CN108876422B - Method and device for information popularization, electronic equipment and computer readable medium - Google Patents
Method and device for information popularization, electronic equipment and computer readable medium Download PDFInfo
- Publication number
- CN108876422B CN108876422B CN201710322279.9A CN201710322279A CN108876422B CN 108876422 B CN108876422 B CN 108876422B CN 201710322279 A CN201710322279 A CN 201710322279A CN 108876422 B CN108876422 B CN 108876422B
- Authority
- CN
- China
- Prior art keywords
- information
- model
- request information
- data
- ranking
- 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
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]
- G06Q30/0631—Item recommendations
-
- 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
- G06Q30/02—Marketing; Price estimation or determination; Fundraising
- G06Q30/0241—Advertisements
- G06Q30/0251—Targeted advertisements
- G06Q30/0263—Targeted advertisements based upon Internet or website rating
Landscapes
- Business, Economics & Management (AREA)
- Accounting & Taxation (AREA)
- Finance (AREA)
- Strategic Management (AREA)
- Engineering & Computer Science (AREA)
- Development Economics (AREA)
- General Physics & Mathematics (AREA)
- Economics (AREA)
- Marketing (AREA)
- Physics & Mathematics (AREA)
- General Business, Economics & Management (AREA)
- Theoretical Computer Science (AREA)
- Game Theory and Decision Science (AREA)
- Entrepreneurship & Innovation (AREA)
- Management, Administration, Business Operations System, And Electronic Commerce (AREA)
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The application discloses a method and a device for information popularization, electronic equipment and a computer readable medium. The method comprises the following steps: acquiring request information in real time, wherein the request information comprises user information and popularization information; mapping the request information into a natural language to generate first data; inputting the first data into a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model; and promoting the promotion information in the request information according to the sorting scores. The method, the device, the electronic equipment and the computer readable medium for information popularization can reduce work such as feature selection and modeling, avoid a large amount of feature engineering, increase generalization capability of the sequencing method, and simplify model complexity by adopting the same network processing for all features.
Description
Technical Field
The invention relates to the field of big data information processing, in particular to a method and a device for information popularization, electronic equipment and a computer readable medium.
Background
The sorting method is an information promotion, and can be a core technology in a recommendation/advertisement system, for example, the sorting method directly influences the display sequence of recommended articles or advertisements, and further influences the click of a user and subsequent purchase conversion. The good ranking method may allow better quality recommendations/advertisements to be ranked in the top position, thereby promoting user clicks and consumption. Therefore, the ranking algorithm is a very critical ring in the recommendation/advertisement system. Currently, the ranking scheme adopted in recommendation/advertisement systems mainly comprises two steps or modules: firstly, extracting and processing relevant features; the candidate recommendations or advertisements are then ranked. In the feature extraction stage, features of relevant dimensions, such as user dimension features, context features and the like, are extracted from the recommendation/advertisement request, and then a series of feature engineering processing is performed.
At present, most sequencing methods need to do a large amount of feature work to process features in a feature extraction stage, model the features, and do different processing aiming at different features, which is too complex. The various model algorithms employed in the ranking stage also have corresponding disadvantages. The mainstream model algorithm based on shallow machine learning is a linear model essentially, only linear transformation of features can be learned, and extraction and expression of many features are not sufficient, especially some non-linear features. To introduce a non-linear transformation in the ordering, such as logistic regression, feature combinations are typically performed on the basis of existing features. This also leads to two disadvantages: first, many additional feature engineering needs to be performed, such as feature screening, determining how features are combined, etc.; secondly, the feature combination can enable the sorting method to learn good memory capacity, can well process frequently-appearing features and combinations in training data, but can also cause the shortage of generalization capacity of the sorting method and the shortage of processing of newly-appearing features and combinations.
Therefore, a new method, apparatus, electronic device and computer readable medium for information promotion is needed.
The above information disclosed in this background section is only for enhancement of understanding of the background of the invention and therefore it may contain information that does not constitute prior art that is already known to a person of ordinary skill in the art.
Disclosure of Invention
In view of this, the present invention provides a method, an apparatus, an electronic device, and a computer-readable medium for information popularization, which can reduce the work of feature selection and modeling, avoid a large amount of feature engineering, increase the generalization capability of the ranking method, and simplify the model complexity because all features are processed by the same network.
Additional features and advantages of the invention will be set forth in the detailed description which follows, or may be learned by practice of the invention.
According to an aspect of the present invention, there is provided a method for information promotion, the method including: acquiring request information in real time, wherein the request information comprises user information and popularization information; mapping the request information into a natural language to generate first data; inputting the first data into a ranking model to determine a ranking score, wherein the ranking model is a deep convolutional neural network model; and promoting the promotion information in the request information according to the sequencing scores.
In an exemplary embodiment of the present disclosure, further comprising: and establishing a sequencing model through the history request information and the history request result.
In an exemplary embodiment of the present disclosure, establishing a ranking model by historical request information and historical request results includes: and training the deep convolutional neural network model through the historical request information and the historical request result to obtain a sequencing model.
In an exemplary embodiment of the present disclosure, a ranking model includes: a model structure corresponding to each layer of the deep convolutional neural network model; and the weight corresponding to the model structure of each layer.
In an exemplary embodiment of the present disclosure, the first data is input to a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model comprising: performing one-hot encoding processing on the first data to generate an encoding matrix; performing convolution processing on the coding matrix to extract a plurality of characteristic matrixes; extracting combined features through a plurality of feature matrixes; and determining a ranking score by combining the features with the activation function.
In an exemplary embodiment of the present disclosure, the first data is input to a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model, further comprising: and performing dimension reduction processing on the feature matrix.
In an exemplary embodiment of the present disclosure, performing one-hot encoding processing on first data to generate an encoding matrix includes: mapping each character in the first data into a column vector; and generating an encoding matrix by the column vectors.
In an exemplary embodiment of the present disclosure, the one-hot encoding processing is performed on the first data to generate an encoding matrix, further including: determining a character table through the first data; and designating an index number of the character in the character table.
In an exemplary embodiment of the present disclosure, convolving the encoding matrix to extract a plurality of feature extraction matrices includes: and performing convolution processing on the coding matrix in a table look-up mode to extract a plurality of feature extraction matrixes.
In an exemplary embodiment of the present disclosure, the convolving the encoding matrix by looking up a table to extract a plurality of feature extraction matrices includes: recording the position with the column vector value of 1 in the coding matrix in a lookup table; during convolution processing, acquiring an element value of a corresponding position of a convolution core through an index value of the corresponding position in a lookup table; and adding the element values to obtain a convolution processing result.
According to an aspect of the present invention, there is provided an apparatus for information promotion, the apparatus including: the information module is used for acquiring request information in real time, wherein the request information comprises user information and popularization information; the data module is used for mapping the request information into a natural language to generate first data; the sorting module is used for inputting the first data into a sorting model to determine a sorting score, and the sorting model is a deep convolutional neural network model; and the promotion module is used for promoting the promotion information in the request information according to the sequencing scores.
In an exemplary embodiment of the present disclosure, further comprising: and the model module is used for establishing a sequencing model through the history request information and the history request result.
According to an aspect of the present invention, there is provided an electronic apparatus including: one or more processors; storage means for storing one or more programs; when executed by one or more processors, cause the one or more processors to implement a method as any one of the above.
According to an aspect of the invention, a computer-readable medium is proposed, on which a computer program is stored, which program, when being executed by a processor, carries out the method as any of the above.
According to the method, the device, the electronic equipment and the computer readable medium for information popularization, the work of feature selection, modeling and the like can be reduced, a large amount of feature engineering is avoided, the generalization capability of the sequencing method is improved, all features are processed by the same network, and the complexity of the model is simplified.
It is to be understood that both the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the invention, as claimed.
Drawings
The above and other objects, features and advantages of the present invention will become more apparent by describing in detail exemplary embodiments thereof with reference to the attached drawings. The drawings described below are only some embodiments of the invention and other drawings may be derived from those drawings by a person skilled in the art without inventive effort.
FIG. 1 is a system architecture illustrating a method for information dissemination in accordance with an exemplary embodiment.
FIG. 2 is a flow diagram illustrating a method for information dissemination in accordance with an example embodiment.
Fig. 3 is a system flow diagram illustrating a method for information dissemination in accordance with another exemplary embodiment.
Fig. 4 is a schematic diagram of a method for information dissemination shown in accordance with another example embodiment.
Fig. 5 is a schematic diagram illustrating convolution processing in a method for information dissemination according to another exemplary embodiment.
FIG. 6 is a block diagram illustrating an apparatus for information dissemination in accordance with an exemplary embodiment.
FIG. 7 is a block diagram illustrating an electronic device in accordance with another example embodiment.
DETAILED DESCRIPTION OF EMBODIMENT (S) OF INVENTION
Example embodiments will now be described more fully with reference to the accompanying drawings. Example embodiments may, however, be embodied in many different forms and should not be construed as limited to the embodiments set forth herein; rather, these embodiments are provided so that this disclosure will be thorough and complete, and will fully convey the concept of example embodiments to those skilled in the art. The same reference numerals denote the same or similar parts in the drawings, and thus, a repetitive description thereof will be omitted.
Furthermore, the described features, structures, or characteristics may be combined in any suitable manner in one or more embodiments. In the following description, numerous specific details are provided to provide a thorough understanding of embodiments of the invention. One skilled in the relevant art will recognize, however, that the invention may be practiced without one or more of the specific details, or with other methods, components, devices, steps, and so forth. In other instances, well-known methods, devices, implementations or operations have not been shown or described in detail to avoid obscuring aspects of the invention.
The block diagrams shown in the figures are functional entities only and do not necessarily correspond to physically separate entities. I.e. these functional entities may be implemented in the form of software, or in one or more hardware modules or integrated circuits, or in different networks and/or processor means and/or microcontroller means.
The flow charts shown in the drawings are merely illustrative and do not necessarily include all of the contents and operations/steps, nor do they necessarily have to be performed in the order described. For example, some operations/steps may be decomposed, and some operations/steps may be combined or partially combined, so that the actual execution sequence may be changed according to the actual situation.
It will be understood that, although the terms first, second, third, etc. may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one element from another. Thus, a first component discussed below could be termed a second component without departing from the teachings of the disclosed concepts. As used herein, the term "and/or" includes any and all combinations of one or more of the associated listed items.
It will be appreciated by those skilled in the art that the drawings are merely schematic representations of exemplary embodiments, and that the blocks or flow charts in the drawings are not necessarily required to practice the present invention and are, therefore, not intended to limit the scope of the present invention.
The following detailed description of exemplary embodiments of the disclosure refers to the accompanying drawings.
Fig. 1 illustrates an exemplary system architecture 100 to which the methods and apparatus for information dissemination of the present application may be applied.
As shown in fig. 1, the system architecture 100 may include terminal devices 101, 102, 103, a network 104, and a server 105. The network 104 serves as a medium for providing communication links between the terminal devices 101, 102, 103 and the server 105. Network 104 may include various connection types, such as wired, wireless communication links, or fiber optic cables, among others.
The user may use the terminal devices 101, 102, 103 to interact with the server 105 via the network 104 to receive or send messages or the like. The terminal devices 101, 102, 103 may have various communication client applications installed thereon, such as a shopping application, a web browser application, a search application, an instant messaging tool, a mailbox client, social platform software, and the like.
The terminal devices 101, 102, 103 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop portable computers, desktop computers, and the like.
The server 105 may be a server providing various services, such as a background management server (for example only) providing support for shopping-like websites browsed by users using the terminal devices 101, 102, 103. The backend management server may analyze and perform other processing on the received data such as the product information query request, and feed back a processing result (for example, target push information, product information — just an example) to the terminal device.
It should be noted that the web page generation method provided in the embodiment of the present application is generally executed by the server 105, and accordingly, the web page generation apparatus is generally disposed in the server 105.
It should be understood that the number of terminal devices, networks, and servers in fig. 1 is merely illustrative. There may be any number of terminal devices, networks, and servers, as desired for implementation.
FIG. 2 is a flow diagram illustrating a method for information dissemination in accordance with an example embodiment.
As shown in fig. 2, in S202, request information is obtained in real time, where the request information includes user information and promotion information. The request information may be, for example: "request { name: wangming, age: 25, gene: male, interest: automobile, ad: audi a4, price: 125$ } ", where" name: wangming, age: 25, gene: male, interest: car "is user information," ad: audi a4, price: 125$ "is the promotion information. The request information for the same user may include a plurality of pieces, and the present invention is not limited by the number.
In S204, the request information is mapped to a natural language to generate first data. In the prior art, it is common practice (a method not mapped to natural language) to extract individual features from the request information, and for example, feature values such as "wangxing", "25", "male", "automobile", "audi a 4", "125 $" are extracted from the above request information, and then modeling processing is performed according to whether the feature values belong to a continuous feature (for example, price) or a category feature (for example, gender, interest, etc.). The method of mapping the request into the natural language does not need to extract the characteristics and further process the characteristics, and only needs to simply map the whole request into a flat character string. For example, the request is directly mapped to a string: "name: wangming, age: 25, gene: male, interest: automobile, ad: audi a4, price: 125$ ". For example, to facilitate uniform processing, chinese may be replaced with pinyin, and the segmentors of the respective fields are uniformly replaced, so as to obtain a more concise character string, and the processed character string may be used as first data for subsequent processing: "name walking analyzing | age 25| gen der nan | interest qiche | ad aodi A4| price125 $". This treatment can bring about two main advantages: firstly, the processing of the characteristics becomes simple, and additional characteristic engineering is not needed; secondly, the whole character string contains all the information carried in the original request and is completely submitted to the subsequent convolutional neural network for processing, so that the problem of information loss is avoided.
In S206, the first data is input to a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model. And inputting the first data into a built sequencing model for sequencing scoring, wherein the sequencing model can be built by adopting a deep convolutional neural network. Deep Convolutional Neural Networks (CNNs), which are one of artificial neural networks, have been the focus of research in the field of speech analysis and image recognition. The weight sharing network structure of the system is more similar to a biological neural network, the complexity of a network model is reduced, and the number of weights is reduced. The method has the advantages that the method is more obvious when the input of the network is multidimensional data, so that the data can be directly used as the input of the network, and the complex characteristic extraction and data reconstruction processes in the traditional recognition algorithm are avoided. Convolutional networks are a multi-layered perceptron specifically designed to recognize two-dimensional shapes, the structure of which is highly invariant to translation, scaling, tilting, or other forms of deformation. CNNs is the first learning algorithm to truly successfully train multi-layer network structures. It utilizes spatial relationships to reduce the number of parameters that need to be learned to improve the training performance of the general forward BP algorithm. CNNs are proposed as a deep learning architecture to minimize the preprocessing requirements of the data.
In S208, the promotion information in the request information is promoted according to the ranking score.
According to the method for information popularization, the request message is converted into the natural language and is input into the sequencing model constructed by the deep convolutional neural network, the grading mode of the popularization message is obtained, the work of feature selection, modeling and the like can be reduced, a large number of feature engineering is avoided, the generalization capability of the sequencing method is improved, all features are processed by the same network, and the complexity of the model is simplified.
It should be clearly understood that the present disclosure describes how to make and use particular examples, but the principles of the present disclosure are not limited to any details of these examples. Rather, these principles can be applied to many other embodiments based on the teachings of the present disclosure.
In an exemplary embodiment of the present disclosure, further comprising: and establishing a sequencing model through the history request information and the history request result. The method comprises the following steps: and training the deep convolutional neural network model through the historical request information and the historical request result to obtain a sequencing model. A ranking model comprising: a model structure corresponding to each layer of the deep convolutional neural network model; and the weight corresponding to the model structure of each layer.
Fig. 3 is a system flow diagram illustrating a method for information dissemination in accordance with another exemplary embodiment.
As shown in FIG. 3, real-time access by an online user requests a server, which screens out candidate advertisements (ad-1, ad-2 … …) according to a policy, and after screening out the candidate advertisements, continues to request a predictor (predictor), which may, for example, load a ranking model to predict scores for the candidate advertisements and then rank them according to the scores. And providing the sequencing result to a server side, and recommending the advertisement to the user by the server side according to the sequencing.
Fig. 4 is a schematic diagram of a method for information dissemination shown in accordance with another example embodiment.
In an exemplary embodiment of the present disclosure, the ranking model is built by a deep convolutional neural network, and the ranking model mainly includes 5 types of structures: an input layer, a convolutional layer, a pooling layer, a fully-connected layer, and an output layer, wherein between the input layer and the fully-connected layer, a number of convolutional and pooling layers are alternately stacked, and the last output layer is preceded by a number of fully-connected layers.
1) Input layer
In an exemplary embodiment of the present disclosure, performing one-hot encoding processing on first data to generate an encoding matrix includes: mapping each character in the first data into a column vector; and generating an encoding matrix by the column vectors. In an exemplary embodiment of the present disclosure, the one-hot encoding processing is performed on the first data to generate an encoding matrix, and the method further includes: determining a character table through the first data; and designating an index number of the character in the character table. The input processed by the convolutional neural network is usually in a matrix form, because the character string to which the request is mapped needs to be encoded in a matrix form. Specifically, the number of all possible characters in the request string is counted first, and a character table is formed, for example: "abcdefghijklmnopqrstuvwyz 0123456798" has a character table length of 36. Each character is then assigned an index number (index) in order, for example: and an index corresponding to a is 1, an index corresponding to b is 2, and so on. And each character can be subjected to one-hot coding according to the index value to obtain a sparse column vector consisting of 0 and 1. For example: a is encoded as a column vector [ 1000.. cndot.)] T B is encoded as a column vector [ 0100.. cndot.. cndot..)] T Each column vector has only index corresponding to position 1, and other positions are all 0. Thus, each character is represented by a unique column vector, and the characters in the whole character string are replaced by the corresponding column vectors, so that the coding matrix of the whole character string is obtained and is sparse. Because the characters are in one-to-one correspondence with their column vectors, the entire encoding process does not result in any loss of information.
According to the method for information promotion, the promotion information is uniformly mapped into the natural language. The information contained in the whole advertisement request is uniformly mapped into a section of text, the information contained in the request is not distinguished and independently extracted, feature modeling is not needed, and a large amount of feature engineering is avoided. The mapped text is directly used as input and submitted to the model for training, and information loss caused by characteristic engineering is avoided.
2) Convolutional layer
In general, local pixels in the graph are closely related, and distant pixels are less correlated. Therefore, each neuron in the network only needs to sense the local part, and then the local information is integrated by the neurons at higher layers to obtain the global information. A convolution kernel is equivalent to a linear transformation on a local view field, a certain specific potential information contained in the local view field is further extracted, the same convolution kernel is used for convolving different local view fields in the whole matrix, a new feature matrix can be obtained, and each position in the new matrix represents a feature value extracted from a certain local view field. One convolution kernel can only be used for extracting one feature and only one new feature matrix can be obtained, and different convolution kernels are used for extracting different kinds of features and obtaining a plurality of different feature matrices. And adopting a plurality of different convolution kernels in each convolution layer, further extracting a plurality of different feature matrixes, and submitting the new feature matrixes obtained by convolution to a higher-layer network by taking the new feature matrixes obtained by convolution as input.
According to the method for information popularization, the mapped text request is processed by utilizing the deep convolutional neural network. And performing one-hot coding on each character in the text request, coding a sparse column vector by each character, splicing the column vectors of all the characters together in sequence to obtain the sparse matrix code of the whole advertisement request, and extracting key information from the coding matrix by using a deep convolutional neural network.
As shown in fig. 5, the request character string is processed at the input layer, and each character is mapped into a column vector, and the length of the column vector is the length of the character table. Therefore, in order to observe the complete information carried by each character in a local view, the length of the convolution kernel needs to be equal to the length of the column vector, and each convolution kernel can be convolved to the whole column vector of the character in the convolution local view. This results in only the input layers having a two-dimensional feature matrix, and the feature matrices output by the first convolutional layer and subsequent convolutional layers having one-dimensional features, so that the convolution kernels of the other convolutional layers are one-dimensional except for the convolution kernel of the first convolutional layer. The convolution operation is equivalent to performing linear transformation on local input, in order to introduce nonlinearity, the output of the convolution operation needs to pass through a nonlinear activation function, and the output value of the activation function is used as the characteristic value in the new characteristic matrix. The activation function may be determined, for example, during the ranking model building process.
3) Pooling layer
The main purpose of the pooling layer is to reduce the dimension of the network following the human visual system, e.g. the mean (or maximum) of the features over an area of the image can be calculated and this mean of the features used to replace all features in this area. These summary statistical features not only can greatly reduce feature dimensionality (compared to using all extracted features), but also improve model results (not easily overfit).
4) Full connection layer
The convolutional layers extract local features by continuously convolving the local input field, and finally the higher convolutional layers can obtain the features of the whole global input by convolving the local field, and the features are highly aggregated features. The combination of the features can be further learned through a plurality of full connection layers, and higher-level combination features are extracted.
5) Output layer
The output layer takes the combined characteristics of the output of the previous full-connection layer as input, carries out linear transformation on the input characteristics through a group of weight parameters, and then gives a score through an activation function, so as to be used as the quality of the current candidate advertisement. In the embodiment of the invention, the activation function and the weight parameter are determined in the process of constructing the model through historical data.
In an exemplary embodiment of the present disclosure, convolving the encoding matrix to extract a plurality of feature extraction matrices includes: and performing convolution processing on the coding matrix in a table look-up mode to extract a plurality of feature extraction matrixes. In an exemplary embodiment of the present disclosure, the convolving the encoding matrix by looking up a table to extract a plurality of feature extraction matrices includes: recording the position with the column vector value of 1 in the coding matrix in a lookup table; during convolution processing, acquiring an element value of a corresponding position of a convolution core through an index value of the corresponding position in a lookup table; and adding the element values to obtain a convolution processing result.
In practical situations, a large amount of data is processed in real time, and the requirement on response time is more strict, so in the embodiment of the invention, an optimization strategy is provided for convolution operation of a sparse matrix, that is, a lookup table mode is adopted to process sparse convolution. The specific method comprises the following steps: taking a convolution operation as an example, the convolution kernel is an 8 x 5 matrix, and the sparse matrix to be convolved is as follows,
1 | 0 | 0 | 0 | 0 |
0 | 0 | 1 | 0 | 0 |
0 | 0 | 0 | 0 | 0 |
0 | 0 | 0 | 0 | 0 |
0 | 0 | 0 | 0 | 0 |
0 | 1 | 0 | 0 | 1 |
0 | 0 | 0 | 0 | 0 |
0 | 0 | 0 | 1 | 0 |
the matrix is very sparse, only five positions with 1 values are effective in the convolution operation, and the positions with 0 values do not affect the final convolution result. However, 8 × 5-40 multiplications and 40 additions are required to obtain the convolution result, which is very inefficient. Especially, a large amount of computation time is consumed when a large number of sparse matrix convolution operations are involved, and most of the computation does not influence the final convolution result. To avoid unnecessary calculations, this can be handled, for example, by means of a look-up table. Specifically, each location (index) having a column vector value of 1 is recorded in a look-up table. For example, the sparse matrix may be expressed as:
8 | 1 | 6 | 2 | 8 | 6 |
the first element in the table indicates the length of the original column vector as 8, and the position (index) of the value 1 in each column vector of the original matrix is indicated in sequence from the second element in the table. Taking the second element '1' as an example, it can be deduced that the length of the first column vector of the original matrix is 8, the first element value of the column vector is 1, and the remaining elements are all 0. When convolution operation is carried out by using a convolution kernel, the element value of the corresponding position of the convolution kernel is searched by using the index value of the corresponding position in the lookup table, and then the corresponding value is taken out and added. Thus, only 5 times of inquiry and 5 times of addition operation are needed, and the calculation complexity is only 1/8.
According to the method for information popularization, general convolution operation is performed on a dense matrix, each element of the matrix can participate in the convolution operation, for a sparse matrix, an element with a value of 0 cannot influence the final convolution result, and if the element is processed in a convolution mode of the general dense matrix, the calculation complexity is high, especially for a highly sparse matrix. The invention provides a sparse convolution method based on a table look-up mode, which can greatly reduce the calculation complexity of convolution operation and improve the operation efficiency of the whole model.
Those skilled in the art will appreciate that all or part of the steps implementing the above embodiments are implemented as computer programs executed by a CPU. The computer program, when executed by the CPU, performs the functions defined by the method provided by the present invention. The program may be stored in a computer readable storage medium, which may be a read-only memory, a magnetic or optical disk, or the like.
Furthermore, it should be noted that the above-mentioned figures are only schematic illustrations of the processes involved in the method according to exemplary embodiments of the invention, and are not intended to be limiting. It will be readily understood that the processes shown in the above figures are not intended to indicate or limit the chronological order of the processes. In addition, it is also readily understood that these processes may be performed synchronously or asynchronously, e.g., in multiple modules.
The following are embodiments of the apparatus of the present invention that may be used to perform embodiments of the method of the present invention. For details which are not disclosed in the embodiments of the apparatus of the present invention, reference is made to the embodiments of the method of the present invention.
FIG. 6 is a block diagram illustrating an apparatus for information promotion, according to an example embodiment.
The information module 602 is configured to obtain request information in real time, where the request information includes user information and popularization information.
The data module 604 is configured to map the request information to a natural language to generate first data.
The ranking module 606 is configured to input the first data into a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model.
The promotion module 608 is configured to promote promotion information in the request information according to the ranking score.
In an exemplary embodiment of the present disclosure, further comprising: the model module (not shown in the figure) is used for establishing a sequencing model through the historical request information and the historical request result.
According to the device for information popularization, the request message is converted into the natural language and is input into the sequencing model constructed by the deep convolutional neural network, the grading mode of the popularization message is obtained, the work of feature selection, modeling and the like can be reduced, a large number of feature engineering is avoided, the generalization capability of the sequencing method is improved, all features are processed by the same network, and the model complexity is simplified.
FIG. 7 is a block diagram illustrating an electronic device in accordance with another example embodiment.
Referring now to FIG. 7, shown is a schematic diagram of an electronic device 700 suitable for use in implementing embodiments of the present application. The electronic device shown in fig. 7 is only an example, and should not bring any limitation to the functions and the scope of use of the embodiments of the present application.
As shown in fig. 7, the computer system 700 includes a Central Processing Unit (CPU)701, which can perform various appropriate actions and processes in accordance with a program stored in a Read Only Memory (ROM)702 or a program loaded from a storage section 708 into a Random Access Memory (RAM) 703. In the RAM 703, various programs and data necessary for the operation of the system 700 are also stored. The CPU701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input/output (I/O) interface 705 is also connected to bus 704.
The following components are connected to the I/O interface 705: an input portion 706 including a keyboard, a mouse, and the like; an output section 707 including a display such as a Cathode Ray Tube (CRT), a Liquid Crystal Display (LCD), and the like, and a speaker; a storage section 708 including a hard disk and the like; and a communication section 709 including a network interface card such as a LAN card, a modem, or the like. The communication section 509 performs communication processing via a network such as the internet. A drive 710 is also connected to the I/O interface 705 as needed. A removable medium 711 such as a magnetic disk, an optical disk, a magneto-optical disk, a semiconductor memory, or the like is mounted on the drive 710 as necessary, so that a computer program read out therefrom is mounted into the storage section 708 as necessary.
In particular, according to an embodiment of the present disclosure, the processes described above with reference to the flowcharts may be implemented as computer software programs. For example, embodiments of the present disclosure include a computer program product comprising a computer program embodied on a computer readable medium, the computer program comprising program code for performing the method illustrated in the flow chart. In such an embodiment, the computer program can be downloaded and installed from a network through the communication section 709, and/or installed from the removable medium 711. The computer program executes the above-described functions defined in the system of the present application when executed by the Central Processing Unit (CPU) 701.
It should be noted that the computer readable medium shown in the present application may be a computer readable signal medium or a computer readable storage medium or any combination of the two. A computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the foregoing. More specific examples of the computer readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing. In the present application, a computer readable storage medium may be any tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. In this application, however, a computer readable signal medium may include a propagated data signal with computer readable program code embodied therein, for example, in baseband or as part of a carrier wave. Such a propagated data signal may take any of a variety of forms, including, but not limited to, electro-magnetic, optical, or any suitable combination thereof. A computer readable signal medium may also be any computer readable medium that is not a computer readable storage medium and that can communicate, propagate, or transport a program for use by or in connection with an instruction execution system, apparatus, or device. Program code embodied on a computer readable medium may be transmitted using any appropriate medium, including but not limited to: wireless, wire, fiber optic cable, RF, etc., or any suitable combination of the foregoing.
The flowchart and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the specified logical function(s). It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams or flowchart illustration, and combinations of blocks in the block diagrams or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The units described in the embodiments of the present application may be implemented by software or hardware. The described units may also be provided in a processor, and may be described as: a processor includes a transmitting unit, an obtaining unit, a determining unit, and a first processing unit. The names of these units do not in some cases constitute a limitation to the unit itself, and for example, the sending unit may also be described as a "unit sending a picture acquisition request to a connected server".
As another aspect, the present application also provides a computer-readable medium, which may be contained in the apparatus described in the above embodiments; or may be separate and not incorporated into the device. The computer readable medium carries one or more programs which, when executed by a device, cause the device to comprise: acquiring request information in real time, wherein the request information comprises user information and popularization information; mapping the request information into a natural language to generate first data; inputting the first data into a ranking model to determine a ranking score, wherein the ranking model is a deep convolutional neural network model; and promoting the promotion information in the request information according to the sequencing scores.
Those skilled in the art will appreciate that the modules described above may be distributed in the apparatus according to the description of the embodiments, or may be modified accordingly in one or more apparatuses unique from the embodiments. The modules of the above embodiments may be combined into one module, or further split into multiple sub-modules.
Through the above description of the embodiments, those skilled in the art will readily understand that the exemplary embodiments described herein may be implemented by software, or by software in combination with necessary hardware. Therefore, the technical solution according to the embodiment of the present invention can be embodied in the form of a software product, which can be stored in a non-volatile storage medium (which can be a CD-ROM, a usb disk, a removable hard disk, etc.) or on a network, and includes several instructions to enable a computing device (which can be a personal computer, a server, a mobile terminal, or a network device, etc.) to execute the method according to the embodiment of the present invention.
From the above detailed description, those skilled in the art can readily appreciate that the method, apparatus, electronic device, and computer-readable medium for information dissemination according to embodiments of the present invention have one or more of the following advantages.
According to the method for information popularization, the request message is converted into the natural language and is input into the sequencing model constructed by the deep convolutional neural network, the grading mode of the popularization message is obtained, the work of feature selection, modeling and the like can be reduced, a large number of feature engineering is avoided, the generalization capability of the sequencing method is improved, all features are processed by the same network, and the complexity of the model is simplified.
According to the method for information popularization, the accuracy of the sorting method can be improved, the clicks and consumption of the user are increased, the conversion rate is improved, and the GMV is increased. The work of feature selection, modeling and the like is reduced, and a large amount of feature engineering is avoided. The generalization capability of the sequencing method is increased, all the characteristics are processed by the same network, and the complexity of the model is simplified.
Exemplary embodiments of the present invention are specifically illustrated and described above. It is to be understood that the invention is not limited to the precise construction, arrangements, or instrumentalities described herein; on the contrary, the invention is intended to cover various modifications and equivalent arrangements included within the spirit and scope of the appended claims.
In addition, the structures, the proportions, the sizes, and the like shown in the drawings of the present specification are only used for matching with the contents disclosed in the specification, so as to be understood and read by those skilled in the art, and are not used for limiting the limit conditions which the present disclosure can implement, so that the present disclosure has no technical essence, and any modification of the structures, the change of the proportion relation, or the adjustment of the sizes, should still fall within the scope which the technical contents disclosed in the present disclosure can cover without affecting the technical effects which the present disclosure can produce and the purposes which can be achieved. In addition, the terms "above", "first", "second" and "a" as used in the present specification are for the sake of clarity only, and are not intended to limit the scope of the present disclosure, and changes or modifications of the relative relationship may be made without substantial technical changes and modifications.
Claims (13)
1. A method for information promotion, comprising:
acquiring request information in real time, wherein the request information comprises user information and popularization information;
mapping the request information into a natural language to generate first data; wherein mapping the request information to a natural language comprises: directly mapping the request information into a character string without feature extraction, wherein Chinese is replaced by pinyin;
inputting the first data into a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model; and
promoting the promotion information in the request information according to the sequencing scores;
wherein the first data is input into a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model comprising:
performing one-hot encoding processing on the first data to generate an encoding matrix;
performing convolution processing on the coding matrix to extract a plurality of feature matrices;
extracting combined features through the feature matrixes; and
determining the ranking score by combining the features with an activation function.
2. The method of claim 1, further comprising:
and establishing the sequencing model through history request information and history request results.
3. The method of claim 2, wherein the establishing the ranking model from historical request information and historical request results comprises:
and training a deep convolutional neural network model through the historical request information and the historical request result to obtain the sequencing model.
4. The method of claim 3, wherein the order model comprises:
a model structure corresponding to each layer of the deep convolutional neural network model; and
the model structure of each layer corresponds to a weight.
5. The method of claim 1, wherein the first data is input to a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model, further comprising:
and performing dimension reduction processing on the feature matrix.
6. The method of claim 1, wherein the performing a one-hot encoding process on the first data to generate an encoding matrix comprises:
mapping each character in the first data into a column vector; and
and generating the coding matrix through the column vectors.
7. The method according to claim 1, wherein said one-hot encoding said first data to generate an encoding matrix, further comprises:
determining a character table through the first data; and
specifying an index number for a character in the character table.
8. The method of claim 1, wherein said convolving said coding matrix to extract a plurality of eigen matrices comprises:
and performing convolution processing on the coding matrix in a table look-up mode to extract a plurality of feature extraction matrixes.
9. The method of claim 8, wherein convolving the coding matrix by table lookup to extract a plurality of feature extraction matrices comprises:
recording the position with the column vector value of 1 in the coding matrix in a lookup table;
during the convolution processing, acquiring an element value of a corresponding position of the convolution core through an index value of the corresponding position in the lookup table;
and adding the element values to obtain the convolution processing result.
10. An apparatus for information dissemination, comprising:
the information module is used for acquiring request information in real time, wherein the request information comprises user information and popularization information;
a data module for mapping the request information to a natural language to generate first data; wherein mapping the request information to a natural language comprises: directly mapping the request information into a character string without feature extraction, wherein Chinese is replaced by pinyin;
a ranking module for inputting the first data into a ranking model to determine a ranking score, the ranking model being a deep convolutional neural network model; and
the promotion module is used for promoting the promotion information in the request information according to the sequencing scores;
wherein the ordering module is to:
performing one-hot encoding processing on the first data to generate an encoding matrix;
performing convolution processing on the coding matrix to extract a plurality of feature matrices;
extracting combined features through the feature matrixes; and
determining the ranking score by combining the features with an activation function.
11. The apparatus of claim 10, further comprising:
and the model module is used for establishing the sequencing model through the history request information and the history request result.
12. An electronic device, comprising:
one or more processors;
storage means for storing one or more programs;
when executed by the one or more processors, cause the one or more processors to implement the method of any one of claims 1-9.
13. A computer-readable medium, on which a computer program is stored, which, when being executed by a processor, carries out the method according to any one of claims 1-9.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710322279.9A CN108876422B (en) | 2017-05-09 | 2017-05-09 | Method and device for information popularization, electronic equipment and computer readable medium |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710322279.9A CN108876422B (en) | 2017-05-09 | 2017-05-09 | Method and device for information popularization, electronic equipment and computer readable medium |
Publications (2)
Publication Number | Publication Date |
---|---|
CN108876422A CN108876422A (en) | 2018-11-23 |
CN108876422B true CN108876422B (en) | 2022-08-12 |
Family
ID=64287898
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710322279.9A Active CN108876422B (en) | 2017-05-09 | 2017-05-09 | Method and device for information popularization, electronic equipment and computer readable medium |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN108876422B (en) |
Families Citing this family (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN109816438B (en) * | 2019-01-11 | 2020-10-30 | 珠海天燕科技有限公司 | Information pushing method and device |
CN110321487A (en) * | 2019-07-01 | 2019-10-11 | 上海硕恩网络科技股份有限公司 | A kind of accurate label recommendations system and its workflow |
CN113743972B (en) * | 2020-08-17 | 2024-06-18 | 北京沃东天骏信息技术有限公司 | Method and device for generating article information |
Family Cites Families (5)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101075320A (en) * | 2006-05-16 | 2007-11-21 | 申凌 | System and method for issuing and inquiring information |
CN102903008B (en) * | 2011-07-29 | 2016-05-18 | 国际商业机器公司 | For the method and system of computer question and answer |
CN102346899A (en) * | 2011-10-08 | 2012-02-08 | 亿赞普(北京)科技有限公司 | Method and device for predicting advertisement click rate based on user behaviors |
CN104598611B (en) * | 2015-01-29 | 2018-03-23 | 百度在线网络技术(北京)有限公司 | The method and system being ranked up to search entry |
CN104915386B (en) * | 2015-05-25 | 2018-04-27 | 中国科学院自动化研究所 | A kind of short text clustering method based on deep semantic feature learning |
-
2017
- 2017-05-09 CN CN201710322279.9A patent/CN108876422B/en active Active
Also Published As
Publication number | Publication date |
---|---|
CN108876422A (en) | 2018-11-23 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
Guo et al. | Efficient hyperparameter optimization for convolution neural networks in deep learning: A distributed particle swarm optimization approach | |
CN112085565B (en) | Deep learning-based information recommendation method, device, equipment and storage medium | |
CN109840730B (en) | Method and device for data prediction | |
CN112119388A (en) | Training image embedding model and text embedding model | |
CN111737578B (en) | Recommendation method and system | |
EP4242955A1 (en) | User profile-based object recommendation method and device | |
JP2019079302A (en) | Sales activity support system, sales activity support method and sales activity support program | |
CN108876422B (en) | Method and device for information popularization, electronic equipment and computer readable medium | |
CN112074828A (en) | Training image embedding model and text embedding model | |
CN112801425B (en) | Method and device for determining information click rate, computer equipment and storage medium | |
CN109189922B (en) | Comment evaluation model training method and device | |
WO2023284516A1 (en) | Information recommendation method and apparatus based on knowledge graph, and device, medium, and product | |
CN113705698B (en) | Information pushing method and device based on click behavior prediction | |
CN111340522A (en) | Resource recommendation method, device, server and storage medium | |
CN115730125A (en) | Object identification method and device, computer equipment and storage medium | |
CN111489196B (en) | Prediction method and device based on deep learning network, electronic equipment and medium | |
CN112449217B (en) | Method and device for pushing video, electronic equipment and computer readable medium | |
US11049041B2 (en) | Online training and update of factorization machines using alternating least squares optimization | |
CN111768218A (en) | Method and device for processing user interaction information | |
CN114298118B (en) | Data processing method based on deep learning, related equipment and storage medium | |
CN114429384A (en) | Intelligent product recommendation method and system based on e-commerce platform | |
CN113378033A (en) | Training method and device for recommendation model | |
CN116992158B (en) | Recommendation method and device for resource allocation, storage medium and electronic device | |
CN116911913B (en) | Method and device for predicting interaction result | |
EP4398128A1 (en) | Recommendation method and related device |
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 |