A recommendation method, a terminal, and a storage medium

Abstract

The application discloses a recommendation method, a terminal and a storage medium. The recommendation method comprises the following steps: acquiring a first object set, acquiring a first comment of a first user on each first object and a second comment of a second user on each first object according to the first object set; acquiring a first sentiment parameter corresponding to each first comment according to each first comment, and acquiring a second sentiment parameter corresponding to each second comment according to each second comment; determining a target similarity between the first user and the second user according to the first sentiment parameter and the second sentiment parameter; and acquiring a recommended object corresponding to the first user according to the target similarity. The application acquires the sentiment parameter of the user on the object, and combines the sentiment parameter of the user to acquire the similarity between the users, so that the obtained similarity is more in line with the actual situation, and the recommendation effect is better.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of electronic commerce, in particular to a recommendation method, a terminal and a storage medium. BACKGROUND

[0002] In recent years, the commodity recommendation system has developed rapidly in the field of electronic commerce. In order to make a suitable match between a large number of commodities and a large number of users and promote the occurrence of transactions, many network platforms use a recommendation system to recommend suitable commodities or content to users. However, in the prior art, the recommendation methods used by many network platforms only obtain the similarity of two users based on the similarity of the ratings of the commodities by the two users, and then recommend the commodities purchased by the user with high similarity to the target user, that is, when the ratings of two users for a plurality of commodities are similar, it is considered that the similarity of the two users is high, and then the other user purchases or rates other products highly to one of the users. However, due to the different standards of users for ratings, in this recommendation method, the user may be recommended with commodities that the user does not like, and the recommendation effect is poor.

[0003] Therefore, the prior art still needs to be improved and enhanced. SUMMARY

[0004] The present application provides a recommendation method, a terminal and a storage medium, which aims to solve the problem of poor recommendation effect caused by obtaining the similarity of users only according to the similarity of the ratings of commodities by users in the prior art.

[0005] In order to solve the above technical problems, the technical scheme adopted by the present application is as follows:

[0006] A recommendation method, wherein the recommendation method comprises:

[0007] obtaining a first object set, the first object set comprising each first object evaluated by a first user and a second user, and obtaining each first comment of the first user on each first object and each second comment of the second user on each first object according to the first object set;

[0008] obtaining each first sentiment parameter corresponding to each first comment according to each first comment, and obtaining each second sentiment parameter corresponding to each second comment according to each second comment;

[0009] determining a target similarity of the first user and the second user according to each first sentiment parameter and each second sentiment parameter;

[0010] obtaining a recommended object corresponding to the first user according to the target similarity.

[0011] The recommendation method, wherein the respective first sentiment parameters corresponding to the respective first comments and the respective second sentiment parameters corresponding to the respective second comments are obtained according to the respective first comments and the respective second comments respectively.

[0012] The respective first sentiment parameters and the respective second sentiment parameters output by the preset model are obtained by inputting the respective first comments and the respective second comments into the trained preset model respectively.

[0013] The recommendation method, wherein the target similarity of the first user and the second user is determined according to the respective first sentiment parameters and the respective second sentiment parameters.

[0014] The respective first ratings of the first user to the respective first objects and the respective second ratings of the second user to the respective second objects are obtained, and a rating similarity of the first user and the second user is obtained according to the respective first ratings and the respective second ratings.

[0015] The target similarity is determined according to the respective first sentiment parameters, the respective second sentiment parameters and the rating similarity.

[0016] The recommendation method, wherein the target similarity is determined according to the respective first sentiment parameters, the respective second sentiment parameters and the rating similarity.

[0017] A first similarity of the first user and the second user is obtained according to the respective first sentiment parameters, the respective second sentiment parameters and the rating similarity.

[0018] A second similarity of the first user and the second user is obtained according to a second object set and a third object set, wherein the second object set comprises respective third objects evaluated by the first user, and the third object set comprises respective fourth objects evaluated by the second user.

[0019] The target similarity is obtained according to the first similarity and the second similarity.

[0020] The recommendation method, wherein the second similarity of the first user and the second user is obtained according to a second object set and a third object set.

[0021] Respective first attribute parameters corresponding to the respective third objects and respective second attribute parameters corresponding to the respective fourth objects are obtained respectively.

[0022] The second similarity is obtained according to the respective first attribute parameters and the respective second attribute parameters.

[0023] The recommendation method, wherein the target similarity is obtained according to the first similarity and the second similarity comprises:

[0024] The target similarity is obtained according to the first similarity, the first preset weight corresponding to the first similarity, the second similarity and the second preset weight corresponding to the second similarity.

[0025] The recommendation method, wherein the target similarity is obtained according to the first similarity and the second similarity comprises:

[0026] The first preset weight and the second preset weight are determined according to subjective weighting method and objective weighting method.

[0027] The recommendation method, wherein the target similarity is obtained according to the first similarity and the second similarity comprises:

[0028] The target similarity of the first user and at least one second user is obtained respectively, and a target user is determined according to each target similarity.

[0029] The target object corresponding to the target user is obtained, and the recommendation object is obtained according to the target object.

[0030] A terminal, wherein the terminal comprises a processor and a storage medium connected with the processor in communication, the storage medium is adapted to store a plurality of instructions, and the processor is adapted to call the instructions in the storage medium to execute the steps of the above-mentioned recommendation method.

[0031] A storage medium, wherein 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 steps of the above-mentioned recommendation method.

[0032] Beneficial effects: compared with the prior art, the present application provides a recommendation method, a terminal and a storage medium, the recommendation method obtains the emotional parameters of the user to the object according to the comments of the user to the object, and obtains the similarity between the users in combination with the emotional parameters of the user, so that the obtained similarity is more in line with the actual situation and the recommendation effect is better. BRIEF DESCRIPTION OF DRAWINGS

[0033] Figure 1 The flow chart of the embodiment of the recommendation method provided by the present application;

[0034] Figure 2 The schematic diagram of the object in the recommendation method provided by the present application;

[0035] Figure 3Sub-step flow in an embodiment of the recommendation method provided by the present application Figure 1 ;

[0036] Figure 4 Sub-step flow in an embodiment of the recommendation method provided by the present application Figure 2 ;

[0037] Figure 5 Flowchart of acquiring the first preset weight and the second preset weight in the recommendation method provided by the present application

[0038] Figure 6 Structure principle diagram of an embodiment of the terminal provided by the present application DETAILED DESCRIPTION

[0039] To make the objectives, technical solutions and effects of the present application clearer and more explicit, the present application is further described in detail below with reference to the drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain the present application and do not limit the present application.

[0040] The recommendation method provided by the present application can be applied in a terminal, which can be but is not limited to various personal computers, notebook computers, mobile phones, tablet computers, vehicle-mounted computers and portable wearable devices. The terminal can acquire a corresponding recommendation object of a user according to the recommendation method.

[0041] Embodiment one

[0042] Please refer to Figure 1 , Figure 1 Flowchart of an embodiment of the recommendation method provided by the present application. The recommendation method comprises the following steps:

[0043] S100, acquiring a first object set, the first object set comprising each first object evaluated by a first user and a second user, and acquiring each first comment of the first user on each first object and each second comment of the second user on each first object according to the first object set.

[0044] In the embodiment, the object recommendation is made to the first user according to the similarity between the first user and the second user, and the object can be a real physical item such as an electrical appliance or a daily necessity, or a virtual item such as a movie, a song or a question. Specifically, a first object set is acquired, which is a set of objects evaluated by the first user and the second user, that is, the first object set comprises each first object evaluated by the first user and the second user. Specifically, the evaluation comprises a comment and a score, that is, the first user and the second user have both scored and commented on each first object, as shown in Figure 2 ​Figure 2 The middle circle part is a score, and the square part is a comment. The first user and the second user can be distinguished by the ID, IP address, etc. of the user, that is, the same ID on the network platform is regarded as the same user or the same IP address is regarded as the same user, etc.

[0045] After obtaining the first object set, a first comment of the first user on each first object and a second comment of the second user on each first object are obtained respectively, and the first comment and the second comment are in a text format.

[0046] As shown in Figure 1 After obtaining the first comment and the second comment, the recommendation method further comprises:

[0047] S200, according to the first comment and the second comment, a first emotion parameter corresponding to each first comment and a second emotion parameter corresponding to each second comment are obtained respectively.

[0048] In the present application, the emotion parameter is data reflecting the emotion contained in the text, that is, the first emotion parameter corresponding to the first comment reflects the emotion contained in the first comment, for example, like or hate. In the prior art, the recommendation method used by many network platforms is only based on the similarity of the scores of two users on a product to obtain the similarity of the two users, and then recommend the products purchased by the user with high similarity to the target user, that is, when two users have similar scores on multiple products, it is considered that the similarity of the two users is high, and then one of the users is recommended to purchase other products purchased or scored by the other user. However, due to different standards of users for scoring, for example, the scores of users with high tolerance will be high, and the scores of users with low tolerance will be low. For the same product, even if the scores of two users are consistent, the degree of liking of the two users for the product is likely to be different, so in this recommendation method, the user may be recommended a product that he does not like, and the recommendation effect is poor. In this embodiment, the emotion parameter of the user to the object is obtained according to the comment of the user to the object. It is not difficult to see that, compared with the score, the comment has more content and can reflect the emotion of the user to the object. In this embodiment, according to the first comment and the second comment, the first emotion parameter corresponding to each first comment and the second emotion parameter corresponding to each second comment are obtained respectively, which comprises:

[0049] The first comment and the second comment are input into a preset model trained, and the first emotion parameter and the second emotion parameter output by the preset model are obtained.

[0050] Specifically, in the present embodiment, the respective first sentiment parameters and the respective second sentiment parameters are obtained through a preset model that is pre-trained. The preset model can be constructed according to an existing natural language processing model, such as Bi-LSTM (Bi-directional Long Short-Term Memory) and the like. The preset model is trained through a plurality of sets of sample data, each set of sample data including a sample text and a sentiment parameter corresponding to the sample text. After training is completed, the preset model can output a sentiment parameter corresponding to an input text. Specifically, whether the preset model is trained can be determined by pre-setting a training target, for example, setting a target accuracy of a sentiment parameter, obtaining an accuracy of a sentiment parameter output by the preset model after training, and determining that the preset model is trained when the accuracy reaches the target accuracy. Whether the preset model is trained can also be determined by a pre-set loss function. Determining the training degree of a model by using a loss function is a common means in the field, and will not be described herein again. After the respective first comments and the respective second comments are obtained, the respective first comments and the respective second comments are respectively input into the preset model that is trained, and the preset model outputs the respective first sentiment parameters and the respective second sentiment parameters.

[0051] In a possible implementation, the sentiment parameter includes a sentiment bias value. In the present embodiment, in order to facilitate calculation, the sentiment bias value is set to a value in the range of (0, 1), for example, the first sentiment parameter in (0, 0.3) reflects that the sentiment of the corresponding first comment is negative, that is, the sentiment of the first user to the corresponding first object is negative, the first sentiment parameter in [0.3, 0.7] reflects that the sentiment of the corresponding first comment is neutral, and the first sentiment parameter in (0.7, 1) reflects that the sentiment of the corresponding first comment is positive. The sentiment bias value can be related to the intensity of the sentiment, for example, the greater the sentiment bias value, the more positive the corresponding sentiment.

[0052] In order to prevent the sentiment bias value from being neutralized in the subsequent calculation process and losing sensitivity to the intensity of the sentiment, in another possible implementation, the sentiment parameter includes a sentiment bias value and a separate sentiment intensity value, and the sentiment intensity value is a fixed value, for example, 0.2 represents weak, 0.3 represents general, and 0.5 represents strong. For example, the first sentiment parameters are 0.1275 and 0.1, which are in the range of (0, 0.3), and then the sentiment of the corresponding first comment is weakly negative.

[0053] It should be understood by those skilled in the art that the setting of the emotional bias value and the emotional intensity value described above is only an example, and those skilled in the art can set the emotional bias value and the emotional intensity value differently according to the spirit of the present application.

[0054] S300, determining a target similarity between the first user and the second user according to the respective first emotional parameters and the respective second emotional parameters.

[0055] As described above, the respective first emotional parameters respectively reflect the emotions of the first user to the respective first objects, and the respective second emotional parameters respectively reflect the emotions of the second user to the respective first objects. In this embodiment, the target similarity between the first user and the second user is determined in combination with the respective first emotional parameters and the respective second emotional parameters, which can improve the accuracy of the target similarity, and thus the recommended object of the first user obtained according to the target similarity is more in line with the preferences of the first user, and the recommendation effect is improved.

[0056] Specifically, as shown in Figure 3 determining the target similarity between the first user and the second user according to the respective first emotional parameters and the respective second emotional parameters comprises:

[0057] S310, obtaining respective first scores of the first user to the respective first objects and respective second scores of the second user to the respective second objects, and obtaining a score similarity between the first user and the second user according to the respective first scores and the respective second scores.

[0058] As described above, the first objects are objects that are evaluated by the first user and the second user, and the evaluation includes comments and scores. In this embodiment, the score similarity between the first user and the second user is obtained according to the respective first scores of the first user to the respective first objects and the respective second scores of the second user to the respective first objects, and then the target similarity is obtained according to the first emotional parameters and the second emotional parameters and the score similarity.

[0059] Specifically, the score similarity can be obtained by a plurality of similarity calculation methods, such as a Cosine Similarity method, a Jaccard Similarity method, etc. Taking the Cosine Similarity method as an example, the first scores can be represented by a multi-dimensional vector, for example, the first scores are 3, 3.5, 2, …, which can be represented by a vector (3, 3.5, 2, …). The second scores can be represented by a multi-dimensional vector, for example, the second scores are 4, 1, 3, …, which can be represented by a vector (4, 1, 3, …). It should be noted that the arrangement order of each component in the vector representation of the first scores is consistent with the arrangement order of each component in the vector representation of the second scores, that is, the vector components at the same position correspond to the scores of the same first object. The score similarity can be obtained by the formula The score similarity can be obtained, where sim(x, y) is the score similarity, r x,s is the s component of the vector corresponding to the first scores, that is, the first score corresponding to the first object s, r y,s is the s component of the vector corresponding to the first scores, that is, the second score corresponding to the first object s, S xy is the first vector set.

[0060] S320, determining the target similarity according to the first emotional parameters, the second emotional parameters, and the score similarity.

[0061] After obtaining the score similarity, the target similarity is obtained in combination with the first emotional parameters and the second emotional parameters. Specifically, as shown in Figure 4 the determining the target similarity according to the first emotional parameters, the second emotional parameters, and the score similarity includes:

[0062] S321, obtaining the first similarity between the first user and the second user according to the first emotional parameters, the second emotional parameters, and the score similarity.

[0063] In this embodiment, the emotional similarity is obtained according to the first emotional parameters and the second emotional parameters, and the first similarity is obtained according to the emotional similarity and the score similarity.

[0064] The emotion similarity can be obtained by a plurality of similarity methods, such as a Cosine Similarity method, a Jaccard Similarity method, etc. It is worth noting that when the emotion parameters include the emotion bias value and the emotion intensity value, the emotion similarity corresponds to bias similarity and intensity similarity.

[0065] After obtaining the emotion similarity, the first similarity is obtained according to a preset formula. Specifically, when the emotion parameters only include the emotion bias value, the preset formula is S1=S q1 *A1+S f *A2, wherein S1 is the first similarity, S q1 is the emotion similarity, S f is the score similarity, A1 and A2 are weight values of the emotion similarity and the score similarity respectively, and A1 and A2 are preset. A person skilled in the art can set A1 and A2 according to actual conditions, for example, set A1 to 0.15 and A2 to 0.85.

[0066] When the emotion parameters include the emotion bias value and the emotion intensity value, the preset formula is S1=S q1 *A1+S q2 *A2+S f *A3, wherein S1 is the first similarity, S q1 is the bias similarity, S q2 is the intensity similarity, S f is the score similarity, A1, A2 and A3 are weight values of the bias similarity, the intensity similarity and the score similarity respectively, and A1 and A2 are preset. A person skilled in the art can set A1, A2 and A3 according to actual conditions, for example, set A1 to 0.1, A2 to 0.05 and A3 to 0.85.

[0067] As can be seen from the foregoing description, the first similarity combines the rating similarity and the sentiment similarity of the objects evaluated by the first user and the second user. In a possible implementation, the first similarity can be directly used as the target similarity. However, since there are a large number of objects on the existing network platform, the number of objects evaluated by the first user and the second user is limited in many cases. Therefore, it is sometimes not accurate to determine the target similarity only according to the first objects evaluated by the first user and the second user. Therefore, in this embodiment, a second similarity of the first user and the second user is obtained by combining the goods evaluated by the first user and the goods evaluated by the second user, and the target similarity is obtained by combining the first similarity and the second similarity. Specifically, the target similarity determined according to the first sentiment parameters, the second sentiment parameters and the rating similarity further includes:

[0068] S322, obtaining a second similarity of the first user and the second user according to a second object set and a third object set.

[0069] The second object set includes each third object evaluated by the first user, and the third object set includes each fourth object evaluated by the second user.

[0070] The second similarity of the first user and the second user obtained according to the second object set and the third object set includes:

[0071] S3221, obtaining each first attribute parameter corresponding to each third object and each second attribute parameter corresponding to each fourth object, respectively.

[0072] In the present application, the attributes corresponding to each third object and the attributes corresponding to each fourth object are obtained respectively, and the first attribute parameters corresponding to each third object and the fourth attribute parameters corresponding to each fourth object are obtained according to the attributes corresponding to each third object and the attributes corresponding to each fourth object. Specifically, in the present application, the attribute is the characteristic of the object, for example, for a movie object, the attributes can include actors, directors, themes, etc., and for a mobile phone object, the attributes can include camera pixels, screen size, battery life, etc. The attribute parameter is a vectorized representation of the attributes possessed by the object. Specifically, a preset attribute can be set in advance, and for each preset attribute, a vectorized representation is used, that is, each preset attribute corresponds to an attribute vector. Therefore, the attribute parameter corresponding to the object is a vector composed of the attribute vectors corresponding to the preset attributes included by the object. For example, for a movie object, the attributes include actor A, director C, and comedy. The attribute vector corresponding to actor A is (x, y, z), a is the vectorized representation corresponding to actor A, c is the vectorized representation corresponding to director C, and z is the vectorized representation corresponding to "comedy". Alternatively, the preset attributes are arranged in a certain order. For each object, if A attribute is included, the vector component corresponding to A attribute is set to 1, and if A attribute is not included, the vector component corresponding to A attribute is set to 0. In this way, a multi-dimensional vector composed of 0 and 1 is obtained as the attribute parameter corresponding to the object.

[0073] S3222, obtaining the second similarity according to the first attribute parameters and the second attribute parameters.

[0074] After obtaining the first attribute parameters and the second attribute parameters, the second similarity is obtained according to the first attribute parameters and the second attribute parameters. Specifically, the second similarity can be obtained by a plurality of similarity methods, such as the Cosine Similarity method, the Jaccard Similarity method, etc.

[0075] As can be seen from the foregoing description, the respective first attribute parameters and the respective second attribute parameters reflect the attributes of the objects evaluated by the first user and the attributes of the objects evaluated by the second user, that is, the respective first attribute parameters and the respective second attribute parameters reflect the attributes of interest of the first user and the second user, and then the second similarity obtained according to the respective first attribute parameters and the respective second attribute parameters reflects the similarity of the attributes of interest of the first user and the second user. Since the second similarity is not limited to the objects evaluated by the first user and the second user, the second similarity can be obtained according to more objects, and the accuracy of the target similarity obtained in the application is further improved.

[0076] Please refer again to Figure 4 The determining the target similarity according to the respective first emotion parameters, the respective second emotion parameters and the score similarity further includes:

[0077] S323, obtaining the target similarity according to the first similarity, the first preset weight corresponding to the first similarity, the second similarity and the second preset weight corresponding to the second similarity.

[0078] After obtaining the first similarity and the second similarity, the target similarity is obtained by combining the first similarity and the second similarity. In this embodiment, the first preset weight corresponding to the first similarity and the second preset weight corresponding to the second similarity are pre-set, and then the target similarity is obtained according to the formula: S=S1*X1+S2*X2, wherein S is the target similarity, S1 is the first similarity, X1 is the first preset weight, S2 is the second similarity, and X2 is the second preset weight.

[0079] The determining the target similarity according to the first similarity and the second similarity includes:

[0080] The first preset weight and the second preset weight are determined according to subjective weighting method and objective weighting method.

[0081] Specifically, in the prior art, there are subjective weighting methods and objective weighting methods for obtaining weights. The subjective weighting method (such as the Delphi expert method) has the advantage that experts can reasonably determine weights according to actual decision-making problems and their own knowledge and experience, so that the weights are not contrary to the actual importance. However, the subjective weighting method has strong subjectivity and poor objectivity, and increases the burden on decision analysts, so it has great limitations in application. The objective weighting method (such as the principal component analysis method) determines weights according to the relationship between data, and has strong objectivity, but does not consider the subjective imagination of decision makers, so the determined weights may not be consistent with people's subjective desires or actual situations. In this embodiment, in order to overcome the defects of the subjective weighting method and the objective weighting method, the two are combined, that is, the first preset weight and the second preset weight are determined according to the subjective weighting method and the objective weighting method.

[0082] In one possible implementation manner, as shown in Figure 5 , the Delphi expert method and the PCA (principal component analysis method) are adopted to determine the first preset weight and the second preset weight. Specifically, first, a plurality of groups of weight data obtained by a plurality of industry experts assigning weights to the to-be-weighted indexes (that is, the first similarity and the second similarity) are obtained, each group of weight data including a first expert weight and a second expert weight, as shown in Table 1:

[0083] Expert 1 Expert 2 Expert 3 Expert 4 Expert 5 First expert weight 0.6 0.5 0.7 0.7 0.9 Second expert weight 0.4 0.5 0.3 0.3 0.1

[0084] Table 1

[0085] Then, the PCA method is used to reduce the dimension of the plurality of groups of weight data to obtain the first preset weight and the second preset weight.

[0086] As shown in Figure 1 , after the target similarity is obtained, the method further includes the steps of:

[0087] S400, obtaining a recommended object corresponding to the first user according to the target similarity.

[0088] After the target similarity of the first user and the second user is obtained, the first user can be recommended according to the target similarity. Specifically, the step of obtaining a recommended object corresponding to the first user according to the target similarity includes:

[0089] S410, obtaining the target similarity of the first user and at least one second user respectively, and determining a target user according to each target similarity.

[0090] In the object recommendation for the first user, at least one second user can be acquired, the target similarity between the first user and the at least one second user is acquired respectively, that is, at least one target similarity is acquired, and then the target user for recommending the first user is determined according to each target similarity, for example, the second user corresponding to the highest target similarity in each target similarity can be acquired as the target user, or a similarity threshold is set in advance, and the second user corresponding to the target similarity greater than the similarity threshold can be acquired as the target user.

[0091] In the object recommendation for the first user, at least one second user can be acquired, the target similarity between the first user and the at least one second user is acquired respectively, that is, at least one target similarity is acquired, and then the target user for recommending the first user is determined according to each target similarity, for example, the second user corresponding to the highest target similarity in each target similarity can be acquired as the target user, or a similarity threshold is set in advance, and the second user corresponding to the target similarity greater than the similarity threshold can be acquired as the target user.

[0092] In the object recommendation for the first user, at least one second user can be acquired, the target similarity between the first user and the at least one second user is acquired respectively, that is, at least one target similarity is acquired, and then the target user for recommending the first user is determined according to each target similarity, for example, the second user corresponding to the highest target similarity in each target similarity can be acquired as the target user, or a similarity threshold is set in advance, and the second user corresponding to the target similarity greater than the similarity threshold can be acquired as the target user.

[0093] In the object recommendation for the first user, at least one second user can be acquired, the target similarity between the first user and the at least one second user is acquired respectively, that is, at least one target similarity is acquired, and then the target user for recommending the first user is determined according to each target similarity, for example, the second user corresponding to the highest target similarity in each target similarity can be acquired as the target user, or a similarity threshold is set in advance, and the second user corresponding to the target similarity greater than the similarity threshold can be acquired as the target user.

[0094] It should be understood that, although each step in the flowchart shown in the drawings of the present application specification is shown in sequence according to the arrow, these steps are not necessarily executed in sequence according to the arrow. Unless otherwise specified herein, the execution of these steps is not strictly limited in sequence, and these steps can be executed in other sequences. Moreover, at least part of the steps in the flowchart can include multiple sub-steps or multiple stages, which are not necessarily executed at the same time, but can be executed at different times, and the execution sequence of these sub-steps or stages is not necessarily sequential, but can be executed in rotation or alternation with other steps or at least part of the sub-steps or stages of other steps.

[0095] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer readable storage medium, and when executed, can include the processes of the above-mentioned embodiment methods. Any reference to memory, storage, database or other medium used in the embodiments provided by the present application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. As an illustration but not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM), etc.

[0096] Embodiment two

[0097] Based on the above-mentioned embodiments, the present application further provides a terminal, the principle block diagram of which can be shown as Figure 6 The terminal includes a processor, a memory, a network interface, a display screen, and a temperature sensor connected through a system bus. The processor of the terminal is used to provide computing and control capabilities. The memory of the terminal includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operating system and the computer program in the non-volatile storage medium to run. The network interface of the terminal is used to communicate with external terminals through network connection. The computer program is executed by the processor to implement a recommendation method. The display screen of the terminal can be a liquid crystal display screen or an electronic ink display screen. The temperature sensor of the terminal is pre-set inside the terminal and is used to detect the current running temperature of the internal device.

[0098] Those skilled in the art can understand that the principle block diagram shown in Figure 6 The principle block diagram shown in the above-mentioned embodiments is only a block diagram of part of the structure related to the present application scheme, and does not constitute a limitation on the terminal to which the present application scheme is applied. The specific terminal can include more or less components than those shown in the figure, or combine certain components, or have a different component arrangement.

[0099] In one embodiment, a terminal is provided, comprising a memory and a processor, the memory storing a computer program, and the processor executing the computer program to implement at least the following steps:

[0100] obtaining a first object set comprising respective first objects evaluated by a first user and a second user, obtaining respective first comments of the first user on the respective first objects and respective second comments of the second user on the respective first objects according to the first object set;

[0101] obtaining respective first sentiment parameters corresponding to the respective first comments according to the respective first comments, and obtaining respective second sentiment parameters corresponding to the respective second comments according to the respective second comments;

[0102] determining a target similarity of the first user and the second user according to the respective first sentiment parameters and the respective second sentiment parameters;

[0103] obtaining a recommended object corresponding to the first user according to the target similarity.

[0104] The obtaining of the respective first sentiment parameters corresponding to the respective first comments and the respective second sentiment parameters corresponding to the respective second comments according to the respective first comments and the respective second comments comprises:

[0105] inputting the respective first comments and the respective second comments into a pre-trained model to obtain the respective first sentiment parameters and the respective second sentiment parameters output by the pre-trained model.

[0106] The determining of the target similarity of the first user and the second user according to the respective first sentiment parameters and the respective second sentiment parameters comprises:

[0107] obtaining respective first scores of the first user on the respective first objects and respective second scores of the second user on the respective second objects, and obtaining a score similarity of the first user and the second user according to the respective first scores and the respective second scores;

[0108] determining the target similarity according to the respective first sentiment parameters, the respective second sentiment parameters, and the score similarity.

[0109] The determining of the target similarity according to the respective first sentiment parameters, the respective second sentiment parameters, and the score similarity comprises:

[0110] obtaining a first similarity between the first user and the second user according to the respective first emotional parameters, the respective second emotional parameters and the score similarity;

[0111] obtaining a second similarity between the first user and the second user according to a second object set and a third object set, wherein the second object set comprises respective third objects that are evaluated by the first user, and the third object set comprises respective fourth objects that are evaluated by the second user;

[0112] obtaining the target similarity according to the first similarity and the second similarity.

[0113] The obtaining of the second similarity between the first user and the second user according to the second object set and the third object set comprises:

[0114] obtaining respective first attribute parameters corresponding to the respective third objects and respective second attribute parameters corresponding to the respective fourth objects respectively;

[0115] obtaining the second similarity according to the respective first attribute parameters and the respective second attribute parameters.

[0116] The obtaining of the target similarity according to the first similarity and the second similarity comprises:

[0117] obtaining the target similarity according to the first similarity, a first preset weight corresponding to the first similarity, the second similarity and a second preset weight corresponding to the second similarity.

[0118] The obtaining of the target similarity according to the first similarity and the second similarity comprises:

[0119] determining the first preset weight and the second preset weight according to a subjective weighting method and an objective weighting method.

[0120] The obtaining of the recommended object corresponding to the first user according to the target similarity comprises:

[0121] obtaining respective target similarities between the first user and at least one second user, and determining a target user according to the respective target similarities;

[0122] obtaining a target object corresponding to the target user, and obtaining the recommended object according to the target object.

[0123] Embodiment three

[0124] The application further provides a storage medium storing one or more programs, which can be executed by one or more processors to implement the steps of the recommendation method in the above embodiment I.

[0125] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that they can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacements for some technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A recommendation method characterized by comprising: The recommendation method comprises: obtaining a first object set comprising respective first objects evaluated by a first user and a second user, obtaining respective first comments of the first user on the respective first objects and respective second comments of the second user on the respective first objects according to the first object set; obtaining respective first sentiment parameters corresponding to the respective first comments according to the respective first comments, and obtaining respective second sentiment parameters corresponding to the respective second comments according to the respective second comments; determining a target similarity of the first user and the second user according to the respective first sentiment parameters and the respective second sentiment parameters; the determining of the target similarity of the first user and the second user according to the respective first sentiment parameters and the respective second sentiment parameters comprises: obtaining respective first scores of the first user on the respective first objects and respective second scores of the second user on the respective first objects, and obtaining a score similarity of the first user and the second user according to the respective first scores and the respective second scores; obtaining a first similarity of the first user and the second user according to the respective first sentiment parameters, the respective second sentiment parameters and the score similarity; obtaining a sentiment similarity according to the respective first sentiment parameters and the respective second sentiment parameters, and obtaining the first similarity according to the sentiment similarity and the score similarity, wherein the sentiment parameters comprise sentiment bias values and sentiment intensity values, and the sentiment similarity comprises bias similarity and intensity similarity; obtaining a second similarity of the first user and the second user according to a second object set and a third object set, wherein the second object set comprises respective third objects evaluated by the first user, and the third object set comprises respective fourth objects evaluated by the second user; obtaining the target similarity according to the first similarity and the second similarity; obtaining a recommended object corresponding to the first user according to the target similarity.

2. The recommendation method of claim 1, wherein, the obtaining of the respective first sentiment parameters corresponding to the respective first comments and the respective second sentiment parameters corresponding to the respective second comments according to the respective first comments and the respective second comments respectively comprises: inputting the respective first comments and the respective second comments into a pre-set model trained respectively to obtain the respective first sentiment parameters and the respective second sentiment parameters output by the pre-set model.

3. The recommendation method of claim 1, wherein, the obtaining of the second similarity of the first user and the second user according to the second object set and the third object set comprises: obtaining respective first attribute parameters corresponding to the respective third objects and respective second attribute parameters corresponding to the respective fourth objects respectively; obtaining the second similarity according to the respective first attribute parameters and the respective second attribute parameters.

4. The recommendation method of claim 1, wherein, the obtaining of the target similarity according to the first similarity and the second similarity comprises: The target similarity is obtained according to the first similarity, the first similarity corresponding first preset weight, the second similarity and the second similarity corresponding second preset weight.

5. The recommendation method of claim 4, wherein, The method further comprises: The first preset weight and the second preset weight are determined according to subjective weighting method and objective weighting method.

6. The recommendation method of claim 1, wherein, The method further comprises: The target object corresponding to the target user is obtained, and the recommended object is obtained according to the target object. The target similarity between the first user and at least one second user is obtained respectively, and a target user is determined according to each target similarity.

7. A terminal, characterized by comprising: The terminal comprises a processor and a storage medium connected with the processor, the storage medium is adapted to store a plurality of instructions, and the processor is adapted to call the instructions in the storage medium to execute the steps of the recommendation method according to any one of claims 1-6.

8. A storage medium, characterized by 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 steps of the recommendation method according to any one of claims 1-6.

Images