CN106339510B - Click estimation method and device based on artificial intelligence - Google Patents

Abstract

The present application proposes an artificial intelligence-based click estimation method and device, wherein the method includes: obtaining the attribute information of the entity to be recommended according to the query statement input by the user; Carry out word segmentation processing to determine the features of the query statement and the entity to be recommended; use the preset deep neural network model to determine the click-through rate of the entity to be recommended according to the query statement and the features of the entity to be recommended. Therefore, the method integrates the extracted features into the deep neural network model to estimate the click-through rate of the recommended entity, which improves the accuracy of the click-through rate estimation, enables the recommendation system to provide services for users accurately, and improves the accuracy of the click-through rate. The quality of service of the recommendation system improves the user experience.

Description

Click estimation method and device based on artificial intelligence

technical field

The present application relates to the field of network technology, in particular to an artificial intelligence-based click estimation method and device.

Background technique

Artificial Intelligence (Artificial Intelligence), the English abbreviation is AI. It is a new technical science that studies and develops theories, methods, technologies and application systems for simulating, extending and expanding human intelligence. Artificial intelligence is a branch of computer science that attempts to understand the essence of intelligence and produce a new intelligent machine that can respond in a manner similar to human intelligence. Research in this field includes robotics, language recognition, image recognition, natural language processing and expert systems, etc.

Click through rate (CTR) estimation is one of the classic problems in the application of big data technology. An important aspect of click-through rate estimation is to find out the most suitable advertisement or recommended product and present it to users. At present, in the recommendation fields such as advertising and finance, the click-through rate of the recommended product is usually estimated by the logistic regression (LR) model, and the query statement entered by the user and the feature value of the recommended entity obtained are calculated. Linear weighting and non-linear operations can determine the click-through rate of the entity to be recommended.

However, in the form of determining the click-through rate of the entity to be recommended by using the LR model, the accuracy of the click-through rate of the entity to be recommended is low due to limitations of the LR model.

Contents of the invention

This application aims to solve one of the technical problems in the related art at least to a certain extent.

For this reason, the first purpose of this application is to propose a click estimation method based on artificial intelligence. This method integrates the extracted features into the deep neural network model to estimate the click rate of the entity to be recommended, which improves the The accuracy of click-through rate estimation enables the recommendation system to accurately provide services for users, improves the service quality of the recommendation system, and improves user experience.

The second purpose of the present application is to propose an artificial intelligence-based click estimation device.

The third purpose of the present application is to propose an artificial intelligence-based click estimation device.

The fourth objective of the present application is to provide a non-transitory computer-readable storage medium.

The fifth object of the present application is to propose a computer program product.

In order to achieve the above purpose, the embodiment of the first aspect of the present application proposes a click prediction method based on artificial intelligence, which includes: obtaining the attribute information of the entity to be recommended according to the query sentence input by the user; The attribute information of the entity is separately processed to determine the characteristics of the query statement and the entity to be recommended; using the preset deep neural network model, according to the query statement and the characteristics of the entity to be recommended, determine the click of the entity to be recommended Rate.

According to the artificial intelligence-based click estimation method of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended according to the query sentence input by the user, the query sentence and the attribute information of the entity to be recommended can be respectively subjected to word segmentation processing to determine The query statement and the characteristics of the entity to be recommended are then used to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended using a preset deep neural network model. Therefore, by integrating the extracted features into the deep neural network model, the click-through rate of the entity to be recommended is estimated, which improves the accuracy of the click-through rate estimation, enables the recommendation system to provide services for users accurately, and improves the recommendation system. The quality of service of the system improves the user experience.

In order to achieve the above purpose, the embodiment of the second aspect of the present application proposes an artificial intelligence-based click prediction device, including: an acquisition module, which is used to acquire the attribute information of the entity to be recommended according to the query sentence input by the user; a word segmentation module, It is used to separately perform word segmentation processing on the query statement and the attribute information of the entity to be recommended, and determine the characteristics of the query statement and the entity to be recommended; the processing module is used to use the preset deep neural network model, according to the query statement and the characteristics of the entity to be recommended, and determine the click rate of the entity to be recommended.

According to the artificial intelligence-based click estimation device of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended according to the query sentence input by the user, the query sentence and the attribute information of the entity to be recommended can be respectively subjected to word segmentation processing to determine The query statement and the characteristics of the entity to be recommended are then used to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended using a preset deep neural network model. Therefore, by integrating the extracted features into the deep neural network model, the click-through rate of the entity to be recommended is estimated, which improves the accuracy of the click-through rate estimation, enables the recommendation system to provide services for users accurately, and improves the recommendation system. The quality of service of the system improves the user experience.

To achieve the above purpose, the embodiment of the third aspect of the present application proposes an artificial intelligence-based click estimation device, including:

processor;

memory for storing processor-executable instructions;

Wherein, the processor is configured to execute the artificial intelligence-based click estimation method in the first aspect above.

To achieve the above purpose, the embodiment of the fourth aspect of the present application proposes a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the mobile terminal, the mobile terminal can execute a Such as the artificial intelligence-based click estimation method in the first aspect above.

In order to achieve the above purpose, the embodiment of the fifth aspect of the present application proposes a computer program product, when the instruction processor in the computer program product executes, an artificial intelligence-based click prediction as in the first aspect above is performed. estimation method.

Description of drawings

The above and/or additional aspects and advantages of the present invention will become apparent and easy to understand from the following description of the embodiments in conjunction with the accompanying drawings, wherein:

Fig. 1 is a flowchart of an artificial intelligence-based click estimation method according to an embodiment of the present application;

FIG. 2 is a schematic flow chart of estimating the click-through rate using a deep neural network model provided by the embodiment of the present application;

FIG. 3 is a schematic diagram of the process of determining the vector corresponding to the query sentence provided by the embodiment of the present application;

Fig. 4 is a schematic diagram of the process of determining the click-through rate according to the query statement and the vector of the entity to be recommended;

FIG. 5 is a flowchart of an artificial intelligence-based click estimation method according to another embodiment of the present application;

FIG. 6 is a schematic diagram of the click label estimation process provided by the embodiment of the present application;

FIG. 7 is a schematic structural diagram of an artificial intelligence-based click estimation device according to an embodiment of the present application;

FIG. 8 is a schematic structural diagram of an artificial intelligence-based click estimation device according to another embodiment of the present application.

Detailed ways

Embodiments of the present application are described in detail below, examples of which are shown in the drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below by referring to the figures are exemplary, and are intended to explain the present application, and should not be construed as limiting the present application.

The artificial intelligence-based click estimation method and device according to the embodiments of the present application will be described below with reference to the accompanying drawings.

FIG. 1 is a flow chart of an artificial intelligence-based click estimation method according to an embodiment of the present application.

As shown in Figure 1, the artificial intelligence-based click estimation method includes:

Step 101, according to the query sentence input by the user, attribute information of the entity to be recommended is obtained.

Specifically, the executor of the artificial intelligence-based click estimation method provided in the embodiment of the present invention is the artificial intelligence-based click estimation device provided by this application, which can be configured in any search and recommendation service system to When a user searches, the click rate of the entity to be recommended is estimated.

Wherein, the attribute information of the entity to be recommended includes at least one of the following information: the name of the entity to be recommended, the reason for recommendation, and the identifier of the entity to be recommended. For example, if the query sentence entered by the user is "what are the interesting places near Shaanxi", the click estimation device searches the database through query and determines that one of the entities to be recommended is "Shaanxi Xiangshan". At the same time, in the entity database, it also includes The unique identifier "sxxs" corresponding to "Shaanxi Xiangshan" and the corresponding recommended "provincial scenic spot". Then, the attribute information of the entity to be recommended "Shaanxi Xiangshan" corresponding to the query sentence "what interesting places are there near Shaanxi" may include "Shaanxi Xiangshan", "sxxs" and "provincial scenic spot".

It is understandable that the more content included in the attribute information of the entity to be recommended, that is, the more accurate the portrait of the entity to be recommended is, the more accurate the click rate estimation of the entity to be recommended will be.

Step 102, perform word segmentation processing on the query statement and attribute information of the entity to be recommended, and determine the characteristics of the query statement and the entity to be recommended.

During specific implementation, in order to make the obtained query statement and the features of the entity to be recommended as accurate as possible, the click estimation device may adopt various word segmentation methods to perform word segmentation processing on the query statement and the attribute information of the entity to be recommended. For example, word segmentation processing is performed on the query statement and the attribute information of the entity to be recommended according to different granularities of word segmentation, and word segmentation of different granularities respectively included in the query statement and the entity to be recommended is determined.

For example, for the query sentence "what are the interesting places near Shaanxi", according to the fine-grained word segmentation (basicword segment), the result obtained is Shaanxi|nearby|has|what|interesting|places, and the coarse-grained word segmentation ( phraseword segment) the result obtained is Shaanxi|near|near|has|what|interesting places. For a given entity entity, we also perform word segmentation at two granularities and the result is: Shaanxi|Xiangshan. At the same time, we can get the unique identifier id of the entity in the entity database, and a recommendation reason of the entity=provincial scenic spot. Therefore, we can get the two granularity segmentation words of the recommendation reason respectively: provincial | scenic spot | scenic spot | area, and provincial | scenic spot | scenic spot. Therefore, in the context of search and recommendation, we can select query features including fine-grained word segmentation results and coarse-grained word segmentation results; entity features include coarse-grained word segmentation results of the entity itself, id, and recommendation reasons Coarse and fine-grained word segmentation results.

Step 103, using a preset Deep Neural Network (DNN for short) model to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended.

Specifically, the click prediction device can train a preset deep neural network model by combining a large number of query sentences and corresponding entities to be recommended with a deep neural network, which is used to make recommendations based on the characteristics of query sentences and entities to be recommended. The click-through rate of the entity is estimated.

In specific implementation, when using the preset deep neural network model to estimate the click-through rate of the entity to be recommended, it can mainly be realized through the steps shown in FIG. 2 . FIG. 2 is a schematic flow chart of estimating the click-through rate by using the deep neural network model provided by the embodiment of the present application.

Step 201 , according to word segmentations of different granularities included in the query sentence and the entity to be recommended, and using a preset mapping relationship between word segmentation and vectors, determine vectors corresponding to the query sentence and the entity to be recommended respectively.

Specifically, firstly, each word segment of query and entity is mapped to the embedding layer in the DNN model through the form of vector mapping, and each word segment corresponds to an embedding representation. Among them, for the same word, its corresponding embedding representation is unique. Then add the embedding vectors of each word segmentation in the fine-grained word segmentation to obtain an embedding representation of the fine-grained word segmentation of the query, as shown in Figure 3, Figure 3 is the process of determining the corresponding vector of the query statement provided by the embodiment of the present application schematic diagram.

It should be noted that, in addition to the method of vector addition, the embedding representation of a feature can also be obtained by vector convolution and other methods. For other query and entity features, we also use the same method to obtain embedding representations of different features. For the entity id, we also map each id to a separate embedding to obtain the embedding representation of the entity id.

It can be understood that after embedding mapping, the number of vectors corresponding to query and entity is related to the type of word segmentation granularity for query and entity and whether the word segmentation corresponding to different word segmentation granularity is the same. For example, if two granularities of word segmentation are used for query and entity respectively, and the word segmentation obtained after the two granularities of word segmentation is different, then after embedding mapping, the number of vectors corresponding to query is 2, and the number of vectors corresponding to entity After the name and recommendation reason are segmented at two granularities, the obtained word segmentation is the same, so the number of vectors corresponding to the entity is 3, respectively, the entity name corresponds to 1 vector, the entity id corresponds to 1 vector, and the recommendation reason of the entity corresponds to 1 vectors.

Step 202, using a first preset operation rule to perform calculations on the vectors corresponding to the query statement and the entity to be recommended, to determine the click-through rate of the entity to be recommended.

Specifically, after determining the vectors corresponding to the query and the entity, the click estimation device can determine the click rate of the entity through certain calculations.

Usually in the DNN model, the final embedding vector representation of query and entity can be determined through linear transformation and nonlinear transformation, that is, the above step 202, specifically:

Carrying out linear transformation and nonlinear transformation to the vector corresponding to the query statement and the entity to be recommended respectively, and determining the new vector corresponding to the query statement and the entity to be recommended respectively;

The click rate of the entity to be recommended is determined according to the inner product of the query statement and the new vector corresponding to the entity to be recommended.

For example, FIG. 4 is a schematic diagram of the process of determining the click rate according to the query statement and the vector of the entity to be recommended. As shown in Figure 4, after obtaining the embedding representations of various features, we add the embedding vectors of all the features of the query to obtain the unique embedding representation of the query, and add the embedding vectors of all the features of the entity to obtain the unique embedding representation of the entity . Then after a linear transformation and a nonlinear transformation, a new vector representation of query and entity is obtained. Next, we calculate the inner product of the two vectors and map it to a value between 0 and 1, which is the CTR value.

It should be noted that the above-mentioned nonlinear transformation refers to any nonlinear operation, such as cosine operation, sine operation, tangent operation, cotangent operation, etc., which is not limited in this embodiment.

It can be understood that the click estimation device can estimate the click-through rate of each entity to be recommended corresponding to the query sentence input by the user through the above method, so as to recommend to the user according to the click-through rate of each entity to be recommended .

According to the artificial intelligence-based click estimation method of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended according to the query sentence input by the user, the query sentence and the attribute information of the entity to be recommended can be respectively subjected to word segmentation processing to determine The query statement and the characteristics of the entity to be recommended are then used to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended using a preset deep neural network model. Therefore, by integrating the extracted features into the deep neural network model, the click-through rate of the entity to be recommended is estimated, which improves the accuracy of the click-through rate estimation, enables the recommendation system to provide services for users accurately, and improves the recommendation system. The quality of service of the system improves the user experience.

From the above analysis, it can be seen that the click estimation device can estimate the click rate of the entity to be recommended by using the preset DNN model according to the query statement and the characteristics of the entity to be recommended. In actual implementation, since the click-through rate of the entity to be recommended is related to the displayed position, after determining the click-through rate of the entity to be recommended, it can also be combined with the display position of the recommended entity to determine whether the user clicks on the recommended entity. For estimation, the artificial intelligence-based click estimation method provided in the embodiment of the present application will be further described below in conjunction with FIG. 5 .

FIG. 5 is a flowchart of an artificial intelligence-based click estimation method according to another embodiment of the present application.

As shown in Figure 5, the artificial intelligence-based click estimation method includes the following steps:

Step 501, according to the query sentence input by the user, attribute information of the entity to be recommended is acquired.

Step 502: Perform word segmentation processing on the query statement and the attribute information of the entity to be recommended according to different word segmentation granularities, and determine the word segmentation of different granularities included in the query statement and the entity to be recommended respectively.

Step 503 , according to the word segmentations of different granularities included in the query sentence and the entity to be recommended, and using the preset mapping relationship between word segmentation and vectors, determine the vectors corresponding to the query sentence and the entity to be recommended respectively.

Step 504, perform linear transformation and nonlinear transformation on the vectors corresponding to the query statement and the entity to be recommended, respectively, and determine new vectors corresponding to the query statement and the entity to be recommended respectively.

Step 505: Determine the click-through rate of the entity to be recommended according to the inner product of the query statement and the new vector corresponding to the entity to be recommended.

Step 506, determine the display position of the entity to be recommended.

Specifically, the click estimation device may determine the display position (entity display position) of the entity to be recommended in various ways. For example, the display position of the entity to be recommended is determined according to the user's historical behavior log; or, the display position of the entity to be recommended is determined according to the click rate of the entity to be recommended; or, the display position of the entity to be recommended is assigned by random allocation and many more.

Step 507: Determine the click label of the entity to be recommended according to the click rate and the display position of the entity to be recommended.

Specifically, the click estimation device may still use the DNN model to determine the click label of the entity to be recommended according to the click rate and the display position of the entity to be recommended. That is, the above step 507 includes:

determining a correction vector corresponding to the entity to be recommended according to the click-through rate and the display position of the entity to be recommended;

The correction vector is calculated by using a second preset operation rule to determine the click label corresponding to the entity to be recommended.

Wherein, the calculation method included in the second preset operation rule may be the same as or different from the operation method included in the first preset operation rule, which is not limited in this embodiment.

For example, FIG. 6 is a schematic diagram of a click label estimation process provided by the embodiment of the present application. As shown in Figure 6, first determine the vector representations corresponding to the CTR score and entity display position in the form of vector mapping, and then concatenate the vectors of the CTR score and entity display position as the correction vector of the entity to be recommended, and then correct the The vector is subjected to linear transformation and nonlinear transformation to determine the click label corresponding to the entity to be recommended.

It should be noted that the click estimation device can process the correction vector of the entity to be recommended through linear transformation and sub-linear transformation, and then determine the click label corresponding to the entity to be recommended according to the obtained vector value. For example, if the obtained If the final vector value is greater than 0.5, the corresponding click label is "easy to click"; if the obtained vector value is less than 0.2, the corresponding click label is "difficult to click"; if the obtained vector value is less than 0.5 , and a value greater than 0.2, the corresponding click label is "low probability of click", etc.

In the artificial intelligence-based click prediction method of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended corresponding to the query statement input by the user, firstly, different word segmentation granularities are used to analyze the query statement and the attribute information of the recommended entity. Carry out word segmentation processing to determine the word segmentation of different granularities included in the query statement and the entity to be recommended, and then determine the vector representations corresponding to the word segmentation of the query statement and the entity to be recommended through vector mapping, and then determine through linear transformation and nonlinear transformation Click rate of the entity to be recommended, and then predict whether the entity to be recommended is clicked based on the display position of the entity to be recommended. As a result, the DNN model is used to predict the click-through rate and click tags of the recommended entity, which improves the accuracy of the predicted click-through rate of the recommended entity, improves the service quality of the recommendation system, and improves the user experience.

In order to realize the above embodiments, the present application also proposes an artificial intelligence-based click estimation device.

FIG. 7 is a schematic structural diagram of an artificial intelligence-based click estimation device according to an embodiment of the present application.

As shown in Figure 7, the artificial intelligence-based click estimation device includes:

An acquisition module 71, configured to acquire the attribute information of the entity to be recommended according to the query sentence input by the user;

The word segmentation module 72 is used to perform word segmentation processing on the query sentence and the attribute information of the entity to be recommended, and determine the characteristics of the query sentence and the entity to be recommended;

The processing module 73 is configured to use a preset deep neural network model to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended.

Wherein, the attribute information of the entity to be recommended includes at least one of the following information: a name of the entity to be recommended, a reason for recommendation, and an identifier of the entity to be recommended.

In a possible implementation form of this embodiment, the above word segmentation module 72 is specifically used for:

According to different granularities of word segmentation, word segmentation processing is performed on the query statement and the attribute information of the entity to be recommended respectively, and word segmentation of different granularities included in the query statement and the entity to be recommended are respectively determined.

Correspondingly, the above processing module 73 includes:

The determination unit is used to determine the corresponding vectors of the query statement and the entity to be recommended according to the word segmentation of different granularities included in the query statement and the entity to be recommended respectively, using the preset mapping relationship between the word segmentation and the vector;

The calculation unit is configured to use a first preset calculation rule to perform calculations on the vectors corresponding to the query statement and the entity to be recommended, so as to determine the click-through rate of the entity to be recommended.

In a possible implementation form of this embodiment, the computing unit is specifically configured to:

Carrying out linear transformation and nonlinear transformation to the vector corresponding to the query statement and the entity to be recommended respectively, and determining the new vector corresponding to the query statement and the entity to be recommended respectively;

The click rate of the entity to be recommended is determined according to the inner product of the query statement and the new vector corresponding to the entity to be recommended.

It should be noted that the foregoing explanations of the embodiment of the artificial intelligence-based click estimation method are also applicable to the artificial intelligence-based click estimation device of this embodiment, and details are not repeated here.

According to the artificial intelligence-based click estimation device of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended according to the query sentence input by the user, the query sentence and the attribute information of the entity to be recommended can be respectively subjected to word segmentation processing to determine The query statement and the characteristics of the entity to be recommended are then used to determine the click-through rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended using a preset deep neural network model. Therefore, by integrating the extracted features into the deep neural network model, the click-through rate of the entity to be recommended is estimated, which improves the accuracy of the click-through rate estimation, enables the recommendation system to provide services for users accurately, and improves the recommendation system. The quality of service of the system improves the user experience.

FIG. 8 is a schematic structural diagram of an artificial intelligence-based click estimation device according to another embodiment of the present application.

As shown in Figure 8, on the basis of the above-mentioned Figure 7, the artificial intelligence-based click estimation device also includes:

The first determining module 81 is configured to determine the display position of the entity to be recommended;

The second determination module 82 is configured to determine the click label of the entity to be recommended according to the click rate and the display position of the entity to be recommended.

Specifically, the above-mentioned first determination module 81 is specifically used for:

According to the historical behavior log of the user, the display position of the entity to be recommended is determined.

Correspondingly, the above-mentioned second determination module 82 is specifically used for:

determining a correction vector corresponding to the entity to be recommended according to the click-through rate and the display position of the entity to be recommended;

The correction vector is calculated by using a second preset operation rule to determine the click label corresponding to the entity to be recommended.

It should be noted that the foregoing explanations of the embodiment of the artificial intelligence-based click estimation method are also applicable to the artificial intelligence-based click estimation device of this embodiment, and details are not repeated here.

In the artificial intelligence-based click prediction device of the embodiment of the present application, after obtaining the attribute information of the entity to be recommended corresponding to the query statement input by the user, firstly, different word segmentation granularities are used to analyze the query statement and the attribute information of the recommended entity. Carry out word segmentation processing to determine the word segmentation of different granularities included in the query statement and the entity to be recommended, and then determine the vector representations corresponding to the word segmentation of the query statement and the entity to be recommended through vector mapping, and then determine through linear transformation and nonlinear transformation Click rate of the entity to be recommended, and then predict whether the entity to be recommended is clicked based on the display position of the entity to be recommended. As a result, the DNN model is used to predict the click-through rate and click tags of the recommended entity, which improves the accuracy of the predicted click-through rate of the recommended entity, improves the service quality of the recommendation system, and improves the user experience.

Based on the above-mentioned embodiments, another embodiment of the present application provides an artificial intelligence-based click estimation device, including:

processor;

memory for storing processor-executable instructions;

Wherein, the processor is configured to perform the following operations: acquire the attribute information of the entity to be recommended according to the query statement input by the user;

Perform word segmentation processing on the query statement and the attribute information of the entity to be recommended, and determine the characteristics of the query statement and the entity to be recommended;

Using a preset deep neural network model, according to the query statement and the characteristics of the entity to be recommended, determine the click rate of the entity to be recommended.

Further, the embodiment of the present application also provides a non-transitory computer-readable storage medium. When the instructions in the storage medium are executed by the processor of the mobile terminal, the mobile terminal can The click estimation method for .

Further, the embodiment of the present application also provides a computer program product, and when the instruction processor in the computer program product executes, an artificial intelligence-based click estimation method as shown in the above-mentioned embodiments is executed.

In the description of this specification, descriptions with reference to the terms "one embodiment", "some embodiments", "example", "specific examples", or "some examples" mean that specific features described in connection with the embodiment or example , structure, material or characteristic is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms are not necessarily directed to the same embodiment or example. Furthermore, the described specific features, structures, materials or characteristics may be combined in any suitable manner in any one or more embodiments or examples. In addition, those skilled in the art can combine and combine different embodiments or examples and features of different embodiments or examples described in this specification without conflicting with each other.

In addition, the terms "first" and "second" are used for descriptive purposes only, and cannot be interpreted as indicating or implying relative importance or implicitly specifying the quantity of indicated technical features. Thus, the features defined as "first" and "second" may explicitly or implicitly include at least one of these features. In the description of the present application, "plurality" means at least two, such as two, three, etc., unless otherwise specifically defined.

Any process or method descriptions in flowcharts or otherwise described herein may be understood to represent a module, segment or portion of code comprising one or more executable instructions for implementing custom logical functions or steps of a process , and the scope of preferred embodiments of the present application includes additional implementations in which functions may be performed out of the order shown or discussed, including in substantially simultaneous fashion or in reverse order depending on the functions involved, which shall It should be understood by those skilled in the art to which the embodiments of the present application belong.

The logic and/or steps represented in the flowcharts or otherwise described herein, for example, can be considered as a sequenced listing of executable instructions for implementing logical functions, can be embodied in any computer-readable medium, For use with instruction execution systems, devices, or devices (such as computer-based systems, systems including processors, or other systems that can fetch instructions from instruction execution systems, devices, or devices and execute instructions), or in conjunction with these instruction execution systems, devices or equipment used. For the purposes of this specification, a "computer-readable medium" may be any device that can contain, store, communicate, propagate or transmit a program for use in or in conjunction with an instruction execution system, device or device. More specific examples (non-exhaustive list) of computer-readable media include the following: electrical connection with one or more wires (electronic device), portable computer disk case (magnetic device), random access memory (RAM), Read Only Memory (ROM), Erasable and Editable Read Only Memory (EPROM or Flash Memory), Fiber Optic Devices, and Portable Compact Disc Read Only Memory (CDROM). In addition, the computer-readable medium may even be paper or other suitable medium on which the program can be printed, since the program can be read, for example, by optically scanning the paper or other medium, followed by editing, interpretation or other suitable processing if necessary. The program is processed electronically and stored in computer memory.

It should be understood that each part of the present application may be realized by hardware, software, firmware or a combination thereof. In the embodiments described above, various steps or methods may be implemented by software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented by any one or combination of the following techniques known in the art: Discrete logic circuits, ASICs with suitable combinational logic gates, programmable gate arrays (PGAs), field programmable gate arrays (FPGAs), etc.

Those of ordinary skill in the art can understand that all or part of the steps carried by the methods of the above embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium. During execution, one or a combination of the steps of the method embodiments is included.

In addition, each functional unit in each embodiment of the present application may be integrated into one processing module, each unit may exist separately physically, or two or more units may be integrated into one module. The above-mentioned integrated modules can be implemented in the form of hardware or in the form of software function modules. If the integrated modules are realized in the form of software function modules and sold or used as independent products, they can also be stored in a computer-readable storage medium.

The storage medium mentioned above may be a read-only memory, a magnetic disk or an optical disk, and the like. Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and should not be construed as limitations on the present application, and those skilled in the art can make the above-mentioned The embodiments are subject to changes, modifications, substitutions and variations.

Claims (12)

1. A click pre-estimation method based on artificial intelligence is characterized by comprising the following steps:

acquiring attribute information of an entity to be recommended according to a query statement input by a user;

performing word segmentation processing on the attribute information of the query statement and the entity to be recommended respectively, and determining the characteristics of the query statement and the entity to be recommended;

determining the click rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended by using a preset deep neural network model;

determining the click rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended by using a preset deep neural network model, wherein the determining comprises the following steps:

Determining vectors corresponding to the query statement and the entity to be recommended respectively by utilizing a preset mapping relation between the participles and the vectors according to the participles with different granularities respectively included by the query statement and the entity to be recommended;

calculating vectors corresponding to the query statement and the entity to be recommended respectively by using a first preset operation rule, and determining the click rate of the entity to be recommended;

the method for calculating the vectors corresponding to the query statement and the entity to be recommended respectively by using the first preset operation rule comprises the following steps:

carrying out linear transformation and nonlinear transformation on vectors respectively corresponding to the query statement and the entity to be recommended, and determining new vectors respectively corresponding to the query statement and the entity to be recommended;

and determining the click rate of the entity to be recommended according to the inner products of the new vectors respectively corresponding to the query statement and the entity to be recommended.

2. the method of claim 1, wherein the performing word segmentation processing on the attribute information of the query statement and the entity to be recommended respectively to determine the feature values of the query statement and the entity to be recommended comprises:

And according to different word cutting particle sizes, performing word segmentation processing on the attribute information of the query sentence and the entity to be recommended respectively, and determining word segments with different particle sizes, which are respectively included by the query sentence and the entity to be recommended.

3. the method of claim 1, wherein after determining the click-through rate of the entity to be recommended, further comprising:

Determining the display position of the entity to be recommended;

And determining the click label of the entity to be recommended according to the click rate and the display position of the entity to be recommended.

4. the method of claim 3, wherein the determining the presentation location of the entity to be recommended comprises:

And determining the display position of the entity to be recommended according to the historical behavior log of the user.

5. The method of claim 3, wherein the determining the click label of the entity to be recommended according to the click rate and the presentation position of the entity to be recommended comprises:

Determining a correction vector corresponding to the entity to be recommended according to the click rate and the display position of the entity to be recommended;

Calculating the correction vector by using a second preset calculation rule, and determining a click label corresponding to the entity to be recommended;

the calculating the correction vector by using a second preset calculation rule, and determining the click label corresponding to the entity to be recommended includes:

and performing linear transformation and nonlinear transformation on the correction vector, and determining a click label corresponding to the entity to be recommended.

6. The method according to any of claims 1-5, wherein the attribute information of the entity to be recommended comprises at least one of the following information: the name of the entity to be recommended, the recommendation reason and the identification of the entity to be recommended.

7. the utility model provides a click and predict device based on artificial intelligence which characterized in that includes:

the acquisition module is used for acquiring attribute information of the entity to be recommended according to the query statement input by the user;

The word segmentation module is used for respectively carrying out word segmentation on the attribute information of the query statement and the entity to be recommended and determining the characteristics of the query statement and the entity to be recommended;

the processing module is used for determining the click rate of the entity to be recommended according to the query statement and the characteristics of the entity to be recommended by utilizing a preset deep neural network model;

Wherein the processing module comprises:

The determining unit is used for determining vectors corresponding to the query statement and the entity to be recommended respectively according to the participles with different granularities respectively included by the query statement and the entity to be recommended by utilizing a preset mapping relation between the participles and the vectors;

The operation unit is used for operating the vectors corresponding to the query statement and the entity to be recommended respectively by utilizing a first preset operation rule, and determining the click rate of the entity to be recommended;

Wherein, the arithmetic unit is specifically configured to:

Carrying out linear transformation and nonlinear transformation on vectors respectively corresponding to the query statement and the entity to be recommended, and determining new vectors respectively corresponding to the query statement and the entity to be recommended;

and determining the click rate of the entity to be recommended according to the inner products of the new vectors respectively corresponding to the query statement and the entity to be recommended.

8. The apparatus of claim 7, wherein the word segmentation module is specifically configured to:

And according to different word cutting particle sizes, performing word segmentation processing on the attribute information of the query sentence and the entity to be recommended respectively, and determining word segments with different particle sizes, which are respectively included by the query sentence and the entity to be recommended.

9. The apparatus of claim 7, further comprising:

The first determination module is used for determining the display position of the entity to be recommended;

and the second determining module is used for determining the click label of the entity to be recommended according to the click rate and the display position of the entity to be recommended.

10. the apparatus of claim 9, wherein the first determining module is specifically configured to:

And determining the display position of the entity to be recommended according to the historical behavior log of the user.

11. The apparatus of claim 10, wherein the second determining module is specifically configured to:

determining a correction vector corresponding to the entity to be recommended according to the click rate and the display position of the entity to be recommended;

Calculating the correction vector by using a second preset calculation rule, and determining a click label corresponding to the entity to be recommended;

the calculating the correction vector by using a second preset calculation rule, and determining the click label corresponding to the entity to be recommended includes:

And performing linear transformation and nonlinear transformation on the correction vector, and determining a click label corresponding to the entity to be recommended.

12. The apparatus according to any of claims 7-10, wherein the attribute information of the entity to be recommended includes at least one of the following information: the name of the entity to be recommended, the recommendation reason and the identification of the entity to be recommended.