CN117112607A

CN117112607A - Content searching method and related device

Info

Publication number: CN117112607A
Application number: CN202210510614.9A
Authority: CN
Inventors: 牛明
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2022-05-11
Filing date: 2022-05-11
Publication date: 2023-11-24

Abstract

The embodiment of the application discloses a content searching method and a related device, which are applied to an artificial intelligence or vehicle-mounted scene and are used for converting the characteristic of a throwing characteristic value of candidate content in a candidate content set into a throwing characteristic; the method comprises the steps of interactively fusing a first content coding vector obtained by the release characteristic and the content characteristic of the coding candidate content into a second content coding vector, wherein at least one-dimensional component in the second content coding vector represents the fusion characteristic of the release characteristic and the content characteristic; and searching candidate contents in the candidate content set to obtain target contents through the second content coding vector and the target object coding vector obtained by coding the object characteristics of the target object. According to the method, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, and the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, so that the content searching effect is improved. Wherein the content is, for example, an advertisement and the bid is, for example, a bid for placement.

Description

Content searching method and related device

Technical Field

The present application relates to the field of data processing, and in particular, to a method and related apparatus for searching content.

Background

After the content is delivered by the content delivery person, the content platform stores a plurality of contents. In the process of browsing the content platform by the object, the content platform needs to search a plurality of contents by considering the object characteristics of the object so as to display the contents with better correlation with the object to the object later, and then feedback of the object is obtained; therefore, the content search effect is particularly important.

In the related art, in the content searching method, the object feature and the content feature are generally respectively encoded to obtain two equal-dimension vectors, and a plurality of contents are searched based on the two equal-dimension vectors to obtain the contents with better correlation with the object.

In fact, when a content distributor distributes content, the distribution bid of the content is an important item of information, and the content is required to be reflected on the content searching effect; however, the above-described content search method does not consider the influence of the bid for placement of the content on the content search effect, that is, the content search effect is not sensitive to the bid for placement of the content, resulting in poor content search effect.

Disclosure of Invention

In order to solve the technical problems, the application provides a content searching method and a related device, which enhance the influence effect of the release characteristic value of the candidate content in the content searching process, facilitate the release characteristic value of the candidate content to be effectively reflected on the content searching effect, and effectively improve the sensitivity of the content searching effect to the release characteristic value of the candidate content, thereby improving the content searching effect.

The embodiment of the application discloses the following technical scheme:

in one aspect, the present application provides a method of content searching, the method comprising:

performing feature conversion processing on the release feature values of the candidate contents in the candidate content set to obtain release features of the candidate contents;

performing interactive fusion processing on the release characteristics and the first content coding vectors of the candidate contents to obtain second content coding vectors of the candidate contents; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristics of the release characteristics and the content characteristics;

searching candidate contents in the candidate content set based on the second content coding vector and a target object coding vector of a target object to obtain target contents; the target object coding vector is obtained by coding the object characteristics of the target object.

In another aspect, the present application provides an apparatus for content searching, the apparatus comprising: the device comprises a feature conversion unit, a fusion unit and a search unit;

the feature conversion unit is used for carrying out feature conversion processing on the release feature values of the candidate content in the candidate content set to obtain release features of the candidate content;

The fusion unit is used for carrying out interactive fusion processing on the release characteristics and the first content coding vectors of the candidate contents to obtain second content coding vectors of the candidate contents; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristics of the release characteristics and the content characteristics;

the searching unit is used for searching candidate contents in the candidate content set based on the second content coding vector and a target object coding vector of a target object to obtain target contents; the target object coding vector is obtained by coding object characteristics of the target object.

In another aspect, the present application provides an apparatus for content searching, the apparatus comprising a processor and a memory:

the memory is used for storing program codes and transmitting the program codes to the processor;

the processor is configured to execute the method of content searching described in the above aspect according to instructions in the program code.

In another aspect, embodiments of the present application provide a computer-readable storage medium storing a computer program that, when executed by a processor, performs the method of content searching described in the above aspect.

In another aspect, embodiments of the present application provide a computer program product comprising a computer program or instructions; the method of content searching as described in the above aspects is performed when the computer program or instructions are executed by a processor.

According to the technical scheme, the characteristic of the release characteristic value of the candidate content in the candidate content set is converted into the release characteristic; interactively fusing the throwing characteristic and the first content coding vector of the candidate content into a second content coding vector, wherein the first content coding vector is obtained by coding the content characteristic of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristic of the throwing characteristic and the content characteristic; and searching candidate contents in the candidate content set through the second content coding vector and a target object coding vector of the target object to obtain target contents, wherein the target object coding vector is obtained by coding object characteristics of the target object. Based on the method, the method realizes interactive fusion of the release characteristic obtained by the release characteristic value of the characteristic conversion candidate content and the first content coding vector obtained by the content characteristic of the coding candidate content, so that the second content coding vector after interactive fusion can strengthen the fusion of the release characteristic; and further, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, so that the release characteristic value of the candidate content is effectively reflected on the content searching effect, the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, and the content searching effect is improved.

Drawings

In order to more clearly illustrate the embodiments of the application or the technical solutions of the prior art, the drawings which are used in the description of the embodiments or the prior art will be briefly described, it being obvious that the drawings in the description below are only some embodiments of the application, and that other drawings can be obtained from these drawings without inventive effort for a person skilled in the art.

Fig. 1 is a schematic structural diagram of an advertisement searching method in the related art according to an embodiment of the present application;

fig. 2 is an application scenario schematic diagram of a method for searching content according to an embodiment of the present application;

FIG. 3 is a flowchart of a method for searching content according to an embodiment of the present application;

FIG. 4 is a schematic diagram of an interactive fusion process of discrete placement features and first advertisement encoding vectors according to an embodiment of the present application;

FIG. 5 is a schematic diagram of an interactive fusion process of continuous placement features and first advertisement encoding vectors according to an embodiment of the present application;

fig. 6 is a schematic diagram of an apparatus for searching content according to an embodiment of the present application;

fig. 7 is a schematic structural diagram of a server according to an embodiment of the present application;

Fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application.

Detailed Description

Embodiments of the present application are described below with reference to the accompanying drawings.

In the present stage, a content distributor may, for example, throw an advertisement on an advertisement platform by an advertiser, see a schematic structural diagram of an advertisement searching method in the related art provided in fig. 1, respectively encode object features to obtain object encoding vectors through a double-tower structure, encode advertisement features to obtain advertisement encoding vectors, and perform inner product operation based on the object encoding vectors and the advertisement encoding vectors to obtain inner products, and search a plurality of advertisements based on the inner products according to a maximum inner product searching algorithm to obtain advertisements with better correlation with objects.

According to research, when an advertiser puts advertisements, the putting bid of the advertisements is important information and needs to be reflected on the advertisement searching effect; however, the above-described advertisement search method does not consider the influence of the bid-out price of the advertisement on the advertisement search effect, that is, the advertisement search effect is insensitive to the bid-out price of the advertisement, resulting in poor advertisement search effect.

In view of this, the present application provides a method and related device for searching content, in which the feature conversion candidate advertisement is put in bid to obtain put features, and the interactive fusion is implemented with the first advertisement coding vector obtained by coding the candidate advertisement, so that the second advertisement coding vector after interactive fusion can strengthen the fused put features; and further, the influence effect of the bid for putting the candidate advertisement is enhanced in the advertisement searching process, so that the bid for putting the candidate advertisement is effectively reflected on the advertisement searching effect, the sensitivity of the advertisement searching effect to the bid for putting the candidate advertisement is effectively improved, and the advertisement searching effect is improved.

In order to facilitate understanding of the technical scheme of the present application, the method for searching content provided by the embodiment of the present application is described below in conjunction with an actual application scenario.

Referring to fig. 2, fig. 2 is an application scenario illustrating a method for searching content according to an embodiment of the present application. In the application scenario shown in fig. 2, a terminal device 201 and a server 202 are included, wherein the terminal device 201 is used as a device for target object, and the server 202 is used as a device for content search.

The target object uses the terminal equipment 201 to browse the content platform, and the server 202 performs feature conversion processing on the release feature values of the candidate content in the candidate content set to obtain release features of the candidate content; the candidate content in the candidate content set is obtained by storing the content platform after the content player puts in the content platform. For example, when the target object is object a, the content is an advertisement, and the putting feature value is an putting bid, the object a uses the terminal device 201 to browse the advertisement platform, and the server 202 can feature-convert the putting bid of the candidate advertisement in the candidate advertisement set to obtain the putting feature of the candidate advertisement.

The server 202 performs interactive fusion processing on the delivery characteristics and the first content coding vectors of the candidate contents to obtain second content coding vectors of the candidate contents; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents a fusion characteristic of the delivery characteristic and the content characteristics. For example, based on the above example, server 202 may interactively fuse the placement characteristics of the candidate advertisement with a first advertisement encoding vector of the candidate advertisement to a second advertisement encoding vector of the candidate advertisement, wherein the first advertisement encoding vector is derived by encoding the advertisement characteristics of the candidate advertisement and wherein at least one-dimensional component of the second advertisement encoding vector represents the fused characteristics of the placement characteristics and the advertisement characteristics.

The server 202 searches candidate contents in the candidate content set based on the second content coding vector and the target object coding vector of the target object to obtain target content; the target object code vector is obtained by performing coding processing on the object characteristics of the target object. Correspondingly, the server 202 transmits the target content to the terminal device 201, so that the terminal device 201 presents the target content through the content platform. For example, based on the above example, server 202 may search for a candidate advertisement in the candidate advertisement set by the second content encoding vector of the candidate advertisement and the object encoding vector of object a, where the object encoding vector of object a is obtained by encoding the object feature of object a.

Therefore, interactive fusion is realized between the release characteristic obtained by the release characteristic value of the characteristic conversion candidate content and the first content coding vector obtained by the content characteristic of the coding candidate content, so that the second content coding vector after interactive fusion can be enhanced to be fused into the release characteristic; and further, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, so that the release characteristic value of the candidate content is effectively reflected on the content searching effect, the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, and the content searching effect is improved.

The method comprises the steps that interactive fusion is achieved between a throwing feature obtained by throwing bid of feature conversion candidate advertisements and a first advertisement coding vector obtained by coding advertisement features of the candidate advertisements, so that a second advertisement coding vector after interactive fusion can be strengthened to be fused into the throwing feature; and further, the influence effect of the bid for putting the candidate advertisement is enhanced in the advertisement searching process, so that the bid for putting the candidate advertisement is effectively reflected on the advertisement searching effect, the sensitivity of the advertisement searching effect to the bid for putting the candidate advertisement is effectively improved, and the advertisement searching effect is improved.

It can be understood that, in the method for searching content provided by the present application, the object features of the target object relate to relevant data such as users, and when the above embodiments of the present application are applied to specific products or technologies, individual permissions or individual agreements of users need to be obtained, and the collection, use and processing of relevant data need to comply with relevant laws and regulations and standards of relevant countries and regions.

The method for searching content provided by the embodiment of the application is realized based on artificial intelligence (Artificial Intelligence, AI), wherein the artificial intelligence is the theory, method, technology and application system which simulate, extend and expand human intelligence by using a digital computer or a machine controlled by the digital computer, sense environment, acquire knowledge and acquire an optimal result by using the knowledge. In other words, artificial intelligence is an integrated technology of computer science that attempts to understand the essence of intelligence and create a new intelligent machine that can react in a manner similar to human intelligence. Artificial intelligence, i.e. research on design principles and implementation methods of various intelligent machines, enables the machines to have functions of sensing, reasoning and decision.

The artificial intelligence technology is a comprehensive subject, and relates to the technology with wide fields, namely the technology with a hardware level and the technology with a software level. Artificial intelligence infrastructure technologies generally include technologies such as sensors, dedicated artificial intelligence chips, cloud computing, distributed storage, big data processing technologies, operation/interaction systems, mechatronics, and the like. The artificial intelligence software technology mainly comprises computer vision technology, voice processing technology, natural language processing (Nature Language processing, NLP) technology, machine Learning (ML)/deep learning, automatic driving, intelligent traffic and other directions.

In the embodiment of the application, the artificial intelligence software technology mainly comprises the natural language processing technology, machine learning/deep learning and other directions. For example, the technology of text processing, semantic understanding and the like in natural language processing can be involved, and various artificial neural networks and the like in machine learning/deep learning can be involved.

The method for searching the content provided by the application can be applied to the equipment with data processing capability for searching the content, such as a server and terminal equipment. The server may be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, a cloud server providing cloud computing service, or the like, but is not limited thereto; terminal devices include, but are not limited to, cell phones, tablets, computers, smart cameras, smart voice interaction devices, smart appliances, vehicle terminals, aircraft, and the like. The terminal device and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein.

The content search device may be capable of performing natural language processing, an important direction in the fields of computer science and artificial intelligence. It is studying various theories and methods that enable effective communication between a person and a computer in natural language. Natural language processing is a science that integrates linguistics, computer science, and mathematics. Thus, the research in this field will involve natural language, i.e. language that people use daily, so it has a close relationship with the research in linguistics. Natural language processing techniques typically include text processing, semantic understanding, machine translation, robotic questions and answers, knowledge graph spectroscopy, and the like. In the embodiment of the application, the device for searching the content can obtain the first content coding vector of the candidate content by coding the content characteristics of the candidate content through text processing, semantic understanding and other technologies in natural language processing, obtain the target object coding vector by coding the object characteristics of the target object and the like.

The content search device may be machine learning capable. Machine learning is a multi-domain interdisciplinary, involving multiple disciplines such as probability theory, statistics, approximation theory, convex analysis, algorithm complexity theory, and the like. It is specially studied how the computer simulates or implements the learning behavior of human beings to acquire new knowledge or skills, and reorganizes the existing knowledge structure to continuously improve the performance of the computer. Machine learning is the core of artificial intelligence, a fundamental approach to making computers intelligent, which is applied throughout various areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, confidence networks, reinforcement learning, transfer learning, learning by learning, teaching learning, and the like. In the embodiment of the application, the coding processing and the like in the content searching method mainly relate to application to an artificial neural network, and coding is realized through the artificial neural network.

The method for searching content provided by the embodiment of the application can also relate to a blockchain, wherein data such as processing parameters related to the interactive fusion processing and the coding processing can be stored on the blockchain.

The method for searching the content provided by the application can be applied to various scenes, including but not limited to cloud technology, artificial intelligence, or vehicle-mounted scenes.

The method for searching the content provided by the embodiment of the application is specifically described below by using the server as the device for searching the content.

Referring to fig. 3, a flowchart of a method for searching content according to an embodiment of the present application is shown. As shown in fig. 3, the method of content search includes the steps of:

s301: and performing feature conversion processing on the release feature values of the candidate contents in the candidate content set to obtain release features of the candidate contents.

In the related art, after a content distributor distributes content to a content platform, an object-based content searching method simply performs coding processing on object features and content features to obtain two equal-dimension vectors, and searches a plurality of contents based on the two equal-dimension vectors to obtain content with good relevance to an object.

According to research, when a content distributor distributes content, the distribution bid of the content is an important information, and needs to be reflected on the content searching effect; however, the above-described content search method does not consider the influence of the bid for placement of the content on the content search effect, that is, the content search effect is not sensitive to the bid for placement of the content, resulting in poor content search effect.

In the embodiment of the application, on the basis of the content searching method, the influence of the content delivery bid on the content searching effect is also required to be considered, so that the content searching effect has sensitivity to the content delivery bid; based on the above, after the content is delivered by the content delivery person, the delivered content is used as the candidate content to form a candidate content set, and the delivering bid of the candidate content is used as the delivering characteristic value of the candidate content, the delivering characteristic value characteristic of the candidate content is firstly required to be converted into the characteristic which is convenient for coding processing, and then the characteristic is used as the delivering characteristic of the candidate content.

The input characteristic values of the candidate content can be divided into discrete type and continuous type according to the data types; therefore, in the specific implementation of S301, the feature of the candidate content may be converted into the discrete feature of the candidate content, or the feature of the candidate content may be converted into the continuous feature of the candidate content, which is specifically implemented as follows:

in a first specific implementation manner, the feature conversion processing of the release feature value of the candidate content into discrete release features actually means that: firstly, acquiring a release characteristic value of candidate content and a release mode of the candidate content; the release mode refers to a display mode, a click mode or a conversion mode, further, the conversion mode is divided into a shallow conversion mode or a deep conversion mode, the display mode represents that candidate contents are deducted according to release characteristic values according to display, the click mode represents that the candidate contents are deducted according to clicking to release characteristic values, the conversion mode represents that a conversion target is reached to release the characteristic values, the shallow conversion mode represents that the shallow conversion target is reached to release the characteristic values, and the deep conversion mode represents that the deep conversion target is reached to release the characteristic values. Then, under the condition of different partition sets corresponding to different release modes, according to the partition set corresponding to the release mode, the release characteristic value of the candidate content is divided into the corresponding partition, so that discrete release characteristics serving as the release characteristics of the candidate content can be obtained. Thus, the present application provides a possible implementation manner, where the data type of the delivering feature is discrete, and S301 may include, for example, the following S3011-S3012:

S3011: and acquiring the release characteristic value of the candidate content and the release mode of the candidate content.

In the specific implementation of S3011, considering that different delivery characteristic values of different candidate contents generally meet long-tail distribution, that is, higher delivery characteristic values in different delivery characteristic values are less but have importance; based on this, in order to avoid adverse effects caused by the fact that the higher delivery characteristic value is an abnormal delivery characteristic value, the content delivery person takes the set delivery characteristic value of the candidate content as an initial delivery characteristic value, and after the initial delivery characteristic value of the candidate content is obtained, a second transformation process is further required to be performed on the initial delivery characteristic value of the candidate content so as to obtain the delivery characteristic value of the candidate content. And the fluctuation among the release characteristic values of the candidate contents in the candidate content set is smaller than the fluctuation among the initial release characteristic values of the candidate contents in the candidate content set. Thus, the present application provides a possible implementation manner, and the step of obtaining the serving feature value of the candidate content in S3011 may include, for example, the following S1-S2:

s1: and obtaining initial delivery characteristic values of the candidate contents.

S2: performing second transformation processing on the initial release characteristic value to obtain a release characteristic value; the volatility between the delivery characteristic values of the candidate contents in the candidate content set is smaller than the volatility between the initial delivery characteristic values of the candidate contents in the candidate content set.

The second transformation process may be, for example, log (1+x) data transformation process, where x represents an initial placement feature value of the candidate content; that is, S2 may include, for example: and carrying out log (1+x) data transformation processing on the initial release characteristic value x of the candidate content to obtain the release characteristic value of the candidate content. Of course, in the embodiment of the present application, the second transformation process may also be other implementation manners, for example, other log function transformation processes, etc., which are not described herein.

As an example, when the candidate content is a candidate advertisement and the putting characteristic value is a putting bid, obtaining the initial putting bid of the candidate advertisement as x, and performing log (1+x) data transformation on the initial putting bid x of the candidate advertisement to obtain the putting bid of the candidate advertisement as x'.

As another example, the candidate content is a candidate advertisement, and the special is placedWhen the sign value is the bid, the initial bid of the candidate advertisement is x ₁ And x ₂ Initial impression bid x for candidate advertisement ₁ And x ₂ Log (1+x) data transformation is respectively carried out to obtain that the putting bid of the candidate advertisement is x ₁ ' and x ₂ ＇。

In the implementation of S3011, the delivery mode of the candidate content includes a presentation mode, a click mode, or a conversion mode, and the conversion mode includes a shallow conversion mode or a deep conversion mode. Wherein the presentation mode may be, for example, a pay Per presentation (CPC) mode, the Click mode may be, for example, a pay Per Click (Cost Per Thousand Impression, CPM) mode, and the conversion mode may be, for example, a pay Per presentation (Optimization Cost Per Thousand Impression, OCPM) optimized based on the conversion objective or a pay Per Click (Optimization Cost Per Click, OCPC) optimized based on the conversion objective. Correspondingly, the shallow conversion mode may be, for example, OCPM based on a shallow conversion target or OCPC based on a shallow conversion target, and the deep conversion mode may be, for example, OCPM based on a deep conversion target or OCPC based on a deep conversion target.

S3012: and dividing the release characteristic value based on the partition set corresponding to the release mode to obtain the release characteristic.

In the specific implementation of S3012, because the differences such as the mean value and the variance corresponding to the candidate content in the candidate content set in different delivery modes are different, the partition granularity of different partition sets corresponding to different delivery modes is different, in order to divide the delivery characteristic value of the candidate content into more accurate partitions, it may be determined based on the actual situation that the partition granularity of the partition set corresponding to the display mode is smaller than the partition granularity of the partition set corresponding to the click mode, and the partition granularity of the partition set corresponding to the conversion mode is not uniform.

As an example, the partition granularity of the partition set corresponding to the CPM mode is smaller than the partition granularity of the partition set corresponding to the CPC mode, and the partition granularity of the partition sets corresponding to the OCPM and OCPC modes is not uniform. If the delivery mode of the candidate advertisement is CPM mode (or CPC mode, shallow conversion target-based)OCPM, OCPC based on shallow conversion target), the obtained bid x 'of the candidate advertisement is divided by a partition set corresponding to CPM mode (or CPC mode, OCPM based on shallow conversion target, OCPC based on shallow conversion target), and the bid x' of the candidate advertisement is obtained to obtain the bid feature y of the candidate advertisement. If the putting mode of the candidate advertisement is OCPM based on the deep conversion target (or OCPC based on the deep conversion target), the obtained putting bid of the candidate advertisement is x ₁ ' and x ₂ ' offer x for placement of a candidate advertisement through a set of partitions corresponding to OCPM based on deep conversion objectives (or OCPC based on deep conversion objectives) ₁ ' and x ₂ ' dividing to obtain release characteristic y ₁ And y ₂ 。

In a second specific implementation manner, the feature conversion processing of the release feature value of the candidate content into the continuous release feature is actually that: firstly, acquiring the release characteristic values of candidate contents, then judging whether the number of the release characteristic values of the candidate contents is single or multiple, and if the number of the release characteristic values of the candidate contents is single, directly converting the release characteristic values of the candidate contents into continuous release characteristics serving as the release characteristics of the candidate contents; if the number of the candidate content is multiple, the multiple release characteristic values of the candidate content are fused and then converted into continuous release characteristics, and the continuous release characteristics are used as the release characteristics of the candidate content. Thus, the present application provides a possible implementation, where the data type of the drop feature is continuous, S301 may include, for example, the following S3013-S3015:

s3013: and obtaining the release characteristic value of the candidate content.

Similarly, referring to the specific implementation manner of the step of obtaining the release characteristic value of the candidate content in S3011, when S3013 is specifically implemented, S3013 may include, for example: acquiring an initial delivery characteristic value of the candidate content; performing second transformation processing on the initial release characteristic value to obtain a release characteristic value; the volatility between the casting characteristic values of the candidate contents in the candidate content set is smaller than the volatility between the initial casting characteristic values of the candidate contents in the candidate content set.

S3014: and if the release characteristic values of the candidate contents are single, performing first transformation processing on the release characteristic values to obtain release characteristics.

When the release mode of the candidate content is a display mode, a click mode or a shallow conversion mode, the release characteristic value of the candidate content corresponds to a single target, and the release characteristic value of the candidate content is single. Therefore, the application provides a possible implementation manner, and when the release mode of the candidate content is a display mode, a click mode or a shallow conversion mode, the release characteristic value of the candidate content is single.

The first transformation process may be, for example, a normalized transformation process based on a continuous system, a log function transformation process, or the like; that is, S3014 may include, for example: and if the release characteristic values of the candidate content are single, carrying out continuous standardized transformation processing or log function transformation processing on the release characteristic values to obtain release characteristics of the candidate content. Of course, in the embodiment of the present application, the second transformation process may also be other implementation manners, for example, other function transformation processes, etc., which are not described herein.

S3015: and if the number of the release characteristic values of the candidate content is multiple, carrying out fusion processing and first transformation processing on the multiple release characteristic values to obtain release characteristics.

When the release mode of the candidate content is a deep conversion mode, the release characteristic values of the candidate content correspond to multiple targets, and the release characteristic values of the candidate content are multiple. Therefore, the application provides a possible implementation mode, and when the putting mode is a deep conversion mode, the putting characteristic values of the candidate content are multiple.

The fusion process may be, for example, a weighting process; that is, S3015 may include, for example: if the number of the release characteristic values of the candidate content is multiple, weighting the release characteristic values, and then carrying out continuous standardized transformation processing or log function transformation processing to obtain the release characteristic of the candidate content. However, in the embodiment of the present application, the fusion process may be implemented in other ways, which are not described herein.

As one example, the candidate content is a candidate advertisement, which is placedWhen the characteristic value is the bid, if the bid of the candidate advertisement is CPM mode, CPC mode, OCPM based on shallow conversion target or OCPC based on shallow conversion target, the obtained bid of the candidate advertisement is x ', and continuous standardized transformation processing or log function transformation processing is performed on the bid of the candidate advertisement x', so as to obtain the bid characteristic y of the candidate advertisement. If the putting mode of the candidate advertisement is OCPM based on the deep conversion target or OCPC based on the deep conversion target, the putting bid of the obtained candidate advertisement is x ₁ ' and x ₂ ' bid x for placement of candidate advertisement ₁ ' and x ₂ The' first weighting processing and then standardized transformation processing or log function transformation processing based on continuous mode are carried out, and the putting feature y of the candidate advertisement is obtained. Wherein the weighting process can be, for example, alpha x ₁ ＇+βx ₂ The values of α and β are set and adjusted according to practical application requirements, for example, α=0.5 and β=0.5.

S302: performing interactive fusion processing on the throwing characteristics and the first content coding vectors of the candidate contents to obtain second content coding vectors of the candidate contents; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents a fusion characteristic of the release characteristic and the content characteristic.

In the embodiment of the present application, after the release feature of the candidate content is obtained in S301, the release feature of the candidate content and the first content coding vector obtained by coding the content feature of the candidate content are further required to be subjected to an interactive fusion process, so as to obtain a second content coding vector after the interactive fusion. At least one-dimensional component in the second content coding vector after interactive fusion represents fusion characteristics of the release characteristics and the content characteristics, and compared with the vector obtained by only splicing the release characteristics of the candidate content and the first content coding vector, each one-dimensional component in the vector obtained by splicing the candidate content only represents release characteristics or content characteristics, the second content coding vector after interactive fusion can strengthen the fusion of the release characteristics; and further, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, so that the release characteristic value of the candidate content is effectively reflected on the content searching effect, the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, and the content searching effect is improved.

Corresponding to the specific implementation manner of S301, the feature of the release feature value of the candidate content may be converted into a discrete release feature, or the feature of the release feature value of the candidate content may be converted into a continuous release feature; in the specific implementation of S302, the interactive fusion processing manner of the discrete type release feature and the first content coding vector is also different from the interactive fusion processing manner of the continuous type release feature and the first content coding vector, and the specific implementation manner is as follows:

the first specific implementation manner, the interactive fusion processing manner of the discrete release feature and the first content coding vector refers to: firstly, coding discrete type throwing features to obtain throwing feature coding vectors, wherein the dimension requirements of the throwing feature coding vectors are kept matched with the dimension requirements of the first content coding vectors; and fusing the release characteristic coding vector to the first content coding vector in a vector fusion mode to obtain a second content coding vector. In the mode, the second content coding vector is obtained by interactively fusing the release characteristic coding vector and the first content coding vector, the release characteristic coding vector is obtained by coding the release characteristic of the candidate content, and the first content coding vector is obtained by coding the content characteristic of the candidate content; thus, the second content encoding vector interworks the delivery characteristics of the candidate content and the content characteristics of the candidate content. That is, the present application provides one possible implementation, S302 may include, for example, the following S3021-S3022:

S3021: coding the release characteristics to obtain release characteristic coding vectors of the release characteristics; the dimensions of the put feature encoding vector match the dimensions of the first content encoding vector.

When S3021 is specifically implemented, firstly, simple encoding processing, for example, embedding processing, is required to be performed on discrete type release features, so as to obtain a plurality of release feature embedded vectors, and similarly, the dimensions of the release feature embedded vectors need to be kept matched with the dimensions of the first content encoded vector; then, complex coding processing, such as feature interaction processing, pooling processing and the like, is further needed to be carried out on the plurality of release feature embedded vectors, feature interaction processing is carried out on the plurality of release feature embedded vectors to obtain release feature interaction vectors, and pooling processing is carried out on the release feature interaction vectors to obtain release feature coding vectors. Thus, the present application provides one possible implementation, S3021 may include, for example, the following S3-S5:

s3: embedding the release characteristics to obtain a plurality of release characteristic embedding vectors of the release characteristics; the dimensions of the put feature embedding vector match the dimensions of the first content encoding vector.

S4: performing feature interaction processing on the multiple release feature embedding vectors to obtain release feature interaction vectors; the dimensions of the put feature interaction vector are matched with the dimensions of the first content encoding vector.

S5: and carrying out pooling treatment on the throwing feature interaction vector to obtain a throwing feature coding vector.

In order to improve accuracy of the input feature coding vector, the pooling processing may be pooling processing based on weights, where the weights are set and adjusted according to actual application requirements, for example, the weights may be obtained by adaptively adjusting model parameters in a training process of the input feature coding model of the candidate content.

S3022: and carrying out vector fusion processing on the input feature code vector and the first content code vector to obtain a second content code vector, wherein the dimension of the second content code vector is matched with the dimension of the first content code vector.

The vector fusion process may be, for example, a vector summation process; that is, S3022 may include, for example: and carrying out bit-wise addition processing on the release characteristic coding vector and the first content coding vector to obtain a second content coding vector, wherein each dimension component in the second content coding vector represents the fusion characteristic of the release characteristic and the content characteristic. Of course, in the embodiment of the present application, the vector fusion process may also be other implementation manners, for example, a bitwise multiplication process, etc., which are not described herein.

As an example, when the candidate content is a candidate advertisement and the bid is a bid, see a schematic diagram of an interactive fusion process of a discrete bid feature and a first advertisement coding vector shown in fig. 4. The right model is a characteristic coding model of the candidate advertisement, and comprises an embedding layer, a characteristic interaction layer and a pooling layer, wherein after the characteristic of the candidate advertisement is converted into a discrete type characteristic, the discrete type characteristic is input into the embedding layer for embedding, and a plurality of characteristic embedding vectors are output; embedding a plurality of input features into a vector input feature interaction layer to perform feature interaction processing, and outputting input feature interaction vectors; and inputting the throwing characteristic interaction vector into a pooling layer, and outputting the throwing characteristic coding vector of the candidate advertisement. The intermediate model is an advertisement feature coding model of the candidate advertisement, the model comprises an embedding layer, a preset network layer and a full connection layer, the advertisement features of the candidate advertisement are input into the model, and a first advertisement coding vector of the candidate advertisement is output through the embedding layer, the preset network layer and the full connection layer. And carrying out vector fusion processing on the putting feature code vector and the first advertisement code vector above the two models, for example, carrying out bit-by-bit addition processing to obtain a second advertisement code vector of the candidate advertisement, wherein each dimension component in the second advertisement code vector represents fusion features of the putting feature and the advertisement feature.

In a second specific implementation manner, the interactive fusion processing manner of the continuous release feature and the first content coding vector refers to: firstly, fusing the release characteristics and the first content coding vector in a scalar fusion processing mode to obtain a scalar which is used as a fused scalar; and splicing the fusion scalar to the first content coding vector to obtain a second content coding vector. In the mode, the second content coding vector is obtained by splicing a fusion scalar and a first content coding vector, the fusion scalar is obtained by interactively fusing the release characteristic and the first content coding vector, and the first content coding vector is obtained by coding the content characteristic of the candidate content; thus, the second content encoding vector interactively fuses the serving feature of the candidate content and the content feature of the candidate content. Thus, the present application provides one possible implementation, S302 may include, for example, the following S3023-S3024:

s3023: and scalar fusion processing is carried out on the release characteristics and the first content coding vector, so as to obtain a fusion standard quantity.

In the specific implementation of S3023, first, the first content encoding vector needs to be processed into a content scalar by a scalar-based conversion processing manner; then, based on the content scalar, the content scalar and the delivery feature are fused to obtain a fused scalar. Thus, the present application provides one possible implementation, S3023, for example, may include the following S6-S7:

S6: and performing scalar-based conversion processing on the first content coding vector to obtain a content scalar.

Wherein the scalar-based transformation process may be, for example, vector compression; that is, S6 may include, for example: and carrying out vector compression on the first content coding vector in the form of a class logistic regression model to obtain a content scalar. Of course, the scalar-based conversion process in the embodiment of the present application may also be other implementations, for example, a bit-wise averaging process, etc., which are not described herein.

S7: and carrying out fusion processing on the content scalar and the release characteristic to obtain a fusion scalar.

The fusion process may be, for example, a summation process, and further, may be a weighting process; that is, S7 may include, for example: and weighting the content scalar and the release characteristic to obtain a fusion scalar. Of course, the scalar-based conversion process in the embodiment of the present application may also be other implementations, such as a product-seeking process, and so on, which are not described herein.

S3024: and performing splicing processing on the fusion scalar and the first content coding vector to obtain a second content coding vector.

In the implementation of S3024, for example, the dimension of the first content encoding vector is k, and the fusion scalar is spliced to the first content encoding vector to obtain a second content encoding vector, where the dimension of the second content encoding vector is k+1, and the spliced one-dimensional component in the second content encoding vector represents the fusion feature of the delivery feature and the content feature.

As an example, when the candidate content is a candidate advertisement and the bid feature value is a bid, see a schematic diagram of an interactive fusion process of a continuous bid feature and a first advertisement coding vector shown in fig. 5. The right model is an advertisement feature coding model of the candidate advertisement, the model comprises an embedding layer, a preset network layer and a full connection layer, advertisement features of the candidate advertisement are input into the model, and a first advertisement coding vector of the candidate advertisement is output through the embedding layer, the preset network layer and the full connection layer. After converting the bid feature of the candidate advertisement into continuous bid feature, compressing the first content coding vector in the form of a logistic regression model to obtain an advertisement scalar; weighting advertisement scalar and continuous type throwing characteristics to obtain a fusion scalar; and performing splicing processing on the fusion scalar and the first advertisement coding vector to obtain a second advertisement coding vector of the candidate advertisement, wherein the spliced one-dimensional component in the second advertisement coding vector represents the fusion characteristics of the release characteristic and the advertisement characteristic.

S303: searching candidate contents in the candidate content set based on the second content coding vector and the target object coding vector of the target object to obtain target contents; the target object encoding vector is obtained by encoding the object feature of the target object.

In the embodiment of the present application, after the second content coding vector of the candidate content is obtained in S302, since the second content coding vector of the candidate content can strengthen the impression feature, the second content coding vector of the candidate content needs to be used to replace the first content coding vector of the candidate content, and the target object coding vector obtained by combining the object features of the encoding target object is searched for the candidate content in the candidate content set, and the effect of the impression feature of the candidate content is enhanced in the content searching process, so that the target content can be obtained. The method is convenient for the bid of the candidate advertisement to be effectively reflected on the advertisement searching effect, so that the sensitivity of the advertisement searching effect to the bid of the candidate advertisement is effectively improved, and the advertisement searching effect is improved.

When S303 is specifically implemented, first, a preset operation may be performed on the second content encoding vector and the target object encoding vector of the target object in a preset operation manner, where the obtained operation result is used as an operation result corresponding to the candidate content; then, on the basis of presetting a search condition for the operation result as a preset search condition, the target content can be obtained from the candidate content search in the candidate content set by the operation result and the preset search condition. Thus, the present application provides one possible implementation, S303 may include, for example, the following S3031-S3032:

S3031: and carrying out preset operation based on the second content coding vector and the target object coding vector of the target object to obtain an operation result corresponding to the candidate content.

The preset operation may be, for example, an inner product operation; that is, S3031 may include, for example: and performing inner product operation based on the second content coding vector and the target object coding vector of the target object to obtain inner products corresponding to the candidate content. Of course, in the embodiment of the present application, the second transformation process may also be other implementation manners, for example, vector distance operation, etc., which are not described herein.

S3032: and searching candidate contents in the candidate content set based on the operation result and a preset search condition to obtain target contents.

The preset search condition may be, for example, selecting candidate contents corresponding to the first N operation results after sorting from high to low; that is, S3032 may include, for example: sequencing the operation results corresponding to the candidate contents in the candidate content set from high to low; selecting candidate contents corresponding to the first N ordered operation results as target contents; n is a positive integer, and N is more than or equal to 2.

In addition, in the embodiment of the present application, the preset search condition may be, for example, selecting candidate content greater than or equal to a preset threshold value; that is, S3032 may include, for example: and searching candidate contents which are larger than or equal to a preset threshold value in the candidate content set as target contents based on the operation result and the preset threshold value.

As an example, when the candidate content is a candidate advertisement, the left model in fig. 4 and 5 is a target object coding model of the target object, where the model includes an embedding layer, a preset network layer and a full connection layer, and the object feature of the target object is input into the model, and the target object coding vector of the target object is output through the embedding layer, the preset network layer and the full connection layer on the basis of the above example about S302; on the basis, carrying out inner product operation on the second advertisement coding vector of the candidate advertisement and the target object coding vector of the target object to obtain an inner product corresponding to the candidate advertisement; ordering the inner products corresponding to the candidate advertisements in the candidate advertisement set from high to low; selecting candidate advertisements corresponding to the first N inner products after sequencing as target advertisements; n is positive integer, N is more than or equal to 2.

It should be noted that, in the specific implementation of S302 when the data type corresponding to the delivery feature is continuous, S6-S7 and S3024, for the step of obtaining the target object encoding vector of the target object, the step of obtaining the second content encoding vector with reference to the candidate content is required. Firstly, coding object features to obtain an initial object coding vector; then, the initial object code vector needs to be processed into an object scalar by a scalar-based conversion processing mode; then, the object scalar is spliced to the initial object code vector on the basis of the object scalar, and the target object code vector is obtained. Thus, the present application provides a possible implementation manner, and the step of obtaining the target object coding vector may include, for example, the following S8-S10:

S8: and carrying out coding processing on the object characteristics of the target object to obtain an initial object coding vector.

S9: and performing scalar-based conversion processing on the initial object coding vector to obtain an object scalar.

S10: and performing splicing processing on the object scalar and the initial object coding vector to obtain a target object coding vector.

For example, the dimension of the initial object code vector is k, and the object scalar is stitched to the initial object code vector to obtain the target object code vector, where the dimension of the target object code vector is k+1.

In addition, in the embodiment of the present application, in order to further avoid that S3024 obtains a large difference between the fused scalar in the second content encoding vector and the numerical value of other dimensions in the second content encoding vector, and further avoid that S10 obtains a large difference between the object scalar in the target object encoding vector and the numerical value of other dimensions in the target object encoding vector, the second content encoding vector and the target object encoding vector need to be subjected to normalization processing, so that both the regularized second content encoding vector and the regularized target object encoding vector conform to a regularized constraint. Based on this, when S303 is implemented, the regularized second content coding vector needs to be utilized, and the target object coding vector after regularization is combined with the candidate content in the candidate content set to search for the target content. Thus, the present application provides a possible implementation, and the method may further comprise, for example, the following S11-S12:

S11: and regularizing the second content coding vector to obtain a regularized second content coding vector.

S12: regularizing the target object coding vector to obtain the regularized target object coding vector.

Corresponding to the above S11-S12, S303 may include, for example: searching candidate contents in the candidate content set based on the regularized second content coding vector and the regularized target object coding vector to obtain target contents.

According to the content searching method provided by the embodiment, the characteristic of the release characteristic value of the candidate content in the candidate content set is converted into the release characteristic; interactively fusing the release characteristic and the first content coding vector of the candidate content into a second content coding vector, wherein the first content coding vector is obtained by coding the content characteristic of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristic of the release characteristic and the content characteristic; and searching candidate contents in the candidate content set through the second content coding vector and a target object coding vector of the target object to obtain target contents, wherein the target object coding vector is obtained by coding object characteristics of the target object. Based on the method, the method realizes interactive fusion of the release characteristic obtained by the release characteristic value of the characteristic conversion candidate content and the first content coding vector obtained by the content characteristic of the coding candidate content, so that the second content coding vector after interactive fusion can be enhanced to be fused into the release characteristic; and further, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, so that the release characteristic value of the candidate content is effectively reflected on the content searching effect, the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, and the content searching effect is improved.

The method for searching the content provided by the embodiment of the application can be applied to an advertisement system, and the specific implementation mode of the method for searching the advertisement is as follows:

one specific implementation mode is as follows: firstly, aiming at candidate advertisements in a candidate advertisement set, acquiring initial bid and mode of putting the candidate advertisements, and carrying out log (1+x) data transformation on the initial bid of the candidate advertisements to obtain the bid of the candidate advertisements, wherein x represents initial bid price; and dividing the bid for delivering the candidate advertisement according to a partition set corresponding to the CPM mode (or CPC mode, OCPM based on a shallow conversion target, OCPC based on a shallow conversion target, OCPM based on a deep conversion target and OCPC based on a deep conversion target) and the CPM mode (or CPC mode, OCPM based on a shallow conversion target, OCPC based on a shallow conversion target, OCPM based on a deep conversion target and OCPC based on a deep conversion target) to obtain the delivering characteristic of the candidate advertisement.

Then, inputting the putting features of the candidate advertisements into an embedding layer for embedding treatment, and outputting a plurality of putting feature embedding vectors; embedding a plurality of release features into a vector input feature interaction layer to perform feature interaction processing, and outputting release feature interaction vectors; inputting the input feature interaction vector into a pooling layer, and outputting a candidate advertisement input feature coding vector; and carrying out bit-wise addition processing on the advertisement feature coding vector of the candidate advertisement and the first advertisement coding vector of the candidate advertisement to obtain a second advertisement coding vector of the candidate advertisement, wherein the first advertisement coding vector of the candidate advertisement is obtained by inputting advertisement features of the candidate advertisement into an embedded layer, a preset network layer and a full connection layer for coding processing and then outputting, and each dimension component in the second advertisement coding vector represents fusion features of the advertisement feature and the advertisement feature.

Finally, carrying out inner product operation on a second advertisement coding vector of the candidate advertisement and a target object coding vector of the target object to obtain an inner product corresponding to the candidate advertisement, wherein the target object coding vector of the target object is obtained by inputting object characteristics of the target object into an embedded layer, a preset network layer and a full connection layer for coding processing and then outputting; ordering the inner products corresponding to the candidate advertisements in the candidate advertisement set from high to low; selecting candidate advertisements corresponding to the first N inner products after sequencing as target advertisements; n is a positive integer, and N is more than or equal to 2.

Another specific implementation mode is as follows: firstly, aiming at candidate advertisements in a candidate advertisement set, acquiring initial bid of the candidate advertisements, and performing log (1+x) data transformation on the initial bid of the candidate advertisements to obtain bid of the candidate advertisements, wherein x represents the initial bid of the candidate advertisements; if the putting mode of the candidate advertisement is a CPM mode, a CPC mode, OCPM based on a shallow conversion target or OCPC based on the shallow conversion target, carrying out continuous standardized transformation processing or log function transformation processing on the putting bid of the candidate advertisement to obtain the putting characteristic of the candidate advertisement; and if the putting mode of the candidate advertisement is OCPM based on the deep conversion target or OCPC based on the deep conversion target, weighting the putting bid of the candidate advertisement, and then carrying out continuous standardized transformation processing or log function transformation processing to obtain the putting characteristic of the candidate advertisement.

Then, inputting advertisement characteristics of the candidate advertisements into an embedding layer, a preset network layer and a full connection layer for coding treatment, and outputting a first advertisement coding vector of the candidate advertisements; vector compression is carried out on the first content coding vector in a form of a class logistic regression model to obtain an advertisement scalar; weighting advertisement scalar and continuous type throwing characteristics to obtain a fusion scalar; and performing splicing processing on the fusion scalar and the first advertisement coding vector to obtain a second advertisement coding vector of the candidate advertisement, wherein spliced one-dimensional components in the second advertisement coding vector all represent fusion characteristics of the release characteristic and the advertisement characteristic.

Finally, inputting object characteristics of the target object into an embedding layer, a preset network layer and a full connection layer for coding processing in the target object coding vector of the target object, outputting an initial object coding vector, performing scalar-based conversion processing on the initial object coding vector to obtain an object scalar, and performing splicing processing on the object scalar and the initial object coding vector to obtain the target object coding vector; regularization treatment is carried out on the second advertisement coding vector of the candidate advertisement to obtain a regularized second advertisement coding vector, and regularization treatment is carried out on the target object coding vector to obtain a regularized target object coding vector; performing inner product operation on the regularized second advertisement coding vector and the regularized target object coding vector to obtain inner products corresponding to the candidate advertisements; sequencing the inner products corresponding to the candidate advertisements in the candidate advertisement set from high to low; selecting candidate advertisements corresponding to the first N inner products after sequencing as target advertisements; n is a positive integer, and N is more than or equal to 2.

Aiming at the content searching method provided by the embodiment, the embodiment of the application also provides a content searching device.

Referring to fig. 6, fig. 6 is a schematic diagram of an apparatus for searching content according to an embodiment of the present application. As shown in fig. 6, the apparatus 600 for content search includes: a feature conversion unit 601, a fusion unit 602, and a search unit 603;

the feature conversion unit 601 is configured to perform feature conversion processing on a delivery feature value of a candidate content in the candidate content set, so as to obtain a delivery feature of the candidate content;

the fusion unit 602 is configured to perform interactive fusion processing on the delivery feature and the first content coding vector of the candidate content, so as to obtain a second content coding vector of the candidate content; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristics of the release characteristics and the content characteristics;

a search unit 603 for searching candidate contents in the candidate content set for obtaining target contents based on the second content encoding vector and the target object encoding vector of the target object; the target object encoding vector is obtained by encoding the object feature of the target object.

As a possible implementation manner, the data type of the delivering feature is discrete, and the feature conversion unit 601 includes: a first acquisition subunit, a dividing subunit;

the first acquisition subunit is used for acquiring the release characteristic value of the candidate content and the release mode of the candidate content;

the dividing subunit is used for dividing the throwing characteristic value based on the partition set corresponding to the throwing mode to obtain throwing characteristics;

the fusion unit 602 includes: a first coding subunit and a first fusion subunit;

the first coding subunit is used for carrying out coding processing on the release characteristics to obtain release characteristic coding vectors of the release characteristics; the dimension of the input feature code vector is matched with the dimension of the first content code vector;

and the first fusion subunit is used for carrying out vector fusion processing on the release characteristic coding vector and the first content coding vector to obtain a second content coding vector, and the dimension of the second content coding vector is matched with the dimension of the first content coding vector.

As one possible implementation, the coding subunit includes: the device comprises an embedding module, a characteristic interaction module and a pooling module;

the embedding module is used for carrying out embedding processing on the throwing features to obtain a plurality of throwing feature embedding vectors of the throwing features; the dimension of the release feature embedding vector is matched with the dimension of the first content coding vector;

The feature interaction module is used for carrying out feature interaction processing on the plurality of release feature embedded vectors to obtain release feature interaction vectors; the dimension of the characteristic interaction vector is matched with the dimension of the first content coding vector;

and the pooling module is used for pooling the input feature interaction vector to obtain the input feature coding vector.

As one possible implementation, the put mode includes a show mode, a click mode, or a conversion mode, and the conversion mode includes a shallow conversion mode or a deep conversion mode; the partition granularity of the partition set corresponding to the display mode is smaller than that of the partition set corresponding to the click mode, and the partition granularity of the partition set corresponding to the conversion mode is uneven.

As a possible implementation manner, the data type of the put feature is continuous, and the feature conversion unit 601 includes: the second acquisition subunit, the first transformation subunit and the second transformation subunit;

the second acquisition subunit is used for acquiring the release characteristic values of the candidate contents;

the first transformation subunit is used for performing first transformation processing on the release characteristic values if the release characteristic values of the candidate contents are single, so as to obtain release characteristics;

The second transformation subunit is used for carrying out fusion processing and first transformation processing on the multiple release characteristic values if the release characteristic values of the candidate content are multiple, so as to obtain release characteristics;

the fusion unit 602 includes: a second fusion subunit and a first splice subunit;

the second fusion subunit is used for performing scalar fusion processing on the release characteristics and the first content coding vector to obtain a fusion scalar;

and the first splicing subunit is used for carrying out splicing processing on the fusion scalar and the first content coding vector to obtain a second content coding vector.

As one possible implementation, the second fusion subunit includes: a conversion module and a fusion module;

the conversion module is used for carrying out scalar-based conversion processing on the first content coding vector to obtain a content scalar;

and the fusion module is used for carrying out fusion processing on the content scalar and the release characteristic to obtain a fusion scalar.

As a possible implementation manner, the apparatus further includes: a coding unit; the encoding unit includes: a second coding subunit, a conversion subunit, and a second splicing subunit;

the second coding subunit is used for carrying out coding processing on the object characteristics of the target object to obtain an initial object coding vector;

A conversion subunit, configured to perform scalar-based conversion processing on the initial object encoding vector, to obtain an object scalar;

and the second splicing subunit is used for carrying out splicing processing on the object scalar and the initial object coding vector to obtain a target object coding vector.

As a possible implementation manner, the apparatus further includes: a first regularization unit and a second regularization unit;

the first regularization unit is used for regularizing the second content coding vector to obtain a regularized second content coding vector;

the second regularization unit is used for regularizing the target object coding vector to obtain a regularized target object coding vector;

a search unit 603 for:

searching candidate contents in the candidate content set based on the regularized second content coding vector and the regularized target object coding vector to obtain target contents.

As a possible implementation manner, when the release mode of the candidate content is a display mode, a click mode or a shallow conversion mode, the release characteristic value of the candidate content is single; and when the putting mode is a deep conversion mode, the putting characteristic values of the candidate contents are multiple.

As one possible implementation, the first acquisition subunit or the second acquisition subunit includes: an acquisition module and a transformation module;

the acquisition module is used for acquiring initial release characteristic values of the candidate contents;

the transformation module is used for carrying out second transformation processing on the initial release characteristic value to obtain a release characteristic value; the volatility between the delivery characteristic values of the candidate contents in the candidate content set is smaller than the volatility between the initial delivery characteristic values of the candidate contents in the candidate content set.

As a possible implementation, the search unit 603 includes: an operation subunit and a search subunit;

an operation subunit, configured to perform a preset operation based on the second content encoding vector and the target object encoding vector of the target object, and obtain an operation result corresponding to the candidate content;

and the searching subunit is used for searching candidate contents in the candidate content set based on the operation result and a preset searching condition to obtain target contents.

The content searching device provided by the embodiment converts the characteristic of the release characteristic value of the candidate content in the candidate content set into the release characteristic; interactively fusing the release characteristic and the first content coding vector of the candidate content into a second content coding vector, wherein the first content coding vector is obtained by coding the content characteristic of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristic of the release characteristic and the content characteristic; searching candidate contents in the candidate content set to obtain target contents through the second content coding vector and a target object coding vector of the target object, wherein the target object coding vector is obtained by coding object characteristics of the target object. Based on the method, the method realizes the interactive fusion of the release characteristic obtained by the release characteristic value of the characteristic conversion candidate content and the first content coding vector obtained by the content characteristic of the coding candidate content, so that the second content coding vector after the interactive fusion can be enhanced to be fused into the release characteristic; and further, the influence effect of the release characteristic value of the candidate content is enhanced in the content searching process, so that the release characteristic value of the candidate content is effectively reflected on the content searching effect, the sensitivity of the content searching effect to the release characteristic value of the candidate content is effectively improved, and the content searching effect is improved.

For the method of searching content described above, the embodiment of the present application further provides a device for searching content, so that the method of searching content described above is actually implemented and applied, and the computer device provided by the embodiment of the present application will be described from the perspective of hardware materialization.

Referring to fig. 7, fig. 7 is a schematic diagram of a server structure according to an embodiment of the present application, where the server 700 may have a relatively large difference due to different configurations or performances, and may include one or more central processing units (Central Processing Units, CPU) 722 (e.g., one or more processors) and a memory 732, and one or more storage media 730 (e.g., one or more mass storage devices) storing application programs 742 or data 744. Wherein memory 732 and storage medium 730 may be transitory or persistent. The program stored in the storage medium 730 may include one or more modules (not shown), each of which may include a series of instruction operations on a server. Still further, the central processor 722 may be configured to communicate with the storage medium 730 and execute a series of instruction operations on the server 700 in the storage medium 730.

The Server 700 may also include one or more power supplies 726, one or more wired or wireless network interfaces 750, one or more input/output interfaces 758, and/or one or more operating systems 741, such as Windows Server ^TM ，Mac OS X ^TM ，Unix ^TM ，Linux ^TM ， FreeBSD ^TM Etc.

The steps performed by the server in the above embodiments may be based on the server structure shown in fig. 7.

Wherein, the CPU 722 is configured to perform the following steps:

performing interactive fusion processing on the throwing characteristics and the first content coding vectors of the candidate contents to obtain second content coding vectors of the candidate contents; the first content coding vector is obtained by coding the content characteristics of the candidate content, and at least one-dimensional component in the second content coding vector represents the fusion characteristics of the release characteristics and the content characteristics;

searching candidate contents in the candidate content set based on the second content coding vector and the target object coding vector of the target object to obtain target contents; the target object encoding vector is obtained by encoding the object characteristics of the target object.

Optionally, the CPU 1222 may also perform method steps of any specific implementation of the method for content searching in an embodiment of the present application.

Referring to fig. 8, fig. 8 is a schematic structural diagram of a terminal device according to an embodiment of the present application. For convenience of explanation, only those portions of the embodiments of the present application that are relevant to the embodiments of the present application are shown, and specific technical details are not disclosed, please refer to the method portions of the embodiments of the present application. The terminal equipment can be any terminal equipment including a mobile phone, a tablet personal computer, a PDA and the like, and takes the terminal equipment as the mobile phone as an example:

fig. 8 is a block diagram showing a part of the structure of a mobile phone related to a terminal device provided by an embodiment of the present application. Referring to fig. 8, the mobile phone includes: radio Frequency (RF) circuitry 810, memory 820, input unit 830, display unit 840, sensor 850, audio circuitry 860, wireless fidelity (WiFi) module 870, processor 880, and power supply 890. Those skilled in the art will appreciate that the handset configuration shown in fig. 8 is not limiting of the handset and may include more or fewer components than shown, or may combine certain components, or may be arranged in a different arrangement of components.

The following describes the components of the mobile phone in detail with reference to fig. 8:

The RF circuit 810 may be used for receiving and transmitting signals during a message or a call, and in particular, after receiving downlink information of a base station, it is processed by the processor 880; and transmitting the uplink design data to the base station. Typically, the RF circuitry 810 includes, but is not limited to, an antenna, at least one amplifier, a transceiver, a coupler, a low noise amplifier (Low Noise Amplifier, LNA for short), a duplexer, and the like. In addition, the RF circuitry 810 may also communicate with networks and other devices via wireless communications. The wireless communication may use any communication standard or protocol, including but not limited to global system for mobile communications (Global System of Mobile communication, GSM), general packet radio service (General Packet Radio Service, GPRS), code division multiple access (Code Division Multiple Access, CDMA), wideband code division multiple access (Wideband Code Division Multiple Access, WCDMA), long term evolution (Long Term Evolution, LTE), email, short message service (Short Messaging Service, SMS), etc.

The memory 820 may be used to store software programs and modules, and the processor 880 may implement various functional applications and data processing of the cellular phone by running the software programs and modules stored in the memory 820. The memory 820 may mainly include a storage program area that may store an operating system, application programs required for at least one function (such as a sound playing function, an image playing function, etc.), and a storage data area; the storage data area may store data (such as audio data, phonebook, etc.) created according to the use of the handset, etc. In addition, memory 820 may include high-speed random access memory, and may also include non-volatile memory, such as at least one magnetic disk storage device, flash memory device, or other volatile solid-state storage device.

The input unit 830 may be used to receive input numeric or character information and to generate key signal inputs related to user settings and function controls of the handset. In particular, the input unit 830 may include a touch panel 831 and other input devices 832. The touch panel 831, also referred to as a touch screen, may collect touch operations thereon or thereabout by a user (e.g., operations of the user on the touch panel 831 or thereabout using any suitable object or accessory such as a finger, stylus, etc.), and actuate the corresponding connection device according to a predetermined program. Alternatively, the touch panel 831 may include two portions of a touch detection device and a touch controller. The touch detection device detects the touch azimuth of a user, detects a signal brought by touch operation and transmits the signal to the touch controller; the touch controller receives touch information from the touch detection device and converts it into touch point coordinates, which are then sent to the processor 880 and can receive commands from the processor 880 and execute them. In addition, the touch panel 831 may be implemented in various types of resistive, capacitive, infrared, surface acoustic wave, and the like. The input unit 830 may include other input devices 832 in addition to the touch panel 831. In particular, other input devices 832 may include, but are not limited to, one or more of a physical keyboard, function keys (e.g., volume control keys, switch keys, etc.), a trackball, mouse, joystick, etc.

The display unit 840 may be used to display information input by a user or information provided to the user and various menus of the mobile phone. The display unit 840 may include a display panel 841, and optionally, the display panel 841 may be configured in the form of a liquid crystal display (Liquid Crystal Display, LCD), an Organic Light-Emitting Diode (OLED), or the like. Further, the touch panel 831 may overlay the display panel 841, and when the touch panel 831 detects a touch operation thereon or thereabout, the touch operation is transferred to the processor 880 to determine the type of touch event, and the processor 880 then provides a corresponding visual output on the display panel 841 according to the type of touch event. Although in fig. 8, the touch panel 831 and the display panel 841 are implemented as two separate components to implement the input and input functions of the mobile phone, in some embodiments, the touch panel 831 and the display panel 841 may be integrated to implement the input and output functions of the mobile phone.

The handset may also include at least one sensor 850, such as a light sensor, a motion sensor, and other sensors. Specifically, the light sensor may include an ambient light sensor and a proximity sensor, wherein the ambient light sensor may adjust the brightness of the display panel 841 according to the brightness of the ambient light, and the proximity sensor may turn off the display panel 841 and/or the backlight when the mobile phone moves to the ear. As one of the motion sensors, the accelerometer sensor can detect the acceleration in all directions (generally three axes), can detect the gravity and the direction when the accelerometer sensor is stationary, and can be used for recognizing the application of the gesture of a mobile phone (such as horizontal and vertical screen switching, related games and magnetometer gesture calibration), vibration recognition related functions (such as pedometer and knocking) and the like; other sensors such as gyroscopes, barometers, hygrometers, thermometers, infrared sensors, etc. that may also be configured with the handset are not described in detail herein.

Audio circuitry 860, speaker 861, microphone 862 may provide an audio interface between the user and the handset. The audio circuit 860 may transmit the received electrical signal converted from audio data to the speaker 861, and the electrical signal is converted into a sound signal by the speaker 861 to be output; on the other hand, microphone 862 converts the collected sound signals into electrical signals, which are received by audio circuit 860 and converted into audio data, which are processed by audio data output processor 880 for transmission to, for example, another cell phone via RF circuit 810, or for output to memory 820 for further processing.

WiFi belongs to a short-distance wireless transmission technology, and a mobile phone can help a user to receive emails, browse webpages, access streaming media and the like through a WiFi module 870, so that wireless broadband Internet access is provided for the user. Although fig. 8 shows a WiFi module 870, it is understood that it does not belong to the necessary constitution of the handset, and can be omitted entirely as needed within the scope of not changing the essence of the invention.

The processor 880 is a control center of the mobile phone, connects various parts of the entire mobile phone using various interfaces and lines, and performs various functions of the mobile phone and processes data by running or executing software programs and/or modules stored in the memory 820 and calling data stored in the memory 820, thereby controlling the mobile phone as a whole. In the alternative, processor 880 may include one or more processing units; preferably, the processor 880 may integrate an application processor that primarily handles operating systems, user interfaces, applications, etc., with a modem processor that primarily handles wireless communications. It will be appreciated that the modem processor described above may not be integrated into the processor 880.

The handset further includes a power supply 890 (e.g., a battery) for powering the various components, which may be logically connected to the processor 880 via a power management system, as well as performing functions such as managing charge, discharge, and power consumption via the power management system.

Although not shown, the mobile phone may further include a camera, a bluetooth module, etc., which will not be described herein.

In an embodiment of the present application, the memory 820 included in the mobile phone may store program codes and transmit the program codes to the processor.

The processor 880 included in the mobile phone may perform the method of searching content provided in the above embodiment according to the instructions in the program code.

The embodiment of the application also provides a computer readable storage medium for storing a computer program for executing the method for searching contents provided by the above embodiment.

Embodiments of the present application also provide a computer program product or computer program comprising computer instructions stored in a computer readable storage medium. The computer instructions are read from a computer-readable storage medium by a processor of a computer device, and executed by the processor, cause the computer device to perform the method of content searching provided in various alternative implementations of the above aspects.

Those of ordinary skill in the art will appreciate that: all or part of the steps for implementing the above method embodiments may be implemented by hardware related to program instructions, where the above program may be stored in a computer readable storage medium, and when the program is executed, the program performs steps including the above method embodiments; and the aforementioned storage medium may be at least one of the following media: read-Only Memory (ROM), random access Memory (Random Access Memory, RAM), magnetic or optical disk, and the like.

It should be noted that, in the present specification, each embodiment is described in a progressive manner, and identical and similar parts of each embodiment are referred to each other, and each embodiment mainly describes differences from other embodiments. In particular, for the apparatus and system embodiments, since they are substantially similar to the method embodiments, the description is relatively simple, and references to the details of the method embodiments are only provided. The apparatus and system embodiments described above are merely illustrative, in which elements illustrated as separate elements may or may not be physically separate, and elements shown as elements may or may not be physical elements, may be located in one place, or may be distributed over a plurality of network elements. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment. Those of ordinary skill in the art will understand and implement the present invention without undue burden.

The foregoing is only one specific embodiment of the present application, but the scope of the present application is not limited thereto, and any changes or substitutions easily contemplated by those skilled in the art within the technical scope of the present application should be included in the scope of the present application. Therefore, the protection scope of the present application should be subject to the protection scope of the claims.

Claims

1. A method of content searching, the method comprising:

2. The method according to claim 1, wherein the data type of the delivering feature is discrete, the feature conversion processing is performed on the delivering feature value of the candidate content in the candidate content set, so as to obtain the delivering feature of the candidate content, including:

acquiring a release characteristic value of the candidate content and a release mode of the candidate content;

dividing the release characteristic value based on the partition set corresponding to the release mode to obtain the release characteristic;

the interactive fusion processing is carried out on the release characteristic and the first content coding vector of the candidate content to obtain a second content coding vector of the candidate content, which comprises the following steps:

coding the release characteristics to obtain release characteristic coding vectors of the release characteristics; the dimension of the release feature coding vector is matched with the dimension of the first content coding vector;

and carrying out vector fusion processing on the release characteristic coding vector and the first content coding vector to obtain the second content coding vector, wherein the dimension of the second content coding vector is matched with the dimension of the first content coding vector.

3. The method according to claim 2, wherein the encoding the release feature to obtain a release feature encoding vector of the release feature comprises:

embedding the release characteristics to obtain a plurality of release characteristic embedding vectors of the release characteristics; the dimension of the release feature embedding vector is matched with the dimension of the first content coding vector;

performing feature interaction processing on the plurality of release feature embedded vectors to obtain release feature interaction vectors; the dimension of the release feature interaction vector is matched with the dimension of the first content coding vector;

and carrying out pooling treatment on the release characteristic interaction vector to obtain the release characteristic coding vector.

4. The method of claim 2, wherein the drop mode comprises a show mode, a click mode, or a conversion mode, the conversion mode comprising a shallow conversion mode or a deep conversion mode; the partition granularity of the partition set corresponding to the display mode is smaller than that of the partition set corresponding to the click mode, and the partition granularity of the partition set corresponding to the conversion mode is uneven.

5. The method according to claim 1, wherein the data type of the delivering feature is continuous, the feature conversion processing is performed on the delivering feature value of the candidate content in the candidate content set, so as to obtain the delivering feature of the candidate content, including:

Acquiring a release characteristic value of the candidate content;

if the release characteristic values of the candidate content are single, performing first transformation processing on the release characteristic values to obtain the release characteristics;

if the number of the release characteristic values of the candidate content is multiple, fusion processing and first transformation processing are carried out on the multiple release characteristic values, so that the release characteristic is obtained;

scalar fusion processing is carried out on the release characteristics and the first content coding vector to obtain a fusion scalar;

and performing splicing processing on the fusion scalar and the first content coding vector to obtain the second content coding vector.

6. The method of claim 5, wherein scalar fusion processing of the cast feature and the first content encoding vector to obtain a fused scalar comprises:

performing scalar-based conversion processing on the first content coding vector to obtain a content scalar;

and carrying out fusion processing on the content scalar and the release characteristic to obtain the fusion scalar.

7. The method of claim 6, wherein the obtaining the target object code vector comprises:

performing coding processing on the object characteristics to obtain an initial object coding vector;

performing scalar-based conversion processing on the initial object coding vector to obtain an object scalar;

and performing splicing processing on the object scalar and the initial object coding vector to obtain the target object coding vector.

8. The method of claim 7, wherein the method further comprises:

regularizing the second content coding vector to obtain a regularized second content coding vector;

regularizing the target object coding vector to obtain a regularized target object coding vector;

the searching candidate content in the candidate content set based on the second content coding vector and the target object coding vector of the target object to obtain target content includes:

and searching candidate contents in the candidate content set based on the regularized second content coding vector and the regularized target object coding vector to obtain the target content.

9. The method according to claim 5, wherein when the putting mode of the candidate content is a display mode, a click mode or a shallow conversion mode, the putting feature value of the candidate content is single; and when the release mode is a deep conversion mode, the release characteristic values of the candidate content are multiple.

10. The method according to claim 2 or 5, wherein the obtaining the delivery characteristic value of the candidate content includes:

acquiring an initial release characteristic value of the candidate content;

performing second transformation processing on the initial release characteristic value to obtain the release characteristic value; and the fluctuation among the release characteristic values of the candidate contents in the candidate content set is smaller than the fluctuation among the initial release characteristic values of the candidate contents in the candidate content set.

11. The method of claim 1, wherein the searching for candidate content in the candidate content set based on the second content encoding vector and a target object encoding vector of a target object to obtain target content comprises:

performing preset operation based on the second content coding vector and a target object coding vector of a target object to obtain an operation result corresponding to the candidate content;

And searching candidate contents in the candidate content set based on the operation result and a preset search condition to obtain the target content.

12. An apparatus for content searching, the apparatus comprising: the device comprises a feature conversion unit, a fusion unit and a search unit;

the searching unit is used for searching candidate contents in the candidate content set based on the second content coding vector and a target object coding vector of a target object to obtain target contents; the target object coding vector is obtained by coding the object characteristics of the target object.

13. A computer device, the device comprising a processor and a memory:

the processor is configured to perform the method of content searching of any of claims 1-11 according to instructions in the program code.

14. A computer readable storage medium for storing a computer program which, when executed by a processor, performs the method of content searching according to any of claims 1-11.

15. A computer program product comprising a computer program or instructions; the method of content searching of any of claims 1-11, when executed by a processor.