CN112598483A - Migration object generation and object recommendation method and device for target platform - Google Patents
Migration object generation and object recommendation method and device for target platform Download PDFInfo
- Publication number
- CN112598483A CN112598483A CN202011627705.8A CN202011627705A CN112598483A CN 112598483 A CN112598483 A CN 112598483A CN 202011627705 A CN202011627705 A CN 202011627705A CN 112598483 A CN112598483 A CN 112598483A
- Authority
- CN
- China
- Prior art keywords
- vector
- platform
- target
- recommendation
- model
- 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 90
- 230000005012 migration Effects 0.000 title claims abstract description 69
- 238000013508 migration Methods 0.000 title claims abstract description 69
- 239000013598 vector Substances 0.000 claims abstract description 359
- 238000012549 training Methods 0.000 claims abstract description 58
- 230000002441 reversible effect Effects 0.000 claims abstract description 47
- 230000006399 behavior Effects 0.000 claims description 37
- 230000006870 function Effects 0.000 claims description 31
- 230000008569 process Effects 0.000 claims description 23
- 238000012795 verification Methods 0.000 claims description 16
- 230000002457 bidirectional effect Effects 0.000 claims description 15
- 238000004590 computer program Methods 0.000 claims description 15
- 238000012512 characterization method Methods 0.000 claims description 3
- 230000000295 complement effect Effects 0.000 description 5
- 238000010586 diagram Methods 0.000 description 5
- 230000014509 gene expression Effects 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 238000009825 accumulation Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000009795 derivation Methods 0.000 description 2
- 239000000047 product Substances 0.000 description 2
- 239000013589 supplement Substances 0.000 description 2
- 230000001360 synchronised effect Effects 0.000 description 2
- 238000012360 testing method Methods 0.000 description 2
- 230000007704 transition Effects 0.000 description 2
- 230000004913 activation Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 238000000605 extraction Methods 0.000 description 1
- 230000008676 import Effects 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000012800 visualization Methods 0.000 description 1
Images
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]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F40/00—Handling natural language data
- G06F40/30—Semantic analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- 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
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Business, Economics & Management (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Data Mining & Analysis (AREA)
- Finance (AREA)
- Artificial Intelligence (AREA)
- Accounting & Taxation (AREA)
- Evolutionary Computation (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Software Systems (AREA)
- Strategic Management (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- General Business, Economics & Management (AREA)
- Biophysics (AREA)
- Marketing (AREA)
- Economics (AREA)
- Development Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application relates to a method and a device for generating and recommending a migration object of a target platform, computer equipment and a storage medium. The method comprises the steps of obtaining a reference vector adopted by a reference platform when object recommendation is carried out, training a preset coding model according to the reference vector and a corresponding reverse semantic meaning to obtain a vector generation model, inputting each platform object provided by a target platform into the vector generation model to obtain a target vector corresponding to each platform object, determining each target vector as a migration object of the target platform, and carrying out object recommendation by a recommendation system of the target platform, so that the target platform can realize accurate recommendation of platform objects such as commodities and/or services provided by the platform at an online initial stage, the cold start problem of the target platform is solved, the recommendation system of the target platform can carry out accurate and effective real-time recommendation at a cold start stage, the cold start stage is perfectly transited, and user experience brought by the target platform can be improved.
Description
Technical Field
The present application relates to the field of data processing technologies, and in particular, to a method and an apparatus for generating a migration object and recommending an object on a target platform, a computer device, and a readable storage medium.
Background
The recommendation system has a plurality of recommendation models, but all the recommendation models face the problem of cold start, namely when a user does not have any behavior interaction on objects such as commodities at the initial stage of online of a certain platform, the recommendation system does not know the preference of the user and does not know how to perform corresponding recommendation on the user in the platform. At this time, the conventional scheme often makes some hot spot recommendations for the user, or asks the preference of the user in advance, so that the user selects a tag of the user, and then recommends the user according to the tag group. When a certain amount of user behaviors are accumulated, the recommendation system can perform modeling based on the user behaviors, so that more accurate user recommendation is realized. In the accumulation process, a period of time is needed, sometimes even several months, and in the accumulation process, accurate recommendation is often difficult to perform by a recommendation system of a new platform, so that the pertinence of a recommended object is poor, the accuracy is low, and the user experience is influenced.
Disclosure of Invention
In view of the foregoing, it is desirable to provide a migration object generation method, an object recommendation method, an apparatus, a computer device, and a readable storage medium for a target platform, which can perform accurate recommendation on a relevant online platform at an early stage.
A method of migrating object generation for a target platform, the method comprising:
acquiring each reference vector adopted by a reference platform when object recommendation is carried out; the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to user behaviors during the operation of the reference platform;
training a preset coding model according to each reference vector and the reverse semantic meaning corresponding to each reference vector to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse-deducing semantics are semantics obtained by reverse deduction according to corresponding reference vectors;
inputting each platform object provided by a target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determining each target vector as a migration object of the target platform so as to allow a recommendation system of the target platform to recommend the object; wherein the target platform is of the same type as the reference platform.
In one embodiment, the training a preset coding model according to the respective reference vectors and the semantics corresponding to the respective reference vectors to obtain a vector generation model includes:
inputting each reverse semantic meaning into the preset coding model, and acquiring output vectors respectively obtained by the preset coding model aiming at each reverse semantic meaning;
and when the process of obtaining the output vector by the preset coding model reaches a preset training standard, determining the vector generation model according to the current model parameters of the preset coding model.
As an embodiment, the preset coding model includes a bidirectional coding representation model and a full connection layer; the step of inputting each reverse-thrust semantic into the preset coding model and acquiring an output vector of the preset coding model respectively obtained aiming at each reverse-thrust semantic comprises the following steps:
inputting each reverse semantic into the bidirectional coding representation model, and acquiring a coding vector which is respectively output by the bidirectional coding representation model aiming at each reverse semantic;
inputting the coding vector into a full-connection layer for fitting to obtain an output vector; wherein the output vector has a dimension that is the same as the dimension of the reference vector.
Specifically, the fully-connected layer comprises a first fully-connected layer and a second fully-connected layer; the inputting the coding vector into a full-link layer for fitting to obtain an output vector comprises:
inputting the coding vector into a first full-connection layer to perform first fitting to obtain an initial fitting vector;
and inputting the initial fitting vector into the second full-connection layer to perform second fitting to obtain an output vector, and enabling the dimension of the output vector to be equal to that of the reference vector.
As an embodiment, before determining the vector generation model according to the current model parameter of the preset coding model when the process of obtaining the output vector by the preset coding model reaches the preset training standard, the method further includes:
dividing each reference vector and the reverse semantic meaning corresponding to each reference vector into a training set and a verification set;
training the preset coding model by adopting the training set to obtain an initial coding model, and obtaining a loss function of the initial coding model to obtain a training loss function;
inputting the reverse semantic meaning of the verification set into the initial coding model for vector output, and acquiring a current loss function to obtain a verification loss function;
and after the training loss function is converged, when the verification loss function and the training loss function are equal, judging that the process of obtaining the output vector by the preset coding model reaches a preset training standard.
In one embodiment, before obtaining the reference vectors used by the reference platform in object recommendation, the method further includes:
reading each user behavior detected in a set time period of a reference platform to generate a behavior sequence; the behavior sequence records the operation objects of the users on the reference platform and the relevance among the operation objects;
and inputting the behavior sequence into a vector coding model to obtain a reference vector adopted by a recommendation system of the reference platform when recommending an object.
In particular, the vector coding model is an item2vec model.
A migration object generation apparatus of a target platform, the apparatus comprising:
the acquisition module is used for acquiring each reference vector adopted by the reference platform when object recommendation is carried out; the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to user behaviors during the operation of the reference platform;
the training module is used for training a preset coding model according to the reference vectors and the reverse semantic meanings corresponding to the reference vectors to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse-deducing semantics are semantics obtained by reverse deduction according to corresponding reference vectors;
and the determining module is used for inputting each platform object provided by the target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determining each target vector as a migration object of the target platform so as to allow a recommendation system of the target platform to recommend the object.
An object recommendation method for a target platform, the method comprising:
generating a migration object of a target platform by adopting the migration object generation method of the target platform in any embodiment;
importing the migration object into a target recommendation system, and enabling the target recommendation system to perform object recommendation on the target platform; the target recommendation system is a recommendation system of a target platform.
In one embodiment, the importing the migration object into a target recommendation system, so that the target recommendation system performs object recommendation on the target platform includes:
importing the migration object into a target recommendation system, and enabling the target recommendation system to read each target vector;
acquiring a user selection object currently received by a target platform, and generating a current operation vector of the user selection object by adopting the target recommendation system; wherein the current operation vector is a low-dimensional dense vector generated by the target recommendation system for the user selection object;
and in the target recommendation system, determining a recommendation object according to the current operation vector and the target vector.
Specifically, the determining a recommended object according to the current operation vector and the target vector includes:
searching vectors with Euclidean distance smaller than a set distance from the current operation vector in the target vector to obtain recommended vectors;
and determining the platform object corresponding to the recommendation vector as the recommendation object.
Specifically, after the platform object corresponding to the recommendation vector is determined as the recommendation object, the method further includes:
if the number of the recommended vectors is multiple, respectively obtaining Euclidean distances between each recommended vector and the current operation vector;
sorting recommendation objects corresponding to the recommendation vectors respectively according to the Euclidean distance between the recommendation vectors and the current operation vector from small to large;
and recommending the recommended objects on the target platform according to the arrangement sequence of the recommended objects.
An object recommendation apparatus of a target platform, the apparatus comprising:
a migration object generation module, configured to generate a migration object of a target platform by using a migration object generation apparatus of the target platform according to any embodiment of the foregoing description;
the object recommendation module is used for importing the migration object into a target recommendation system so that the target recommendation system carries out object recommendation on the target platform; the target recommendation system is a recommendation system of a target platform.
A computer device includes a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements a migrated object generation method of a target platform in any of the above embodiments or an object recommendation method of the target platform in any of the above embodiments when executing the computer program.
A computer-readable storage medium, on which a computer program is stored, which, when executed by a processor, implements a migration object generation method of a target platform in any of the above embodiments, or an object recommendation method of the target platform in any of the above embodiments.
The method, the device, the computer equipment and the readable storage medium for generating and recommending the migration object of the target platform acquire each reference vector adopted by the reference platform when recommending the object, train the preset coding model according to each reference vector and the reverse semantic meaning corresponding to each reference vector to acquire the vector generation model, input each platform object provided by the target platform into the vector generation model to acquire the target vector corresponding to each platform object respectively, determine each target vector as the migration object of the target platform for the recommendation system of the target platform to recommend the object, enable the target platform to realize the accurate recommendation of the platform objects such as goods and/or services provided by the platform at the initial stage of online, solve the problem of cold start faced by the target platform, and enable the recommendation system of the target platform to perform accurate and effective real-time recommendation at the stage of cold start, therefore, the cold start stage is perfectly transited, and the user experience brought by the target platform is improved.
Drawings
FIG. 1 is a flowchart illustrating a method for generating migrated objects for a target platform according to an embodiment;
FIG. 2 is a diagram illustrating an exemplary operation of a pre-defined coding model;
FIG. 3 is a block diagram that illustrates the architecture of a migrated object generation apparatus for a target platform in one embodiment;
FIG. 4 is a flowchart illustrating a method for object recommendation of a target platform according to an embodiment;
FIG. 5 is a block diagram of an object recommendation device of a target platform in one embodiment;
FIG. 6 is a diagram illustrating an internal structure of a computer device according to an embodiment.
Detailed Description
In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.
Reference herein to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment can be included in at least one embodiment of the application. The appearances of the phrase in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. It is explicitly and implicitly understood by one skilled in the art that the embodiments described herein can be combined with other embodiments.
The migration object generation method of the target platform and/or the object recommendation method of the target platform can be applied to an intelligent terminal for operating the target platform. The intelligent terminal can obtain each reference vector adopted by the reference platform when object recommendation is carried out, a preset coding model is trained according to each reference vector and the reverse semantic meaning corresponding to each reference vector to obtain a vector generation model, then each platform object provided by the target platform is input into the vector generation model to obtain a target vector corresponding to each platform object, each target vector is determined to be a migration object of the target platform and is used for a recommendation system of the target platform to carry out object recommendation, so that the target platform can realize accurate recommendation of platform objects such as commodities and/or services provided by the platform at the on-line initial stage, the cold start problem of the target platform is solved, and the user experience corresponding to the target platform is improved. The reference platform is an online platform which can accurately recommend the recommendation system according to various operation behaviors of the user on the reference platform and the accumulated sufficient user behavior data, and the type of the online platform is the same as that of the target platform, such as a government affair handling platform in a certain region or a certain online shopping platform. The target platform is an online platform to be online or just online, such as a government affair handling platform required to be introduced in a certain area or a shopping platform required to be online by a certain company, and the like. The intelligent terminal can communicate with other intelligent terminals through a network, for example, the intelligent terminal can obtain each reference vector adopted by the reference platform when object recommendation is performed from other intelligent terminals, execute the migration object generation method of the target platform to obtain the migration object of the target platform, so that a recommendation system in the target platform operated by the intelligent terminal performs object recommendation or sends the migration object of the target platform to other intelligent terminals, so that the target platforms operated by other intelligent terminals perform objects and the like. The intelligent terminal running the migration object generation method of the target platform and/or the object recommendation method of the target platform may be, but is not limited to, various personal computers, notebook computers, smart phones, or tablet computers.
In an embodiment, as shown in fig. 1, a migration object generation method for a target platform is provided, which is described by taking an example that the method is applied to an intelligent terminal for operating the target platform, and includes the following steps:
s10, acquiring each reference vector adopted by the reference platform when object recommendation is carried out; and the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to the user behavior during the operation of the reference platform.
Specifically, the type of the reference platform is the same as the type of the target platform, that is, the same type of online platform. In general, a reference platform accumulates enough user behavior data, so that a recommendation system of the reference platform can accurately recommend various operation behaviors of a user on the reference platform. For example, if the target platform is a government affair handling platform which needs to be introduced into a certain region, the reference platform may be a government affair handling platform which has been operated for a certain period of time in another region and handles the same kind of business with the target platform, and the reference platform may accurately recommend the next business object which needs to be handled by the user after reading the business object selected by the user on the platform; if the target platform is a shopping platform on which a company needs to go online, the reference platform can be an online platform of which the type of the sold commodity is the same as that of the commodity sold by the target platform, and the reference platform can accurately recommend the commodity which the user may need to purchase next after reading the commodity input or selected by the user on the platform.
S20, training a preset coding model according to the reference vectors and the reverse semantic meaning corresponding to the reference vectors to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse semantic is a semantic obtained by reverse derivation according to the corresponding reference vector.
The reverse semantic meaning can be obtained by carrying out reverse derivation on a relevant semantic meaning detection model according to reference vectors, and each reference vector has corresponding feedback semantic meaning. With respect to a reference platform, the feedback semantics of a certain reference vector is the object name of the reference vector corresponding to the reference platform.
The preset coding model can be a simple two-layer neural network, and specifically can include a bert model and a plurality of fully-connected layer models. The vector generation model obtained by training the preset coding model by adopting each reference vector and the reverse semantic meaning corresponding to each reference vector can generate the low-dimensional dense vector (target vector) corresponding to each platform object provided by the target platform according to the generation rule from each platform object (reverse semantic meaning) provided by the reference platform to the corresponding low-dimensional dense vector (reference vector).
S30, inputting each platform object provided by a target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determining each target vector as a migration object of the target platform for a recommendation system of the target platform to recommend the object; wherein the target platform is of the same type as the reference platform.
In the actual use process, although the types of the reference platform and the target platform are the same, the expressions of the two platforms of the same type to the same thing may have differences, taking a government affair handling platform as an example, the name of an affair for subsidizing the ID card on a government affair handling platform is 'subsidy ID card', and the name of an affair for subsiding the ID card on another government affair handling platform may be 'ID card subsidy', so that if the target platform directly uses the reference vector adopted by the reference platform to recommend the platform object in the target platform, the problem of inaccurate recommendation result may occur, in this embodiment, the preset coding model is trained according to the reference vector of the reference platform and the reverse-push semantics obtained by the reverse-push of the reference vector, so that the vector generation model can obtain the vector generation model according to the generation rule from each platform object (reverse-push semantics) provided by the reference platform to the corresponding low-density vector (reference vector), and generating low-dimensional dense vectors corresponding to each platform object provided by the target platform respectively so as to obtain the target vectors required by the target platform for object recommendation, so that a recommendation system of the target platform can adopt the target vectors for accurate object recommendation.
The method for generating the migration object of the target platform comprises the steps of obtaining each reference vector adopted by the reference platform when object recommendation is carried out, training a preset coding model according to each reference vector and the reverse-thrust semantics corresponding to each reference vector to obtain a vector generation model, inputting each platform object provided by the target platform into the vector generation model to obtain the target vector corresponding to each platform object respectively, determining each target vector as the migration object of the target platform for a recommendation system of the target platform to carry out object recommendation, enabling the target platform to realize accurate recommendation of platform objects such as goods and/or services provided by the target platform at the initial stage of online, solving the problem of cold start faced by the target platform, enabling the recommendation system of the target platform to carry out accurate and effective real-time recommendation at the cold start stage, and further perfectly transiting the cold start stage, and the user experience brought by the target platform is improved.
In an embodiment, the training a preset coding model according to the respective reference vectors and semantics corresponding to the respective reference vectors to obtain a vector generation model includes:
inputting each reverse semantic meaning into the preset coding model, and acquiring output vectors respectively obtained by the preset coding model aiming at each reverse semantic meaning;
and when the process of obtaining the output vector by the preset coding model reaches a preset training standard, determining the vector generation model according to the current model parameters of the preset coding model.
The training standard may be set according to a feature of a preset coding model, for example, set as a corresponding loss function convergence or meet other criteria, and a vector generation model determined when a process of an output vector obtained by the preset coding model for each reverse semantic reaches the preset training standard may generate a target vector required by the target platform according to a generation rule of a reference vector in the reference platform.
As an embodiment, the preset coding model includes a bidirectional coding representation model and a full connection layer; the step of inputting each reverse-thrust semantic into the preset coding model and acquiring an output vector of the preset coding model respectively obtained aiming at each reverse-thrust semantic comprises the following steps:
inputting each reverse semantic into the bidirectional coding representation model, and acquiring a coding vector which is respectively output by the bidirectional coding representation model aiming at each reverse semantic;
inputting the coding vector into a full-connection layer for fitting to obtain an output vector; wherein the output vector has a dimension that is the same as the dimension of the reference vector.
Specifically, the fully-connected layer comprises a first fully-connected layer and a second fully-connected layer; the inputting the coding vector into a full-link layer for fitting to obtain an output vector comprises:
inputting the coding vector into a first full-connection layer to perform first fitting to obtain an initial fitting vector;
and inputting the initial fitting vector into the second full-connection layer to perform second fitting to obtain an output vector, and enabling the dimension of the output vector to be equal to that of the reference vector.
The Bidirectional coding characterization model may be a bert model (Bidirectional Encoder Representations from transforms), and preferably, the Bidirectional coding characterization model may adopt a bert small model with a more simplified structure, which is effectively simplified in structure relative to the bert base. In the process of training the preset coding model, a short text representing reverse semantic meaning can be transmitted into a bert small model, a CLS vector is obtained, downstream task training is carried out by adopting a full connection layer in the CLS vector, and a required vector generation model can be obtained.
In an example, the working process of the preset coding model may refer to fig. 2, as shown in fig. 2, the size of each partial structure output vector of the preset coding model may be set, and the corresponding training process may include: feeding each back-pushing semantic into a bert small model, wherein the bert small model respectively generates 256 x 1 coding vectors aiming at each back-pushing semantic, inputting the 256 x 1 coding vectors into a first full-connection layer, performing 128-dimensional dense layer fitting on the layer to obtain 128 x 1 initial fitting vectors, then inputting the 128 x 1 initial fitting vectors into a second full-connection layer, and fitting again to obtain 50 1 output vectors, so that the preset coding model can output 50-dimensional testing vectors (low-dimensional dense vectors) corresponding to each back-pushing semantic.
As an embodiment, before determining the vector generation model according to the current model parameter of the preset coding model when the process of obtaining the output vector by the preset coding model reaches the preset training standard, the method further includes:
dividing each reference vector and the reverse semantic meaning corresponding to each reference vector into a training set and a verification set;
training the preset coding model by adopting the training set to obtain an initial coding model, and obtaining a loss function of the initial coding model to obtain a training loss function;
inputting the reverse semantic meaning of the verification set into the initial coding model for vector output, and acquiring a current loss function to obtain a verification loss function;
and after the training loss function is converged, when the verification loss function and the training loss function are equal, judging that the process of obtaining the output vector by the preset coding model reaches a preset training standard.
The embodiment can accurately judge whether the preset coding model is trained or not, so that the determined vector generation model can generate the required low-dimensional dense vector for various platform objects according to the corresponding rule.
In an embodiment, before obtaining each reference vector that is adopted by the reference platform when performing object recommendation, the method further includes:
reading each user behavior detected in a set time period of a reference platform to generate a behavior sequence; the behavior sequence records the operation objects of the users on the reference platform and the relevance among the operation objects;
and inputting the behavior sequence into a vector coding model to obtain a reference vector adopted by a recommendation system of the reference platform when recommending an object.
The set time period may be set according to the type of the reference platform, for example, may be set to a time period within a half year before a time point of generating the required reference vector.
Specifically, the vector coding model may be an item2vec model, and the item2vec model may be obtained by pre-training.
When the vector coding model can be an item2vec model, the essence of the recommendation system in the reference platform is a vector similarity recommendation model based on item2 vec. In the actual use process, after the recommendation system of the reference platform is online, enough user behavior interaction data are collected, the user behavior interaction data are converted into behavior sequences, objects such as articles in the behavior sequences are subjected to embedding modeling, according to the item2vec modeling principle, the similarity degree of the articles which tend to be interacted by the user at the same time on embedding vectors is similar, and therefore when the recommendation system of the reference platform executes a recommendation task, a group of most similar non-interacted articles can be searched for through the previously interacted objects of the user to recommend.
In one example, the migration object generation process of a corresponding target platform is explained in detail by taking a government affairs handling platform of a certain provincial level project as a reference platform. In the government affair transaction platform, records of government affairs of 5 months can be accumulated, interaction data of all users and affairs in the period is extracted, a behavior sequence that each user interacts with affairs in 5 months is generated, and examples of the behavior sequence can include: the userid is [ 'supplement identity card', 'supplement social security card', … … ]. And then testing the related sequence by using the initial item2vec model, debugging parameters and modeling to obtain a required item2vec model pair, so that the item2vec model can map the names of all workable matters into a low-dimensional dense vector with 50 dimensions for a corresponding government affair transaction platform to recommend the matters.
Compared with the reference platform, the target platform of the same type of government affair transaction platform which is newly on-line or to be on-line may not be completely consistent with the expression of the same item, for example, "ID card complement" and "complement ID card" may both represent the item of the user's complement ID card, although the "ID card complement" and "complement ID card" have the same meaning in the text sense, the two expressions cannot obtain the same vector expression in the item2vec model. To address this issue, the present example compresses 50-dimensional transaction vectors to 2-dimensional using t-SNE algorithm (dimension reduction algorithm for high-dimensional vector extraction low-dimensional feature) to visualize the trained transaction vectors, the visualized vector visualization showing: the relationship between items in the fiscal teacher's document examination (upper left) is very dense, and also the relationship between items in the motor vehicle driving license, the relationship between items in laws and regulations, the relationship between items in the entrance of a residence, the relationship between items in the public accumulation fund and the like are dense. It can thus be found that: these transaction vectors trained from the user behavior sequence are also semantically related to a high degree. Thus, this example uses the pre-training model of bert small to make a pre-designed coding model that includes two layers of neural networks as shown in fig. 2. As shown in fig. 2, feeding the transaction name of the reference platform into the bert small, and taking out the cls vector in the result vector of the bert small to obtain a 256 × 1 vector (encoding vector); then, the layer is subjected to a 128-dimensional dense layer full-connection hidden layer fitting to obtain an initial fitting vector of 128 × 1, and then a full-connection output layer of 50 × 1 is carried out to obtain an output vector of 50 × 1. Wherein the hidden layer's activation function may use relu. The optimization target of the preset coding model is the cosine similarity between the previously trained 50-dimensional reference vector of item2vec and the 50-dimensional vector output by using a bert small +2 layer full connection. In other words, the word semantics contained in the pre-trained language model may be used, and it is desirable to extract some information that can be mapped into a vector that can be used as a recommendation. The input of the model is the name of the recommended item (such as the reverse semantic meaning of a reference vector), a 50-dimensional embedding vector (such as the reference vector) is output, the vector similarity of the item which tends to be associated and clicked by a user is very close, a required vector generation model is determined, the target vector of each platform object in the target platform is obtained, the migration object of the target platform is determined, the migration representation of item2vec in the recommendation system is completed at low cost, then the recommendation system of the target platform can read the service selected by the user in real time, the embedding vector of the selected service is generated, and real-time recommendation is carried out according to the embedding vector of the selected service and the migration object.
According to the method, the relevance described by the platform objects such as recommended articles and the like is excavated by fully utilizing the pre-training language model, and the relevance vector representation for recommendation is excavated, so that the target platform can be accurately recommended at the initial stage of online, the target platform can be in a perfect transition cold start stage, and more effective real-time recommendation can be performed at the cold start stage.
In one embodiment, as shown in fig. 3, there is provided a migration object generation apparatus of a target platform, including:
the acquisition module 10 is configured to acquire each reference vector adopted by the reference platform when object recommendation is performed; the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to user behaviors during the operation of the reference platform;
the training module 20 is configured to train a preset coding model according to the reference vectors and the reverse semantic meanings corresponding to the reference vectors to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse-deducing semantics are semantics obtained by reverse deduction according to corresponding reference vectors;
the determining module 30 is configured to input each platform object provided by the target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determine each target vector as a migration object of the target platform, so that the recommendation system of the target platform can recommend the object.
In one embodiment, the training module comprises:
the output vector acquisition module is used for inputting each reverse semantic into the preset coding model and acquiring output vectors which are respectively obtained by the preset coding model aiming at each reverse semantic;
and the vector generation model determining module is used for determining the vector generation model according to the current model parameters of the preset coding model when the process of obtaining the output vector by the preset coding model reaches the preset training standard.
As an embodiment, the preset coding model includes a bidirectional coding representation model and a full connection layer; the output vector acquisition module is further to:
inputting each reverse semantic into the bidirectional coding representation model, and acquiring a coding vector which is respectively output by the bidirectional coding representation model aiming at each reverse semantic;
inputting the coding vector into a full-connection layer for fitting to obtain an output vector; wherein the output vector has a dimension that is the same as the dimension of the reference vector.
Specifically, the fully-connected layer comprises a first fully-connected layer and a second fully-connected layer; the output vector acquisition module is further to:
inputting the coding vector into a first full-connection layer to perform first fitting to obtain an initial fitting vector;
and inputting the initial fitting vector into the second full-connection layer to perform second fitting to obtain an output vector, and enabling the dimension of the output vector to be equal to that of the reference vector.
As an embodiment, the training module further includes:
the division module is used for dividing each reference vector and the reverse semantic meaning corresponding to each reference vector into a training set and a verification set;
a loss function obtaining module, configured to train the preset coding model by using the training set to obtain an initial coding model, and obtain a loss function of the initial coding model to obtain a training loss function; inputting the reverse semantic meaning of the verification set into the initial coding model for vector output, and acquiring a current loss function to obtain a verification loss function;
and the judging module is used for judging that the process of obtaining the output vector by the preset coding model reaches a preset training standard when the verification loss function and the training loss function are equal after the training loss function is converged.
In an embodiment, the migration object generating apparatus of the target platform further includes:
the behavior sequence generation module is used for reading each user behavior detected in a set time period of the reference platform and generating a behavior sequence; the behavior sequence records the operation objects of the users on the reference platform and the relevance among the operation objects;
and the reference vector generation module is used for inputting the behavior sequence into a vector coding model to obtain a reference vector adopted by the recommendation system of the reference platform when recommending the object.
For specific limitations of the migration object generation apparatus of the target platform, reference may be made to the above limitations of the migration object generation method of the target platform, which are not described herein again. The modules in the migration object generation apparatus of the target platform may be implemented in whole or in part by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In one embodiment, as shown in fig. 4, there is provided an object recommendation method for a target platform, including the following steps:
s600, generating a migration object of the target platform by adopting the migration object generation method of the target platform in any embodiment;
s700, importing the migration object into a target recommendation system, and enabling the target recommendation system to recommend the object on the target platform; the target recommendation system is a recommendation system of a target platform.
According to the object recommendation method for the target platform, the migration object of the target platform is generated by adopting the migration object generation method for the target platform, the migration object is led into the target recommendation system, and the target recommendation system accurately recommends the object at the initial stage of online of the target platform, so that the target platform can perfectly transition to a cold start stage, and more effective real-time recommendation is carried out at the cold start stage.
In an embodiment, the importing the migration object into a target recommendation system, so that the target recommendation system performs object recommendation on the target platform includes:
importing the migration object into a target recommendation system, and enabling the target recommendation system to read each target vector;
acquiring a user selection object currently received by a target platform, and generating a current operation vector of the user selection object by adopting the target recommendation system; wherein the current operation vector is a low-dimensional dense vector generated by the target recommendation system for the user selection object;
and in the target recommendation system, determining a recommendation object according to the current operation vector and the target vector.
Specifically, the determining a recommended object according to the current operation vector and the target vector includes:
searching vectors with Euclidean distance smaller than a set distance from the current operation vector in the target vector to obtain recommended vectors;
and determining the platform object corresponding to the recommendation vector as the recommendation object.
In the actual use process, after the target platform is online, the user can enable the target platform to detect the current user selection object by inputting commodities to the target platform or clicking corresponding matters on the target platform and the like according to the requirements of the user, so that the target recommendation system can generate the current operation vector of the user selection object, and accurately recommend other objects which need to be operated next by the user according to the current operation vector. The target recommendation system can comprise vector coding models such as an item2vec model and the like so as to accurately generate the required current operation vector for the user to select the object, and further ensure the accuracy of object recommendation performed by the target platform.
Preferably, after the determining the platform object corresponding to the recommendation vector as the recommendation object, the method further includes:
if the number of the recommended vectors is multiple, respectively obtaining Euclidean distances between each recommended vector and the current operation vector;
sorting recommendation objects corresponding to the recommendation vectors respectively according to the Euclidean distance between the recommendation vectors and the current operation vector from small to large;
and recommending the recommended objects on the target platform according to the arrangement sequence of the recommended objects.
According to the embodiment, the recommendation objects arranged in the front can be preferentially recommended, namely, the platform objects which are more likely to be needed by the user are preferentially recommended, so that the user requirements are responded to the maximum extent, and the user experience brought by the target platform can be further improved.
In one embodiment, as shown in fig. 5, there is provided a migration object generation apparatus of a target platform, including:
a migration object generation module 600, configured to generate a migration object of a target platform by using a migration object generation apparatus of the target platform according to any embodiment of the foregoing description;
the object recommendation module 700 is configured to import the migrated object into a target recommendation system, so that the target recommendation system performs object recommendation on the target platform; the target recommendation system is a recommendation system of a target platform.
In one embodiment, the object recommendation module includes:
the target vector reading module is used for importing the migration object into a target recommendation system so that the target recommendation system reads each target vector;
the operation vector generation module is used for acquiring a user selection object currently received by a target platform and generating a current operation vector of the user selection object by adopting the target recommendation system; wherein the current operation vector is a low-dimensional dense vector generated by the target recommendation system for the user selection object;
and the recommended object determining module is used for determining a recommended object in the target recommending system according to the current operation vector and the target vector.
Specifically, the recommended object determining module is further configured to:
searching vectors with Euclidean distance smaller than a set distance from the current operation vector in the target vector to obtain recommended vectors;
and determining the platform object corresponding to the recommendation vector as the recommendation object.
Specifically, the object recommendation module further includes:
the Euclidean distance acquisition module is used for respectively acquiring Euclidean distances between each recommendation vector and the current operation vector if the recommendation vectors are multiple;
the recommended object sorting module is used for sorting the recommended objects corresponding to the recommended vectors according to the Euclidean distance between the recommended vectors and the current operation vector from small to large;
and the sequence recommending module is used for recommending the recommended objects on the target platform according to the arrangement sequence of the recommended objects.
For specific limitations of the object recommendation device of the target platform, reference may be made to the above limitations of the object recommendation method of the target platform, which are not described herein again. The modules in the object recommendation device of the target platform may be implemented in whole or in part by software, hardware, and a combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.
In an embodiment, a computer device is provided, which includes a memory, a processor, and a computer program stored in the memory and executable on the processor, and when the processor executes the computer program, the processor implements the method for generating a migrated object of a target platform according to any of the above embodiments or the method for recommending an object of the target platform according to any of the above embodiments.
Specifically, the internal structure of the computer device may be as shown in fig. 6. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement the migration object generation method of the target platform according to any of the above embodiments or the object recommendation method of the target platform according to any of the above embodiments. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.
Those skilled in the art will appreciate that the architecture shown in fig. 6 is merely a block diagram of some of the structures associated with the disclosed aspects and is not intended to limit the computing devices to which the disclosed aspects apply, as particular computing devices may include more or less components than those shown, or may combine certain components, or have a different arrangement of components.
In one embodiment, a computer-readable storage medium is provided, on which a computer program is stored, and the computer program is executed by a processor to implement the migration object generation method of the target platform according to any of the above embodiments or the object recommendation method of the target platform according to any of the above embodiments.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.
It should be noted that the terms "first \ second \ third" referred to in the embodiments of the present application merely distinguish similar objects, and do not represent a specific ordering for the objects, and it should be understood that "first \ second \ third" may exchange a specific order or sequence when allowed. It should be understood that "first \ second \ third" distinct objects may be interchanged under appropriate circumstances such that the embodiments of the application described herein may be implemented in an order other than those illustrated or described herein.
The terms "comprising" and "having" and any variations thereof in the embodiments of the present application are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product, or device that comprises a list of steps or modules is not limited to the listed steps or modules but may alternatively include other steps or modules not listed or inherent to such process, method, product, or device.
Reference herein to "a plurality" means two or more. "and/or" describes the association relationship of the associated objects, meaning that there may be three relationships, e.g., a and/or B, which may mean: a exists alone, A and B exist simultaneously, and B exists alone. The character "/" generally indicates that the former and latter associated objects are in an "or" relationship.
The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.
Claims (15)
1. A method for generating a migration object of a target platform is characterized by comprising the following steps:
acquiring each reference vector adopted by a reference platform when object recommendation is carried out; the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to user behaviors during the operation of the reference platform;
training a preset coding model according to each reference vector and the reverse semantic meaning corresponding to each reference vector to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse-deducing semantics are semantics obtained by reverse deduction according to corresponding reference vectors;
inputting each platform object provided by a target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determining each target vector as a migration object of the target platform so as to allow a recommendation system of the target platform to recommend the object; wherein the target platform is of the same type as the reference platform.
2. The method of claim 1, wherein the training of the preset coding model according to the respective reference vectors and the semantics corresponding to the respective reference vectors to obtain a vector generation model comprises:
inputting each reverse semantic meaning into the preset coding model, and acquiring output vectors respectively obtained by the preset coding model aiming at each reverse semantic meaning;
and when the process of obtaining the output vector by the preset coding model reaches a preset training standard, determining the vector generation model according to the current model parameters of the preset coding model.
3. The method according to claim 2, wherein the preset coding model comprises a bidirectional coding characterization model and a full connection layer; the step of inputting each reverse-thrust semantic into the preset coding model and acquiring an output vector of the preset coding model respectively obtained aiming at each reverse-thrust semantic comprises the following steps:
inputting each reverse semantic into the bidirectional coding representation model, and acquiring a coding vector which is respectively output by the bidirectional coding representation model aiming at each reverse semantic;
inputting the coding vector into a full-connection layer for fitting to obtain an output vector; wherein the output vector has a dimension that is the same as the dimension of the reference vector.
4. The method of claim 3, wherein the fully-connected layer comprises a first fully-connected layer and a second fully-connected layer; the inputting the coding vector into a full-link layer for fitting to obtain an output vector comprises:
inputting the coding vector into a first full-connection layer to perform first fitting to obtain an initial fitting vector;
and inputting the initial fitting vector into the second full-connection layer to perform second fitting to obtain an output vector, and enabling the dimension of the output vector to be equal to that of the reference vector.
5. The method according to claim 2, wherein before the determining the vector generation model according to the current model parameters of the preset coding model when the process of obtaining the output vector by the preset coding model reaches the preset training standard, the method further comprises:
dividing each reference vector and the reverse semantic meaning corresponding to each reference vector into a training set and a verification set;
training the preset coding model by adopting the training set to obtain an initial coding model, and obtaining a loss function of the initial coding model to obtain a training loss function;
inputting the reverse semantic meaning of the verification set into the initial coding model for vector output, and acquiring a current loss function to obtain a verification loss function;
and after the training loss function is converged, when the verification loss function and the training loss function are equal, judging that the process of obtaining the output vector by the preset coding model reaches a preset training standard.
6. The method according to any one of claims 1 to 5, wherein the obtaining of the reference vectors used by the reference platform in object recommendation further comprises:
reading each user behavior detected in a set time period of a reference platform to generate a behavior sequence; the behavior sequence records the operation objects of the users on the reference platform and the relevance among the operation objects;
and inputting the behavior sequence into a vector coding model to obtain a reference vector adopted by a recommendation system of the reference platform when recommending an object.
7. The method of claim 6, wherein the vector coding model is an item2vec model.
8. A migration object generation apparatus of a target platform, comprising:
the acquisition module is used for acquiring each reference vector adopted by the reference platform when object recommendation is carried out; the reference vector is a low-dimensional dense vector generated by a recommendation system of the reference platform according to user behaviors during the operation of the reference platform;
the training module is used for training a preset coding model according to the reference vectors and the reverse semantic meanings corresponding to the reference vectors to obtain a vector generation model; the preset coding model is used for generating a low-dimensional dense vector of the input object; the reverse-deducing semantics are semantics obtained by reverse deduction according to corresponding reference vectors;
and the determining module is used for inputting each platform object provided by the target platform into the vector generation model to obtain a target vector corresponding to each platform object, and determining each target vector as a migration object of the target platform so as to allow a recommendation system of the target platform to recommend the object.
9. An object recommendation method for a target platform is characterized by comprising the following steps:
generating a migration object of a target platform by adopting the migration object generation method of the target platform according to any one of claims 1 to 7;
importing the migration object into a target recommendation system, and enabling the target recommendation system to perform object recommendation on the target platform; the target recommendation system is a recommendation system of a target platform.
10. The method of claim 9, wherein importing the migrated object into a target recommendation system, such that the target recommendation system makes object recommendations on the target platform comprises:
importing the migration object into a target recommendation system, and enabling the target recommendation system to read each target vector;
acquiring a user selection object currently received by a target platform, and generating a current operation vector of the user selection object by adopting the target recommendation system; wherein the current operation vector is a low-dimensional dense vector generated by the target recommendation system for the user selection object;
and in the target recommendation system, determining a recommendation object according to the current operation vector and the target vector.
11. The method of claim 10, wherein determining a recommended object from the current operation vector and the target vector comprises:
searching vectors with Euclidean distance smaller than a set distance from the current operation vector in the target vector to obtain recommended vectors;
and determining the platform object corresponding to the recommendation vector as the recommendation object.
12. The method of claim 11, wherein after determining the platform object corresponding to the recommendation vector as the recommendation object, further comprising:
if the number of the recommended vectors is multiple, respectively obtaining Euclidean distances between each recommended vector and the current operation vector;
sorting recommendation objects corresponding to the recommendation vectors respectively according to the Euclidean distance between the recommendation vectors and the current operation vector from small to large;
and recommending the recommended objects on the target platform according to the arrangement sequence of the recommended objects.
13. An object recommendation apparatus for a target platform, comprising:
a migration object generation module, configured to generate a migration object of a target platform by using the migration object generation apparatus of the target platform according to claim 8;
the object recommendation module is used for importing the migration object into a target recommendation system so that the target recommendation system carries out object recommendation on the target platform; the target recommendation system is a recommendation system of a target platform.
14. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor implements the method for generating migrated objects for a target platform of any of claims 1 to 7 or the method for recommending objects for a target platform of any of claims 9 to 12 when executing the computer program.
15. A computer-readable storage medium on which a computer program is stored, the computer program, when executed by a processor, implementing the migration object generation method of the target platform of any one of claims 1 to 7 or the object recommendation method of the target platform of any one of claims 9 to 12.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011627705.8A CN112598483A (en) | 2020-12-30 | 2020-12-30 | Migration object generation and object recommendation method and device for target platform |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202011627705.8A CN112598483A (en) | 2020-12-30 | 2020-12-30 | Migration object generation and object recommendation method and device for target platform |
Publications (1)
Publication Number | Publication Date |
---|---|
CN112598483A true CN112598483A (en) | 2021-04-02 |
Family
ID=75206483
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202011627705.8A Pending CN112598483A (en) | 2020-12-30 | 2020-12-30 | Migration object generation and object recommendation method and device for target platform |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN112598483A (en) |
Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130265852A1 (en) * | 2012-04-05 | 2013-10-10 | Westerngeco L.L.C. | Converting a first acquired data subset to a second acquired data subset |
CN110377714A (en) * | 2019-07-18 | 2019-10-25 | 泰康保险集团股份有限公司 | Text matching technique, device, medium and equipment based on transfer learning |
CN110910218A (en) * | 2019-11-21 | 2020-03-24 | 南京邮电大学 | Multi-behavior migration recommendation method based on deep learning |
CN111061700A (en) * | 2019-11-12 | 2020-04-24 | 山大地纬软件股份有限公司 | Hospitalizing migration scheme recommendation method and system based on similarity learning |
-
2020
- 2020-12-30 CN CN202011627705.8A patent/CN112598483A/en active Pending
Patent Citations (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20130265852A1 (en) * | 2012-04-05 | 2013-10-10 | Westerngeco L.L.C. | Converting a first acquired data subset to a second acquired data subset |
CN110377714A (en) * | 2019-07-18 | 2019-10-25 | 泰康保险集团股份有限公司 | Text matching technique, device, medium and equipment based on transfer learning |
CN111061700A (en) * | 2019-11-12 | 2020-04-24 | 山大地纬软件股份有限公司 | Hospitalizing migration scheme recommendation method and system based on similarity learning |
CN110910218A (en) * | 2019-11-21 | 2020-03-24 | 南京邮电大学 | Multi-behavior migration recommendation method based on deep learning |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN108874992B (en) | Public opinion analysis method, system, computer equipment and storage medium | |
US11032422B1 (en) | Heuristic sales agent training assistant | |
CN109582876B (en) | Tourist industry user portrait construction method and device and computer equipment | |
CN111431742B (en) | Network information detection method, device, storage medium and computer equipment | |
CN108228873A (en) | Object recommendation, publication content delivery method, device, storage medium and equipment | |
US20190354887A1 (en) | Knowledge graph based learning content generation | |
CN112711937A (en) | Template recommendation method, device, equipment and storage medium | |
CN111192025A (en) | Occupational information matching method and device, computer equipment and storage medium | |
CN109617781B (en) | Instant communication message monitoring method and device, computer equipment and storage medium | |
CN111192153B (en) | Crowd relation network construction method, device, computer equipment and storage medium | |
CN113377936A (en) | Intelligent question and answer method, device and equipment | |
CN111309887B (en) | Method and system for training text key content extraction model | |
US20230009527A1 (en) | User Presence Detection for Authentication Question Generation | |
CN110955608B (en) | Test data processing method, device, computer equipment and storage medium | |
CN110858353B (en) | Method and system for obtaining case judge result | |
CN112949297A (en) | Intention recognition method, device, computer equipment and computer readable storage medium | |
CN111475628B (en) | Session data processing method, apparatus, computer device and storage medium | |
CN112200538A (en) | Data processing method, device, equipment and storage medium | |
CN110362662A (en) | Data processing method, device and computer readable storage medium | |
CN110442713A (en) | Abstract generation method, apparatus, computer equipment and storage medium | |
CN110532229B (en) | Evidence file retrieval method, device, computer equipment and storage medium | |
US11269896B2 (en) | System and method for automatic difficulty level estimation | |
CN117786086A (en) | Reply text generation method, reply text generation device, computer equipment and readable storage medium | |
CN113806568B (en) | Multimedia resource recommendation method and device, electronic equipment and storage medium | |
Liu et al. | [Retracted] Deep Learning and Collaborative Filtering‐Based Methods for Students’ Performance Prediction and Course Recommendation |
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 |