CN107341187B

CN107341187B - Search processing method, device, equipment and computer storage medium

Info

Publication number: CN107341187B
Application number: CN201710422717.9A
Authority: CN
Inventors: 杜宏伟
Original assignee: Nubia Technology Co Ltd
Current assignee: Nubia Technology Co Ltd
Priority date: 2017-06-07
Filing date: 2017-06-07
Publication date: 2021-08-10
Anticipated expiration: 2037-06-07
Also published as: CN107341187A

Abstract

The embodiment of the invention discloses a search processing method, a device and a computer storage medium, wherein the method comprises the following steps: searching in a preset resource library according to the search word to obtain resource identifiers of N resources; acquiring a preference attribute value of each resource in the N resources at the current moment; inputting the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model to correspondingly obtain N predicted resource conversion rates; the resource conversion rate model is obtained by training based on preference attribute values of M resources in the resource library at historical time; sequencing the N resource identifications according to the N predicted resource conversion rates; and outputting the N sorted resource identifications as search results. The technical scheme provided by the embodiment of the invention can enable the sequencing of the search results to be more in line with the hope of the user, thereby improving the viewing efficiency of the search results.

Description

Search processing method, device, equipment and computer storage medium

Technical Field

The present invention relates to the field of information, and in particular, to a search processing method, device, and computer storage medium.

Background

With the development of the internet and the popularization of mobile terminals, obtaining information from the internet has become an important way for people to obtain information in daily life and work. Because the information quantity of resources stored in the internet is huge, how to effectively and quickly acquire the required target resources from the massive resources becomes a difficult thing, a search engine is a query tool for providing information search functions for users, and the emergence of the search engine is an effective method for solving the problem that the target resources are difficult to acquire. Search engines are in common use in today's society because of their great practical and commercial value in today's information society.

Currently, a search engine may determine a search result corresponding to a search term according to a preset matching manner and a matching degree condition, for example, the matching manner may be that the matching degree between the search term and a resource is determined according to the number of times that the search term appears in description information corresponding to each resource in a certain resource library, and the search engine ranks identifiers of resources satisfying the matching degree condition according to the corresponding matching degrees and feeds the ranked identifiers as search results back to a user.

Since the matching degree between the search terms and the resources is fixed and unchanged, the ranking of each resource identifier in the search results is also unchanged, which results in that the resource identifier corresponding to the target resource which the user most wants to search is ranked behind in the search results, and when the number of resources meeting the matching degree condition is large, the efficiency of the user for viewing the search results and acquiring the target resource is reduced, which results in the reduction of the satisfaction degree of the user on the search engine.

Disclosure of Invention

In view of this, embodiments of the present invention are expected to provide a search processing method, a search processing apparatus, and a computer storage medium, which can improve the efficiency of a user in viewing search results.

In order to achieve the above purpose, the technical solution of the embodiment of the present invention is realized as follows:

in a first aspect, an embodiment of the present invention provides a search processing method, including:

searching in a preset resource library according to the search word to obtain resource identifiers of N resources, wherein N is an integer greater than or equal to 2;

acquiring a preference attribute value of each resource in the N resources at the current moment;

inputting the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model to correspondingly obtain N predicted resource conversion rates; the resource conversion rate model is obtained by training preference attribute values of M resources in the resource library at a historical moment earlier than the current moment; m is an integer greater than or equal to 2;

sequencing the N resource identifications according to the N predicted resource conversion rates;

and outputting the N sorted resource identifications as search results.

In a second aspect, an embodiment of the present invention provides a search processing apparatus, including:

the acquisition module is configured to search in a preset resource library according to the search word to obtain resource identifiers of N resources, wherein N is an integer greater than or equal to 2;

the obtaining module is further configured to obtain a preference attribute value of each resource of the N resources at the current moment;

the processing module is configured to input the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model, and correspondingly obtain N predicted resource conversion rates; the resource conversion rate model is obtained by training preference attribute values of M resources in the resource library at a historical moment earlier than the current moment; m is an integer greater than or equal to 2;

the processing module is further configured to rank the N resource identifiers according to the N predicted resource conversion rates;

and the output module is configured to output the sequenced N resource identifications as search results.

In a third aspect, an embodiment of the present invention provides a server, where the server includes: a memory, a processor and a search handler stored on the memory and operable on the processor, the search handler when executed by the processor implementing the steps of the search processing method according to any of the first aspects.

In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, on which a search processing program is stored, and the search processing program, when executed by a processor, implements the steps of the search processing method according to any one of the first aspects.

In the embodiment of the invention, the resource identifiers of N resources are obtained by searching in a preset resource library according to the search word, the preference attribute values of the N resources at the current moment are input into a resource conversion rate model to correspondingly obtain N predicted resource conversion rates, the resource conversion rate model is obtained by training according to the preference attribute values of M resources at the historical moment, the N resource identifiers are sequenced according to the N predicted resource conversion rates, and the sequenced N resource identifiers are output as the search result, because the resource conversion rate model can reflect the attention degree of the user to various preference attributes when the user searches the resources, namely in the resource conversion rate model, the preference attribute coefficient of the preference attribute which is more important by the user is higher, the search result can put the resource identifier which is most suitable for the user to select in the earlier sequencing, the search result is more in line with the search habit of the user, so that the efficiency of the user for checking the search result can be improved, and the satisfaction degree of the user on the search engine can be improved.

Drawings

Fig. 1 is a schematic hardware configuration of a mobile terminal implementing various alternative embodiments of the present invention;

FIG. 2 is a first flowchart illustrating a search processing method according to an embodiment of the present invention;

FIG. 3 is a second flowchart illustrating a search processing method according to an embodiment of the present invention;

FIG. 4A is a first schematic diagram of a graphical user interface of a search processing method according to an embodiment of the present invention;

FIG. 4B is a diagram of a graphical user interface of a search processing method according to an embodiment of the present invention;

FIG. 5 is a schematic structural diagram of a user log collection system in the search processing method according to an embodiment of the present invention;

FIG. 6 is a schematic processing flow chart of log information analysis processing in the search processing method according to the embodiment of the present invention;

FIG. 7 is a schematic diagram of a processing flow of search result optimization processing in the search processing method according to an embodiment of the present invention;

FIG. 8 is a schematic structural diagram of a search processing apparatus according to an embodiment of the present invention;

fig. 9 is a schematic structural diagram of a server in an embodiment of the present invention.

Detailed Description

The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the drawings in the embodiments of the present invention.

A mobile terminal implementing various embodiments of the present invention will now be described with reference to fig. 1. In the following description, suffixes such as "module", "part", or "unit" used to represent elements are used only for facilitating the description of the present invention, and do not have a specific meaning per se. Thus, "module" and "component" may be used in a mixture.

The mobile terminal may be implemented in various forms. For example, the terminal described in the present invention may include a mobile terminal such as a mobile phone, a smart phone, a notebook computer, a digital broadcast receiver, a Personal Digital Assistant (PDA), a PAD computer (PAD), a Portable Multimedia Player (PMP), a navigation device, etc., and a stationary terminal such as a digital TV, a desktop computer, etc. In the following, it is assumed that the terminal is a mobile terminal. However, it will be understood by those skilled in the art that the configuration according to the embodiment of the present invention can be applied to a fixed type terminal in addition to elements particularly used for moving purposes.

Fig. 1 is a schematic hardware configuration of a mobile terminal implementing various alternative embodiments of the present invention.

The mobile terminal 100 may include a wireless communication unit 110, an audio/video (a/V) input unit 120, a user input unit 130, a sensing unit 140, an output unit 150, a memory 160, an interface unit 170, a controller 180, and a power supply unit 190, etc. Fig. 1 illustrates a mobile terminal having various components, but it is to be understood that not all illustrated components are required to be implemented, and that more or fewer components may alternatively be implemented, the elements of which will be described in detail below.

The wireless communication unit 110 typically includes one or more components that allow radio communication between the mobile terminal 100 and a wireless communication system or network. For example, the wireless communication unit may include at least one of a broadcast receiving module 111, a mobile communication module 112, a wireless internet module 113, a short-range communication module 114, and a location information module 115.

The broadcast receiving module 111 receives a broadcast signal and/or broadcast associated information from an external broadcast management server via a broadcast channel. The broadcast channel may include a satellite channel and/or a terrestrial channel. The broadcast management server may be a server that generates and transmits a broadcast signal and/or broadcast associated information or a server that receives a previously generated broadcast signal and/or broadcast associated information and transmits it to a terminal. The broadcast signal may include a TV broadcast signal, a radio broadcast signal, a data broadcast signal, and the like. Also, the broadcast signal may further include a broadcast signal combined with a TV or radio broadcast signal. The broadcast associated information may also be provided via the mobile communication network, and in this case, the broadcast associated information may be received by the mobile communication module 112. The broadcast signal may exist in various forms, for example, it may exist in the form of an Electronic Program Guide (EPG) of Digital Multimedia Broadcasting (DMB), an Electronic Service Guide (ESG) of digital video broadcasting-handheld (DVB-H), and the like. The broadcast receiving module 111 may receive a signal broadcast by using various types of broadcasting systems. In particular, the broadcast receiving module 111 may receive digital broadcasting by using a digital broadcasting system such as a data broadcasting system of multimedia broadcasting-terrestrial (DMB-T), digital multimedia broadcasting-satellite (DMB-S), digital video broadcasting-handheld (DVB-H), forward link media (MediaFLO @), terrestrial digital broadcasting integrated service (ISDB-T), and the like. The broadcast receiving module 111 may be constructed to be suitable for various broadcasting systems that provide broadcast signals as well as the above-mentioned digital broadcasting systems. The broadcast signal and/or broadcast associated information received via the broadcast receiving module 111 may be stored in the memory 160 (or other type of storage medium).

The mobile communication module 112 transmits and/or receives radio signals to and/or from at least one of a base station (e.g., access point, node B, etc.), an external terminal, and a server. Such wireless telecommunication signals may include voice call signals, video call signals, or various types of data transmitted and/or received in accordance with text and/or multimedia messages.

The wireless internet module 113 supports wireless internet access of the mobile terminal. The module may be internally or externally coupled to the terminal. The wireless internet access technology to which the module relates may include WLAN (wireless LAN) (Wi-Fi), Wibro (wireless broadband), Wimax (worldwide interoperability for microwave access), HSDPA (high speed downlink packet access), and the like.

The short-range communication module 114 is a module for supporting short-range communication. Some examples of short-range communication technologies include bluetooth (TM), Radio Frequency Identification (RFID), infrared data association (IrDA), Ultra Wideband (UWB), zigbee (TM), and the like.

The location information module 115 is a module for checking or acquiring location information of the mobile terminal. A typical example of the location information module 115 is a Global Positioning System (GPS). According to the current technology, the location information module 115 of the GPS calculates distance information and accurate time information from three or more satellites and applies triangulation to the calculated information, thereby accurately calculating three-dimensional current location information according to longitude, latitude, and altitude. Currently, a method for calculating position and time information uses three satellites and corrects an error of the calculated position and time information by using another satellite. In addition, the location information module 115 of the GPS can calculate speed information by continuously calculating current location information in real time.

The a/V input unit 120 is used to receive an audio or video signal. The a/V input unit 120 may include a camera 121 and a microphone 122, and the camera 121 processes image data of still pictures or video obtained by an image capturing apparatus in a video capturing mode or an image capturing mode. The processed image frames may be displayed on the display unit 151. The image frames processed by the cameras 121 may be stored in the memory 160 (or other storage medium) or transmitted via the wireless communication unit 110, and two or more cameras 121 may be provided according to the construction of the mobile terminal. The microphone 122 may receive sounds (audio data) via the microphone 122 in a phone call mode, a recording mode, a voice recognition mode, or the like, and is capable of processing such sounds into audio data. The processed audio (voice) data may be converted into a format output transmittable to a mobile communication base station via the mobile communication module 112 in case of a phone call mode. The microphone 122 may implement various types of noise cancellation (or suppression) algorithms to cancel (or suppress) noise or interference generated in the course of receiving and transmitting audio signals.

The user input unit 130 may generate key input data to control various operations of the mobile terminal according to a command input by the user. The user input unit 130 allows a user to input various types of information, and may include a keyboard, dome sheet, touch pad (e.g., a touch-sensitive member that detects a change in resistance, pressure, capacitance, etc. due to being touched), wheel, joystick, etc. In particular, when the touch pad is superimposed on the display unit 151 in the form of a layer, a touch screen may be formed.

The sensing unit 140 detects a current state of the mobile terminal 100 (e.g., an open or closed state of the mobile terminal 100), a position of the mobile terminal 100, presence or absence of contact (i.e., touch input) by a user with the mobile terminal 100, an orientation of the mobile terminal 100, acceleration or deceleration movement and direction of the mobile terminal 100, and the like, and generates a command or signal for controlling an operation of the mobile terminal 100. For example, when the mobile terminal 100 is implemented as a slide-type mobile phone, the sensing unit 140 may sense whether the slide-type phone is opened or closed. In addition, the sensing unit 140 can detect whether the power supply unit 190 supplies power or whether the interface unit 170 is coupled with an external device. The sensing unit 140 may include a proximity sensor 141 as will be described below in connection with a touch screen.

The interface unit 170 serves as an interface through which at least one external device is connected to the mobile terminal 100. For example, the external device may include a wired or wireless headset port, an external power supply (or battery charger) port, a wired or wireless data port, a memory card port, a port for connecting a device having an identification module, an audio input/output (I/O) port, a video I/O port, an earphone port, and the like. The identification module may store various information for authenticating a user using the mobile terminal 100 and may include a User Identity Module (UIM), a Subscriber Identity Module (SIM), a Universal Subscriber Identity Module (USIM), and the like. In addition, a device having an identification module (hereinafter, referred to as an "identification device") may take the form of a smart card, and thus, the identification device may be connected with the mobile terminal 100 via a port or other connection means. The interface unit 170 may be used to receive input (e.g., data information, power, etc.) from an external device and transmit the received input to one or more elements within the mobile terminal 100 or may be used to transmit data between the mobile terminal and the external device.

In addition, when the mobile terminal 100 is connected with an external cradle, the interface unit 170 may serve as a path through which power is supplied from the cradle to the mobile terminal 100 or may serve as a path through which various command signals input from the cradle are transmitted to the mobile terminal. Various command signals or power input from the cradle may be used as signals for recognizing whether the mobile terminal is accurately mounted on the cradle. The output unit 150 is configured to provide output signals (e.g., audio signals, video signals, alarm signals, vibration signals, etc.) in a visual, audio, and/or tactile manner. The output unit 150 may include a display unit 151, an audio output module 152, and the like.

The display unit 151 may display information processed in the mobile terminal 100. For example, when the mobile terminal 100 is in a phone call mode, the display unit 151 may display a User Interface (UI) or a Graphical User Interface (GUI) related to a call or other communication (e.g., text messaging, multimedia file downloading, etc.). When the mobile terminal 100 is in a video call mode or an image capturing mode, the display unit 151 may display a captured image and/or a received image, a UI or GUI showing a video or an image and related functions, and the like.

Meanwhile, when the display unit 151 and the touch pad are overlapped with each other in the form of a layer to form a touch screen, the display unit 151 may serve as an input device and an output device. The display unit 151 may include at least one of a Liquid Crystal Display (LCD), a thin film transistor LCD (TFT-LCD), an Organic Light Emitting Diode (OLED) display, a flexible display, a three-dimensional (3D) display, and the like. Some of these displays may be configured to be transparent to allow a user to view from the outside, which may be referred to as transparent displays, and a typical transparent display may be, for example, a TOLED (transparent organic light emitting diode) display or the like. Depending on the particular desired implementation, the mobile terminal 100 may include two or more display units (or other display devices), for example, the mobile terminal may include an external display unit (not shown) and an internal display unit (not shown). The touch screen may be used to detect a touch input pressure as well as a touch input position and a touch input area.

The audio output module 152 may convert audio data received by the wireless communication unit 110 or stored in the memory 160 into an audio signal and output as sound when the mobile terminal is in a call signal reception mode, a call mode, a recording mode, a voice recognition mode, a broadcast reception mode, or the like. Also, the audio output module 152 may provide audio output related to a specific function performed by the mobile terminal 100 (e.g., a call signal reception sound, a message reception sound, etc.). The audio output module 152 may include a speaker, a buzzer, and the like.

The alarm unit 153 may provide an output to notify the mobile terminal 100 of the occurrence of an event. Typical events may include call reception, message reception, key signal input, touch input, and the like. In addition to audio or video output, the alarm unit 153 may provide output in different ways to notify the occurrence of an event. For example, the alarm unit 153 may provide an output in the form of vibration. The alarm unit 153 may also provide an output notifying the occurrence of an event via the display unit 151 or the audio output module 152.

The memory 160 may store software programs or the like for processing and controlling operations performed by the controller 180, or may temporarily store data that has been output or is to be output.

The memory 160 may include at least one type of storage medium including a flash memory, a hard disk, a multimedia card, a card-type memory, a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a Read Only Memory (ROM), an Electrically Erasable Programmable Read Only Memory (EEPROM), a Programmable Read Only Memory (PROM), a magnetic memory, a magnetic disk, an optical disk, and the like. Also, the mobile terminal 100 may cooperate with a network storage device that performs a storage function of the memory 160 through a network connection.

The controller 180 generally controls the overall operation of the mobile terminal. For example, the controller 180 performs control and processing related to voice calls, data communications, video calls, and the like. In addition, the controller 180 may include a multimedia module 181 for reproducing (or playing back) multimedia data, and the multimedia module 181 may be constructed within the controller 180 or may be constructed separately from the controller 180. The controller 180 may perform a pattern recognition process to recognize a handwriting input or a picture drawing input performed on the touch screen as a character or an image.

The power supply unit 190 receives external power or internal power and provides appropriate power required to operate the respective elements and components under the control of the controller 180.

The various embodiments described herein may be implemented in a computer-readable medium using, for example, computer software, hardware, or any combination thereof. For a hardware implementation, the embodiments described herein may be implemented using at least one of an Application Specific Integrated Circuit (ASIC), a Digital Signal Processor (DSP), a Digital Signal Processing Device (DSPD), a Programmable Logic Device (PLD), a Field Programmable Gate Array (FPGA), a processor, a controller, a microcontroller, a microprocessor, an electronic unit designed to perform the functions described herein, and in some cases, such embodiments may be implemented in the controller 180. For a software implementation, the implementation such as a process or a function may be implemented with a separate software module that allows performing at least one function or operation. The software codes may be implemented by software applications (or programs) written in any suitable programming language, which may be stored in the memory 160 and executed by the controller 180.

Up to now, the mobile terminal has been described in terms of its functions. Hereinafter, a slide-type mobile terminal among various types of mobile terminals, such as a folder-type, bar-type, swing-type, slide-type mobile terminal, and the like, will be described as an example for the sake of brevity. Accordingly, the present invention can be applied to any type of mobile terminal, and is not limited to a slide type mobile terminal.

The mobile terminal 100 as shown in fig. 1 may be configured to operate with communication systems such as wired and wireless communication systems and satellite-based communication systems that transmit data via frames or packets.

The communication system in which the mobile terminal of the present invention is capable of operating may use different air interfaces and/or physical layers. For example, the air interface used by the communication system includes, for example, Frequency Division Multiple Access (FDMA), Time Division Multiple Access (TDMA), Code Division Multiple Access (CDMA), and Universal Mobile Telecommunications System (UMTS) (in particular, Long Term Evolution (LTE)), global system for mobile communications (GSM), and the like. By way of non-limiting example, the following description relates to a CDMA communication system, but such teachings are equally applicable to other types of systems.

Based on the above mobile terminal hardware structure and communication system, the present invention provides various embodiments of the method.

In the embodiment of the invention, the optimization of the search engine can be considered from the use angle of the vast users, so that the search engine can continuously optimize the search results according to the use habits of the vast users, and the search results can better fit the expectations of the users, thereby better meeting the requirements of the vast users. According to the embodiment of the invention, the machine learning multiple linear regression algorithm is introduced in the sequencing process of the search engine, so that the sequencing result of the search engine can be continuously self-optimized along with the use of a user. According to the technical scheme of the embodiment of the invention, the purpose of continuous self-optimization of the search engine according to the use of the user is achieved by continuously collecting the use log information of the user on the search engine, extracting useful data which can be used for training the machine learning multivariate linear regression model from the use log information, training the machine sequencing model based on the multivariate thread regression algorithm, and applying the machine sequencing model to the sequencing process before the search result is output.

Based on the above purpose, the embodiment of the present invention provides the following search processing method with self-optimization capability.

Example one

The execution subject of the embodiment of the present invention may be a server that provides a search engine service. The server may execute the steps of the search processing method provided by the embodiment of the present invention after receiving a search request with search terms from a client. Fig. 2 is a first flowchart illustrating a search processing method according to an embodiment of the present invention. As shown in fig. 2, the steps of the embodiment of the present invention include:

s101: and searching in a preset resource library according to the search word to obtain resource identifiers of the N resources.

The server providing the search engine service may store a preset resource library, where the resource library may store at least M resources, each resource may have a resource identifier, and M is an integer greater than or equal to 2, and N may be an integer greater than or equal to 2.

In the embodiment of the invention, the server can provide a plurality of search matching algorithms to obtain the search result corresponding to the search word. For example, a similarity-inverse document frequency (TF-IDF) search model may be used to search in the resource library, so as to obtain resource identifiers corresponding to N resources whose search terms satisfy the preset matching degree. The embodiment of the invention does not limit the search matching algorithm for obtaining N resources by searching in the resource library.

S102: and acquiring the preference attribute value of each resource in the N resources at the current moment.

In order to make the search result closer to the user requirement, that is, to set the resource identifier of the resource that the user may most want to see at the position in the search result ranked earlier as much as possible, the embodiment of the present invention may select the resource conversion rate as an evaluation index for evaluating the user satisfaction, select an influence factor that may influence the resource conversion rate, for example, a preference attribute value of one or more attributes related to the resource that some users may care about during the search, and establish the resource conversion rate model based on the multiple linear regression algorithm.

In an embodiment of the present invention, the preference attribute value related to the resource may be a representation of some attributes of the resource itself, and/or usage data of the resource used by a large number of users.

In an example, the resource that the user wants to search may be an Application (APP), the preset resource library may be an application library, and the characterization value of the attribute of the application itself may be: the application is a logical value of an attribute of official application issued by enterprises or organizations, the size and magnitude of an installation package, whether a language used by the application is a logical value of an attribute of Chinese, whether the application is loaded with advertisements, and the like, and the application data of the application can be the application download amount, the application use times, the application uninstallation times, the application quality estimation value, and the like of a mass of users in a preset time length.

It should be noted that the application quality estimation value may be obtained according to an application quality evaluation factor and a preset application quality evaluation function, and the application quality evaluation factor may be the number of times of application uninstallation and/or the number of times of application use and/or the application use time. For example, the application quality evaluation function may be set as a function in which the application quality estimation value is in inverse proportion to the number of application uninstallations, that is, the higher the number of application uninstallations, the lower the application quality estimation value, and similarly, the application quality evaluation function may be set as a function in which the application quality estimation value is in direct proportion to the number of application usages and the application usage time, that is, the more the number of application usages and the longer the usage time, the higher the application quality estimation value.

Correspondingly, the resource conversion rate of the application may be a ratio of the number of times of downloading the application to the number of times of exposing the application within a preset detection duration, where the number of times of exposing the application may be the number of times that the identifier corresponding to the application appears in any search result within the detection duration.

In another example, the resource that the user wants to search may be a post, and the preset repository may be a post repository, and the preference attribute value of the post includes at least one of: the click frequency of the post and the quality attribute of the post; the post quality attribute is obtained according to a post quality factor and a preset post quality evaluation function, wherein the post quality factor is post browsing time and/or post clicking times.

Accordingly, the resource conversion rate of the post may be a ratio of the number of post clicks to the number of post exposure within a preset detection duration, where the number of post exposure may be the number of times that the identifier corresponding to the post appears in any search result within the detection duration.

In this embodiment of the present invention, the manner of obtaining the preference attribute value of each resource in the N resources at the current time may include: receiving user log information sent by at least one client within a preset detection duration before the current time, wherein the user log information comprises a user search behavior event, a user search result event and a user click search result event within the detection duration before the current time; and analyzing according to the user log information to obtain preference attribute values of the M resources at the current moment.

It should be noted that the client may set a buried point event, and the buried point event may include, but is not limited to, a user search behavior event, a user view search result event, a user click search result event, and the like. If a preset embedded point event is triggered in the process of using a search engine by a user, the client collects user log information according to the preset embedded point event, and can analyze search words, search results, resource identification click times/resource downloading times, resource using/browsing time, resource unloading times and the like according to the user log information. The client side sends the collected user log information to the server, and the server checks, filters and stores the received user information in a persistent mode.

S103: inputting the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model to correspondingly obtain N predicted resource conversion rates; the resource conversion rate model is obtained by training preference attribute values of M resources in the resource library at a historical moment earlier than the current moment.

Wherein M is an integer greater than or equal to 2. The historical time may be a time earlier than the current time by a preset detection time period.

In the embodiment of the present invention, the resource conversion rate model may be configured to determine the predicted resource conversion rate of the ith resource at time t according to the preference attribute value of the ith resource at time t and the pre-obtained attribute coefficient set.

For example, the resource conversion rate model may be the following formula (1)

y_i＝a₁*G₁(x_1i)+a₂*G₂(x_2i)+……+a_S*G_S(x_Si) (1)；

Wherein x is_1i、x_2i、……、x_SiS offsets for ith resourceGood attribute value, a₁、a₂、……、a_SFor sets of attribute coefficients corresponding to the S preference attribute values, G₁、G₂、……、G_SS normalization processing functions, y, corresponding to S preference attribute values respectively_iPredicting the resource conversion rate corresponding to the ith resource; i is an integer greater than or equal to 1 and less than or equal to M; and S is an integer greater than or equal to 1, and S attribute coefficients in the attribute coefficient group correspond to S preference attribute values one by one.

It should be noted that some data input into the resource conversion rate model may be used as a data set for model training and testing after being normalized, because some values of the preference attribute values may have very large differences, such as "application download amount", some download amounts may reach several tens of millions, and some download amounts may only be hundreds of times, and if the data is directly used for training without being processed, the influence of the data on the result becomes large. A simple normalization method may be to log the preference attribute values, e.g., log (application download size). The normalization functions corresponding to the preference attribute values may be the same or different. In other embodiments of the present invention, some preference attribute values may be directly input into the resource conversion rate model without normalization.

In the embodiment of the present invention, the attribute coefficient group in the resource conversion rate model for calculating the predicted resource conversion rate of the N resources at the current time may be an optimal attribute coefficient group obtained by performing model training according to training data before inputting the preference attribute value of each resource of the N resources at the current time into the resource conversion rate model and obtaining the N predicted resource conversion rates correspondingly.

The embodiment of the invention can comprise the following steps of the model training process:

acquiring preference attribute values and resource conversion rates of M resources at historical time as training data; and performing parameter fitting according to the M training data and a target cost function corresponding to the resource conversion rate model to obtain an attribute coefficient group which can enable the cost value of the target cost function to be smaller than a preset cost value range.

The cost value of the target cost function can be obtained by performing difference comparison according to the predicted resource conversion rate of the M resources at the historical time and the resource conversion rate of the M resources at the historical time, wherein the predicted resource conversion rates of the M resources at the historical time are obtained by inputting preference attribute values of the M resources at the historical time into resource conversion rate models adopting different attribute coefficient groups respectively.

For example, the target cost function may be the following formula (2):

wherein Z is the cost value, p_iIs the resource conversion rate of the ith resource.

Substituting equation (1) into equation (2) may result in equation (3):

and (4) obtaining cost values corresponding to different attribute coefficient groups according to the formula (3) and the training data, and selecting a group of attribute coefficient groups with the minimum cost value from the optimal attribute coefficient group.

In other embodiments of the present invention, the target cost function may also adopt other difference comparison methods, which is not limited in this embodiment of the present invention. For example, the following formula (4) may be adopted for the target cost function:

s104: and sequencing the N resource identifications according to the N predicted resource conversion rates.

And the N predicted resource conversion rates are obtained by respectively inputting preference attribute values of the N resources at the current moment into a resource conversion rate model adopting an optimal attribute coefficient group. In the embodiment of the present invention, the N resource identifiers may be sorted according to the N predicted resource conversion rates from high to low.

S105: and outputting the N sorted resource identifications as search results.

Wherein the server may send the search results to the client initiating the search.

In the embodiment of the invention, the resource identifiers of N resources are obtained by searching in a preset resource library according to the search word, the preference attribute values of the N resources at the current moment are input into a resource conversion rate model to obtain N predicted resource conversion rates correspondingly, the resource conversion rate model is obtained by training according to the preference attribute values of M resources at the historical moment, the N resource identifiers are sequenced according to the N predicted resource conversion rates, the sequenced N resource identifiers are output as the search result, because the resource conversion rate model can reflect the attention degree of the user to various preference attributes when the user searches the resources, namely in the resource conversion rate model, the preference attribute coefficient of the preference attribute which is more important by the user is higher, the search result can put the resource identifier which is most suitable for the resource which the user wants to select in the former sequencing, the search result is more in line with the search habit of the user, so that the efficiency of the user for checking the search result can be improved, and the satisfaction degree of the user on a search engine can be improved.

Example two

The embodiment of the invention also provides a search processing method, and fig. 3 is a flow diagram illustrating the search processing method in the embodiment of the invention. As shown in fig. 3, the steps of the embodiment of the present invention may include:

s201: the first client sets a buried point event.

The buried point event can be used for collecting a user search behavior event, a user search result event, and a user click search result event.

S202: the first client collects user log information when a buried point event is triggered.

When the first client initiates searching and checks a searching result, a preset embedded point event is triggered. User log information may include, but is not limited to, search terms, search results, click resource identification, resource browsing time, and the like.

S203: the first client performs log analysis on user log information in a preset detection duration to obtain a single-user preference attribute value corresponding to the first client.

It should be noted that the preference attributes used for training the resource conversion rate model may include two types of attributes, one type is a massive user preference attribute that can be analyzed from the statistical information on the server side, and the other type is a single user preference attribute that is analyzed from the user log information on one or more clients using the search engine. In other embodiments of the present invention, the first client may identify the single-user preference attribute value of the single-user preference attribute reported to the server by using the user identifier or the device identifier corresponding to the first client. The server can obtain the summarized single-user preference attribute value for the single-user preference attribute values reported by the clients.

For example, the single user preference attributes may include resource quality attributes, category preference attributes, and the like. The resource quality attribute value corresponding to the resource quality attribute may be a quality score of the resource determined according to the resource browsing time, the resource usage times, the resource download times, the resource offload times, and the like included in the user log information. Illustratively, the longer the user browses the resource, the more times the resource is used, the more times the resource is downloaded, the less times the resource is unloaded, and the higher the attribute value of the resource quality attribute may be.

The classification preference attribute value corresponding to the classification preference attribute may be used to characterize whether a large number of users searching for the resource search the resource through a plurality of resource classifications or search the resource under a specific resource classification. The resource classification may be, for example, a game class, a financing class, a reading class. When the mass users search the resource under a certain classification, the association degree of the resource and the specific classification is possibly higher, then the classification preference attribute value of the resource and the specific classification is higher, and when the mass users search the resource under a plurality of classifications, the association degree of the resource and the specific classification is possibly lower, then the classification preference attribute value of the resource and the specific classification is lower.

For the first client, the category preference attribute value in the single user preference attribute values reported by the first client may be a degree of association between the resource obtained by analyzing the log used by the user on the first client and the specific category. Exemplarily, if the resource class designated in the search request sent by the first client is class a, the search result includes resource a and resource B, and the user finally selects to download resource a at the first client, the first client may report that the association degree between resource a and class a is high, and the association degree between resource B and class a is low, that is, the class preference attribute value corresponding to resource a and class a is high, and the class preference attribute value corresponding to resource B and class a is low.

It should be further noted that, when the resource conversion rate model is established by using the classification preference attribute value, if the attribute coefficient corresponding to the classification preference attribute value in the model is larger, the correlation degree between the resource and the resource classification is represented by the user more prominently, and if the corresponding attribute coefficient is smaller, the correlation degree between the resource and the application classification is represented by the user less prominently, that is, the user more hopes to search the resource under a wider classification. Therefore, when the search engine specifies the resource classification, the user can conveniently acquire the resources of other application classifications related to the search terms.

S204: and the first client sends the single user preference attribute value corresponding to the first client to the server.

The server collects the single-user preference attribute value corresponding to the first client and the single-user preference attribute values sent by other clients to obtain the collected single-user preference attribute values.

S205: the first client sends a search request carrying search words to the server.

Wherein, the search request can also comprise a specified resource classification.

S206: and the server searches in the resource library by adopting a similarity retrieval model to obtain resource identifications corresponding to N resources of which the search terms meet the preset matching degree.

And searching results obtained by adopting the similarity retrieval model comprise resource identifications corresponding to the N resources which are arranged according to the similarity of the N resources.

S207: the server obtains a preference attribute value of each resource of the N resources at the current moment.

The preference attribute value can be obtained by summarizing the preference attribute values of the single users reported by one or more clients within a preset time before the current time, and can also be determined according to the preference attribute values of the mass users recorded by the server.

S208: the server inputs the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model corresponding to the first client, and correspondingly obtains N predicted resource conversion rates, wherein the resource conversion rate model corresponding to the first client is obtained by training preference attribute values of M resources at historical moments, wherein the preference attribute values of the M resources appear in the search results of the first client in the resource library.

The M resources are part or all of resources appearing in the search result of the first client, that is, resources exposed on the first client, and furthermore, a resource conversion rate of at least one of the M resources may not be zero, that is, at least one of the M resources is downloaded or browsed on the first client.

In other embodiments of the present invention, a single-user preference attribute value among preference attribute values of M resources at a historical time may be obtained by only selecting and summarizing the single-user preference attribute values reported by the first client. For example, the single-user preference attribute values reported by the first client within the predicted detection duration before the historical time may be selected to be summarized. Correspondingly, when the resource conversion rate model corresponding to the first client is subjected to model training, the massive user preference attribute values at the historical moment and the single user preference attribute value corresponding to the first client can be used as training data. It should be noted that the training process for the resource conversion rate model corresponding to the first client is similar to the training process for the resource conversion rate model used by the client without distinction, and accordingly, the optimal preference attribute coefficient group adopted in the finally determined resource conversion rate model corresponding to the first client should include the attribute coefficients corresponding to the massive user preference attributes and the single user preference attributes.

After the optimal preference attribute coefficient group in the resource conversion rate model corresponding to the first client is determined, N corresponding predicted resource conversion rates can be calculated for N resources, respectively.

S209: and the server sequences the N resource identifications according to the N predicted resource conversion rates.

Wherein, the server can sort the N resource identifications in a sequence from high to low.

S210: and the server outputs the sequenced N resource identifiers as search results to the first client.

It should be noted that the execution sequence between the steps S202 to S205 and the steps S205 and S210 in the embodiment of the present invention is not limited to the sequence shown in fig. 3. For example, the client may collect corresponding user log information while initiating a search action each time, that is, the client performs a collection process and a reporting process of the user log information, and may perform the collection process and the reporting process simultaneously with the server receiving the search word, determining the search result, and outputting the search result. For another example, the client may collect and report the user log information collected within the preset detection duration to the server, and may collect and report the user log information collected at the next detection duration to the server when the next detection duration after the preset detection duration is finished. In addition, user log information collected by the client at the user's last search action may affect search results for search actions subsequent to the last search action.

In the embodiment of the invention, the server adopts the resource conversion rate model of the first client to sort the resource identifiers of the N resources matched with the search terms requested to be searched by the first client, and the resource conversion rate model of the first client can reflect the degree of importance of the user using the first client on various preference attributes when searching the resources, namely in the resource conversion rate model of the first client, the preference attribute coefficient of the preference attribute which is more important by the user is higher, so that the sorting of the N resource identifiers output as the search results is more suitable for the search habits of the user, the efficiency of the user for checking the search results can be improved, and the satisfaction degree of the user on a search engine can be improved.

Further, the embodiment of the present invention may set different detection durations, and obtain the attribute values of the massive user preference attributes and/or the attribute values of the single user preference attributes corresponding to a certain client in each detection duration, so as to train and obtain the resource conversion rate model of the undistinguished client or the differentiated client corresponding to the end time of each detection duration. Therefore, the resource conversion rate model can reflect the change of the attention degree of the mass users/single users to various preference attributes along with the time.

Fig. 4A is a first schematic diagram of a user graphical interface of a search processing method according to an embodiment of the present invention. Fig. 4A illustrates a scenario in which a client collects user log information, and the client may be a browser or an APP as illustrated in fig. 4A. The browser or APP provides a search engine service, and when a user needs to use the search engine, a search word, such as "discount", can be input in a search bar of the client, and the client is triggered to send a search request to the server. Illustratively, the client may also provide the option to search for resource types, e.g., the resource categories may be applications, posts, papers, etc. At this time, the client may collect user search events when the user initiates a search request, and obtain data such as search terms and resource classifications of the search.

Fig. 4B is a schematic diagram of a graphical user interface of the search processing method according to the embodiment of the present invention. Fig. 4B shows a search result sent by the server to the client, where the ranking of the resource identifiers in the search result is optimized according to the resource conversion rate model, and for example, when the user is more concerned about whether the resource is in chinese language and whether the resource is an official application, the resource identifier of the resource with higher comprehensive conformity degree of the two attributes appears at a position higher in the search result.

Other technical solution details and technical effects of the embodiment of the present invention are similar to those of the method shown in fig. 2, and reference may be made to the related descriptions in other embodiments of the present invention for other technical solution details and technical effects.

EXAMPLE III

The overall implementation flow of the search processing method provided by the embodiment of the invention can comprise three processing processes of user log information collection, log information analysis and search result optimization.

Fig. 5 is a schematic structural diagram of a user log collection system in the search processing method according to the embodiment of the present invention. As shown in fig. 5, the user log collection system may include two parts, a client 51 and a log server 52.

The four processing modules on the client 51 related to the user log collection processing procedure are: the system comprises a preset buried point event module 501, a buried point event triggering module 502, a user log information obtaining module 503 and a data packaging and sending module 504.

The preset buried point event module is used for presetting buried point events, and the buried point events mainly include but are not limited to user search behavior events, user search result viewing events, user search result clicking events and the like. If a user triggers a preset buried point event in the process of using a search engine, the buried point event triggering module 502 may trigger the user log information obtaining module 503 to collect user log information according to the user preset buried point event, where the user log information may include, but is not limited to, a search word, a search result, a resource click identifier, a resource browsing time, and the like. After completing the collection of the user log information, the user log information obtaining module 503 packages and sends the collected user information to the log server 52 through the data packaging and sending module 504. Illustratively, the client 51 may transmit the user log information or the analysis result of the user log information through a hypertext Transfer Protocol (HTTP) Protocol.

The three processing modules on the log server 52 related to the user log collection processing procedure are: a data checking module 505, a data filtering module 506 and a data storage module 507.

The data checking module 505 is configured to check data correctness of the received user log information, the data filtering module 506 is configured to remove abnormal data, and the data storage module 507 is configured to perform normalized storage on the user log information to facilitate persistent storage and management.

In other embodiments of the present invention, the log server 52 and the server shown in fig. 3 may be separate servers or may be combined.

Fig. 6 is a schematic processing flow diagram of log information analysis processing in the search processing method according to the embodiment of the present invention. As shown in fig. 6, the steps of the log information analysis processing flow may include: s601 reads the user log information and S602 analyzes the user log information, and the output result of the user log information analysis is the attribute values of various preference attributes. Such as dynamic download or click through, resource quality, or other attribute values of preference attributes. The log information analysis process may be executed by the log server 52 or a server storing a repository.

The following description will be given taking an example in which the log information analysis flow is executed by a server storing a repository.

The server storing the repository may be provided with a log information analysis module, and when executing S601, the log information analysis module reads the user log information stored in the log server 52, then executes S602 to perform log analysis, and extracts data required for machine training from the read user log information. Log information analysis entails extracting preference attribute values for attributes of some dimensions that may affect resource conversion rates. Taking application search as an example, the user log information includes the downloaded application name, the downloaded application classification, the downloading time, and the downloaded user identifier which are finally downloaded in the process of using the search engine by a large number of users, and the data reflecting the downloading preference of the users, such as the dynamic downloading amount of the applications in a period of time, can be counted and analyzed.

The search result optimization process trains a model using the result of the user log information analysis as training data of a ranking model for ranking resource identifiers of search results, and then performs ranking optimization on a search engine according to the trained ranking model according to the result of search term similarity matching. Illustratively, the ranking model in embodiments of the present invention may be a resource conversion model with results labeled as resource conversion. Fig. 7 is a schematic processing flow diagram of search result optimization processing in the search processing method according to the embodiment of the present invention. As shown in fig. 7, the search result optimization process includes three stages of sub-processes, which are respectively: a data preparation process 71, a model training process 72, and a rank optimization process 73.

The data preparation process 71 is used to prepare a data set for training and testing the resource conversion rate model, and the preparation of the training data is divided into two parts, namely determining the ranking data and normalizing the data.

Wherein, the process of determining the ranking data requires determining which data can be used for training and testing the ranking model. The preparation of model training data is related to the specific search service and the collected and extracted data. For example, the application search service may select "application download amount", "application quality", "official" as training data, and select "download conversion rate" as a result flag of the training data.

Some data output by user log information processing can be used as a data set for training and testing a sequencing model after normalization processing, because the value difference of user log data can be very large, such as application download amount, some download amount can reach tens of millions, and some download amount only has hundreds of times, and if the data is not processed and is directly used for training, the influence of the data on the result is increased. The simplest normalization method can take the log value directly, such as log (application download size).

The model training process 72 may include determining a model algorithm 721, determining a cost function 722, fitting parameters 723, and an optimization algorithm 724.

The deterministic model algorithm 721 refers to an algorithm determined for training of the ranking model, and here, a machine-learned multiple linear regression algorithm is used as a training algorithm of the ranking model. For example, in the case of application search, the "application download amount" and "official application attribute" and "application quality attribute" may be selected as the "application download amount" and "official application attributeVariables in the multiple linear regression algorithm are marked by 'application download conversion rate' as the result of the training sample, and a₁、a₂、 a₃As coefficients for the variables, the following training models were obtained:

application download conversion rate ═ a₁Log (application download) + a₂Whether official application attribute + a₃Applying a quality attribute;

wherein, the application download conversion rate is equal to the ratio of the application download times to the application exposure times. The application exposure times refer to the times that the application identifier corresponding to the application appears in any search result within a preset time length, and the application download times can be obtained according to the times that the application is downloaded and recorded on the server. I.e. the application download conversion rate can be obtained from actual statistical data.

Thus, the above multivariate linear model training process translates into data computation a through user log information₁、a₂And a₃The value of (c).

In determining the cost function 722, the algorithm for the ranking determined above, i.e., under the assumption of a₁、 a₂And a₃The ranking score of the search result may be calculated in certain cases, but in order to evaluate the difference between the calculated result and the actual user desired result as described above, a cost function is introduced, which represents the difference between the predicted ranking result and the actual user desired result. If m represents the number of all or part of the search resources in the application resource pool, the cost function Z can be expressed as formula (5):

wherein x is_1iThe amount of application download, x, for application i_2iWhether an attribute value, x, is officially applied for application i_3iApplication quality attribute value, p, for application i_iIs the actual download conversion rate of application i.

The training process of the ranking model is to determine a₁、a₂And a₃The error between the calculated ordering result and the actual result is minimized, that is, the minimum value of the cost function Z is solved. The process of solving the cost function zeminimums may use a gradient descent method.

The process of fitting the parameters 723 is the solving process of the optimization algorithm 724, and the gradient descent method is taken as an example, namely a group a is randomly determined₁、a₂And a₃Then calculating the value of a cost function, then solving the partial derivative of the cost function, and determining a₁、a₂And a₃Adjusting the parameter, and gradually iterating until a is obtained to make the value of the cost function reach the minimum value₁、a₂And a₃。

In the process of fitting the parameters, the model can be tested by adopting the test set data, and when the predicted resource conversion rate of m resources calculated according to the model is compared with the corresponding actual resource conversion rate data, and the accuracy reaches the preset accuracy, the training of the sequencing model is determined to be finished. The accuracy is, for example, that the difference between the predicted resource conversion rate of 80% of the m resources and the corresponding actual resource conversion rate is smaller than a preset deviation range.

The search optimization process 73 is divided into two stages, a similarity ranking 731 and a machine model ranking 732. The machine model sorting in the sorting optimization process is to perform secondary sorting optimization on the search results of similarity sorting by using a trained sorting model, so that the aim of optimizing the search results is fulfilled.

Similarity ranking 731 is the first stage of the search optimization process, and in this stage, mainly using a common similarity retrieval model, TF-IDF retrieves results matching with search terms carried in user search actions from index data in an index base corresponding to a resource base, and ranks the search results according to similarities. The file in the index repository may be a file corresponding to the application describing the application. For search term t_iFile d in index repository_jThe similarity TF-IDF in (1) can be obtained using the formula (6).

Wherein, tf_i,jFor search terms t_iWith respect to the jth file d_jTerm Frequency (Term Frequency), idf of_i,jFor search terms t_iAgainst document d_jInverse Document Frequency (Inverse Document Frequency), n_i，jFor search terms t_iIn document d_jThe number of occurrences in (a); k is a file d_jThe number of all the words in (a) appear,

as a file d_jThe sum of the number of occurrences of all the words in (b); d is the total number of the files in the index library; i { j: t_i∈d_jIs a term containing a search term t_iThe number of files of (c). For example, file d_jIs 'orange apple banana watermelon', k is 4,

is 6.

Machine model ranking 732 is the second stage of the search optimization process, which uses a trained ranking model to accurately re-rank the Top K, for example, (Top-K) results that are most similar and output from the first stage. In the machine model sorting, data such as application download amount, official application attribute and application quality attribute corresponding to the application in the Top-K search result in the index are used as input data of the secondary sorting machine model, and the output result of the sorting model is used as the basis of the secondary sorting. And outputting the search result of the secondary sorting as a final search result.

According to the search engine self-optimization method provided by the embodiment of the invention, the problem that the search engine cannot be self-optimized is solved by introducing the machine learning-based multiple linear regression ranking model in the ranking process of the search engine, so that the search engine is continuously optimized along with the use of a user. The application of the method can lead the prior search engine to be continuously optimized along with the increase of the use of the user, thereby leading the search result to better meet the requirement of the user.

Other technical solution details and technical effects of the embodiment of the present invention are similar to those of the method shown in fig. 3 to 4B, and reference may be made to the related descriptions in other embodiments of the present invention for other technical solution details and technical effects.

Example four

Fig. 8 is a schematic structural diagram of a search processing apparatus according to an embodiment of the present invention, and as shown in fig. 8, a search processing apparatus 80 according to an embodiment of the present invention includes:

an obtaining module 801 configured to search in a preset resource library according to a search term to obtain resource identifiers of N resources, where N is an integer greater than or equal to 2;

an obtaining module 801, further configured to obtain a preference attribute value of each resource in the N resources at the current time;

a processing module 802, configured to input a preference attribute value of each resource of the N resources at the current time into a resource conversion rate model, and correspondingly obtain N predicted resource conversion rates; the resource conversion rate model is obtained by training based on preference attribute values of M resources in the resource library at a historical moment earlier than the current moment; m is an integer greater than or equal to 2;

a processing module 802 further configured to rank the N resource identifications according to the N predicted resource conversion rates;

an output module 803, configured to output the sorted N resource identifiers as a search result.

In the embodiment of the present invention, the search processing means may be located on the server side.

Other technical solution details and technical effects of the embodiment of the present invention are similar to those of the method shown in fig. 3 to 7, and reference may be made to the relevant description in other embodiments of the present invention for other technical solution details and technical effects.

An embodiment of the present invention further provides a server, fig. 9 is a schematic diagram of a structure of the server in the embodiment of the present invention, and as shown in fig. 9, a server 90 in the embodiment of the present invention includes:

a memory 901, a processor 902 and a search processing program (not shown in fig. 9) stored on the memory 901 and operable on the processor 902, the search processing program, when executed by the processor 902, implementing the steps performed by the server in any of the search processing methods shown in fig. 3 to 7.

In the embodiment of the present invention, the server may further include an interface 903 and a bus 904, where the interface 903 may be used to establish a communication connection with at least one client, and implement interaction of information related to the embodiment of the present invention based on the interface 903 and the client.

The server according to the embodiment of the present invention may be configured to execute any one of the search processing methods shown in fig. 3 to 7, and other technical solution details according to the embodiment of the present invention may refer to the description of the embodiments shown in fig. 3 to 7.

An embodiment of the present invention further provides a computer-readable storage medium, where a search processing program is stored on the computer-readable storage medium, and when the search processing program is executed by a processor, the steps of the search processing method executed by the server shown in fig. 3 to 7 are implemented.

The above-mentioned serial numbers of the embodiments of the present invention are merely for description and do not represent the merits of the embodiments.

Through the above description of the embodiments, those skilled in the art will clearly understand that the above embodiment method can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better embodiment. Based on such understanding, the technical solution of the present invention or portions thereof contributing to the prior art can be essentially embodied in the form of a software product, which is stored in a storage medium (such as ROM/RAM, magnetic disk, optical disk) and includes instructions for enabling a terminal device (such as a mobile phone, a computer, a server, an air conditioner, or a network device) to execute the method described in the embodiments of the present invention.

The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.

These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by the contents of the present specification and drawings, or used directly or indirectly in other related fields, are included in the scope of the present invention.

Claims

1. A method of search processing, the method comprising:

acquiring a preference attribute value of each resource in the N resources at the current moment; the preference attribute value comprises a representation value of an attribute of the resource itself;

sequencing the N resource identifications according to the N predicted resource conversion rates; the method comprises the following steps: sequencing the N resource identifications from high to low according to the values of the N predicted resource conversion rates;

and outputting the N sorted resource identifications as search results.

2. The method according to claim 1, before inputting the preference attribute value of each resource in the N resources at the current time into the resource conversion rate model to obtain N predicted resource conversion rates, comprising:

acquiring preference attribute values and resource conversion rates of the M resources at historical time as training data;

performing parameter fitting according to the M training data and a target cost function corresponding to the resource conversion rate model to obtain an attribute coefficient group which can enable the cost value of the target cost function to be smaller than a preset cost value range;

the cost value of the target cost function is obtained by comparing the predicted resource conversion rate of the M resources at the historical time with the resource conversion rate of the M resources at the historical time, and the predicted resource conversion rate of the M resources at the historical time is calculated by inputting the preference attribute values of the M resources at the historical time into resource conversion rate models adopting different attribute coefficient groups.

3. The method according to claim 1, wherein before the searching in the preset resource library according to the search term to obtain the resource identifiers of N resources, the method comprises: receiving a search request carrying the search word from a first client;

inputting the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model, and correspondingly obtaining N predicted resource conversion rates, including: inputting the preference attribute value of each resource in the N resources at the current moment into a resource conversion rate model corresponding to the first client, and correspondingly obtaining N predicted resource conversion rates; the resource conversion rate model corresponding to the first client is obtained by training preference attribute values of M resources appearing in the search results of the first client at historical time in the resource library.

4. The method of claim 1, wherein the resource is an application, and wherein the preference attribute value of the resource comprises at least one of: application download times, logical values of official application attributes and application quality estimation values; the application quality estimation value is obtained according to an application quality evaluation factor and a preset application quality evaluation function, and the application quality evaluation factor is the application unloading times and/or the application using time;

the resource conversion rate is the ratio of the application download times and the application exposure times of the application in a preset detection duration, and the application exposure times are the times of the application identifier corresponding to the application appearing in any search result in the detection duration.

5. The method of claim 1, wherein the resource is a post, and wherein the preference attribute value of the resource comprises at least one of: click times of posts and estimated values of post quality; the post quality estimation value is obtained according to a post quality factor and a preset post quality evaluation function, wherein the post quality factor is post browsing time and/or post clicking times;

the resource conversion rate is the ratio of the post click frequency and the post exposure frequency of the post in a preset detection duration, and the post exposure frequency is the frequency of the post mark corresponding to the post appearing in any search result in the detection duration.

6. The method according to claim 2, wherein the obtaining preference attribute values and resource conversion rates of the M resources at historical time as training data comprises:

receiving user log information sent by at least one client in a preset detection duration before the historical time, wherein the user log information comprises user searching behavior operation, user searching result operation and user clicking search result operation in the preset detection duration before the historical time;

and analyzing and obtaining preference attribute values and resource conversion rates of the M resources at the historical moment according to the user log information.

7. The method of claim 2, wherein the resource conversion model is

y_i＝a₁*G₁(x_1i)+a₂*G₂(x_2i)+……+a_S*G_S(x_Si)，

Wherein x is_1i、x_2i、……、x_SiS preference attribute values for the ith resource, a₁、a₂、……、a_SFor sets of attribute coefficients corresponding to the S preference attribute values, G₁、G₂、……、G_SS normalization processing functions, y, corresponding to S preference attribute values, respectively_iPredicting the resource conversion rate corresponding to the ith resource; wherein i is an integer greater than or equal to 1 and less than or equal to M; s is an integer greater than or equal to 1;

the target cost function is

8. A search processing apparatus, characterized in that the search processing apparatus comprises:

the obtaining module is further configured to obtain a preference attribute value of each resource of the N resources at the current moment; the preference attribute value comprises a representation value of an attribute of the resource itself;

the processing module is further configured to rank the N resource identifiers according to the N predicted resource conversion rates; the concrete configuration is as follows: sequencing the N resource identifications from high to low according to the values of the N predicted resource conversion rates;

9. A server, characterized in that the server comprises: memory, a processor and a search processing program stored on the memory and executable on the processor, the search processing program, when executed by the processor, implementing the steps of the search processing method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that a search processing program is stored thereon, which when executed by a processor implements the steps of the search processing method according to any one of claims 1 to 7.