CN113947455B - Data recommendation method, system, equipment and medium - Google Patents
Data recommendation method, system, equipment and medium Download PDFInfo
- Publication number
- CN113947455B CN113947455B CN202111195654.0A CN202111195654A CN113947455B CN 113947455 B CN113947455 B CN 113947455B CN 202111195654 A CN202111195654 A CN 202111195654A CN 113947455 B CN113947455 B CN 113947455B
- Authority
- CN
- China
- Prior art keywords
- commodity
- sequence
- commodities
- user
- time
- 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.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 31
- 239000013598 vector Substances 0.000 claims abstract description 125
- 239000011159 matrix material Substances 0.000 claims abstract description 123
- 230000008901 benefit Effects 0.000 claims description 18
- 238000004590 computer program Methods 0.000 claims description 11
- 239000000758 substrate Substances 0.000 claims 1
- 238000004364 calculation method Methods 0.000 abstract description 16
- 230000011218 segmentation Effects 0.000 description 12
- 238000005516 engineering process Methods 0.000 description 10
- 238000013473 artificial intelligence Methods 0.000 description 6
- 230000006870 function Effects 0.000 description 6
- 238000012163 sequencing technique Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 4
- 238000004422 calculation algorithm Methods 0.000 description 3
- 238000010586 diagram Methods 0.000 description 3
- 230000002093 peripheral effect Effects 0.000 description 3
- 230000008569 process Effects 0.000 description 3
- 230000005540 biological transmission Effects 0.000 description 1
- 230000001413 cellular effect Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000013500 data storage Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 230000001419 dependent effect Effects 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 238000000802 evaporation-induced self-assembly Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 230000003993 interaction Effects 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000003058 natural language processing Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 239000007787 solid Substances 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q30/00—Commerce
- G06Q30/06—Buying, selling or leasing transactions
- G06Q30/0601—Electronic shopping [e-shopping]
- G06Q30/0631—Item recommendations
-
- 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
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Development Economics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Strategic Management (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application provides a data recommendation method, a system, equipment and a medium, wherein a first sequence of commodities can be obtained through calculation of similarity between query sentences input by a user and commodity information, and then a second sequence is determined according to the query sentences input by the user, a sentence vector matrix and an auxiliary matrix, wherein the sentence vector matrix and the auxiliary matrix are obtained by adjusting the original sentence vector matrix and the auxiliary matrix based on feedback of commodities selected or browsed by the user after the commodities are recommended last time, and a target sequence of the commodities recommended to the user is determined according to the first sequence and the second sequence, so that commodity recommendation can be adjusted according to feedback of the user, the recommended commodities are more in accordance with the user requirements, and the results of the two sequences are combined to be the recommended commodities of the user, so that the problems that the early recommended commodities do not meet the user requirements due to less feedback of the received user can be prevented.
Description
Technical Field
The application relates to the field of artificial intelligence, in particular to a data recommendation method, a system, equipment and a medium.
Background
The electronic commerce transaction platform becomes an important channel for users to purchase goods, the electronic commerce transaction platform can recommend the goods which are possibly interested to the users according to the user information, the users also frequently search the goods by inputting characters, and at the moment, the goods which are interested to the users are searched according to the matching degree of query sentences input by the users and the goods information. However, in the case where the commodity information is not accurate enough or detailed enough, the recommended commodity is highly likely to be out of compliance with the user's demand.
Therefore, how to improve the recommended goods to better meet the demands of users and the recommended data is more accurate is a problem to be solved urgently.
Disclosure of Invention
The application provides a data recommendation method, a system, equipment and a medium, which can adjust the next recommended commodity according to the feedback of a user on the recommended commodity, so that the recommended commodity meets the user requirement. And besides, the next recommended commodity is adjusted according to the feedback of the user on the recommended commodity, and another recommended commodity sequence is obtained according to the similarity of the query statement and commodity information input by the user. Finally, the result of integrating the two sequences is that the user recommends the commodity, so that the problem that the early recommended commodity does not meet the user requirement can be avoided due to less feedback of the user.
The object and other objects are achieved by the features in the independent claims. Further implementations are presented in the dependent claims, the description and the figures.
In a first aspect, the present application provides a method comprising: generating commodity vectors of each commodity according to commodity information of each commodity; generating sentence vectors according to the query sentences input by the user at the t time, wherein t is more than 1, and t is a positive integer; according to the similarity between the sentence vector and the commodity vector of each commodity, determining N commodities with the maximum similarity and a first sequence of the N commodities, wherein N is a positive integer, and the first sequence is used for indicating the ordering of the similarity between the commodity vectors of the N commodities and the sentence vector; obtaining a t-th sentence vector matrix according to the sentence vectors of the query sentences input by the user for the previous t-1 times; adjusting an auxiliary matrix of the t-1 th time according to the operation of a user on the commodity recommended by the t-1 th time to obtain an auxiliary matrix of the t time, wherein the auxiliary matrix is used for indicating the attention degree of the user on the commodity recommended by the previous t-1 time; determining a second sequence of the N commodities according to the sentence vector matrix of the t time and the auxiliary matrix of the t time, wherein the second sequence is used for indicating the ordering of the attention degree of the N commodities by the user for the previous t-1 time; and determining a target sequence according to the first sequence and the second sequence, wherein the target sequence represents the order of the commodity recommended by the user.
In a second aspect, the present application provides a system comprising: a generating unit, a determining unit: the generating unit is used for generating commodity vectors of each commodity according to commodity information of each commodity; the generating unit is also used for generating sentence vectors according to the query sentences input by the t-th user, t is more than 1, and t is a positive integer; the determining unit is used for determining N commodities with the maximum similarity and a first sequence of the N commodities according to the similarity between the sentence vector and the commodity vector of each commodity, wherein N is a positive integer, and the first sequence is used for indicating the ordering of the similarity between the commodity vectors of the N commodities and the sentence vector; the determining unit is used for obtaining a t-th sentence vector matrix according to the sentence vectors of the query sentences input by the user for the previous t-1 times; the determining unit is used for adjusting the auxiliary matrix of the t-1 th time according to the operation of the user on the commodity recommended by the t-1 th time to obtain the auxiliary matrix of the t time, wherein the auxiliary matrix of the t time is used for indicating the attention degree of the user on the commodity recommended by the previous t-1 time; the determining unit is further used for determining a second sequence of the N commodities according to the t-1 th sentence vector matrix and the t-1 th auxiliary matrix, wherein the second sequence is used for indicating the ordering of the attention degree of the N commodities by the user for t-1 time; the determining unit is further configured to determine a target sequence according to the first sequence and the second sequence, where the target sequence represents an order of goods recommended by the user.
In a third aspect, the present application provides a computer device, comprising: a processor and a memory, said memory storing a computer program, said processor executing the computer program in said memory to perform the method as described in the first aspect.
In a fourth aspect, the present application provides a computer readable storage medium storing a computer program, characterized in that the computer program, when run on a computer, causes the computer to perform the method as described in the first aspect.
In summary, the data recommendation method, system, device and medium provided in the embodiments of the present application calculate, through the similarity between the query sentence input by the user and the commodity information, a first sequence of the commodity, and then determine a second sequence according to the query sentence input by the user, the sentence vector matrix and the auxiliary matrix, where after the commodity is recommended last time, the sentence vector matrix and the auxiliary matrix are obtained by adjusting the original sentence vector matrix and the auxiliary matrix based on feedback of the commodity selected or browsed by the user, and the target sequence of the commodity recommended to the user is determined according to the first sequence and the second sequence, so that the recommended commodity can be adjusted according to the feedback of the user, the recommended commodity better meets the user requirement, and the result of the two sequences is the recommended commodity of the user, which can prevent the feedback of the user received at an early stage from being less, and avoid the problem that the early recommended commodity does not meet the user requirement.
Drawings
In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings used in the description of the embodiments or the prior art will be briefly described below.
Fig. 1 is a schematic flow chart of a data recommendation method according to an embodiment of the present application;
fig. 2 is a schematic structural diagram of a data recommendation system according to an embodiment of the present application;
fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application.
Detailed Description
The terminology used in the following embodiments of the application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. As used in the specification of the present application and the appended claims, the singular forms "a," "an," "the," and "the" are intended to include the plural forms as well, unless the context clearly indicates to the contrary. It should also be understood that the term "and/or" as used in this disclosure refers to and encompasses any or all possible combinations of one or more of the listed items.
The embodiment of the application can acquire and process the related data based on the artificial intelligence technology. Wherein artificial intelligence (ARTIFICIAL INTELLIGENCE, AI) is the theory, method, technique, and application system that uses a digital computer or a digital computer-controlled machine to simulate, extend, and expand human intelligence, sense the environment, acquire knowledge, and use knowledge to obtain optimal results.
Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises a computer vision technology, a robot technology, a biological recognition technology, a voice processing technology, a natural language processing technology, machine learning/deep learning and other directions.
With the development of technology, the electronic commerce transaction platform has become an important channel for users to purchase goods, and through the electronic commerce transaction platform, users can purchase various goods, such as articles of daily use and clothes, and even through the electronic commerce transaction platform, insurance business can be purchased. The electronic commerce transaction platform can recommend goods which are possibly interested to the user according to the user information, specifically comprises the steps of recommending the goods to the user according to basic information filled by the user, such as age, sex and the like, and the type of the goods which are interested and selected by the user, taking the information as a label, matching the goods information corresponding to the label through a big data algorithm.
However, the user also often searches for the commodity by inputting text, and then the commodity of interest to the user needs to be found out according to the matching degree of the query statement input by the user and the commodity information. However, in the case where the commodity information is not accurate enough or detailed enough, the recommended commodity is highly likely to be out of compliance with the user's demand.
In order to improve the accuracy of recommending commodities, the application provides a commodity recommendation algorithm, wherein the weight of each commodity is determined by calculating the matching degree of inquiry sentences and commodity information to obtain a first sequence; then calculating according to the sentence vector matrix and the auxiliary matrix to obtain new weights of each commodity to obtain a second sequence, wherein the sentence vector matrix and the auxiliary matrix are obtained by adjusting the original sentence vector matrix and the auxiliary matrix based on feedback of commodities selected or browsed by a user and the like after recommending the commodities last time; obtaining a target sequence according to the first sequence and the second sequence, and recommending commodities to the user according to the target sequence; and acquiring feedback of the user based on the recommended commodity, and then adjusting the sentence vector matrix and the auxiliary matrix for the next commodity recommendation.
S110, generating sentence vectors according to the query sentences input by the t-th user, and generating commodity vectors of each commodity according to commodity information of each commodity.
The electronic equipment acquires a query sentence input by a user at the time, performs word segmentation on the query sentence to obtain a word segmentation result, and obtains a word vector based on the word segmentation result; and obtaining a sentence vector X t of the query sentence according to the word vector of each word segmentation result in the query sentence, wherein t represents the acquired times of the query sentence input by the user, and t is a positive integer. The word vector and the sentence vector can be obtained through word2vec, gloVe and other algorithms.
Likewise, the electronic device may further acquire commodity information of each commodity, where the commodity information is used to describe a type, an attribute, and the like of the commodity, and the commodity information may be a long text or a short text. Assuming that M commodities are in total, generating a plurality of word vectors of each commodity according to the result of word segmentation of commodity information of each commodity, and generating sentence vectors of commodity information of each commodity according to the word vectors of each commodity, wherein the sentence vectors are commodity vectors Y i of the commodity, i=1, 2.
In some embodiments, if the commodity vector of the commodity information for each commodity has been calculated in the previous t-1 calculations, then there is no need to calculate the commodity vector for each commodity again. If the number of products is increased in the present calculation as compared with the previous calculation, the number M of corresponding products is also increased, and the product vector of the increased products needs to be calculated. If the commodity information of a part of the commodity is modified in the present calculation, the commodity vector of the commodity whose commodity information is modified needs to be recalculated. It should be understood that the present application is not limited in particular as to whether the acquisition of commodity information is added or modified.
In some embodiments, the word segmentation process may further include removal of related words, human pronouns, and the like. In a specific implementation, the text to be identified can be segmented by adopting a jieba segmentation tool, a Hanlp segmentation device, an LTP segmentation device and other segmentation tools, and the segmentation method is not particularly limited.
In other embodiments, after each word of the text to be matched and the text to be compared is obtained, the method may further include word duplication removing processing to remove repeated words in the word segmentation result; or the words can be screened by a Term Frequency-inverse text Frequency index (Term Frequency-Inverse Document Frequency, TF-IDF) method, and more important words are reserved.
S120, obtaining initial weights of all the commodities according to the similarity between the sentence vectors and the commodity vectors, and determining N commodities and a first sequence of the N commodities.
The electronic device calculates the similarity between the sentence vector X t of the query sentence and the commodity vector Y i of each commodity to obtain the initial weight of each commodity, wherein if the similarity between the sentence vector X t and the commodity vector Y i is higher, the initial weight is larger. The method comprises the steps of adding M commodities in total, determining N commodities with higher initial weights from all the commodities according to the initial weights of the M commodities, and sequencing the N commodities from large to small according to the initial weights of the N commodities, wherein N is smaller than M, so that a first sequence of the N commodities is obtained.
The similarity is obtained by calculating the distance between the sentence vector and the commodity vector, and the distance between the sentence vector and the commodity vector can be cosine distance (CosineDistance), euclidean distance (Euclidean Distance), manhattan distance (MANHATTAN DISTANCE), chebyshev distance (Chebyshev Distance), minkowski distance (Minkowski Distance) and the like. The present application is not particularly limited to the distance calculation method. For example, the distance formula D i (cosine distance) between the calculated sentence vector X t and the commodity vector Y i can be referred to the following formula (1):
s130, calculating target weights of the N commodities according to the sentence vector matrix and the auxiliary matrix to obtain a second sequence of the N commodities.
And obtaining a sentence vector matrix A t,n according to sentence vectors of query sentences input by a user for t-1 times, identifying whether the user operates each commodity after recommending the commodity for the user according to t-1 times, and adjusting an auxiliary matrix B t,n obtained by the sentence vector matrix A t-1,n and the auxiliary matrix B t-1,n based on the operation of the user, wherein n=1, 2. And according to the sentence vector matrix A t,n and the auxiliary matrix B t,n, calculating to obtain target weights of N commodities, and sequencing the target weights of the N commodities from large to small to obtain a second sequence of the N commodities.
The operations of the user may include clicking on the commodity, purchasing the commodity, collecting the commodity, or browsing the commodity for more than a certain period of time, etc., and the benefits may be obtained based on the operations of the user, where the benefits of the N commodities are denoted by r t,n, and r t,n is used to represent the benefits obtained by each commodity. If the user operates the first commodity in the recommended commodities, the first commodity will obtain the benefit r t,n, and the auxiliary matrix B t,n corresponding to the first commodity is added with the product of the benefit r t,n and the sentence vector. While the auxiliary matrix B t,n corresponding to the remaining items will not change.
The generation of the t-th sentence vector matrix A t,n and the auxiliary matrix B t,n from the t-1 th sentence vector matrix A t-1,n and the auxiliary matrix B t-1,n is described in detail below.
The sentence vector matrix is obtained according to the sentence vector of the query sentence input by the user for the previous t-1 times, the auxiliary matrix is obtained by adjusting the auxiliary matrix of the t-1 th time according to the operation of the user on the commodity recommended by the t-1 th time, and the auxiliary matrix is used for representing the attention degree of the user on the commodity recommended by the previous t-1 time. According to the sentence vector matrix A t-1,n, the auxiliary matrix B t-1,n and the selection of the t-1 time user, the calculation mode for generating the sentence vector matrix A t,n can refer to the formula (2), and the calculation mode for generating the auxiliary matrix B t,n can refer to the formula (3).
At,n＝At-1,n+Xt-1Xt-1 T (2)
Bt,n＝Bt-1,n+rt-1,nXt-1 (3)
The sentence vector matrix a t,n is initialized to an identity matrix, the auxiliary matrix B t,n is initialized to a zero matrix, that is, when the input of the user is obtained for the first time, i.e., t=1, the sentence vector matrix a 1,n is the identity matrix, and the auxiliary matrix B 1,n is the zero matrix.
For example, the sentence vector of the query sentence input by the user at time t-1 is X t-1, 6 commodities are recommended to the user at time t-1, if the user operates a certain commodity, the corresponding commodity will obtain 0.6 benefits, and if the user does not operate a certain commodity, the corresponding commodity will not have benefits. Wherein, if the user clicks only the first commodity, r t-1.1 =0.6, that is, when the user recommends the commodity for the time t, the auxiliary matrix B t,1 of the first commodity is the product of the auxiliary matrix B t-1,n of the time t-1 plus the gain 0.6 and the sentence vector X t-1. In addition, the rest 5 commodities do not obtain benefits, namely r t-1.2＝rt-1.3＝rt-1.4＝rt-1.4＝rt-1.6 =0, and according to the formula (3), the auxiliary matrix of the rest 5 commodities is unchanged at the t time.
In some embodiments, the benefits can be further divided into positive feedback benefits r 1 and negative feedback benefits r 2, wherein r 1＞0,r2 < 0 if a user clicks on a commodity, purchases a commodity, collects a commodity, or browses a commodity for more than a certain period of time, the commodity will obtain benefits r t,n as positive feedback benefits r 1. If the user does not perform the above operation on the commodity, the commodity will obtain a negative feedback benefit r t,n as a negative feedback benefit r 2.
In some embodiments, the benefits obtained by different user operations may be different, such as the user purchasing more than the user collecting more than the user browsing more than a certain period of time, the user browsing more than the user clicking on the merchandise, but browsing less than a certain period of time, etc.
The process of calculating the target weights P t,n of N commodities from the sentence vector matrix a t,n and the auxiliary matrix B t,n is described in detail below.
Firstly, calculating according to the sentence vector matrix A t,n and the auxiliary matrix B t,n to obtain a parameter theta t,n, wherein the calculation mode of the parameter theta t,n refers to the following formula (4):
Obtaining a target weight P t,n according to a parameter theta t,n, a sentence vector X t and a sentence vector matrix A t,n, wherein the calculation mode of the target weight P t,n refers to the following formula (5):
Wherein alpha is a superparameter, which determines the model update step size.
After the target weight P t,n of each commodity is obtained through calculation, sorting from large to small is carried out according to the target weight P t,n of each commodity, and a second sequence of N commodities is obtained.
And S140, determining a target sequence of the commodity according to the first sequence and the second sequence, recommending the commodity to a user according to the target sequence, and adjusting the sentence vector matrix and the auxiliary matrix according to user operation.
Determining target sequences of commodities according to the first sequence and the second sequence, recommending the commodities to a user according to the sequence of the target sequences, acquiring feedback of the user based on whether the recommended commodities are clicked, browsed, purchased, collected and the like, adjusting sentence vector matrixes and auxiliary matrixes based on the feedback of the user, and calculating the second sequence when the adjusted sentence vector matrixes and auxiliary matrixes are used for recommending the commodities for the t+1st time.
The target sequence of the commodity is determined based on the first sequence and the second sequence will be described in detail.
Comparing the first sequence of the N commodities with the top K commodities in the second sequence, and taking the second sequence as a target sequence when the top K commodities in the first sequence and the top K commodities in the second sequence are the same. And under the condition that the first K commodities in the second sequence are not identical with the first K commodities in the first sequence, sequencing the first J commodities in the first sequence and the first J commodities in the second sequence to be used as the first 2J commodities in the target sequence, wherein J is less than or equal to K < N, sequencing the second sequence to be used as the 2J+1-N commodities in the target sequence, wherein the first J commodities in the first sequence and the rest commodities in the second sequence are removed.
The case where the first K products in the second sequence are not identical to the first K products in the first sequence will be described in detail below.
The first J products of the first sequence and the first J products of the second sequence are combined and ordered to form the first 2J products of the target sequence, comprising: taking the first J commodities with the largest initial weight in the first sequence as the first J commodities of the target sequence, and taking the first J commodities with the largest target weight in the second sequence as the (J+1) th to (2) th commodities of the target sequence; or the first J commodities with the largest initial weight in the second sequence are used as the first J commodities of the target sequence, and the first J commodities with the largest target weight in the first sequence are used as the (J+1) th to (2) th commodities of the target sequence; or the first J with the largest initial weight in the first sequence and the first J with the largest initial weight in the second sequence are staggered to be used as the (J+1) th to (2) th commodity of the target sequence; and then removing the first J commodities of the first sequence from the second sequence and removing the rest commodities of the first J commodities of the second sequence, wherein the rest commodities are used as 2J+1 to Nth commodities of the target sequence according to the sequence of the second sequence.
For example, the first sequence and the second sequence respectively share 10 commodities, and the top 3 commodities in the first sequence and the second sequence are not identical, namely 6 non-identical commodities, so that the 6 non-identical commodities can be arbitrarily arranged to be the top 6 commodities of the target sequence, then the remaining 4 commodities except for the 6 commodities in the second sequence are arranged to be the last 4 commodities of the target sequence according to the order in the second sequence.
The sentence vector matrix and the auxiliary matrix are adjusted based on feedback of the user as described in detail below.
And acquiring the selection of N commodities based on the target sequence by a user, and adjusting the sentence vector matrix A t,n and the auxiliary matrix B t,n according to the selection of the user. Referring to equation (2) and equation (3), similarly, the adjusted sentence vector matrix a t+1,n and the auxiliary matrix B t+1,n can refer to the following equations (6) and (7):
At+1,n＝At,n+XtXt T (6)
Bt+1,n＝Bt,n+rt,nXt (7)
S150, acquiring a query sentence input by the t+1st time user, and executing steps S110-S140.
The electronic equipment acquires a query sentence input by a (t+1) th time user, and executes the content as shown in the step S110 to obtain a sentence vector X t+1; then, executing the content as described in the step S120, and obtaining the first sequences of the N commodities at the t+1st time and the N commodities at the t+1st time according to the similarity between the sentence vector X t+1 and the commodity vector Y i of each commodity; step S130 is executed again, and a second sequence of the N commodities in the t+1st time is obtained according to the sentence vector matrix A t+1,n and the auxiliary matrix B t+1,n; and then executing the content as shown in the step S140, determining a t+1st target sequence according to the t+1st first sequence and the second sequence, sending the t+1st target sequence to an application program, and adjusting a t+2nd sentence vector matrix and an auxiliary matrix of the N commodities according to user operation.
In some embodiments, if N products obtained in the first sequence of the t+1st time have N products obtained in the first sequence not belonging to the first t times, that is, none of the N products of the first t times appear, the product corresponds to a value of 1 in the sentence vector matrix a t+1,n and corresponds to a value of 0 in the auxiliary matrix B t+1,n when step S130 is executed.
In summary, the commodity data recommending method provided by the application can adjust the weight of the commodity by acquiring the feedback of the user on the commodity recommendation, and the adjusted weight is used for the calculation of the next commodity recommendation, so that the commodity recommendation can be adjusted according to the preference of the user, and the recommended commodity meets the requirement of the user. Besides, the application calculates the similarity between the query statement input by the user and the commodity information to obtain the commodity sequence, compares the commodity sequence with the commodity sequence based on the feedback adjustment of the user, synthesizes the results of the two sequences, and recommends the commodity for the user, thereby preventing the problem that the recommended commodity does not meet the user requirement due to less feedback of the user received at early stage.
In order to solve the problem that the recommended goods do not meet the user's requirements, the present application provides a data recommendation system, as shown in the figure, the data recommendation system 200 includes: the generation unit 210 and the determination unit 220.
The generating unit 210 is configured to generate a commodity vector of each commodity according to the commodity information of each commodity; the generating unit 210 is further configured to generate a sentence vector according to the query sentence input by the t-th user, and the specific calculation manner may refer to the foregoing step S110, which is not described herein.
The determining unit 220 is configured to determine N articles and the first sequences of the N articles according to the similarity between the sentence vector and the article vector of each article, and the specific calculation method may refer to the foregoing step S120, which is not repeated herein.
The determining unit 220 is further configured to determine a second sequence of N products according to a sentence vector matrix and an auxiliary matrix, where the sentence vector matrix is obtained according to a sentence vector of a query sentence input by a user for the previous t-1 time, and the auxiliary matrix is obtained by adjusting the auxiliary matrix for the t-1 th time according to the operation of the user on the product recommended for the t-1 th time, and the auxiliary matrix is used to represent the attention degree of the user on the product recommended for the previous t-1 time. Wherein the operations of the user based on the recommended goods include: the user clicks any one of the recommended commodities, the user browses any one of the recommended commodities for more than a preset time period, the user collects any one of the recommended commodities, and the user purchases any one of the recommended commodities. In some embodiments, in the case of a query sentence entered by a user for the first time, the sentence vector matrix is an identity matrix and the auxiliary matrix is a zero matrix. The specific calculation method may refer to the aforementioned step S130, and will not be described herein.
The determining unit 220 is further configured to determine, according to the first sequence and the second sequence, a target sequence for recommending a commodity to the user, where the second sequence is used as the target sequence when the first K commodities of the first sequence and the first K commodities of the second sequence have the same commodity; and under the condition that the first K commodities in the second sequence are not identical with the first K commodities in the first sequence, sequencing the first J commodities in the first sequence and the first J commodities in the second sequence, wherein J is less than or equal to K < N, taking the first J commodities in the second sequence except the first sequence and the rest commodities except the first J commodities in the second sequence as the 2J+1 to N commodities in the target sequence according to the sequencing of the second sequence. The data recommendation system 200 provided by the application can adjust the weight of the commodity by acquiring the feedback of the user on the commodity recommendation, and the adjusted weight is used for the calculation of the next commodity recommendation, so that the commodity recommendation can be adjusted according to the preference of the user, and the recommended commodity meets the user requirement. Besides, the application calculates the similarity between the query statement input by the user and the commodity information to obtain the commodity sequence, compares the commodity sequence with the commodity sequence based on the feedback adjustment of the user, synthesizes the results of the two sequences, and recommends the commodity for the user, thereby preventing the problem that the recommended commodity does not meet the user requirement due to less feedback of the user received at early stage.
Referring to fig. 3, fig. 3 is a schematic structural diagram of an electronic device according to an embodiment of the present application. The electronic device 300 may be the data recommendation system 200 in the foregoing. As shown in fig. 3, the electronic device 300 includes: processor 310, communication interface 320, and memory 330, with processor 310, communication interface 320, and memory 330 being shown interconnected by internal bus 340.
The processor 310, the communication interface 320 and the memory 330 may be connected by a bus, or may communicate by other means such as wireless transmission. The embodiments of the present application take the connection via bus 340 as an example, where bus 340 may be a peripheral component interconnect standard (PERIPHERAL COMPONENT INTERCONNECT, PCI) bus or an extended industry standard architecture (Extended Industry Standard Architecture, EISA) bus, etc. The bus 340 may be classified into an address bus, a data bus, a control bus, and the like. For ease of illustration, only one thick line is shown in fig. 3, but not only one bus or one type of bus.
The processor 310 may be comprised of one or more general purpose processors, such as a central processing unit (Central Processing Unit, CPU), or a combination of CPU and hardware chips. The hardware chip may be an Application-specific integrated Circuit (ASIC), a programmable logic device (Programmable Logic Device, PLD), or a combination thereof. The PLD may be a complex Programmable Logic device (Complex Programmable Logic Device, CPLD), a Field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA), general-purpose array Logic (GENERIC ARRAY Logic, GAL), or any combination thereof. Processor 310 executes various types of digitally stored instructions, such as software or firmware programs stored in memory 330, that enable electronic device 300 to provide a wide variety of services.
In particular, the processor 310 may be comprised of at least one general purpose processor, such as a central processing unit (Central Processing Unit, CPU), or a combination of CPU and hardware chips. The hardware chip may be an Application-specific integrated Circuit (ASIC), a programmable logic device (Programmable Logic Device, PLD), or a combination thereof. The PLD may be a complex Programmable Logic device (Complex Programmable Logic Device, CPLD), a Field-Programmable gate array (Field-Programmable GATE ARRAY, FPGA), general-purpose array Logic (GENERIC ARRAY Logic, GAL), or any combination thereof. Processor 310 executes various types of digitally stored instructions, such as software or firmware programs stored in memory 330, that enable electronic device 300 to provide a wide variety of services.
Memory 330 may include volatile memory (VolatileMemory), such as random access memory (Random Access Memory, RAM); the Memory 330 may also include a Non-Volatile Memory (Non-Volatile Memory), such as Read-Only Memory (ROM), flash Memory (Flash Memory), hard disk (HARD DISK DRIVE, HDD), or Solid state disk (Solid-state disk-STATE DRIVE, SSD); memory 330 may also include combinations of the above. The memory 330 may store application program codes and program data, among other things. The program code can calculate and obtain a first sequence of the commodity through the similarity between the query statement input by the user and commodity information, then determine a second sequence according to the query statement input by the user, the sentence vector matrix and the auxiliary matrix, wherein after the commodity is recommended last time, the sentence vector matrix and the auxiliary matrix are obtained by adjusting the original sentence vector matrix and the auxiliary matrix based on feedback of the commodity selected or browsed by the user and the like, and determine a target sequence for recommending the commodity to the user according to the first sequence and the second sequence. And may be used to perform other steps described in connection with the embodiment of fig. 1, and are not described in detail herein. The codes of the memory 330 may include codes for implementing functions of the generating unit and the determining unit, where the functions of the generating unit include functions of the generating unit 210 in fig. 2, for example, generating a commodity vector of each commodity according to the commodity information of each commodity, and generating a sentence vector according to a query sentence input by the t-th user, which may be specifically used to execute step S110 and optional steps of the foregoing method, which will not be described herein. The function of the obtaining unit includes the function of the determining unit 220 in fig. 2, for example, determining N articles and the first sequences of the N articles according to the similarity between the sentence vector and the article vector of each article, determining the second sequences of the N articles according to the sentence vector matrix and the auxiliary matrix, which are obtained by adjusting the sentence vector matrix and the auxiliary matrix of the t-1 st time according to the operation of the recommended article by the t-1 st time user, and determining the target sequence of the recommended article for the user according to the first sequences and the second sequences, which are specifically usable for executing the steps S120, S130, S140 and optional steps of the foregoing method, which are not repeated herein.
The communication interface 320 may be a wired interface (e.g., an ethernet interface), may be an internal interface (e.g., a high-speed serial computer expansion bus (PERIPHERAL COMPONENT INTERCONNECT EXPRESS, PCIe) bus interface), a wired interface (e.g., an ethernet interface), or a wireless interface (e.g., a cellular network interface or using a wireless local area network interface) for communicating with other devices or modules.
It should be noted that fig. 3 is only one possible implementation of the embodiment of the present application, and in practical applications, the electronic device may further include more or fewer components, which is not limited herein. For details not shown or described in the embodiment of the present application, reference may be made to the foregoing description of the embodiment of fig. 1, which is not repeated herein. The electronic device shown in fig. 3 may also be a computer cluster formed by a plurality of computing nodes, which is not particularly limited by the present application.
Embodiments of the present application also provide a computer readable storage medium having instructions stored therein that, when executed on a processor, implement the method flow shown in fig. 1.
Embodiments of the present application also provide a computer program product which, when run on a processor, implements the method flow shown in fig. 1.
The above embodiments may be implemented in whole or in part by software, hardware, firmware, or any other combination. When implemented in software, the above-described embodiments may be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When loaded or executed on a computer, produces a flow or function in accordance with embodiments of the present invention, in whole or in part. The computer may be a general purpose computer, a special purpose computer, a computer network, or other programmable apparatus. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another computer-readable storage medium, e.g., from one website, computer, server, or data center by a wired (e.g., coaxial cable, fiber optic, digital subscriber line (Digital Subscriber Line, DSL), or wireless (e.g., infrared, wireless, microwave, etc.) means, the computer-readable storage medium may be any available medium that can be accessed by the computer or a data storage device such as a server, data center, etc., that contains a collection of one or more available media, the available media may be magnetic media (e.g., floppy disk, hard disk, tape), optical media (e.g., high density digital video disc (Digital Video Disc, DVD), or semiconductor media.
While the invention has been described with reference to certain preferred embodiments, it will be understood by those skilled in the art that various changes and substitutions of equivalents may be made and equivalents will be apparent to those skilled in the art without departing from the scope of the invention. Therefore, the protection scope of the invention is subject to the protection scope of the claims.
Claims (7)
1. A data recommendation method, comprising:
generating commodity vectors of each commodity according to commodity information of each commodity;
Generating sentence vectors according to the query sentences input by the user for the t time, wherein t is more than 1, and t is a positive integer;
sorting the similarity between the sentence vector and the commodity vector of each commodity from big to small to obtain a sorted result;
Determining N commodities and a first sequence of the N commodities according to the ordered results, wherein the N commodities are the first N commodities of the ordered results, the first sequence is the ordered results of the N commodities, N is a positive integer, and the first sequence is used for indicating the ordering of the similarity between commodity vectors of the N commodities and the sentence vectors;
obtaining a t-th sentence vector matrix according to the sentence vectors of the query sentences input by the user for the previous t-1 times;
Adjusting an auxiliary matrix of the t-1 th time according to the operation of the user on the commodity recommended by the t-1 th time to obtain an auxiliary matrix of the t time, wherein the auxiliary matrix of the t time is used for indicating the attention degree of the user on the commodity recommended by the previous t-1 time;
Determining a second sequence of the N commodities according to the sentence vector matrix of the t time and the auxiliary matrix of the t time, wherein the second sequence is used for indicating the ordering of the attention degree of the N commodities by the user for the previous t-1 time;
Determining a target sequence according to the first sequence and the second sequence, wherein the target sequence represents the order of the t-th commodity recommended to the user;
the operation of acquiring a first commodity in the commodities ordered by the target sequence by the user comprises the following steps: the user clicks the first commodity, the user browses the first commodity for longer than a certain period of time, the user collects the first commodity or the user purchases the first commodity;
Determining benefits of the first commodity according to the operation of the user on the first commodity in the commodities ordered according to the target sequence, wherein the benefits are used for indicating the attention degree of the user on the first commodity;
And adjusting the auxiliary matrix according to the income of the first commodity, the sentence vector and the auxiliary matrix of the t time, wherein the adjusted auxiliary matrix is used for recommending the commodity of the t+1th time.
2. The method of claim 1, wherein said determining a target sequence from said first sequence and said second sequence comprises:
Under the condition that the first K commodities of the first sequence and the first K commodities of the second sequence do not have the same commodity, any one of the first J commodities of the first sequence and the first J commodities of the second sequence is ordered to be used as the first 2J commodities of the target sequence, wherein J is less than or equal to K < N, and J and K are positive integers;
And taking the first J commodities except the first sequence and the rest commodities except the first J commodities in the second sequence as 2J+1 to Nth commodities in the target sequence according to the sequence of the second sequence.
3. The method of claim 1, wherein said determining a target sequence from said first sequence and said second sequence comprises:
And taking the second sequence as the target sequence when the first K commodities of the first sequence and the first K commodities of the second sequence exist in the same commodities.
4. The method of claim 1, wherein the step of determining the position of the substrate comprises,
Under the condition of the query sentence input by the user for the first time, the sentence vector matrix is an identity matrix, and the auxiliary matrix is a zero matrix.
5. A data recommendation system for performing the method of any one of claims 1-4, the data recommendation system comprising: a generating unit, a determining unit:
The generating unit is used for generating commodity vectors of each commodity according to commodity information of each commodity;
The generating unit is also used for generating sentence vectors according to the query sentences input by the user for the t time, t is more than 1, and t is a positive integer;
The determining unit is used for determining N commodities with the maximum similarity and a first sequence of the N commodities according to the similarity between the sentence vector and the commodity vector of each commodity, wherein N is a positive integer, and the first sequence is used for indicating the ordering of the similarity between the commodity vectors of the N commodities and the sentence vector;
The determining unit is used for obtaining a t-th sentence vector matrix according to the sentence vectors of the query sentences input by the user for t-1 times;
the determining unit is used for adjusting an auxiliary matrix of the t-1 th time according to the operation of the user on the commodity recommended by the t-1 th time to obtain an auxiliary matrix of the t time, wherein the auxiliary matrix of the t time is used for indicating the attention degree of the user on the commodity recommended by the previous t-1 time;
The determining unit is further configured to determine a second sequence of the N commodities according to the sentence vector matrix of the t time and the auxiliary matrix of the t time, where the second sequence is used to indicate the ranking of the attention degree of the N commodities for the previous t-1 time of the user;
the determining unit is further configured to determine a target sequence according to the first sequence and the second sequence, where the target sequence represents the order of the t-th merchandise recommended for the user.
6. A computer device, comprising: a processor and a memory, the memory storing a computer program, the processor executing the computer program in the memory to implement the method of any of claims 1 to 4.
7. A computer readable storage medium storing a computer program, characterized in that the computer program, when run on a computer, causes the computer to perform the method of any one of claims 1 to 4.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111195654.0A CN113947455B (en) | 2021-10-14 | 2021-10-14 | Data recommendation method, system, equipment and medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111195654.0A CN113947455B (en) | 2021-10-14 | 2021-10-14 | Data recommendation method, system, equipment and medium |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113947455A CN113947455A (en) | 2022-01-18 |
| CN113947455B true CN113947455B (en) | 2024-06-11 |
Family
ID=79329727
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111195654.0A Active CN113947455B (en) | 2021-10-14 | 2021-10-14 | Data recommendation method, system, equipment and medium |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113947455B (en) |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107038609A (en) * | 2017-04-24 | 2017-08-11 | 广州华企联信息科技有限公司 | A kind of Method of Commodity Recommendation and system based on deep learning |
| CN109087178A (en) * | 2018-08-28 | 2018-12-25 | 清华大学 | Method of Commodity Recommendation and device |
| CN109214882A (en) * | 2018-07-09 | 2019-01-15 | 西北大学 | A kind of Method of Commodity Recommendation |
| CN109598586A (en) * | 2018-11-30 | 2019-04-09 | 哈尔滨工程大学 | A kind of recommended method based on attention model |
| JP2021039712A (en) * | 2019-09-04 | 2021-03-11 | 株式会社トゥービーソフトＴｏｂｅｓｏｆｔ Ｃｏ．， Ｌｔｄ． | User customization type commodity recommendation device through artificial intelligence-based machine learning |
Family Cites Families (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8635310B2 (en) * | 2010-07-14 | 2014-01-21 | William G. Bartholomay | Business process utilizing systems, devices and methods engendering cooperation among service providers to maximize end user satisfaction |
| US10515400B2 (en) * | 2016-09-08 | 2019-12-24 | Adobe Inc. | Learning vector-space representations of items for recommendations using word embedding models |
-
2021
- 2021-10-14 CN CN202111195654.0A patent/CN113947455B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107038609A (en) * | 2017-04-24 | 2017-08-11 | 广州华企联信息科技有限公司 | A kind of Method of Commodity Recommendation and system based on deep learning |
| CN109214882A (en) * | 2018-07-09 | 2019-01-15 | 西北大学 | A kind of Method of Commodity Recommendation |
| CN109087178A (en) * | 2018-08-28 | 2018-12-25 | 清华大学 | Method of Commodity Recommendation and device |
| CN109598586A (en) * | 2018-11-30 | 2019-04-09 | 哈尔滨工程大学 | A kind of recommended method based on attention model |
| JP2021039712A (en) * | 2019-09-04 | 2021-03-11 | 株式会社トゥービーソフトＴｏｂｅｓｏｆｔ Ｃｏ．， Ｌｔｄ． | User customization type commodity recommendation device through artificial intelligence-based machine learning |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113947455A (en) | 2022-01-18 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110969516B (en) | Commodity recommendation method and device | |
| CN111798273B (en) | Training method of product purchase probability prediction model and purchase probability prediction method | |
| US7885859B2 (en) | Assigning into one set of categories information that has been assigned to other sets of categories | |
| CN109033101B (en) | Label recommendation method and device | |
| CN105956628B (en) | Data classification method and device for data classification | |
| CN112085565A (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
| CN109242002A (en) | High dimensional data classification method, device and terminal device | |
| CN112818218B (en) | Information recommendation method, device, terminal equipment and computer readable storage medium | |
| CN110516033A (en) | A kind of method and apparatus calculating user preference | |
| CN111126442B (en) | Method for generating key attribute of article, method and device for classifying article | |
| Antonio et al. | Sentiment analysis for covid-19 in Indonesia on Twitter with TF-IDF featured extraction and stochastic gradient descent | |
| CN114547307A (en) | Text vector model training method, text matching method, device and equipment | |
| CN112685635B (en) | Item recommendation method, device, server and storage medium based on classification label | |
| CN114139634A (en) | Multi-label feature selection method based on paired label weights | |
| CN111143678B (en) | Recommendation system and recommendation method | |
| CN112800226A (en) | Method for obtaining text classification model, method, device and equipment for text classification | |
| CN111488479A (en) | Hypergraph construction method, hypergraph construction device, computer system and medium | |
| CN104376120B (en) | A kind of information retrieval method and system | |
| CN113947455B (en) | Data recommendation method, system, equipment and medium | |
| CN113704617A (en) | Article recommendation method, system, electronic device and storage medium | |
| CN111144098B (en) | Recall method and device for extended question | |
| CN109886299B (en) | User portrait method and device, readable storage medium and terminal equipment | |
| CN110959157B (en) | Accelerating large-scale similarity computation | |
| CN114706927B (en) | Data batch labeling method based on artificial intelligence and related equipment | |
| CN113781180B (en) | Article recommendation method and device, electronic equipment and 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 |