CN117473035A

CN117473035A - Advertisement recall method, model training method, device and electronic equipment

Info

Publication number: CN117473035A
Application number: CN202311337463.2A
Authority: CN
Inventors: 黄飞; 汪洋; 叶超; 潘学东; 邢越; 冯朝兵; 崔自鑫; 朱坤鸿; 戴琳; 王超; 杨新涛; 梁智豪; 林坤海; 徐一凯; 丁文达; 郭树波; 张贺明; 米琳; 许子豪; 景春臻
Original assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Current assignee: Beijing Baidu Netcom Science and Technology Co Ltd
Priority date: 2023-10-16
Filing date: 2023-10-16
Publication date: 2024-01-30

Abstract

The disclosure provides an advertisement recall method, a model training device and electronic equipment, relates to the technical field of artificial intelligence, and particularly relates to the technical fields of natural language processing, deep learning and large models. The method comprises the following steps: acquiring a target query text; inputting the target query text into a large model, and generating an associated text of the target advertisement to be recalled based on the target query text through the large model; based on the associated text, a targeted advertisement is determined from a plurality of candidate advertisements in an advertisement library. Therefore, the large model can be utilized to directly generate the associated text of the target advertisement to be recalled based on the target query text so as to determine the target advertisement, the advertisement recall precision is improved, and compared with the method in the related art that the advertisement recall is realized by relying on the inverted index table of the advertisement, the method does not need to rely on the inverted index table of the advertisement, greatly reduces the storage space and the computing resource required by the advertisement recall, and is convenient to be deployed on electronic equipment such as terminals.

Description

Advertisement recall method, model training method, device and electronic equipment

Technical Field

The present disclosure relates to the field of artificial intelligence technology, and more particularly, to the field of natural language processing, deep learning, and large model technology, and more particularly, to an advertisement recall method, a model training method, an apparatus, an electronic device, a storage medium, and a computer program product.

Background

At present, along with the continuous development of artificial intelligence technology, a large model has the advantages of good generalization and the like, and is widely applied to the fields of information extraction, text credibility evaluation, machine translation and the like. However, the advertisement recall method in the related art has problems of low advertisement recall accuracy, required memory space and large computing resources.

Disclosure of Invention

The present disclosure proposes an advertisement recall method, a model training method, an apparatus, an electronic device, a storage medium, and a computer program product.

According to a first aspect of the present disclosure, an advertisement recall method is provided, comprising: acquiring a target query text; inputting the target query text into a large model, and generating an associated text of a target advertisement to be recalled based on the target query text through the large model; the targeted advertisement is determined from a plurality of candidate advertisements in an advertisement library based on the associated text.

According to a second aspect of the present disclosure, a model training method is presented, comprising: acquiring a sample query text; inputting the sample query text into a large model, and generating a prediction associated text of a sample advertisement to be recalled based on the sample query text through the large model; training the large model based on the prediction related text.

According to a third aspect of the present disclosure, there is provided an advertisement recall apparatus including: the acquisition module is used for acquiring the target query text; the generation module is used for inputting the target query text into a large model, and generating an associated text of the target advertisement to be recalled based on the target query text through the large model; and the determining module is used for determining the target advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text.

According to a fourth aspect of the present disclosure, there is provided a model training apparatus comprising: the acquisition module is used for acquiring a sample query text; the generation module is used for inputting the sample query text into a large model, and generating a prediction associated text of the sample advertisement to be recalled based on the sample query text through the large model; and the training module is used for training the large model based on the prediction associated text.

According to a fifth aspect of the present disclosure, there is provided an electronic device, comprising: at least one processor; and a memory communicatively coupled to the at least one processor; wherein the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the advertisement recall method set forth in the first aspect or the model training method set forth in the second aspect.

According to a sixth aspect of the present disclosure, a non-transitory computer-readable storage medium storing computer instructions for causing the computer to perform the advertisement recall method set forth in the first aspect above or the model training method set forth in the second aspect above is provided.

According to a seventh aspect of the present disclosure, a computer program product is presented, comprising a computer program which, when executed by a processor, implements the advertisement recall method presented in the first aspect above, or implements the model training method presented in the second aspect above.

It should be understood that the description in this section is not intended to identify key or critical features of the embodiments of the disclosure, nor is it intended to be used to limit the scope of the disclosure. Other features of the present disclosure will become apparent from the following specification.

Drawings

The drawings are for a better understanding of the present solution and are not to be construed as limiting the present disclosure. Wherein:

FIG. 1 is a flow chart of an advertisement recall method according to an embodiment of the disclosure;

FIG. 2 is a flow chart of an advertisement recall method according to another embodiment of the present disclosure;

FIG. 3 is a flow chart of an advertisement recall method according to another embodiment of the present disclosure;

FIG. 4 is a flow chart of an advertisement recall method according to another embodiment of the present disclosure;

FIG. 5 is a flow chart of a model training method according to an embodiment of the disclosure;

FIG. 6 is a schematic diagram of an advertisement recall apparatus according to an embodiment of the present disclosure;

FIG. 7 is a schematic diagram of a model training apparatus according to an embodiment of the disclosure;

fig. 8 is a schematic block diagram of an electronic device of an embodiment of the present disclosure.

Detailed Description

Exemplary embodiments of the present disclosure are described below in conjunction with the accompanying drawings, which include various details of the embodiments of the present disclosure to facilitate understanding, and should be considered as merely exemplary. Accordingly, one of ordinary skill in the art will recognize that various changes and modifications of the embodiments described herein can be made without departing from the scope and spirit of the present disclosure. Also, descriptions of well-known functions and constructions are omitted in the following description for clarity and conciseness.

AI (Artificial Intelligence ) is a technical science that studies, develops theories, methods, techniques and application systems for simulating, extending and expanding human intelligence. At present, the AI technology has the advantages of high automation degree, high accuracy and low cost, and is widely applied.

NLU (Natural Language Processing ) is an important direction in the field of computer science and artificial intelligence to study a computer system that can effectively implement natural language communication, and in particular, a science of software systems therein.

DL (Deep Learning) is a new research direction in the field of ML (Machine Learning), and is an inherent rule and expression hierarchy of Learning sample data, so that a Machine can analyze Learning ability like a person, can recognize data such as characters, images and sounds, and is widely applied to speech and image recognition.

Large models refer to machine learning models where the model has a large parameter size and complexity, requires a large amount of computing resources and memory space to train and store, and often requires distributed computing and special hardware acceleration techniques. The large model has stronger generalization capability and expression capability.

Fig. 1 is a flow chart illustrating an advertisement recall method according to an embodiment of the disclosure. As shown in fig. 1, the method includes:

s101, acquiring a target query text.

It should be noted that, the execution body of the advertisement recall method according to the embodiment of the present disclosure may be a hardware device having data information processing capability and/or software necessary for driving the hardware device to operate. Alternatively, the execution body may include a workstation, a server, a computer, a user terminal, and other intelligent devices. The user terminal comprises, but is not limited to, a mobile phone, a computer, intelligent voice interaction equipment, intelligent household appliances, vehicle-mounted terminals and the like.

It should be noted that the target query text is not limited too much, and may include text composed of at least one language such as chinese and english.

In the embodiment of the disclosure, the target query text is obtained, including the following several possible implementations:

mode 1, generating a target query text based on advertisement recall requirements.

Thus, the method can automatically generate the target query text in consideration of advertisement recall requirements.

It should be noted that, the advertisement recall requirement is not limited too much, for example, the target service class to which the target advertisement to be recalled belongs, the enterprise to which the target advertisement belongs, the delivery parameter of the target advertisement, and the like may be included. The target business category may include education, games, shopping, etc., and the delivery parameters may include a delivery region, a delivery time period, a delivery number, a delivery location on a page, etc.

In some examples, obtaining the advertisement recall requirement includes receiving an advertisement recall requirement sent by a client. It may be appreciated that taking the implementation subject of the advertisement recall method as an example of a server, the client may generate an advertisement recall requirement based on operation information (such as text input by the user, a clicked icon, and voice interaction information) of the user who manipulates the client, and send the advertisement recall requirement to the server, and correspondingly, the server may receive the advertisement recall requirement sent by the client.

In some examples, generating the target query text based on the advertisement recall requirement may include combining a plurality of advertisement recall requirements to generate the target query text. For example, if the advertisement recall request includes "administrative division x" and "electric toothbrush", an advertisement "requesting recall of the electric toothbrush in the region of administrative division x" may be generated.

And 2, receiving a target query text sent by the client.

Therefore, the target query text sent by the client can be received, namely, the target query text can be set by a user who controls the client, so that the target query text is more personalized.

It may be appreciated that taking the execution subject of the advertisement recall method as an example of a server, the client may generate the target query text based on operation information (such as text input by the user, a clicked icon, voice information of the user, etc.) of the user who manipulates the client, and send the target query text to the server, and accordingly, the server may receive the target query text sent by the client.

In some examples, the client may obtain voice information of a user who manipulates the client, and perform voice recognition on the voice information of the user to obtain the target query text.

And 3, acquiring candidate query texts sent by the client and/or data of a user controlling the client, and acquiring target query texts based on the candidate query texts and/or the data of the user.

Therefore, the method can obtain the target query text based on the candidate query text sent by the client and/or the data of the user controlling the client, and improves the comprehensiveness of the target query text.

It should be noted that, the obtaining of the candidate query text sent by the client may refer to the receiving of the related content of the target query text sent by the client in the above embodiment, which is not described herein.

It should be noted that, the data of the user is not limited too much, and may include, for example, administrative division, interest, age, sex, and the like to which the user belongs.

S102, inputting the target query text into a large model, and generating the associated text of the target advertisement to be recalled based on the target query text through the large model.

It should be noted that the large model may be implemented by any large model in the related art, which is not limited herein. For example, a transducer model may be used, and it should be noted that the transducer model is a neural network model based on a self-attention mechanism.

It should be noted that, the number of the target advertisements is at least one, one target advertisement may correspond to a plurality of associated texts, and one associated text may correspond to a plurality of target advertisements. The associated text is not so limited and may include at least one of targeted advertising content, targeted advertising identification, for example.

It should be noted that the targeted advertising content may include at least one of the advertising content in text form in the related art, which is not limited herein. For example, advertisement titles, keywords, floor page segments, service or merchandise introductions, and the like may be included.

It should be noted that the target advertisement identifier is not limited too much, and may include, for example, an identifier of enterprise granularity, an identifier of advertisement granularity, and the like. The enterprise granularity identification is used for identifying enterprises, and the advertisement granularity identification is used for identifying advertisements. The identification of the granularity of the enterprise corresponds to the enterprise one by one, and the identification of the granularity of the advertisement corresponds to the advertisement one by one. For example, the identification of the business granularity may include a business name, a business number, etc., and the identification of the advertisement granularity may include an advertisement name, an advertisement number, etc.

S103, determining target advertisements from a plurality of candidate advertisements in an advertisement library based on the associated text.

It should be noted that the advertisement library includes a large number of candidate advertisements, and the target advertisement is a partial candidate advertisement.

In one embodiment, determining a target advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text includes obtaining a mapping relationship between the candidate text and the candidate advertisement, and determining the target advertisement based on the associated text and the mapping relationship. For example, candidate advertisements associated with the text map may be determined to be targeted advertisements.

It should be noted that, the relevant content of the candidate text may refer to the relevant content of the associated text in the above embodiment, which is not described herein.

In one embodiment, determining a targeted advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text includes matching the associated text with the candidate advertisement and determining a successfully matched candidate advertisement as the targeted advertisement. It should be noted that the matching policy is not limited too much, for example, the matching method may include determining that the matching between the associated text and the candidate advertisement is successful if the candidate text of the candidate advertisement is consistent with the associated text, and/or if the candidate text of the candidate advertisement includes the associated text, and/or if the similarity between the candidate text of the candidate advertisement and the associated text is greater than a set threshold.

According to the advertisement recall method, a target query text is acquired, the target query text is input into a large model, the large model is used for generating an associated text of a target advertisement to be recalled based on the target query text, and the target advertisement is determined from a plurality of candidate advertisements in an advertisement library based on the associated text. Therefore, the large model can be utilized to directly generate the associated text of the target advertisement to be recalled based on the target query text so as to determine the target advertisement, the advertisement recall precision is improved, and compared with the method in the related art that the advertisement recall is realized by relying on the inverted index table of the advertisement, the method does not need to rely on the inverted index table of the advertisement, greatly reduces the storage space and the computing resource required by the advertisement recall, and is convenient to be deployed on electronic equipment such as terminals.

In the above embodiment, in the case where the associated text includes the targeted advertisement content, regarding the determination of the targeted advertisement from the plurality of candidate advertisements in the advertisement library based on the associated text in step S103, as can be further understood with reference to fig. 2, fig. 2 is a flowchart of an advertisement recall method according to another embodiment of the present disclosure, as shown in fig. 2, the method includes:

s201, acquiring target query text.

S202, inputting the target query text into a large model, and generating the associated text of the target advertisement to be recalled based on the target query text through the large model.

For the relevant content of steps S201-S202, refer to the above embodiment, and are not repeated here.

S203, obtaining the mapping relation between the candidate advertisement content and the candidate advertisement.

It should be noted that, the candidate advertisement content may refer to the related content of the target advertisement content in the above embodiment, which is not described herein.

It should be noted that, the mapping relationship between the candidate advertisement content and the candidate advertisement may be established in advance. For example, text information extraction can be performed on the candidate advertisements to obtain candidate advertisement contents, and a mapping relationship between the candidate advertisement contents and the candidate advertisements is established. It should be noted that the text information extraction may be implemented by any text information extraction method in the related art, which is not limited herein, and may include OCR (Optical Character Recognition ) and the like.

It will be appreciated that one candidate advertisement may map a plurality of candidate advertisement content and one candidate advertisement content may map a plurality of candidate advertisements.

S204, determining the target advertisement from the plurality of candidate advertisements based on the target advertisement content and the mapping relation.

In an embodiment of the present disclosure, determining a targeted advertisement from a plurality of candidate advertisements based on the targeted advertisement content and the mapping relationship includes determining the candidate advertisement mapped to the targeted advertisement content as the targeted advertisement.

For example, the candidate advertisement contents include advertisement contents 1 to 3, the candidate advertisement includes advertisements x, y and z, the advertisement content 1 has a mapping relationship with the advertisements x and y, the advertisement content 2 has a mapping relationship with the advertisement z, and the advertisement content 3 has a mapping relationship with the advertisements x and z.

If the targeted advertising content includes advertising content 1, then the advertisements x, y mapped by advertising content 1 may be determined as targeted advertisements.

If the targeted advertising content includes advertising content 2, then advertisement z mapped by advertising content 2 may be determined to be the targeted advertisement.

If the targeted advertising content includes advertising content 3, then the advertisements x, z mapped by advertising content 3 may be determined as targeted advertisements.

In the advertisement recall method provided by the disclosure, under the condition that the associated text comprises target advertisement content, a mapping relation between candidate advertisement content and candidate advertisements is obtained, and the target advertisement is determined from a plurality of candidate advertisements based on the target advertisement content and the mapping relation. Thus, the target advertisement can be determined from the plurality of candidate advertisements in consideration of the target advertisement content and the mapping relationship between the candidate advertisement content and the candidate advertisement.

In the above embodiment, in the case where the associated text includes the targeted advertisement content, regarding the determination of the targeted advertisement from the plurality of candidate advertisements in the advertisement library based on the associated text in step S103, it can be further understood with reference to fig. 3, and fig. 3 is a flowchart of an advertisement recall method according to another embodiment of the present disclosure, as shown in fig. 3, the method includes:

s301, acquiring a target query text.

S302, inputting the target query text into a large model, and generating the associated text of the target advertisement to be recalled based on the target query text through the large model.

S303, obtaining candidate advertisement contents of the candidate advertisements.

For the relevant content of steps S301-S303, refer to the above embodiments, and are not repeated here.

S304, determining target advertisements from a plurality of candidate advertisements based on the candidate advertisement content and the target advertisement content.

In embodiments of the present disclosure, a targeted advertisement is determined from a plurality of candidate advertisements based on candidate advertisement content and targeted advertisement content, including the following several possible implementations:

in embodiment 1, if the candidate advertisement content matches the target advertisement content, or if the candidate advertisement content includes the target advertisement content, the candidate advertisement corresponding to the candidate advertisement content is determined as the target advertisement.

Thus, when the candidate advertisement content matches the target advertisement content or the candidate advertisement content includes the target advertisement content, the candidate advertisement corresponding to the candidate advertisement content is determined as the target advertisement.

For example, taking candidate advertisement content and target advertisement content as advertisement titles, the candidate advertisements comprise advertisements x, y and z, and advertisement titles of the advertisements x, y and z are respectively 1 to 3. If the targeted advertising content includes advertisement title 2, advertisement y corresponding to advertisement title 2 may be determined as the targeted advertisement.

For example, taking candidate advertisement content and target advertisement content as keywords, the candidate advertisement comprises advertisements x, y and z, the keywords of the advertisement x comprise keywords 1 to 3, the keywords of the advertisement y comprise keywords 1, and the keywords of the advertisement z comprise keywords 2.

If the targeted advertisement content includes keyword 1, then the keyword for advertisement x includes keyword 1, i.e., the candidate advertisement content for advertisement x includes the targeted advertisement content, advertisement x may be determined to be a targeted advertisement.

If the targeted advertising content includes keyword 1, then the candidate advertising content for advertisement y is consistent with the targeted advertising content, and advertisement y may be determined to be a targeted advertisement.

And 2, acquiring the similarity between the candidate advertisement content and the target advertisement content, and if the similarity is larger than a set threshold, determining the candidate advertisement corresponding to the candidate advertisement content as the target advertisement, or ordering the plurality of candidate advertisements in a descending order of the similarity, and determining N candidate advertisements before ordering as the target advertisement, wherein N is a positive integer.

Thus, the similarity between the candidate advertisement content and the target advertisement content can be obtained, and the candidate advertisement corresponding to the candidate advertisement content with the larger similarity is determined as the target advertisement.

It should be noted that, the obtaining of the similarity may be implemented by any text similarity obtaining method in the related art, which is not limited herein. For example, a first text feature of the candidate advertisement content may be obtained, and a second text feature of the target advertisement content may be obtained, and a similarity between the first text feature and the second text feature may be obtained as a similarity between the candidate advertisement content and the target advertisement content.

It should be noted that the set threshold is not limited too much, for example, if the value range of the similarity is 0 to 100%, the set threshold may be 80%.

Note that N is not limited too much, and may be 30% of the number of candidate advertisements, for example.

In the advertisement recall method provided by the disclosure, candidate advertisement contents of candidate advertisements are acquired under the condition that the associated text comprises the target advertisement contents, and the target advertisements are determined from a plurality of candidate advertisements based on the candidate advertisement contents and the target advertisement contents. Thus, the target advertisement can be determined from a plurality of candidate advertisements in consideration of the candidate advertisement content and the target advertisement content.

In the above embodiment, in the case that the associated text includes the target advertisement identifier, regarding the determination of the target advertisement from the plurality of candidate advertisements in the advertisement library based on the associated text in step S103, it may be further understood with reference to fig. 4, and fig. 4 is a flowchart of an advertisement recall method according to another embodiment of the present disclosure, as shown in fig. 4, the method includes:

s401, acquiring target query text.

S402, inputting the target query text into a large model, and generating the associated text of the target advertisement to be recalled based on the target query text through the large model.

For the relevant content of steps S401 to S402, refer to the above embodiment, and are not repeated here.

S403, determining a first advertisement from a plurality of candidate advertisements based on the target advertisement identification.

It should be noted that the first advertisement is at least one, and the first advertisement is a partial candidate advertisement.

In an embodiment of the present disclosure, a first advertisement is determined from a plurality of candidate advertisements based on a targeted advertisement identification, including several possible implementations:

in the mode 1, under the condition that the target advertisement identification is the identification of the enterprise granularity, the target enterprise identified by the target advertisement identification is determined, and the candidate advertisement corresponding to the target enterprise is determined as the first advertisement.

Thus, in the case where the target advertisement identification is an identification of the business granularity, the target business may be determined based on the target advertisement identification, and the candidate advertisement corresponding to the target business may be determined as the first advertisement.

It will be appreciated that there is a correspondence between a target business and candidate advertisements, one target business may correspond to at least one candidate advertisement, and one candidate advertisement may correspond to one target business.

For example, the candidate advertisements include advertisements x, y and z, the enterprise granularity identifier corresponding to the advertisements x and y is identifier 1, the enterprise granularity identifier corresponding to the advertisement z is identifier 2, wherein identifier 1 is the identifier of enterprise 1, and identifier 2 is the identifier of enterprise 2.

If the target advertisement identifier includes identifier 1, it may be determined that enterprise 1 identified by identifier 1 is a target enterprise, and advertisements x and y corresponding to enterprise 1 are determined to be first advertisements.

If the target advertisement identifier includes identifier 2, it may be determined that enterprise 2 identified by identifier 2 is the target enterprise, and advertisement z corresponding to enterprise 2 is determined to be the first advertisement.

Mode 2, in the case where the target advertisement identification is an identification of advertisement granularity, determining a candidate advertisement identified by the target advertisement identification as a first advertisement.

Thus, in the case that the target advertisement is identified as the enterprise-granularity identification, the target enterprise can be determined based on the target advertisement identification, and the candidate advertisement corresponding to the target enterprise can be determined as the first broad

For example, candidate advertisements include advertisements x, y, and z, and the advertisements x, y, and z correspond to advertisements with the respective identifiers 1 to 3.

If the targeted advertisement identification includes identification 1, then advertisement x identified by identification 1 may be determined to be the first advertisement.

If the targeted advertisement identification includes identification 2, then advertisement y identified by identification 2 may be determined to be the first advertisement.

If the targeted advertisement identification includes identification 3, then advertisement z identified by identification 3 may be determined to be the first advertisement.

S404, determining a target advertisement based on the first advertisement.

In an embodiment of the present disclosure, determining a targeted advertisement based on a first advertisement may include the following several possible implementations:

mode 1, a first advertisement is determined to be a targeted advertisement.

Therefore, the first advertisement can be directly determined to be the target advertisement, and the method is suitable for application scenes of which the target advertisement is identified as the enterprise granularity identification and the advertisement granularity identification.

Mode 2, when there are a plurality of first advertisements, determining a target traffic class to which the target advertisement belongs based on the target query text, and screening the target advertisement from the plurality of first advertisements based on the target traffic class.

Therefore, under the condition that a plurality of first advertisements are provided, the target business category to which the target advertisement belongs can be determined based on the target query text, so that the target advertisement is screened from the plurality of first advertisements, and the accuracy of the target advertisement is improved.

It should be noted that, for the relevant content of the target service class, reference may be made to the above embodiment, and details are not repeated here.

In one embodiment, determining a target business category to which the target advertisement belongs based on the target query text includes performing semantic analysis on the target query text to determine the target business category. It should be noted that, the semantic analysis may be implemented by any semantic analysis method in the related art, which is not described herein.

In one embodiment, screening the targeted advertisement from the plurality of first advertisements based on the targeted traffic class includes obtaining a candidate traffic class for the first advertisement, and determining the first advertisement as the targeted advertisement if the candidate traffic class for the first advertisement matches the targeted traffic class or if the candidate traffic class for the first advertisement includes the targeted traffic class.

In one embodiment, screening the target advertisement from the plurality of first advertisements based on the target traffic class includes obtaining candidate traffic classes of the first advertisements, obtaining a similarity between the candidate traffic classes and the target traffic class, sorting the plurality of first advertisements in descending order of the similarity, and determining the first advertisements of the first Q advertisements as the target advertisement. Wherein Q is a positive integer.

In the advertisement recall method provided by the disclosure, under the condition that the associated text comprises the target advertisement identifier, a first advertisement is determined from a plurality of candidate advertisements based on the target advertisement identifier, and the target advertisement is determined based on the first advertisement, so that the determination of the target advertisement is realized.

On the basis of any one of the above embodiments, in the case that the target advertisements are plural, the method further includes finely arranging the plurality of target advertisements, and determining the first M target advertisements as the second advertisements to be placed, where M is a positive integer. Therefore, under the condition that the target advertisements are multiple, the multiple target advertisements can be finely arranged to determine the second advertisements to be put, the coarse arrangement stage is omitted, and the advertisement putting efficiency is improved.

It should be noted that the second advertisement is at least one, and the second advertisement is a partial target advertisement.

It should be noted that, the fine-ranking may be implemented by any fine-ranking method in the information delivery technology, which is not limited herein. For example, the delivery effect data of the target advertisements can be obtained, and the plurality of target advertisements are finely ranked based on the delivery effect data. Note that the delivery effect data is not limited to a large amount, and may include CTR (Click-Through-Rate), CVR (Conversion Rate), and the like.

Fig. 5 is a flow chart of a model training method according to an embodiment of the disclosure. As shown in fig. 5, the method includes:

s501, acquiring a sample query text.

It should be noted that, the execution body of the model training method according to the embodiment of the present disclosure may be a hardware device having data information processing capability and/or software necessary for driving the hardware device to operate. Alternatively, the execution body may include a workstation, a server, a computer, a user terminal, and other intelligent devices. The user terminal comprises, but is not limited to, a mobile phone, a computer, intelligent voice interaction equipment, intelligent household appliances, vehicle-mounted terminals and the like.

It should be noted that, the sample query text may refer to the related content of the target query text, which is not described herein.

In an embodiment of the present disclosure, obtaining sample query text includes the following several possible implementations:

mode 1, generating a sample query text based on sample advertisement recall requirements.

And 2, receiving a sample query text sent by the client.

And 3, acquiring sample candidate query texts sent by the client and/or data of sample users controlling the client, and acquiring the sample query texts based on the sample candidate query texts and/or the data of the sample users.

S502, inputting the sample query text into a large model, and generating a prediction associated text of the sample advertisement to be recalled based on the sample query text through the large model.

For the relevant content of step S502, refer to the above embodiment, and will not be described herein.

S503, training the large model based on the prediction related text.

In embodiments of the present disclosure, training a large model based on predictive relevance text includes several possible implementations:

in the mode 1, a labeling associated text of a sample advertisement is obtained, and training is carried out on a large model based on a prediction associated text and the labeling associated text.

Therefore, the prediction associated text and the annotation associated text can be comprehensively considered, the large model can be trained, and the supervised training of the large model can be realized.

Note that, the associated text is noted and the associated text is predicted, and reference may be made to the related content of the associated text in the above embodiment, which is not described herein.

In one embodiment, predicting the associated text includes at least one of predicting advertisement content and predicting advertisement identification.

In one embodiment, annotating the associated text includes annotating at least one of advertisement content and an annotation advertisement identification.

In one embodiment, when the labeling associated text includes labeling advertisement content, obtaining the labeling associated text of the sample advertisement includes extracting text information of the sample advertisement to obtain sample advertisement content of the sample advertisement, and obtaining the labeling advertisement content based on the sample advertisement content.

In one embodiment, in the case that the annotation associated text includes an annotation advertisement identifier, the annotation associated text of the sample advertisement is obtained, including obtaining the sample advertisement identifier of the sample advertisement as the annotation advertisement identifier.

It should be noted that, based on the prediction related text and the labeling related text, training the large model may be implemented by using any model training method in the related art, which is not described herein again. For example, a loss function of the large model can be obtained based on the predictive relevance text and the annotation relevance text, and the large model can be trained based on the loss function. It should be noted that the loss function is not limited too much, and may include CE (Cross Entropy), MSE (Mean Square Error), KL (Kullback-Leibler) divergence, contrast loss function, and the like, for example.

And 2, acquiring the putting effect data of the sample advertisement, and training the large model based on the prediction associated text and the putting effect data.

Therefore, the prediction associated text and the throwing effect data can be comprehensively considered, the large model is trained, and the method is suitable for an online training scene and an offline training scene of the large model.

In one embodiment, obtaining impression data of the sample advertisement includes determining the sample advertisement from a plurality of candidate advertisements in an advertisement library based on the predictive correlation text, and delivering the sample advertisement through an advertisement system to obtain impression data of the sample advertisement. For example, based on behavior data of the client side for the sample advertisement (such as whether to click on the sample advertisement, browse time length for the sample advertisement, whether to purchase goods corresponding to the sample advertisement, etc.), the throwing effect data of the sample advertisement is obtained.

In one embodiment, training the large model based on the predictive relevance text and the delivery effect data includes obtaining a loss function of the large model based on the predictive relevance text and the delivery effect data, and training the large model based on the loss function.

Mode 3, obtaining preference data of the advertisement system and/or the client for the sample advertisement, and training the large model based on the prediction associated text and the preference data.

Therefore, the prediction associated text and the preference data can be comprehensively considered, the large model can be trained, and the method is suitable for a feedback tuning training scene of the large model.

The advertisement system refers to a system for advertisement recall, fine discharge, delivery and the like. The preference data may include preference weights, preference scores, and the like.

In one embodiment, obtaining preference data includes determining a sample advertisement from a plurality of candidate advertisements in an advertisement library based on a predicted associated text, putting the sample advertisement through an advertisement system, obtaining putting effect data of the sample advertisement, and obtaining preference data of the advertisement system for the sample advertisement based on the putting effect data of the sample advertisement.

In one embodiment, obtaining preference data includes determining a sample advertisement from a plurality of candidate advertisements in an advertisement library based on predictive relevance text, delivering the sample advertisement through an advertisement system, and obtaining preference data of a client for the sample advertisement based on behavior data of the client for the sample advertisement.

In one embodiment, training the large model based on the predictive relevance text and the preference data includes deriving a loss function for the large model based on the predictive relevance text and the preference data, and training the large model based on the loss function.

In one embodiment, training the large model based on the predictive relevance text further includes pre-training the large model based on sample text for a plurality of knowledge domains. Therefore, the large model can be pre-trained based on sample texts in multiple knowledge fields, so that the large language model can learn the sample texts in multiple knowledge fields in the pre-training process, and the large model can realize generation of general text data.

It should be noted that the knowledge field is not limited too much, and may include, for example, medicine, meteorology, literature, advertisement, etc.

According to the model training method, a sample query text is acquired, the sample query text is input into a large model, a prediction associated text of a sample advertisement to be recalled is generated through the large model based on the sample query text, and the large model is trained based on the prediction associated text, so that the large model can learn the relation between the sample query text and the prediction associated text in the training process, and the trained large model can generate the associated text of the target advertisement to be recalled based on the target query text, so that advertisement recall is achieved.

In the technical scheme of the disclosure, the related processes of collecting, storing, using, processing, transmitting, providing, disclosing and the like of the personal information of the user accord with the regulations of related laws and regulations, and the public order colloquial is not violated.

According to an embodiment of the disclosure, the disclosure further provides an advertisement recall device for implementing the advertisement recall method.

Fig. 6 is a block diagram of an advertising recall device according to an embodiment of the present disclosure.

As shown in fig. 6, the advertisement recall apparatus 600 includes: an acquisition module 601, a generation module 602 and a determination module 603.

An obtaining module 601, configured to obtain a target query text;

the generation module 602 is configured to input the target query text into a large model, and generate, based on the target query text, an associated text of a target advertisement to be recalled through the large model;

a determining module 603, configured to determine the target advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text.

In one embodiment of the present disclosure, the associated text includes at least one of targeted advertising content, targeted advertising identification.

In one embodiment of the present disclosure, in the case that the associated text includes the targeted advertising content, the determining module 603 is further configured to: obtaining a mapping relation between candidate advertisement content and the candidate advertisement; the targeted advertisement is determined from a plurality of the candidate advertisements based on the targeted advertisement content and the mapping relationship.

In one embodiment of the present disclosure, in the case that the associated text includes the targeted advertising content, the determining module 603 is further configured to: acquiring candidate advertisement contents of the candidate advertisements; the targeted advertisement is determined from a plurality of the candidate advertisements based on the candidate advertisement content and the targeted advertisement content.

In one embodiment of the present disclosure, the determining module 603 is further configured to: and if the candidate advertisement content is consistent with the target advertisement content, or if the candidate advertisement content contains the target advertisement content, determining the candidate advertisement corresponding to the candidate advertisement content as the target advertisement.

In one embodiment of the present disclosure, the determining module 603 is further configured to: obtaining the similarity between the candidate advertisement content and the target advertisement content; if the similarity is larger than a set threshold, determining the candidate advertisement corresponding to the candidate advertisement content as the target advertisement; or sorting the candidate advertisements according to the similarity descending order, and determining the candidate advertisements of the first N of the sorted candidate advertisements as the target advertisements, wherein N is a positive integer.

In one embodiment of the present disclosure, in the case that the associated text includes the target advertisement identification, the determining module 603 is further configured to: determining a first advertisement from a plurality of the candidate advertisements based on the targeted advertisement identification; the targeted advertisement is determined based on the first advertisement.

In one embodiment of the present disclosure, in the case where the targeted advertisement identification is an identification of enterprise granularity, the determining module 603 is further configured to: determining a target enterprise identified by the target advertisement identification; and determining the candidate advertisement corresponding to the target enterprise as the first advertisement.

In one embodiment of the present disclosure, in the case where the targeted advertisement identification is an identification of advertisement granularity, the determining module 603 is further configured to: the candidate advertisement identified by the targeted advertisement identification is determined to be the first advertisement.

In one embodiment of the present disclosure, the determining module 603 is further configured to: the first advertisement is determined to be the targeted advertisement.

In one embodiment of the present disclosure, in a case where the first advertisement is plural, the determining module 603 is further configured to: determining a target business class to which the target advertisement belongs based on the target query text; and screening the target advertisement from a plurality of first advertisements based on the target service class.

In one embodiment of the present disclosure, in a case where the targeted advertisement is plural, the determining module 603 is further configured to: and finely arranging the target advertisements, and determining M target advertisements before sequencing as second advertisements to be put, wherein M is a positive integer.

In one embodiment of the present disclosure, the obtaining module 601 is further configured to: acquiring candidate query text sent by a client and/or controlling data of a user of the client; and obtaining the target query text based on the candidate query text and/or the data of the user.

The advertisement recall device provided by the disclosure obtains a target query text, inputs the target query text into a large model, generates an associated text of a target advertisement to be recalled based on the target query text through the large model, and determines the target advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text. Therefore, the large model can be utilized to directly generate the associated text of the target advertisement to be recalled based on the target query text so as to determine the target advertisement, the advertisement recall precision is improved, and compared with the method in the related art that the advertisement recall is realized by relying on the inverted index table of the advertisement, the method does not need to rely on the inverted index table of the advertisement, greatly reduces the storage space and the computing resource required by the advertisement recall, and is convenient to be deployed on electronic equipment such as terminals.

According to an embodiment of the present disclosure, the present disclosure further provides a model training apparatus, which is configured to implement the above-mentioned model training method.

FIG. 7 is a block diagram of a model training apparatus according to an embodiment of the present disclosure.

As shown in fig. 7, the model training apparatus 700 includes: an acquisition module 701, a generation module 702 and a training module 703.

An obtaining module 701, configured to obtain a sample query text;

A generating module 702, configured to input the sample query text into a large model, and generate, by using the large model, a prediction related text of a sample advertisement to be recalled based on the sample query text;

and the training module 703 is used for training the large model based on the prediction related text.

In one embodiment of the present disclosure, the training module 703 is further configured to: acquiring a labeling associated text of the sample advertisement; training the large model based on the prediction related text and the annotation related text.

In one embodiment of the present disclosure, the training module 703 is further configured to: obtaining the putting effect data of the sample advertisement; and training the large model based on the prediction associated text and the throwing effect data.

In one embodiment of the present disclosure, the training module 703 is further configured to: obtaining preference data of an advertisement system and/or a client for the sample advertisement; training the large model based on the predictive relevance text and the preference data.

According to the model training device, a sample query text is acquired, the sample query text is input to a large model, a prediction associated text of a sample advertisement to be recalled is generated through the large model based on the sample query text, and the large model is trained based on the prediction associated text, so that the large model can learn the relation between the sample query text and the prediction associated text in the training process, and the trained large model can generate the associated text of the target advertisement to be recalled based on the target query text, so that advertisement recall is achieved.

According to embodiments of the present disclosure, the present disclosure also proposes an electronic device, a readable storage medium and a computer program product.

Fig. 8 illustrates a schematic block diagram of an example electronic device 800 that may be used to implement embodiments of the present disclosure. Electronic devices are intended to represent various forms of digital computers, such as laptops, desktops, workstations, personal digital assistants, servers, blade servers, mainframes, and other appropriate computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular telephones, smartphones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions, are meant to be exemplary only, and are not meant to limit implementations of the disclosure described and/or claimed herein.

As shown in fig. 8, the apparatus 800 includes a computing unit 801 that can perform various appropriate actions and processes according to a computer program stored in a Read Only Memory (ROM) 802 or a computer program loaded from a storage unit 808 into a Random Access Memory (RAM) 803. In the RAM 803, various programs and data required for the operation of the device 800 can also be stored. The computing unit 801, the ROM 802, and the RAM 803 are connected to each other by a bus 804. An input/output (I/O) interface 805 is also connected to the bus 804.

Various components in device 800 are connected to I/O interface 805, including: an input unit 806 such as a keyboard, mouse, etc.; an output unit 806, such as various types of displays, speakers, and the like; a storage unit 808, such as a magnetic disk, optical disk, etc.; and a communication unit 809, such as a network card, modem, wireless communication transceiver, or the like. The communication unit 809 allows the device 800 to exchange information/data with other devices via a computer network such as the internet and/or various telecommunication networks.

The computing unit 801 may be a variety of general and/or special purpose processing components having processing and computing capabilities. Some examples of computing unit 801 include, but are not limited to, a Central Processing Unit (CPU), a Graphics Processing Unit (GPU), various specialized Artificial Intelligence (AI) computing chips, various computing units running machine learning model algorithms, a Digital Signal Processor (DSP), and any suitable processor, controller, microcontroller, etc. The computing unit 801 performs the various methods and processes described above, such as the advertisement recall method, the model training method. For example, in some embodiments, the advertisement recall method, the model training method, may be implemented as a computer software program tangibly embodied on a machine-readable medium, such as the storage unit 808. In some embodiments, part or all of the computer program may be loaded and/or installed onto device 800 via ROM 802 and/or communication unit 809. When the computer program is loaded into RAM 803 and executed by computing unit 801, one or more of the steps of the advertisement recall method, model training method described above may be performed. Alternatively, in other embodiments, the computing unit 801 may be configured to perform the advertisement recall method, the model training method, by any other suitable means (e.g., by means of firmware).

Various implementations of the systems and techniques described here above may be implemented in digital electronic circuitry, integrated circuit systems, field Programmable Gate Arrays (FPGAs), application Specific Integrated Circuits (ASICs), application Specific Standard Products (ASSPs), systems On Chip (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and/or combinations thereof. These various embodiments may include: implemented in one or more computer programs, the one or more computer programs may be executed and/or interpreted on a programmable system including at least one programmable processor, which may be a special purpose or general-purpose programmable processor, that may receive data and instructions from, and transmit data and instructions to, a storage system, at least one input device, and at least one output device.

Program code for carrying out methods of the present disclosure may be written in any combination of one or more programming languages. These program code may be provided to a processor or controller of a general purpose computer, special purpose computer, or other programmable data processing apparatus such that the program code, when executed by the processor or controller, causes the functions/operations specified in the flowchart and/or block diagram to be implemented. The program code may execute entirely on the machine, partly on the machine, as a stand-alone software package, partly on the machine and partly on a remote machine or entirely on the remote machine or server.

In the context of this disclosure, a machine-readable medium may be a tangible medium that can contain, or store a program for use by or in connection with an instruction execution system, apparatus, or device. The machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. The machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any suitable combination of the foregoing. More specific examples of a machine-readable storage medium would include an electrical connection based on one or more wires, a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disc read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

To address interactions with a user account, the systems and techniques described here can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to a user account; and a keyboard and pointing device (e.g., a mouse or trackball) through which a user account may present input to the computer. Other kinds of devices may also be used to propose interactions with a user account; for example, feedback presented to the user account may be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user account may be received in any form, including acoustic input, speech input, or tactile input.

The systems and techniques described here can be implemented in a computing system that includes a background component (e.g., as a data server), or that includes a middleware component (e.g., an application server), or that includes a front-end component (e.g., a user account computer having a graphical user account interface or a web browser through which a user account can interact with an implementation of the systems and techniques described here), or any combination of such background, middleware, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: local Area Networks (LANs), wide Area Networks (WANs), and the internet.

The computer system may include a client and a server. The client and server are typically remote from each other and typically interact through a communication network. The relationship of client and server arises by virtue of computer programs running on the respective computers and having a client-server relationship to each other. The server may be a cloud server, a server of a distributed system, or a server incorporating a blockchain.

According to an embodiment of the present disclosure, there is also provided a computer program product including a computer program, wherein the computer program, when executed by a processor, implements the steps of the advertisement recall method and the model training method described in the above embodiments of the present disclosure.

It should be appreciated that various forms of the flows shown above may be used to reorder, add, or delete steps. For example, the steps recited in the present disclosure may be performed in parallel, sequentially, or in a different order, provided that the desired results of the disclosed aspects are achieved, and are not limited herein.

The above detailed description should not be taken as limiting the scope of the present disclosure. It will be apparent to those skilled in the art that various modifications, combinations, sub-combinations and alternatives are possible, depending on design requirements and other factors. Any modifications, equivalent substitutions and improvements made within the spirit and principles of the present disclosure are intended to be included within the scope of the present disclosure.

Claims

1. An advertising recall method comprising:

acquiring a target query text;

inputting the target query text into a large model, and generating an associated text of a target advertisement to be recalled based on the target query text through the large model;

The targeted advertisement is determined from a plurality of candidate advertisements in an advertisement library based on the associated text.

2. The method of claim 1, wherein the associated text comprises at least one of targeted advertising content, targeted advertising identification.

3. The method of claim 2, wherein, in the case where the associated text includes the targeted advertising content, the determining the targeted advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text comprises:

obtaining a mapping relation between candidate advertisement content and the candidate advertisement;

the targeted advertisement is determined from a plurality of the candidate advertisements based on the targeted advertisement content and the mapping relationship.

4. The method of claim 2, wherein, in the case where the associated text includes the targeted advertising content, the determining the targeted advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text comprises:

acquiring candidate advertisement contents of the candidate advertisements;

the targeted advertisement is determined from a plurality of the candidate advertisements based on the candidate advertisement content and the targeted advertisement content.

5. The method of claim 4, wherein the determining the targeted advertisement from a plurality of the candidate advertisements based on the candidate advertisement content and the targeted advertisement content comprises:

and if the candidate advertisement content is consistent with the target advertisement content, or if the candidate advertisement content contains the target advertisement content, determining the candidate advertisement corresponding to the candidate advertisement content as the target advertisement.

6. The method of claim 4, wherein the determining the targeted advertisement from a plurality of the candidate advertisements based on the candidate advertisement content and the targeted advertisement content comprises:

obtaining the similarity between the candidate advertisement content and the target advertisement content;

if the similarity is larger than a set threshold, determining the candidate advertisement corresponding to the candidate advertisement content as the target advertisement; or,

and sorting the candidate advertisements according to the similarity descending order, and determining the candidate advertisements of the first N advertisements as the target advertisements, wherein N is a positive integer.

7. The method of claim 2, wherein, in the case where the associated text includes the targeted advertisement identification, the determining the targeted advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text comprises:

Determining a first advertisement from a plurality of the candidate advertisements based on the targeted advertisement identification;

the targeted advertisement is determined based on the first advertisement.

8. The method of claim 7, wherein, in the event that the targeted advertisement identification is an enterprise-granularity identification, the determining a first advertisement from a plurality of the candidate advertisements based on the targeted advertisement identification comprises:

determining a target enterprise identified by the target advertisement identification;

and determining the candidate advertisement corresponding to the target enterprise as the first advertisement.

9. The method of claim 7, wherein, in the event that the targeted advertisement identification is an identification of advertisement granularity, the determining a first advertisement from a plurality of the candidate advertisements based on the targeted advertisement identification comprises:

the candidate advertisement identified by the targeted advertisement identification is determined to be the first advertisement.

10. The method of claim 7, wherein the determining the targeted advertisement based on the first advertisement comprises:

the first advertisement is determined to be the targeted advertisement.

11. The method of claim 7, wherein, in the event that the first advertisement is a plurality, the determining the targeted advertisement based on the first advertisement comprises:

Determining a target business class to which the target advertisement belongs based on the target query text;

and screening the target advertisement from a plurality of first advertisements based on the target service class.

12. The method of any of claims 1-11, wherein, in the event that the targeted advertisement is a plurality, the method further comprises:

and finely arranging the target advertisements, and determining M target advertisements before sequencing as second advertisements to be put, wherein M is a positive integer.

13. The method of any one of claims 1-11, wherein the method further comprises:

acquiring candidate query text sent by a client and/or controlling data of a user of the client;

and obtaining the target query text based on the candidate query text and/or the data of the user.

14. A model training method, comprising:

acquiring a sample query text;

inputting the sample query text into a large model, and generating a prediction associated text of a sample advertisement to be recalled based on the sample query text through the large model;

training the large model based on the prediction related text.

15. The method of claim 14, wherein the training the large model based on the predictive correlation text comprises:

acquiring a labeling associated text of the sample advertisement;

training the large model based on the prediction related text and the annotation related text.

16. The method of claim 14, wherein the training the large model based on the predictive correlation text comprises:

obtaining the putting effect data of the sample advertisement;

and training the large model based on the prediction associated text and the throwing effect data.

17. The method of claim 14, wherein the training the large model based on the predictive correlation text comprises:

obtaining preference data of an advertisement system and/or a client for the sample advertisement;

training the large model based on the predictive relevance text and the preference data.

18. An advertising recall device comprising:

the acquisition module is used for acquiring the target query text;

the generation module is used for inputting the target query text into a large model, and generating an associated text of the target advertisement to be recalled based on the target query text through the large model;

And the determining module is used for determining the target advertisement from a plurality of candidate advertisements in an advertisement library based on the associated text.

19. The apparatus of claim 18, wherein the associated text comprises at least one of targeted advertising content, targeted advertising identification.

20. The apparatus of claim 19, wherein, in a case where the associated text includes the targeted advertising content, the determining module is further to:

21. The apparatus of claim 19, wherein, in a case where the associated text includes the targeted advertising content, the determining module is further to:

acquiring candidate advertisement contents of the candidate advertisements;

22. The apparatus of claim 21, wherein the means for determining is further configured to:

23. The apparatus of claim 21, wherein the means for determining is further configured to:

24. The apparatus of claim 19, wherein, in a case where the associated text includes the targeted advertisement identification, the determining module is further to:

the targeted advertisement is determined based on the first advertisement.

25. The apparatus of claim 24, wherein, in the event that the targeted advertisement identification is an enterprise-granularity identification, the means for determining is further to:

26. The apparatus of claim 24, wherein, in the event that the targeted advertisement identification is an identification of advertisement granularity, the means for determining is further to:

27. The apparatus of claim 24, wherein the means for determining is further configured to:

the first advertisement is determined to be the targeted advertisement.

28. The apparatus of claim 24, wherein, in the case where the first advertisement is a plurality, the determining module is further configured to:

29. The apparatus of any of claims 18-28, wherein, in a case where the targeted advertisement is a plurality, the determining module is further to:

30. The apparatus of any of claims 18-28, wherein the acquisition module is further to:

31. A model training apparatus comprising:

the acquisition module is used for acquiring a sample query text;

the generation module is used for inputting the sample query text into a large model, and generating a prediction associated text of the sample advertisement to be recalled based on the sample query text through the large model;

and the training module is used for training the large model based on the prediction associated text.

32. The apparatus of claim 31, wherein the training module is further configured to:

acquiring a labeling associated text of the sample advertisement;

33. The apparatus of claim 31, wherein the training module is further configured to:

obtaining the putting effect data of the sample advertisement;

34. The apparatus of claim 31, wherein the training module is further configured to:

35. An electronic device, comprising:

at least one processor; and

a memory communicatively coupled to the at least one processor; wherein,

the memory stores instructions executable by the at least one processor to enable the at least one processor to perform the method of any one of claims 1-17.

36. A non-transitory computer readable storage medium storing computer instructions for causing the computer to perform the method of any one of claims 1-17.

37. A computer program product comprising a computer program which, when executed by a processor, implements the method according to any of claims 1-17.