendation text vector of a preset dimension.
[0101] Afterwards, the text cosine similarity between the description text vector and the recommendation text vector may be determined, and an object whose text cosine similarity is greater than a preset similarity threshold may be determined as a target object, and the target object may be recommended.
[0102] As another optional implementation method, the execution entity of the embodiment of the present application may use a pre-trained large language model (hereinafter referred to as the third large language model for ease of description) to determine whether the object features described in the detailed description text of each object match the object features described in the recommendation text, and recommend the matching objects.
[0103] As an exemplary implementation, the execution subject of the embodiment of the present application may, for each object in the object set, input the detailed description text and recommended text of the object into a preset prompt template (hereinafter referred to as the third prompt template for ease of description) to generate a corresponding prompt text (hereinafter referred to as the third prompt text for ease of description). For example, the third prompt template may be "You are a senior film and television drama operator, responsible for screening film and television series for users based on the user's vague description (or evaluation) of the plot. You are given a plot description of a film and television drama, and the film and television drama will unfold based on the plot description. The criteria for the screening of this film and television drama are: based on the common sense of plot development, the plot description or the plot that meets the user's evaluation. Determine whether the following plot description (here used to write the detailed text of each object) can pass the screening of the fuzzy description (here used to write the recommended text).".
[0104] Afterwards, the third prompt text may be input into a preset third language model, which may determine whether the object features described in the detailed description text of the object match the object features described in the recommendation text based on the third prompt text, and output a matching result.
[0105] Afterwards, the execution subject of the embodiment of the present application may obtain the above matching result, characterize the above matching result as an object whose object features described in the detailed description text match the object features described in the recommendation text, and recommend the object.
[0106] The technical solution provided by the embodiment of the present application obtains a preset recommendation text, the above recommendation text is used to describe the object characteristics of the object to be recommended, extracts a search phrase from the recommendation text, the above search phrase is used as a keyword to search for the object to be recommended, sorts multiple objects in the preset object library according to the above search phrase, obtains an object sequence, obtains an object profile of each object in the above object sequence, the above object profile is a text describing the object content, and determines the object set of the object to be recommended from the object sequence according to the object profile of each object and the above recommendation text, so as to recommend the object in the object set. This technical solution extracts the search phrase from the recommendation text, and based on this, preliminarily sorts and screens the objects in the preset object library from the two dimensions of the search phrase and the object profile, obtains an object sequence that conforms to the recommendation text, and then further determines the object set of the object to be recommended according to the object profile, so that the object to be recommended can be quickly determined, and fast and efficient object recommendation is achieved, saving computing resources and improving user experience.
[0107] join Figure 2 , which is a flow chart of an embodiment of another object recommendation method provided in an embodiment of the present application. Figure 2 The process shown in Figure 1 Based on the process shown in FIG. 1 , it describes how to use the preset first language model to extract search phrases from the recommended text. Figure 2 As shown, the process may include the following steps:
[0108] Step 201: Write the recommended text into a preset first prompt template to obtain a first prompt text.
[0109] Step 202: Input the first prompt text into a preset first language model to obtain a search phrase extracted from the recommended text.
[0110] The following is a unified description of step 201 and step 202:
[0111] The first prompt text is a prompt text used to prompt extraction of a short phrase text with fewer words from a text.
[0112] Correspondingly, the above-mentioned first prompt template is a template used to generate the above-mentioned first prompt text.
[0113] In one embodiment, the user may pre-set a first prompt template. After the executing entity of the embodiment of the present application obtains the recommended text, in order to extract the search phrase from the recommended text, the executing entity of the embodiment of the present application may write the recommended text into the first prompt template to obtain the corresponding first prompt text.
[0114] Afterwards, the first prompt text can be input into a pre-trained first large language model. After obtaining the first prompt text, the large language model can extract a search phrase that describes the object features consistent with those described in the recommended text from the recommended text according to the first prompt text, and output the search phrase. Based on this, the execution subject of the embodiment of the present application can obtain the search phrase output by the first large language model.
[0115] For example, assuming that the first prompt template mentioned above is "Please give me a search phrase for searching the TV drama and film plot that is most relevant to the description below in the browser - recommended text", and continue to assume that the recommended text above is "The protagonist starts as a new recruit and grows into an elite soldier after a series of hardships." Then write the recommended text into the corresponding position of the recommended text in the first prompt template above, and you can get the following first prompt text "Please give me a search phrase for searching the TV drama and film plot that is most relevant to the description below in the browser - the protagonist starts as a new recruit and grows into an elite soldier after a series of hardships."
[0116] Afterwards, the first prompt text is input into the first language model, and the search phrase "new soldiers grow into elite soldiers" output by the first language model can be obtained.
[0117] The technical solution provided by the embodiment of the present application writes the recommended text into a preset first prompt template to obtain a first prompt text, and inputs the first prompt text into a preset first large language model to obtain a search phrase extracted from the recommended text. This technical solution generates a corresponding first prompt text from the recommended text by designing a first prompt template, and generates a search phrase using a large language model, which can ensure that the search phrase is extracted from the recommended text more accurately.
[0118] join Figure 3 , which is a flow chart of an embodiment of another object recommendation method provided in an embodiment of the present application. Figure 3 The process shown in Figure 1 Based on the process shown, it is described how to sort multiple objects in the preset object library according to the search phrase to obtain the object sequence.
[0119] like Figure 3 As shown, the process may include the following steps:
[0120] Step 301: For each object in the object library, the object identifier and the search phrase of the object are input into a preset search engine to obtain the first N search results obtained by the search engine.
[0121] Step 302: Determine the similarity between the object and the recommended text based on the first N search results.
[0122] The following is a unified description of step 301 and step 302:
[0123] The above-mentioned object identifier refers to an identifier used to characterize an object. The identifier can be an object name or an object ID, and the embodiment of the present application does not impose any limitation on this.
[0124] The search results refer to web page data related to the input object and the search phrase retrieved by the search engine. The web page data may include web page images and text. Furthermore, in order to improve efficiency, the web page data may include text.
[0125] In an embodiment of the present application, in order to sort multiple objects in a preset object library, the execution subject of the embodiment of the present application may sort the objects in the above object library according to the similarity between each object and the recommended text.
[0126] As an optional implementation, the execution subject of the embodiment of the present application can input the object identifier and search phrase of each object in the object library into a preset search engine, and obtain the first N search results obtained by the search engine, where N is a positive integer. For example, if the object is a movie and the object identifier is the movie title, the movie title and search phrase can be input into the search engine in the form of <movie title>+<search phrase>.
[0127] Afterwards, the similarity between the object and the recommended text may be determined based on the top N search results.
[0128] As an exemplary implementation, the similarity between the recommended text and each search result can be determined for each object. Since the search results are obtained based on the object identifier and the search phrase, the similarity between the recommended text and each search result can be used as the similarity between the object and the recommended text.
[0129] Based on this, the execution entity of the embodiment of the present application may determine the maximum similarity value as the similarity between the object and the recommended text.
[0130] Specifically, first, in order to reduce the negative impact of irrelevant content on ranking, the execution subject of the embodiment of the present application can delete the search results without object identifiers in the first N search results, and use the remaining search results as target search results.
[0131] Afterwards, the recommended text can be encoded into a recommended text vector of a preset dimension, and each target search result can be encoded into a target search result vector of a preset dimension. Afterwards, for each target search result, the cosine similarity between the target search result vector corresponding to the target search result and the above-mentioned recommended text vector can be determined. So far, multiple cosine similarities corresponding to each object can be obtained. Based on this, the execution body of the embodiment of the present application can use the maximum cosine similarity as the similarity between the object and the recommended text. Among them, the above-mentioned recommended text vector and target search result vector can both be 1024-dimensional embedding vectors obtained by the embedding model.
[0132] It should be noted that if the search result of a certain object searched by the search engine is empty, then it can be determined that the similarity between the object and the recommended text is 0.
[0133] Step 303: Arrange the multiple objects in the object library based on the similarity to obtain an object sequence.
[0134] In one embodiment, the execution subject of the embodiment of the present application may arrange multiple objects in the object library according to the similarity between each object in the object library and the recommended text, thereby obtaining an object sequence.
[0135] Optionally, since it is necessary to select the recommended object that best matches the recommended text for recommendation, the above multiple objects can be arranged in order of similarity from large to small to obtain an object sequence. According to this rule, when determining the recommended object that matches the recommended text, in order to quickly determine the recommended object, the execution subject of the embodiment of the present application can be determined from the first N objects in the above object sequence.
[0136] Furthermore, when the above-mentioned objects are objects such as movies or news, considering that the latest released and highly popular objects (such as movies) are more likely to attract users to watch, the execution entity of the embodiment of the present application can take into account the release time and popularity information of the objects while re-sorting the screening order of the above-mentioned object library, and perform stage sorting according to the release time and similarity.
[0137] As an exemplary implementation, the publishing time and attention of each object may be determined, and the objects in the object library may be arranged in descending order of similarity to obtain an initial object sequence. Thereafter, the order of the objects in the initial object sequence may be adjusted according to the publishing time and attention to obtain an object sequence.
[0138] Specifically, the maximum similarity and the minimum similarity in the initial object sequence may be determined, and the initial object sequence may be divided into a plurality of similarity interval sequences according to the maximum similarity and the minimum similarity.
[0139] As an optional implementation, the maximum similarity and the minimum similarity may be subtracted to obtain a similarity difference, and the similarity difference may be divided by a preset number of similarity interval sequences to obtain an interval length of each similarity interval in the preset number of similarity intervals. Afterwards, the similarities in the initial object sequence may be divided into intervals according to the interval lengths to obtain a preset number of similarity intervals.
[0140] Afterwards, according to the similarity of each object in the initial object sequence, the objects in the initial object sequence may be divided according to the similarity intervals to obtain a preset number of similarity interval sequences.
[0141] Finally, for each of the similarity interval sequences, the objects in the similarity interval sequence may be sorted according to the publishing time and attention level of each object in the similarity interval sequence, thereby obtaining an object sequence.
[0142] As an exemplary implementation, when sorting the objects in the similarity interval, for each similarity interval sequence, the objects in the similarity interval sequence may be arranged in order of release time from front to back to obtain a first object sequence.
[0143] Afterwards, the first object sequence may be divided into multiple first object interval sequences according to the release time of each object in the first object sequence, for example, divided into multiple first object interval sequences according to a time interval of 3 years.
[0144] Finally, for each first object interval sequence, the objects in the first object interval sequence can be reordered in descending order of the attention of each object in the first object interval sequence. According to the above method, an object sequence can be obtained after the objects in the object library are arranged according to similarity, release time, and attention.
[0145] For example, if the above object is a movie, and the release time of the object is the release time of the movie, then according to the above method, the object sequence shown in Table 1 below can be obtained:
[0146] Table 1
[0147] Similarity Showtime heat Movie Title 1-0.95 2023-2020 10009 Video 1 8900 Video 2 2019-2016 … … 0.95-0.90 … … … … … … … 0.5-0.45 … … …
[0148] It can be seen from Table 1 that the similarity interval sequence may include: [0.45, 0.5], (0.5, 0.55]...(0.90, 0.95], (0.95, 1]; the first object interval sequence may include: 2023-2020, 2019-2016... According to the first object interval sequence 2023-2020, it can be seen from Table 1 that the first object interval sequence may include movie 1 and movie 2. Then, according to the popularity of movie 1 and movie 2, the movie 1 and movie 2 can be sorted, and movie 1 is located in movie 2. And so on, the object sequence under the movie name in Table 1 is obtained.
[0149] As another exemplary real-time method, for each similarity interval sequence, the objects in the similarity interval can be first scored according to the release time of each object in the similarity interval sequence to obtain a first score. The release time of each object is inversely proportional to the first score, that is, the closer the release time is to the present, the higher the first score is.
[0150] At the same time, the objects in the similarity interval can be scored according to the attention of each object in the similarity interval sequence to obtain a second score, wherein the attention of the object is proportional to the second score, that is, the higher the attention, the higher the second score.
[0151] According to the above steps, the first score and the second score of each object can be obtained, based on which, the first score and the second score of each object can be preset for calculation to obtain the target score of each object. The above preset calculation can be to calculate the average of the first score and the second score, or to sum the first score and the second score, or other calculation methods, which are not limited in the embodiments of the present application.
[0152] Based on this, the execution subject of the embodiment of the present application can sort the objects in the similarity interval sequence according to the target score of each object to obtain an object sequence.
[0153] The technical solution provided by the embodiment of the present application is to input the object identifier and search phrase of each object in the object library into a preset search engine, obtain the first N search results obtained by the search engine, determine the similarity between the object and the recommended text based on the first N search results, and arrange multiple objects in the object library based on the similarity to obtain an object sequence. This technical solution, by obtaining the search results of each object and the search phrase, and determining the similarity between the object and the recommended text based on the search results, and then arranging multiple objects based on the similarity to obtain an object sequence, realizes the arrangement of multiple objects based on the similarity between the object and the recommended text, thereby realizing accurate and rapid preliminary screening and sorting of multiple objects in the object library, so as to reduce the number of objects in the object library, thereby quickly and accurately determining the objects to be recommended.
[0154] join Figure 4 , which is a flow chart of an embodiment of another object recommendation method provided in an embodiment of the present application. Figure 4 The process shown in Figure 1 Based on the process shown in FIG. 1 , it is described how to determine the object set of the recommended objects from the object sequence according to the object profile and recommendation text of each object. Figure 4 As shown, the process may include the following steps:
[0155] Step 401: For each object in the above object sequence, write the object introduction and recommended text of the object into a preset second prompt template to obtain a second prompt text.
[0156] Step 402: Input the second prompt text into a preset second largest language model, so that the second largest language model determines whether the content of the object profile description matches the content of the recommendation text description based on the second prompt text, and outputs a matching result.
[0157] The following is a unified description of step 401 and step 402:
[0158] The second prompt text is used to prompt whether the content of the object profile description matches the content of the recommendation text description.
[0159] Correspondingly, the second prompt template is a template used to generate the second prompt text.
[0160] In practical applications, the response time of the large language model is proportional to the number of words in the input content. However, due to the diversity of content involved in the object description, it is necessary to input a relatively detailed text (about a thousand words) to determine whether the object meets the recommended text. This filtering step is very time-consuming. Further considering that the above-mentioned determination of the object sequence is based on the similarity calculated based on the vector distance, it may be inaccurate. Therefore, in order to further reduce the time consumption, the execution subject of the embodiment of the present application can use the large language model to pre-screen the object according to the object profile (about 100 words, mainly describing the theme of the object drama) to filter out objects that do not meet the recommended text in a shorter time (for example, the object is a movie, the recommended text is a sweet pet drama, and the film and television drama that can be judged as a military drama through the plot introduction can be directly filtered out).
[0161] In one embodiment, the user can pre-set the second prompt template. The executing entity of the embodiment of the present application can first obtain the object introduction of each object in the object sequence, and for each object, write the object introduction and recommended text corresponding to the object into the above-mentioned second prompt template to obtain the corresponding second prompt text.
[0162] Furthermore, in order to reduce the number of objects to be screened, the executing entity of the embodiment of the present application may determine the sequence consisting of the first M objects in the object sequence as a new object sequence, and for each object in the new object sequence, write the object profile and recommended text into a preset second prompt template to obtain a second prompt text.
[0163] Afterwards, the second prompt text may be input into a pre-trained second large language model. After obtaining the second prompt text, the large language model may determine whether the content of the object profile description matches the content of the recommended text description based on the second prompt text, and output a matching result.
[0164] For example, suppose the second prompt template mentioned above is "You are a senior film and television drama operator, responsible for screening film and television series for users based on the user's vague description (or evaluation) of the plot. You are given a plot background of a film and television drama, and the complete plot of the film and television drama will unfold under this background setting. The criteria for the screening of this film and television drama are: Based on the common sense of plot development, the plot in the vague description or the plot that meets the user's evaluation may occur under the plot background. Determine whether the following plot background (object introduction) can pass the screening of the fuzzy description (recommended text)."
[0165] Let's continue to assume that the above recommended text is "a type of romantic drama, telling the story of the hero and heroine getting married first, and then getting to know each other and falling in love in various things after marriage", and the object profile is "a memoir, a group portrait narrative spanning 20 years, panoramically showing the black-and-white contest and complex human nature under the changing times". After that, the above recommended text and object profile are input into the above second prompt template, and the second prompt text can be obtained: "You are a senior film and television drama operator, responsible for selecting film and television series for users based on the user's vague description (or evaluation) of the plot. Give you a plot background of a film and television drama, and the complete plot of the film and television drama will unfold under this background setting. The standard for the selection of this film and television drama is: according to the common sense of plot development, the plot in the vague description or the plot that meets the user's evaluation may occur under the plot background. Determine whether the following plot background (a memoir, a group portrait narrative spanning 20 years, panoramically showing the black-and-white contest and complex human nature under the changing times) can pass the screening of the vague description (a type of romantic drama, telling the story of the hero and heroine getting married first, and then getting to know each other and falling in love in various things after marriage)."
[0166] Afterwards, the second prompt text is input into the second largest language model, and the second largest language model can determine that the content of the object profile description does not meet the screening of the recommended text, and thus can output the analysis result as "no".
[0167] Step 403: When the content of the object profile description of the object represented by the above matching result matches the content of the recommendation text description, the object is determined as an object to be recommended.
[0168] Step 404: classify the above-mentioned objects to be recommended into a preset object set of objects to be recommended.
[0169] The following is a unified description of step 403 and step 404:
[0170] In an embodiment of the present application, after the object profile and recommended text of each object in the object sequence are input into the second largest language model, the matching result of the large language model for each object can be obtained to screen out the object whose content described in the object profile in the object sequence is consistent with the recommended text, and the object is determined as the object to be recommended, so as to classify the object to be recommended into a preset object set of objects to be recommended.
[0171] The technical solution provided by the embodiment of the present application is to write the object profile and recommended text of each object in the above-mentioned object sequence into a preset second prompt template to obtain a second prompt text, and input the second prompt text into a preset second largest language model so that the second largest language model determines whether the content described in the object profile matches the content described in the recommended text according to the second prompt text, and outputs a matching result. When the above matching result indicates that the content described in the object profile of the object matches the content described in the recommended text, the object is determined as an object to be recommended, and the above-mentioned object to be recommended is included in a preset object set of objects to be recommended. This technical solution further pre-screens the determined object sequence through the object profile to filter out objects whose content described in the object profile does not conform to the recommended text, thereby further quickly determining the object to be recommended.
[0172] join Figure 5 , is a system flow diagram of an object recommendation system provided in an embodiment of the present application. Figure 5 The system block diagram shown in the figure takes the object system as a movie recommendation system and the target recommendation object set as a movie list as an example for explanation. Figure 5 As shown, the system may include modules: a search content generation module, a relevance ranking module and a film list pre-screening module.
[0173] Among them, the above-mentioned search content generation module is a module for designing prompt words and using a large language model to convert the film list description into a phrase for searching for film-related content in a search engine.
[0174] Specifically, the description of a playlist is often a complete sentence used to attract users to watch. If it is directly used for online search, the relevance of the search results may be low. Therefore, it is necessary to convert the playlist description into a phrase that is conducive to search, and send it to the search engine together with the movie title to obtain relevant information. The input of this module is the playlist description, and the search phrase is generated by designing prompt words 1 and using a large language model. In order to ensure the accuracy of the generated phrase, a few-shot method is used in the prompt word to guide the result.
[0175] As for how to use the large language model to extract the search phrase from the description of the playlist, please refer to Figure 2 The process shown will not be repeated here.
[0176] The above relevance ranking module is responsible for re-ranking the films in the film library to reduce the number of films that need to be screened. The specific process is as follows:
[0177] (1) Send the <title> + <search phrase> of each movie in the movie library to the search engine to obtain the top 10 search results.
[0178] (2) Filter search results that do not contain <film title> to reduce the negative impact of irrelevant content on ranking.
[0179] (3) Use the embedding model to obtain the 1024-dimensional embedding vectors of the movie list description and filtered search results.
[0180] (4) Calculate the cosine similarity (0-1) between the embedding vector of the playlist description and the top 10 search results of each movie, and select the maximum similarity as the similarity between the movie and the playlist description. If the snapshot is empty, the similarity is 0.
[0181] (5) Considering that the latest and most popular films are more likely to attract users to watch, this paper takes into account the release time and popularity information of the films when re-sorting the screening order of the film library, and sorts them according to the release time and similarity. The specific process is as follows:
[0182] (a) The film library is re-sorted according to similarity and divided into 10 segments with step as the interval. The step is calculated as follows: step = (highest similarity value - lowest similarity value) / 10.
[0183] (b) Within each segment, the films are divided into three-year intervals according to their release time, and the films in each segment are sorted from high to low according to their popularity.
[0184] As for how to determine the object sequence according to the release time and popularity of each object, the release time can be used as the release time of the object, the popularity of the object can be used as the attention of the object, and the Figure 3 The process shown will not be repeated here.
[0185] Among them, the pre-screening module for the film list is used because the time for the large language model to generate a response is proportional to the number of words in the input content. However, due to the diversity of content involved in the film list description, it is necessary to input a more detailed plot text (about a thousand words) to determine whether the film meets the description. This filtering step is very time-consuming. Considering that the relevance calculated according to the embedding vector distance in the relevance sorting module may be inaccurate, in order to further reduce the time consumption, this method will also use the large language model to pre-screen the film according to the plot synopsis of the film (about 100 words, mainly describing the theme of the film and television drama) to filter out films that are impossible to have the plot described in the film list in a shorter time (for example, the film list is described as a sweet pet drama, but the film and television drama that can be judged as a military drama through the plot synopsis can be directly filtered out).
[0186] As for how to filter out movies that are unlikely to have plots described in the playlist based on the plot synopsis and playlist description of each movie, the movie can be taken as the object, the plot synopsis as the object synopsis, and the playlist description as the reference for the recommended text. Figure 4 The process shown will not be repeated here.
[0187] The object recommendation system provided in the embodiment of the present application may include a search content generation module, a relevance sorting module and a film list pre-screening module, which can accelerate the speed of generating recommended film lists based on a large language model, save computing resources, and optimize the user experience.
[0188] join Figure 6 , is a block diagram of an embodiment of an object recommendation device provided in an embodiment of the present application. Figure 6 As shown, the device may include:
[0189] The first acquisition module 61 is used to acquire a preset recommendation text and extract a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended;
[0190] A sorting module 62, used to sort multiple objects in a preset object library according to the search phrase to obtain an object sequence;
[0191] A second acquisition module 63 is used to acquire an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object;
[0192] The recommendation module 64 is used to determine an object set of objects to be recommended from the object sequence according to the object introduction and the recommendation text of each object, so as to recommend objects in the object set.
[0193] like Figure 7 As shown, it is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application, including a processor 71, a communication interface 72, a memory 73 and a communication bus 74, wherein the processor 71, the communication interface 72, and the memory 73 communicate with each other through the communication bus 74.
[0194] A memory 73, for storing computer programs;
[0195] In one embodiment of the present application, the processor 71 is used to execute the program stored in the memory 73 to implement the object recommendation method provided by any of the above method embodiments, including:
[0196] Obtaining a preset recommendation text, and extracting a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended;
[0197] According to the search phrase, multiple objects in a preset object library are sorted to obtain an object sequence;
[0198] Obtaining an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object;
[0199] According to the object introduction and the recommendation text of each object, an object set of objects to be recommended is determined from the object sequence to recommend objects in the object set.
[0200] An embodiment of the present application further provides a storage medium on which a computer program is stored. When the computer program is executed by a processor, the steps of the object recommendation method provided in any of the aforementioned method embodiments are implemented.
[0201] 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.
[0202] 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, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiments.
[0203] 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 steps may be used.
[0204] 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. An object recommendation method, characterized in that: The method comprises: Obtaining a preset recommendation text, and extracting a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended; According to the search phrase, multiple objects in a preset object library are sorted to obtain an object sequence; Obtaining an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object; According to the object introduction and the recommendation text of each object, an object set of objects to be recommended is determined from the object sequence to recommend objects in the object set.
2. The method according to claim 1, characterized in that The extracting the search phrase from the recommended text comprises: Writing the recommended text into a preset first prompt template to obtain a first prompt text; The first prompt text is input into a preset first language model to obtain a search phrase extracted from the recommended text.
3. The method according to claim 1, characterized in that The step of sorting a plurality of objects in a preset object library according to the search phrase to obtain an object sequence includes: For each object in the object library, input the object identifier of the object and the search phrase into a preset search engine, and obtain the first N search results obtained by the search engine; Determine the similarity between the object and the recommended text according to the first N search results; The multiple objects in the object library are arranged based on the similarity to obtain an object sequence.
4. The method according to claim 3, characterized in that The determining, based on the first N search results, the similarity between the object and the recommended text includes: Deleting search results that do not contain the object identifier in the search results, and using the remaining search results as target search results; Encoding the recommended text into a recommended text vector of a preset dimension, and encoding each of the target search results into a target search result vector of a preset dimension; For each of the target search results, determining the cosine similarity between the target search result vector corresponding to the target search result and the recommended text vector; The maximum cosine similarity is used as the similarity between the object and the recommended text.
5. The method according to claim 3, characterized in that: Arranging the multiple objects in the object library based on the similarity to obtain an object sequence includes: Determine the publishing time and attention level of each object, where the attention level is used to represent the degree to which the object is paid attention to by multiple users; Arranging the objects in the object library in descending order of the similarity to obtain an initial object sequence; According to the release time and the attention degree, the order of objects in the initial object sequence is adjusted to obtain the object sequence.
6. The method according to claim 5, characterized in that The step of adjusting the order of objects in the initial object sequence according to the release time and the attention degree to obtain the object sequence comprises: Determining a maximum similarity and a minimum similarity in the initial object sequence; Calculating a similarity difference between the maximum similarity and the minimum similarity, and dividing the similarity difference by a preset number to obtain an interval length of each similarity interval in the preset number of similarity intervals; Dividing the similarity in the initial object sequence into intervals according to the interval lengths to obtain the preset number of similarity intervals; According to the similarity of each object in the initial object sequence, the objects in the initial object sequence are divided according to the similarity intervals to obtain a preset number of similarity interval sequences; For each of the similarity interval sequences, a first score is given to the objects in the similarity interval according to the release time of each object in the similarity interval sequence to obtain a first score; and a second score is given to the objects in the similarity interval according to the attention of each object in the similarity interval sequence to obtain a second score; Performing a preset calculation on the first score and the second score of each object to obtain a target score for each object; According to the target score, the objects in the similarity interval sequence are sorted to obtain an object sequence.
7. The method according to claim 1, characterized in that The step of determining an object set of objects to be recommended from the object sequence according to the object introduction and the recommendation text of each object includes: For each of the objects in the object sequence, writing the object introduction and the recommended text of the object into a preset second prompt template to obtain a second prompt text; Inputting the second prompt text into a preset second language model, so that the second language model determines whether the content of the object profile description matches the content of the recommendation text description according to the second prompt text, and outputs a matching result; In the case where the matching result indicates that the content of the object profile description of the object matches the content of the recommendation text description, the object is determined as an object to be recommended; The object to be recommended is classified into a preset object set of objects to be recommended.
8. The method according to claim 1, characterized in that The recommending of the objects in the object set includes: Obtaining a detailed description text of each object in the object set, wherein the detailed description text is a text describing the content of the object, and the number of words in the detailed description text is more than that of the object introduction; Encoding the detailed description text into a description text vector of a preset dimension, and encoding the recommended text into a recommended text vector of a preset dimension; Determining the text cosine similarity between the description text vector and the recommendation text vector; An object whose text cosine similarity is greater than a preset similarity threshold is determined as a target object, and the target object is recommended.
9. An object recommendation device, characterized in that: The device comprises: A first acquisition module is used to acquire a preset recommendation text and extract a search phrase from the recommendation text; the recommendation text is used to describe the object characteristics of the object to be recommended, and the search phrase is used as a keyword to search for the object to be recommended; A sorting module, used to sort multiple objects in a preset object library according to the search phrase to obtain an object sequence; A second acquisition module is used to acquire an object introduction of each object in the object sequence, wherein the object introduction is a text describing the content of the object; The recommendation module is used to determine an object set of objects to be recommended from the object sequence according to the object introduction and the recommendation text of each object, so as to recommend objects in the object set.
10. An electronic device, characterized in that: include: A processor and a memory, wherein the processor is used to execute an object recommendation program stored in the memory to implement the object recommendation method according to any one of claims 1 to 8.
11. A storage medium, characterized in that: The storage medium stores one or more programs, and the one or more programs can be executed by one or more processors to implement the object recommendation method according to any one of claims 1 to 8.
Citation Information
Patent Citations
Recommendation method, recommendation device, electronic equipment and storage medium
CN114625862A
Sequence recommendation method and device
CN117909592A
Object searching method and device, computer equipment and storage medium
CN118093792A
Information recommendation method, electronic device, storage medium and program product
CN118193823A
Information display method and device, and computer storage medium
WO2023051062A1
