CN111159563A - Method, device and equipment for determining user interest point information and storage medium - Google Patents
Method, device and equipment for determining user interest point information and storage medium Download PDFInfo
- Publication number
- CN111159563A CN111159563A CN201911418383.3A CN201911418383A CN111159563A CN 111159563 A CN111159563 A CN 111159563A CN 201911418383 A CN201911418383 A CN 201911418383A CN 111159563 A CN111159563 A CN 111159563A
- Authority
- CN
- China
- Prior art keywords
- determining
- item
- interest point
- vector
- information
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING OR CALCULATING; COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/90—Details of database functions independent of the retrieved data types
- G06F16/95—Retrieval from the web
- G06F16/953—Querying, e.g. by the use of web search engines
- G06F16/9535—Search customisation based on user profiles and personalisation
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D10/00—Energy efficient computing, e.g. low power processors, power management or thermal management
Landscapes
- Engineering & Computer Science (AREA)
- Databases & Information Systems (AREA)
- Theoretical Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the invention provides a method, a device, equipment and a storage medium for determining user interest point information, wherein the determining method comprises the following steps: obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item; determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector; and determining each interest point vector as the interest point information of the target user. According to the technical scheme, the interest degree of the user on each seed item is effectively determined by combining the behavior feedback information of the user on different seed items through information processing on the premise of not increasing the processing time, so that the user interest point information with high accuracy is obtained, and further, the stickiness between the product and the user is effectively improved in product recommendation based on the user interest point information with high accuracy.
Description
Technical Field
The present invention relates to the field of information processing technologies, and in particular, to a method, an apparatus, a device, and a storage medium for determining user interest point information.
Background
With the rapid development of the internet field, the explosive growth of information causes that it is more and more difficult for users to obtain interested effective content, and personalized recommendation for users becomes a basic technology which is not available in the internet field, and plays an increasingly important role in recommendation of products such as news, short videos, music and the like.
Generally, personalized recommendation for a user often needs to determine interest points of the user according to historical behaviors of the user, so that effective recommendation of products in which the user is interested is performed. In the traditional determination of the user interest point information, interest determination is mainly performed on the basis of items browsed or purchased by a user, so that similar items can be searched according to the determined interest points and recommended to the user.
However, in the conventional determination of the user interest point information, the basic attribute information of items purchased or browsed by the user is mainly considered for determination, but a certain deviation exists between the interest point determined according to the characteristic information and the actual interest and preference of the user, and information recommendation based on the user interest point with low accuracy is often provided with low recommendation accuracy, so that the stickiness between a product and the user is reduced.
Disclosure of Invention
The embodiment of the invention provides a method, a device, equipment and a storage medium for determining user interest point information, which improve the effective determination of the user interest point information.
In a first aspect, an embodiment of the present invention provides a method for determining user interest point information, including:
obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item;
determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector;
and determining each interest point vector as the interest point information of the target user.
In a second aspect, an embodiment of the present invention provides an apparatus for determining user interest point information, including:
the basic information determining module is used for acquiring at least one seed item corresponding to a target user and determining a behavior feedback vector of the target user relative to each seed item;
an interest point vector determining module, configured to determine at least one interest point vector corresponding to the target user according to the basic feature information of each seed entry and the corresponding behavior feedback vector;
and the target information determining module is used for determining each interest point vector as the interest point information of the target user.
In a third aspect, an embodiment of the present invention provides a device for determining user interest point information, where the device includes:
one or more processors;
storage means for storing one or more programs;
the one or more programs are executed by the one or more processors to cause the one or more processors to implement the methods provided by the above-described embodiments of the present invention.
In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a computer program is stored, and the computer program, when executed by a processor, implements the method provided by the above-mentioned embodiment of the present invention.
The embodiment of the invention provides a method, a device, equipment and a storage medium for determining user interest point information, wherein the determining method comprises the following steps: obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item; determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector; and determining each interest point vector as the interest point information of the target user. According to the technical scheme, the interest degree of the user on each seed item is effectively determined by combining the behavior feedback information of the user on different seed items through information processing on the premise of not increasing the processing time, so that the user interest point information with high accuracy is obtained, and further, the stickiness between the product and the user is effectively improved in product recommendation based on the user interest point information with high accuracy.
Drawings
Fig. 1 is a schematic flowchart of a method for determining user interest point information according to an embodiment of the present invention;
fig. 2 is a schematic flowchart of a method for determining user interest point information according to a second embodiment of the present invention;
FIG. 3 is a flowchart illustrating an exemplary determination of a user interest point vector according to a second embodiment of the present invention;
FIG. 4 is a flowchart illustrating the determination and recommendation of candidate entries according to a second embodiment of the present invention;
FIG. 5 is a flow chart illustrating the determination of a recommendation score in a second alternative embodiment of the present invention;
fig. 6 is a block diagram of a structure of a device for determining information about a point of interest of a user according to a third embodiment of the present invention;
fig. 7 is a schematic diagram of a hardware structure of a device for determining user interest point information according to a fourth embodiment of the present invention.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention more apparent, embodiments of the present invention will be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are only some embodiments of the invention, and not all embodiments. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention. In addition, the embodiments and features of the embodiments of the present invention may be combined with each other without conflict, and the respective embodiments may be mutually referred to and cited.
It should be noted that the application scenarios in the embodiments of the present invention are in various information recommendation fields with resource recommendation requirements, and are exemplarily applicable to recommendation scenarios of resource items such as news, consultation, music, and short video. In the existing recommendation performed by determining the user interest point information, the user interest is often determined only by considering the basic characteristics of the resource items accessed or browsed by the user, so that the accuracy of the determined user interest point is low, and the recommendation accuracy of resource item recommendation is influenced.
According to the method for determining the user interest point information, provided by the embodiment of the invention, besides the self basic characteristic information of the resource items accessed by the user is considered, the behavior feedback vector of the user relative to the resource items is further added, so that the determined user interest point information can reflect the preference degree difference of the user and express the user intention more accurately, and the accuracy of resource recommendation by adopting the user interest point information is higher.
Example one
Fig. 1 is a schematic flowchart of a method for determining user interest point information according to an embodiment of the present invention, where the method may be implemented by a device for determining user interest point information, where the device may be implemented by software and/or hardware, and may be generally integrated in a device for determining user interest point information, where the device for determining user interest point information is equivalent to an execution carrier device of the method for determining user interest point information, and specifically may be a background server with a data processing function and the like for performing service support.
As shown in fig. 1, a method for determining user interest point information according to an embodiment of the present invention specifically includes the following operations:
s101, obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item.
In this embodiment, an object (or audience) of information to be promoted in any information recommendation field with resource recommendation needs is taken as a target user in the embodiment of the present invention. The seed entry can be understood as a resource entry accessed by the target user in any information publishing platform supporting the information recommendation function. In this step, at least one seed entry that the target user has accessed may be obtained.
In addition, behavior feedback information generated by performing various operations on the seed items when the target user accesses each seed item can be acquired. For example, the behavior feedback information may be user behavior information such as whether to click, whether to like, whether to share, whether to comment, whether to collect, and the like. It can be understood that a common information publishing platform generally has a user behavior data acquiring or recording function, so that user behavior data can be acquired, for example, the user behavior data can be acquired through page tracking and event tracking at a webpage end, and the user behavior data can be acquired through a mode of counting key behaviors of a user by implanting corresponding codes at a mobile end.
In this embodiment, the behavior feedback vector may be specifically understood as a vector obtained through behavior feedback information generated when the target user accesses the seed entry, and may be specifically used to characterize a user behavior of the target user with respect to the seed entry. Illustratively, this step may determine the behavior feedback vector by encoding the behavior feedback information and by vectorizing the encoded information.
Optionally, the coding information corresponding to the behavior feedback information may be obtained by searching a given behavior feedback coding table, where the behavior feedback coding table includes each behavior feedback information and corresponding coding information. Or, the coding information corresponding to the behavior feedback information may also be obtained in a one-hot (one-hot) coding manner, which is optional.
S102, determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector.
In this embodiment, the basic characteristic information may be understood as information for characterizing the basic characteristic attribute of the seed entry itself. Illustratively, taking a seed entry as a video, the basic feature information of the video may include: the video's video entry number, video poster number, access user tag, and access user's home location, etc. The interest point vector can be understood as a vector which is extracted from the basic feature information and the behavior feedback vector corresponding to each seed entry and reflects the interest point information of the target user.
In this embodiment, the basic feature information based on the seed entry may be characterized in the form of a basic feature vector, and in this step, the basic feature vector characterizing the basic feature information may be spliced with a behavior feedback vector of the user with respect to the seed entry, and the target user interest point vector is determined by using the spliced vector.
For example, the step may obtain the interest point vector for the target through processing the vector formed after the stitching, and specifically, the step may capture an internal correlation between a basic feature vector and a behavior feedback vector in the vector formed after the stitching through a self-attention network model, thereby obtaining an attention matrix output from the attention network model, where the attention matrix represents a correlation between the attention of the user and the seed entry itself. And finally, summarizing the interest point vectors relative to all the seed items through the interest points corresponding to the seed items by the user, so as to represent the interest points of the user through the interest point vectors.
It should be noted that, in this step, at least one interest point vector may be determined, where the number of the determined interest point vectors may specifically be related to the number of parameter groups of analysis parameters used in analyzing the attention matrix, if only one group of analysis parameters is used for analysis, one interest point vector representing an interest point of the user may be obtained, and if multiple groups of analysis parameters are used, the same number of interest point vectors as the number of the group may be obtained. Each interest point vector obtained in the step can be understood as the representation of the interest point of the user, but because the vector value in each interest point vector is different, the preference of the represented interest point is different.
Compared with the existing method for determining the target user interest point information purely according to the basic characteristic information of the seed entry accessed by the target user, the method provided by the embodiment further takes the behavior feedback information generated when the target user accesses the seed entry as the basis for determining the target user interest point information on the basis of the prior technical scheme, so that the determined user interest point information is more accurate, and the accuracy is higher when the resource recommendation is performed on the corresponding target user according to the user interest point information.
S103, determining each interest point vector as the interest point information of the target user.
In this embodiment, all the determined interest point vectors are used as the interest point information of the target user, and in this embodiment, a plurality of interest point vectors are considered as the interest point information, which is equivalent to acquiring a plurality of interest point preference features of the target user, so that it can be ensured that the determined interest point information more comprehensively represents the interest range of the target user.
According to the method for determining the user interest points, provided by the embodiment of the invention, the interest degree of the user on each seed item is effectively determined by combining the behavior feedback information of the user on different seed items through information processing on the premise of not increasing the processing time, so that the user interest point information with high accuracy is obtained, and further, the stickiness between the product and the user is effectively improved in product recommendation based on the user interest point information with high accuracy.
Example two
Fig. 2 is a schematic flow chart of a method for determining user interest point information according to a second embodiment of the present invention, where the second embodiment is optimized based on the above-mentioned second embodiment, in this embodiment, a behavior feedback vector that determines that the target user corresponds to each seed entry is further optimized as follows: aiming at each seed item, acquiring all behavior feedback information of the target user relative to the seed item; and searching coding information corresponding to each behavior feedback information, and processing each coding information by adopting a dense vector determination mechanism to obtain a behavior feedback vector with a set dimensionality.
Meanwhile, the method further specifically optimizes at least one interest point vector corresponding to the target user determined according to each attention moment array as follows: inputting at least one given first fully-connected network model by taking the whole attention moment array as input data to obtain interest point weight vectors and interest point projection matrixes which are correspondingly output by the first fully-connected network models; determining product vectors of interest point weight vectors and interest point projection matrixes in the first fully-connected networks, and determining the product vectors as interest point vectors corresponding to the target users in the first fully-connected networks; each first full-connection network model is a network model with the same network structure but different full-connection parameters; the total number of the interest point vectors of the target user is the same as the number of the models of the first fully-connected network model.
As shown in fig. 2, a method for determining user interest point information according to a second embodiment of the present invention specifically includes the following steps:
s201, at least one seed item corresponding to the target user is obtained.
For example, at least one seed entry that the target user has accessed may be retrieved from the associated data resource wing.
S202, aiming at each seed entry, all behavior feedback information of the target user relative to the seed entry is obtained.
For example, this step may further obtain all behavior feedback information generated when the target user accesses each seed item from the related data monitoring platform.
Wherein the behavior feedback information of the target user relative to various sub-items comprises at least one of the following: praise behavior feedback, share behavior feedback, comment behavior feedback, and collect behavior feedback.
In this embodiment, the praise behavior feedback is feedback information indicating whether the target user performs praise operation on each seed entry; the sharing behavior feedback is feedback information of whether the target user carries out sharing operation on each seed item; the comment behavior feedback is feedback information of whether the target user carries out comment operation on each seed item; and the collection behavior feedback is feedback information of whether the target user performs collection operation on each seed item.
It can be understood that when a target user accesses or browses a seed item, if a praise operation is performed, it can be understood that the target user has a certain interest preference for the seed item; if the sharing operation is carried out, the target user can be understood to have stronger interest preference on the seed item; if the comment operation is performed, whether the target user is interested in the seed item can be further obtained through comment content classification, for example, the preference degree of the target user for the seed item is obtained according to whether the comment content given by the target user is good comment, bad comment or middle comment; if the collection operation is performed, it can be understood that the target user desires to access or browse the seed item again or even pay long attention, i.e., the target user is very interested in the seed item.
S203, searching coding information corresponding to each behavior feedback information, and processing each coding information by adopting a dense vector determination mechanism to obtain a behavior feedback vector with a set dimensionality.
In this embodiment, behavior feedback information, such as whether behavior information is approved or not, whether behavior information is forwarded or not, and whether behavior information is collected or not, generated when the target user accesses the seed entry may be displayed in a coded form. For example, it may be determined that the coding corresponding to the praise in the behavior feedback information is 000, and the coding corresponding to the favorite is 100.
After obtaining the codes of the behavior feedback information, the embodiment may process each code information by adopting a dense vector determination mechanism to obtain a behavior feedback vector of a set dimension; wherein the dense vector is understood to be a form of vector in which each dimension element of the vector is represented by a double precision floating point type array. Dense vectors are a form of vector representation relative to sparse vectors, illustratively, for vectors (1.0, 0.0, 3.0) represented in dense vector form as [1.0, 0.0, 3.0], and in sparse vector form as (3, [0, 2], [1.0, 3.0]), where 3 is the length of the vector, [0, 2] is the index value of the non-0 dimension in the vector, representing that two elements at positions 0, 2 are non-zero values, and [1.0, 3.0] is the value of the array element in index; the difference between the two is that the dense vector stores all values, including zero values, while the sparse vector stores the index positions and corresponding non-zero values.
It should be noted that the dense vector is only a vector representation form, in this embodiment, after the coding information corresponding to each behavior feedback information of the target user for the seed entry is obtained, the corresponding coding information is generally in a sparse matrix form, and when the amount of the coding information is large, the order of the corresponding sparse matrix is very high, which results in that resources are very occupied during data processing. For this reason, an effective solution is to perform dimension reduction processing on the entire sparse matrix through an embedding (embedding) layer, thereby realizing densification of the encoded information. The dimension reduction principle of the embedded layer is to multiply the sparse matrix and a set mapping matrix, so that the dimension reduction and the densification of the sparse matrix are realized.
Meanwhile, in this embodiment, the set dimension may be determined according to the type of the set behavior feedback information, that is, one dimension corresponds to one behavior feedback information, and one behavior feedback information may represent one characteristic of the behavior fed back by the target user to the seed item.
In this embodiment, the following steps S204 to S207 show a process of determining the interest point vector based on the basic feature information of the seed entry and the behavior feedback vector of the target user.
And S204, constructing a basic feature vector according to the basic feature information of each seed entry.
In this embodiment, for each seed entry, the process of converting the basic feature information of the seed entry into one basic feature vector may be implemented by using a similar method to obtain a corresponding behavior feedback vector according to the behavior feedback information corresponding to the seed entry, for example, the basic feature information of the seed entry may be encoded first, and then a dense vector determination mechanism is used to form a dense basic feature vector corresponding to each seed entry.
S205, splicing the basic characteristic vector and the behavior feedback vector of each seed item to form a characteristic splicing vector of each seed item.
It can be understood that one main purpose of this embodiment is to combine the behavior feedback information generated when the target user accesses the seed item with the basic feature information of the corresponding seed item accessed by the target user to obtain the concatenation information corresponding to each seed item, and then extract the interest point location of the target user on the corresponding seed item and the interest point information such as the corresponding preference degree from each concatenation information. For example, the element values in the behavior feedback vector may be directly spliced into the basic feature vector in sequence in this step, so as to form a feature splicing vector.
And S206, sequentially using the feature splicing vectors as input data, inputting a given self-attention network model, and obtaining an attention matrix corresponding to each seed item output.
It can be understood that the self-attention network model is a network model that can capture internal correlations of data or features, and different behavior feedback information generated by the target user for the accessed seed entry in this embodiment substantially reflects different degrees of attention invested by the target user for the accessed seed entry.
Specifically, in this step, the feature concatenation vector corresponding to each seed entry may be respectively input to the self-attention network model as input data, and thus after the processing of the self-attention network model, an attention matrix corresponding to the output of the network model may be obtained, where the attention matrix is used to represent the attention of the user with respect to the seed entry.
And S207, determining at least one interest point vector corresponding to the target user according to each attention moment array.
In this embodiment, the interest point vectors may represent interest points of the target user with respect to all the seed entries, and in this step, the interest point vectors may be determined by analyzing the attention matrix corresponding to each seed entry, and the interest point vectors having the same number as the set number may be obtained according to the set number of the analysis parameters.
Specifically, in this embodiment, it can be considered that one interest point vector is equivalent to a summary of interest features of the target user with respect to each seed entry, and the interest features of the target user with respect to each seed entry are actually included in the attention matrix of each seed entry. In the step, a full-connection network model capable of performing feature extraction and information integration can be adopted to analyze the attention matrix of the seed entry, the adopted full-connection network model is a network model with a fixed network structure and variable full-connection parameters, and therefore a plurality of full-connection network models with different full-connection parameters can be actually adopted to analyze each attention matrix in the step, and finally an interest point vector relative to a target user is obtained from each full-connection network.
Further, fig. 3 shows an exemplary flowchart of determining a user interest point vector in the second embodiment of the present invention, and as shown in fig. 3, the determining at least one interest point vector corresponding to the target user according to each attention moment array specifically includes the following operations:
and S2071, inputting at least one given first fully-connected network model by taking the whole attention moment matrix as input data, and obtaining interest point weight vectors and interest point projection matrixes correspondingly output by the first fully-connected network model.
Each first full-connection network model is a network model with the same network structure but different full-connection parameters. Specifically, the attention matrix of each seed entry is input into each first fully-connected network model as a whole, or a whole attention matrix is formed by splicing all the attention matrices in rows to serve as input data, wherein each attention matrix in the input data input into the fully-connected network model corresponds to one seed entry, and in the processing based on the first fully-connected network model, it can be considered that firstly, the feature of each attention matrix in the input data is extracted, so that the interest point feature representing a target user relative to the corresponding seed entry can be extracted relative to each attention matrix, and each interest point feature can be represented in the form of a multi-dimensional matrix; then, the weight value occupied by each interest point feature in all the interest point features can be integrally determined, and the multidimensional matrix representing each interest feature can be subjected to dimension reduction processing to obtain a corresponding interest feature vector.
After the first fully-connected network model processes the input data, an interest point weight vector and an interest point projection matrix may be finally obtained, where the interest point weight vector may be regarded as a set of the determined weight values, the number of element values included in the interest point weight vector is the same as the total number of the seed entries, and the interest point projection matrix may be regarded as a set of the interest feature vectors, and the number of row vectors included in the interest point projection matrix is the same as the total number of the seed entries. The interest point weight matrix and the interest point projection matrix obtained in the step represent interest point features of the target user relative to all the seed items and the weight occupied by each interest point feature. In this embodiment, the subjectivity of artificially setting the weight of each seed item according to the behavior feedback information of the target user for each seed item can be effectively avoided by obtaining the interest point weight vector.
S2072, determining product vectors of the interest point weight vectors and the interest point projection matrices in the first fully-connected networks, and determining each product vector as an interest point vector corresponding to the target user in each of the first fully-connected networks.
Illustratively, by multiplying the interest point weight vector in each first fully-connected network by the interest point projection matrix, weighted accumulation of the interest point information of the target user embodied by all the seed entries can be realized, and the obtained interest point vector can fully embody cross information between the interest points of the target user embodied by each seed entry, so that the purpose of more accurately predicting the interest point information of the target user is achieved.
It is understood that the total number of the interest point vectors of the target user is the same as the number of the models of the first fully-connected network model. In the implementation of the step S2071 in this embodiment, it is considered that a plurality of first fully-connected network models with the same network structure and different fully-connected parameters are adopted, and a weight vector of an interest point and a projection matrix of an interest point are obtained for each first fully-connected network, and through the operation in this step, the same number of interest point vectors as the first fully-connected network models can be obtained.
S208, determining each interest point vector as the interest point information of the target user.
In this embodiment, the interest point vector may further be understood as the interest point information of the target user, which is commonly embodied by each seed entry. And the interest point information of the target user consisting of the plurality of interest point vectors can be obtained by obtaining the interest point vectors obtained by the plurality of different first fully-connected network models.
According to the method for determining the user interest point information, provided by the embodiment of the invention, the interest degree of the user on each seed item is effectively determined by combining the behavior feedback information of the user on different seed items through information processing on the premise of not increasing the processing time, so that the user interest point information with high accuracy is obtained, and further, the stickiness between the product and the user is effectively improved in product recommendation based on the user interest point information with high accuracy.
As an optional embodiment of the present invention, the optional embodiment further optimizes including: and determining candidate items from a given item set to be recommended according to the interest point information of the target user and recommending the candidate items to the target user. This step may be specifically executed after S208 described above in this embodiment.
It can be understood that the essence of the determination of the user interest point information lies in resource recommendation, and the present embodiment may predict the subsequent behavior of the user through the user interest point information, so as to discover the resource in which the user is interested.
In this embodiment, the set of items to be recommended may be understood as a set of resource items including all intention recommendation items provided for the target user, where the intention recommendation items included in the set of items to be recommended may be determined according to specific resource recommendation requirements of the resource recommendation platform. The candidate item can be understood as a resource item selected from the item set to be recommended and actually recommended to the target user. Optionally, the candidate items may be determined by scoring all intention recommendation items in the set of items to be recommended, and selecting a set number of intention recommendation items as the candidate items actually recommended to the target user according to the scores.
Further, fig. 4 shows a schematic flow chart of candidate entry determination and recommendation in the second embodiment of the present invention, and as shown in fig. 4, the determining and recommending a candidate entry from a given set of entries to be recommended to the target user according to the interest point information of the target user specifically includes the following operations:
s301, aiming at each item to be recommended in the item set to be recommended, determining an item basic feature vector of the item to be recommended.
In this embodiment, the item to be recommended is an intention recommendation item for the target user, which is determined according to a specific resource recommendation requirement of the resource recommendation platform in the item set to be recommended. The entry basic feature vector may be determined according to the basic feature information of the entry to be recommended, and the specific method for obtaining the entry basic feature vector may refer to the method for obtaining the basic feature vector corresponding to the seed entry, which is not described herein again.
S302, determining cosine similarity values of the entry basic feature vectors relative to the interest point vectors in the interest point information.
In this embodiment, the cosine similarity value is specifically a cosine value of an included angle between the entry basic feature vector and each interest point vector in the interest point information. Cosine similarity measures the difference between two individuals by using the cosine value of the included angle between two vectors in a vector space, and the closer the cosine value is to 1, the closer the included angle is to 0 degree, namely the more similar the two vectors are, which is also called as cosine similarity.
It can be understood that by determining the cosine similarity value of the item basic feature vector relative to each interest point vector in the interest point information, it can be known which item to be recommended is closer to the interest point of the user, and thus, the item to be recommended can be used as an effective basis for resource recommendation of a target user, thereby improving the accuracy of resource recommendation.
S303, determining a recommendation score corresponding to the item to be recommended according to each cosine similarity value.
It can be understood that, after each cosine similarity value, in order to further improve the accuracy of resource recommendation, the cosine similarity value may be further combined with the item to be recommended and some features of the target user, so as to obtain a more accurate recommendation score of the item to be recommended with respect to the target user.
Further, fig. 5 is a schematic flow chart illustrating a process of determining a recommendation score in a second optional embodiment of the present invention, and specifically, as shown in fig. 5, determining a recommendation score corresponding to the to-be-recommended item according to each cosine similarity value includes the following operations:
s3031, determining the user basic feature vector of the target user and the additional feature vector of the item to be recommended.
In this embodiment, in order to obtain a recommendation score of an item to be recommended with respect to a target user, a user basic feature vector corresponding to user basic feature information of the target user and an additional feature vector of the item to be recommended may be determined first.
The user basic feature information may be understood as feature information for characterizing the target user, and exemplarily, the user basic feature information includes a gender, an age, an access account ID, an access device location, and the like of the target user. The user basic feature vector may be understood as a vector obtained by performing dimension reduction and densification processing on the user basic information. The additional feature vector may be understood as a feature attribute or a distinguishing feature attribute characterizing each item to be recommended, and the additional feature vector may be obtained by performing dimension reduction and densification processing on unique feature attribute information of the item to be recommended (for example, when the item to be recommended is a video, information such as resolution of the video may be regarded as the unique feature attribute information).
S3032, the user basic feature vector, the item basic feature vector and the additional feature vector of the item to be recommended and the cosine similarity values are spliced to form an item recommendation feature vector.
In this embodiment, the entry basic feature vector may be understood as a feature attribute that is common to all the entries to be recommended, and in this embodiment, the user basic feature vector, the entry basic feature vector and the additional feature vector of the entry to be recommended, and each cosine similarity value are associated and then scored, so as to jointly serve as an effective basis for resource recommendation for a target user, thereby further improving the accuracy of resource recommendation.
In this embodiment, for each item to be recommended, the item recommendation feature vector may be understood as the item recommendation feature vector obtained by sequentially connecting the user basic feature vector, the item basic feature vector and the additional feature vector of the item to be recommended, and the cosine similarity values in sequence.
S3033, inputting a given second fully-connected network model by taking the item recommendation feature vector as input data, and determining an output value of the second fully-connected network model as a recommendation score of the item to be recommended.
In this embodiment, the second fully-connected network model may be specifically configured to perform some form of weighted summation processing on the item physical examination feature vector, so as to obtain, for an item recommendation feature vector of an item to be recommended, a recommendation score of the recommended item with respect to a target user.
Specifically, the input layer of the second fully-connected network model is equivalent to include nodes with the same number of element values as the number of the item recommendation feature vectors, and meanwhile, the network model may include an implied layer, and the number of the nodes included in the implied layer and the fully-connected parameters of each input node relative to each node in the hidden layer may be predetermined, and finally, through the weighted summation processing performed by the second fully-connected network model, a weighted summation value may be output corresponding to each item recommendation feature vector, where the weighted summation value is equivalent to the recommendation score of the item to be recommended relative to the target user.
Specifically, an entry recommendation weight vector corresponding to the entry recommendation feature vector dimension may be given in the second fully-connected network model, and a product of the entry recommendation feature vector and the entry recommendation weight vector is obtained, so that the method is implemented to perform weighted summation on the user basic feature vector, the entry basic feature vector and the additional feature vector of the entry to be recommended, and each cosine similarity value included in the entry recommendation vector, and determine a recommendation score of the corresponding entry to be recommended according to a value obtained by the weighted summation.
It should be noted that, different from the first fully-connected network model, the second fully-connected network model is used to score the item recommendation vectors, and since the rules of the second fully-connected network model for scoring the item recommendation vectors are consistent, only one second fully-connected network model is required.
S304, determining at least one candidate item according to the recommendation score of each item to be recommended, and recommending each candidate item to the target user.
After the recommendation score of each item to be recommended is determined, the candidate items that can be recommended to the target user can be selected according to the recommendation score.
Specifically, the determining of the at least one candidate item according to the recommendation score of each item to be recommended may also be optimized to sort the items to be recommended from high to low according to the corresponding recommendation score, and use the items to be recommended that are sorted to a preset value as the candidate items; wherein the set value is greater than or equal to 1.
It can be understood that the recommendation score reflects the matching degree of the corresponding item to be recommended and the target user interest point information, and the higher the recommendation score is, the better the recommendation score is; the items to be recommended with the set number before the ranking of the high-selection recommendation score are selected as candidate items to be actually recommended to the target user, so that the high accuracy of resource recommendation can be ensured, and the requirement of a resource recommendation platform on the recommended number of the resource items can be met as much as possible.
The above technical implementation of the optional embodiment provides specific method steps for resource recommendation to the target user according to the interest point information of the target user, so that the method for determining the user interest point information provided by the embodiment of the present invention, on the basis of obtaining high-accuracy user interest point information, by determining an item basic feature vector and an additional feature vector of each item to be recommended in an item set to be recommended and a user basic feature vector of a target user, and determining cosine similarity values of the basic feature vectors of the entries relative to the interest point vectors in the interest point information, thereby, for each item to be recommended, based on the user basic feature vector, the item basic feature vector and the additional feature vector of the item to be recommended and the cosine similarity values, and scoring the items to be recommended, so that the degree of engagement between the items to be recommended and the target user interest point information can be more accurately judged. Therefore, the item to be recommended with high fitness is selected as the candidate item to be recommended to the target user, and the resource recommendation accuracy is further improved.
EXAMPLE III
Fig. 6 is a block diagram of a structure of a device for determining user interest point information according to a third embodiment of the present invention, where the device is suitable for determining an interest point of a target user, and the device may be implemented by software and/or hardware, and may be generally integrated in a device for determining user interest point information, where the device for determining user interest point information is equivalent to an execution carrier device of a method for determining user interest point information, and specifically may be a background server with a data processing function and the like for performing service support. As shown in fig. 6, the apparatus includes: a basic information determination module 31, a point of interest vector determination module 32, and a target information determination module 33.
The basic information determining module 31 is configured to obtain at least one seed entry corresponding to a target user, and determine a behavior feedback vector of the target user with respect to each seed entry;
an interest point vector determining module 32, configured to determine at least one interest point vector corresponding to the target user according to the basic feature information of each seed entry and the corresponding behavior feedback vector;
and a target information determining module 33, configured to determine each of the interest point vectors as the interest point information of the target user.
The device for determining the user interest point information provided by the third embodiment of the invention effectively determines the interest degree of the user on each seed item through information processing in combination with the behavior feedback information of the user on different seed items on the premise of not increasing the processing time, thereby obtaining the user interest point information with high accuracy, and further effectively improving the stickiness between the product and the user in product recommendation based on the user interest point information with high accuracy.
Further, the interest point vector determination module 32 may include:
the basic vector construction unit is used for constructing a basic feature vector according to the basic feature information of each seed entry;
a splicing vector determining unit, configured to splice the basic feature vector and the behavior feedback vector of each seed entry to form a feature splicing vector of each seed entry;
the information processing unit is used for inputting the given self-attention network model by taking the characteristic splicing vectors as input data in sequence to obtain an attention matrix output corresponding to each seed item;
and the vector information determining unit is used for determining at least one interest point vector corresponding to the target user according to each attention moment array.
Further, the vector information determining unit may be specifically configured to input, as input data, the entirety of each attention moment matrix, into a given at least one first fully-connected network model, and obtain an interest point weight vector and an interest point projection matrix that are output by the first fully-connected network model correspondingly; determining product vectors of interest point weight vectors and interest point projection matrixes in the first fully-connected networks, and determining the product vectors as interest point vectors corresponding to the target users in the first fully-connected networks; each first full-connection network model is a network model with the same network structure but different full-connection parameters; the total number of the interest point vectors of the target user is the same as the number of the models of the first fully-connected network model.
Further, the device may further include an information recommendation module, where the information recommendation module may be configured to determine candidate items from a given set of items to be recommended according to the point-of-interest information of the target user, and recommend the candidate items to the target user.
On the basis of the optimization, the information recommendation module may specifically include:
an item feature determining unit, configured to determine, for each item to be recommended in the set of items to be recommended, an item basic feature vector of the item to be recommended;
a similarity value determining unit, configured to determine a cosine similarity value of the entry basic feature vector with respect to each interest point vector in the interest point information;
a recommendation score determining unit, configured to determine, according to each cosine similarity value, a recommendation score corresponding to the item to be recommended;
and the candidate item recommending unit is used for determining at least one candidate item according to the recommendation score of each item to be recommended and recommending each candidate item to the target user.
Further, the recommendation score determining unit may be specifically configured to determine a user basic feature vector of the target user and an additional feature vector of the item to be recommended; splicing the user basic feature vector, the item basic feature vector and the additional feature vector of the item to be recommended and the cosine similarity values to form an item recommendation feature vector; and taking the item recommendation feature vector as input data, inputting a given second fully-connected network model, and determining an output value of the second fully-connected network model as a recommendation score of the item to be recommended.
Further, the candidate item recommending unit may be specifically configured to sort each item to be recommended from high to low according to the corresponding recommendation score, and use the item to be recommended that is ranked at a front set numerical value as the candidate item; wherein the set value is greater than or equal to 1.
Further, the basic information determining module 31 may be specifically configured to, for each seed entry, obtain all behavior feedback information of the target user with respect to the seed entry; and searching coding information corresponding to each behavior feedback information, and processing each coding information by adopting a dense vector determination mechanism to obtain a behavior feedback vector with a set dimensionality.
On the basis of the optimization, the behavior feedback information of the target user relative to various sub-items comprises at least one of the following: praise behavior feedback, share behavior feedback, comment behavior feedback, and collect behavior feedback.
Example four
Fig. 7 is a schematic diagram of a hardware structure of a device for determining user interest point information according to a fourth embodiment of the present invention. As shown in fig. 7, the determining device of the user interest point information may specifically include: a processor 40, a storage device 41, an input device 42, and an output device 43. The number of processors 40 in the device for determining the user's interest point information may be one or more, and one processor 40 is taken as an example in fig. 7. The number of the storage devices 41 in the device for determining the user's interest point information may be one or more, and one storage device 41 is taken as an example in fig. 7. The processor 40, the storage device 41, the input device 42 and the output device 43 of the apparatus for determining the user's interest point information may be connected by a bus or other means, and fig. 7 illustrates the connection by the bus as an example.
The storage device 41 is a computer-readable storage medium, and can be used to store software programs, computer-executable programs, and modules, such as program requests/modules corresponding to the determination method of the user interest point information and/or the image stitching method according to any embodiment of the present invention (for example, the basic information determination module 31, the interest point vector determination module 32, and the target information determination module 33 in the determination device of the user interest point information). The storage device 41 may mainly include a storage program area and a storage data area, wherein the storage program area may store an operating device, an application program required for at least one function; the storage data area may store data created from use of a determination device of user point of interest information, and the like. Further, the storage device 41 may include high speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other non-volatile solid state storage device. In some examples, the storage 41 may further include a memory remotely located from the processor 40, which may be connected to a user point of interest information determination device via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The input device 42 may be used to receive input numeric or character information and generate key signal inputs related to user settings and function controls of the user point of interest determination device, as well as a camera for capturing images and a determination device for capturing user point of interest information for audio in video data. The output means 43 may comprise a determination device for video user point of interest information, such as a display screen, and a determination device for audio user point of interest information, such as a speaker. It should be noted that the specific composition of the input device 42 and the output device 43 can be set according to actual conditions.
The processor 40 executes various functional applications and data processing of the device for determining user's point of interest information by running software programs, requests and modules stored in the storage means 41, i.e. implements the above-described method for determining user's point of interest information.
Specifically, in the embodiment, when the processor 40 executes one or more programs stored in the storage device 41, the following operations are specifically implemented: obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item; determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector; and determining each interest point vector as the interest point information of the target user.
An embodiment of the present invention further provides a computer-readable storage medium, where when a processor of a device for determining user interest point information executes a program in the storage medium, the device for determining user interest point information can execute the method for determining user interest point information according to the foregoing method embodiment. Illustratively, the method for determining the user interest point information includes: obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item; determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector; and determining each interest point vector as the interest point information of the target user.
It should be noted that, as for the embodiments of the apparatus, the device for determining user interest point information, and the storage medium, since they are basically similar to the embodiments of the method, the description is relatively simple, and for relevant points, reference may be made to the partial description of the embodiments of the method.
From the above description of the embodiments, it is obvious for those skilled in the art that the present invention can be implemented by software and necessary general hardware, and certainly, can also be implemented by hardware, but the former is a better embodiment in many cases. Based on such understanding, the technical solutions of the present invention may be embodied in the form of a software product, where the computer software product may be stored in a computer-readable storage medium, such as a floppy disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), a FLASH Memory (FLASH), a hard disk, or an optical disk of a computer, and includes a plurality of devices (which may be a robot, a personal computer, a server, or a network device) requesting to enable a user to perform the method for determining the user interest point information and/or the method for splicing images, according to any embodiment of the present invention.
It should be noted that, in the apparatus for determining information about a point of interest of a user, each unit and each module included in the apparatus are only divided according to functional logic, but are not limited to the above division, as long as the corresponding function can be implemented; in addition, specific names of the functional units are only for convenience of distinguishing from each other, and are not used for limiting the protection scope of the present invention.
It is to be noted that the foregoing is only illustrative of the preferred embodiments of the present invention and the technical principles employed. It will be understood by those skilled in the art that the present invention is not limited to the particular embodiments described herein, but is capable of various obvious changes, rearrangements and substitutions as will now become apparent to those skilled in the art without departing from the scope of the invention. Therefore, although the present invention has been described in greater detail by the above embodiments, the present invention is not limited to the above embodiments, and may include other equivalent embodiments without departing from the spirit of the present invention, and the scope of the present invention is determined by the scope of the appended claims.
Claims (12)
1. A method for determining user interest point information is characterized by comprising the following steps:
obtaining at least one seed item corresponding to a target user, and determining a behavior feedback vector of the target user relative to each seed item;
determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed item and the corresponding behavior feedback vector;
and determining each interest point vector as the interest point information of the target user.
2. The method of claim 1, wherein determining at least one interest point vector corresponding to the target user according to the basic feature information of each seed entry and the corresponding behavior feedback vector comprises:
constructing a basic feature vector according to the basic feature information of each seed entry;
splicing the basic characteristic vector and the behavior feedback vector of each seed item to form a characteristic splicing vector of each seed item;
sequentially using the feature splicing vectors as input data, inputting a given self-attention network model, and obtaining an attention matrix corresponding to each seed item output;
and determining at least one interest point vector corresponding to the target user according to each attention moment array.
3. The method of claim 2, wherein the determining at least one point of interest vector corresponding to the target user according to each attention moment array comprises:
inputting at least one given first fully-connected network model by taking the whole attention moment array as input data to obtain interest point weight vectors and interest point projection matrixes which are correspondingly output by the first fully-connected network models;
determining product vectors of interest point weight vectors and interest point projection matrixes in the first fully-connected networks, and determining the product vectors as interest point vectors corresponding to the target users in the first fully-connected networks;
each first full-connection network model is a network model with the same network structure but different full-connection parameters; the total number of the interest point vectors of the target user is the same as the number of the models of the first fully-connected network model.
4. The method of claim 1, further comprising:
and determining candidate items from a given item set to be recommended according to the interest point information of the target user and recommending the candidate items to the target user.
5. The method according to claim 4, wherein the determining candidate items from a given set of items to be recommended and recommending the candidate items to the target user according to the point-of-interest information of the target user comprises:
determining an item basic feature vector of the item to be recommended for each item to be recommended in the item set to be recommended;
determining cosine similarity values of the entry basic feature vectors relative to the interest point vectors in the interest point information;
determining a recommendation score corresponding to the item to be recommended according to each cosine similarity value;
and determining at least one candidate item according to the recommendation score of each item to be recommended, and recommending each candidate item to the target user.
6. The method according to claim 5, wherein determining the recommendation score corresponding to the item to be recommended according to each cosine similarity value comprises:
determining a user basic feature vector of the target user and an additional feature vector of the item to be recommended;
splicing the user basic feature vector, the item basic feature vector and the additional feature vector of the item to be recommended and the cosine similarity values to form an item recommendation feature vector;
and taking the item recommendation feature vector as input data, inputting a given second fully-connected network model, and determining an output value of the second fully-connected network model as a recommendation score of the item to be recommended.
7. The method of claim 5, wherein the determining at least one candidate item according to the recommendation score of each item to be recommended comprises:
sorting the items to be recommended from high to low according to the corresponding recommendation scores, and taking the items to be recommended which are sorted in a front set numerical value as candidate items;
wherein the set value is greater than or equal to 1.
8. The method of any one of claims 1-7, wherein determining a behavior feedback vector for the target user with respect to each of the seed entries comprises:
aiming at each seed item, acquiring all behavior feedback information of the target user relative to the seed item;
and searching coding information corresponding to each behavior feedback information, and processing each coding information by adopting a dense vector determination mechanism to obtain a behavior feedback vector with a set dimensionality.
9. The method of claim 8, wherein the feedback information of the target user's behavior with respect to various sub-entries comprises at least one of: praise behavior feedback, share behavior feedback, comment behavior feedback, and collect behavior feedback.
10. An apparatus for determining point of interest information of a user, comprising:
the basic information determining module is used for acquiring at least one seed item corresponding to a target user and determining a behavior feedback vector of the target user relative to each seed item;
an interest point vector determining module, configured to determine at least one interest point vector corresponding to the target user according to the basic feature information of each seed entry and the corresponding behavior feedback vector;
and the target information determining module is used for determining each interest point vector as the interest point information of the target user.
11. A device for determining point of interest information of a user, comprising:
one or more processors;
storage means for storing one or more programs;
the one or more programs are executable by the one or more processors to cause the one or more processors to implement the method of any one of claims 1-9.
12. A computer storage medium, comprising: the program when executed by a processor implementing the method of any one of claims 1 to 9.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418383.3A CN111159563B (en) | 2019-12-31 | 2019-12-31 | Method, device, equipment and storage medium for determining user interest point information |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911418383.3A CN111159563B (en) | 2019-12-31 | 2019-12-31 | Method, device, equipment and storage medium for determining user interest point information |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111159563A true CN111159563A (en) | 2020-05-15 |
| CN111159563B CN111159563B (en) | 2024-02-09 |
Family
ID=70560472
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911418383.3A Active CN111159563B (en) | 2019-12-31 | 2019-12-31 | Method, device, equipment and storage medium for determining user interest point information |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111159563B (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112559901A (en) * | 2020-12-11 | 2021-03-26 | 百度在线网络技术（北京）有限公司 | Resource recommendation method and device, electronic equipment, storage medium and computer program product |
| CN113033560A (en) * | 2021-04-23 | 2021-06-25 | 中国建设银行股份有限公司 | Interest point analysis method and device |
| CN113407849A (en) * | 2021-07-15 | 2021-09-17 | 百果园技术(新加坡)有限公司 | Coarse sorting method and device |
| CN113626686A (en) * | 2021-07-19 | 2021-11-09 | 上海泛宥信息科技有限公司 | Automatic pushing method and device based on user data analysis and computer equipment |
| CN113704587A (en) * | 2021-08-31 | 2021-11-26 | 中国平安财产保险股份有限公司 | User adhesion analysis method, device, equipment and medium based on stage division |
| CN114385930A (en) * | 2021-12-28 | 2022-04-22 | 清华大学 | Interest point recommendation method and system |
| CN115190357A (en) * | 2022-07-05 | 2022-10-14 | 三星电子（中国）研发中心 | A method and device for generating a video summary |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108804619A (en) * | 2018-05-31 | 2018-11-13 | 腾讯科技（深圳）有限公司 | Interest preference prediction technique, device, computer equipment and storage medium |
| CN109299327A (en) * | 2018-11-16 | 2019-02-01 | 广州市百果园信息技术有限公司 | Video recommendation method, device, equipment and storage medium |
| CN109783727A (en) * | 2018-12-24 | 2019-05-21 | 东软集团股份有限公司 | Retrieve recommended method, device, computer readable storage medium and electronic equipment |
| CN110119477A (en) * | 2019-05-14 | 2019-08-13 | 腾讯科技（深圳）有限公司 | A kind of information-pushing method, device and storage medium |
| CN110162703A (en) * | 2019-05-13 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Content recommendation method, training method, device, equipment and storage medium |
| CN110347776A (en) * | 2019-07-17 | 2019-10-18 | 北京百度网讯科技有限公司 | Interest point name matching process, device, equipment and storage medium |
-
2019
- 2019-12-31 CN CN201911418383.3A patent/CN111159563B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108804619A (en) * | 2018-05-31 | 2018-11-13 | 腾讯科技（深圳）有限公司 | Interest preference prediction technique, device, computer equipment and storage medium |
| CN109299327A (en) * | 2018-11-16 | 2019-02-01 | 广州市百果园信息技术有限公司 | Video recommendation method, device, equipment and storage medium |
| CN109783727A (en) * | 2018-12-24 | 2019-05-21 | 东软集团股份有限公司 | Retrieve recommended method, device, computer readable storage medium and electronic equipment |
| CN110162703A (en) * | 2019-05-13 | 2019-08-23 | 腾讯科技（深圳）有限公司 | Content recommendation method, training method, device, equipment and storage medium |
| CN110119477A (en) * | 2019-05-14 | 2019-08-13 | 腾讯科技（深圳）有限公司 | A kind of information-pushing method, device and storage medium |
| CN110347776A (en) * | 2019-07-17 | 2019-10-18 | 北京百度网讯科技有限公司 | Interest point name matching process, device, equipment and storage medium |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112559901A (en) * | 2020-12-11 | 2021-03-26 | 百度在线网络技术（北京）有限公司 | Resource recommendation method and device, electronic equipment, storage medium and computer program product |
| CN112559901B (en) * | 2020-12-11 | 2022-02-08 | 百度在线网络技术（北京）有限公司 | Resource recommendation method and device, electronic equipment, storage medium and computer program product |
| CN113033560A (en) * | 2021-04-23 | 2021-06-25 | 中国建设银行股份有限公司 | Interest point analysis method and device |
| CN113407849A (en) * | 2021-07-15 | 2021-09-17 | 百果园技术(新加坡)有限公司 | Coarse sorting method and device |
| CN113407849B (en) * | 2021-07-15 | 2024-06-14 | 百果园技术(新加坡)有限公司 | Coarse ordering method and device |
| CN113626686A (en) * | 2021-07-19 | 2021-11-09 | 上海泛宥信息科技有限公司 | Automatic pushing method and device based on user data analysis and computer equipment |
| CN113704587A (en) * | 2021-08-31 | 2021-11-26 | 中国平安财产保险股份有限公司 | User adhesion analysis method, device, equipment and medium based on stage division |
| CN113704587B (en) * | 2021-08-31 | 2023-06-06 | 中国平安财产保险股份有限公司 | User adhesion analysis method, device, equipment and medium based on stage division |
| CN114385930A (en) * | 2021-12-28 | 2022-04-22 | 清华大学 | Interest point recommendation method and system |
| CN114385930B (en) * | 2021-12-28 | 2024-12-10 | 清华大学 | A method and system for recommending points of interest |
| CN115190357A (en) * | 2022-07-05 | 2022-10-14 | 三星电子（中国）研发中心 | A method and device for generating a video summary |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111159563B (en) | 2024-02-09 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110321422B (en) | Method for training model on line, pushing method, device and equipment | |
| CN112889042B (en) | Identification and application of hyperparameters in machine learning | |
| CN109902708B (en) | Recommendation model training method and related device | |
| CN111159564B (en) | Information recommendation method, device, storage medium and computer equipment | |
| CN111159563B (en) | Method, device, equipment and storage medium for determining user interest point information | |
| CN110909182B (en) | Multimedia resource searching method, device, computer equipment and storage medium | |
| CN109753601B (en) | Method and device for determining click rate of recommended information and electronic equipment | |
| CN113688310B (en) | Content recommendation method, device, equipment and storage medium | |
| CN109993627B (en) | Recommendation method, training method, device and storage medium of recommendation model | |
| CN108304429A (en) | Information recommendation method, device and computer equipment | |
| CN110008397A (en) | A kind of recommended models training method and device | |
| CN104615723B (en) | The determination method and apparatus of query word weighted value | |
| CN113742578A (en) | Data recommendation method and device, electronic equipment and storage medium | |
| WO2024113641A1 (en) | Video recommendation method and apparatus, and electronic device, computer-readable storage medium and computer program product | |
| CN113297471A (en) | Method and device for generating data object label and searching data object and electronic equipment | |
| CN114548401B (en) | Recommendation model training method, device, computing equipment and medium | |
| CN113204703B (en) | Material recommendation method, device, equipment and storage medium | |
| CN113570432B (en) | Resource recommendation determining method, device, equipment and storage medium | |
| CN112035740B (en) | Project usage duration prediction method, device, equipment and storage medium | |
| CN111651643B (en) | Candidate content processing method and related equipment | |
| CN110569447B (en) | Network resource recommendation method and device and storage medium | |
| CN113076450A (en) | Method and device for determining target recommendation list | |
| CN117056619A (en) | Method and device for determining user behavior characteristics | |
| CN112818258B (en) | Social network user searching method based on keywords, computer device and computer readable storage medium | |
| CN117216361A (en) | Recommendation method, recommendation device, electronic equipment and computer readable storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |