Article copywriting generation network training method, article copywriting generation method, and device

Abstract

The embodiments of the present disclosure disclose a method for training an item copy generation network, an item copy generation method, an apparatus, an electronic device, and a computer-readable medium. A specific implementation of the method includes: obtaining item description information for each item in an item set; performing data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set; training an initial first item copy generation network to obtain a trained first item copy generation network; and using a knowledge distillation method to train the initial second item copy generation network to obtain a trained second item copy generation network. This implementation guides the training of the initial second item copy generation network to generate item copies by using the trained first item copy generation network, so that the trained second item copy generation network can accurately and effectively generate item copies based on the item's title information and item attribute information.

Description

Technical Field

[0001] Embodiments of the present disclosure relate to the field of computer technology, and in particular to a network training method for generating article text, an article text generation method, an apparatus, an electronic device, and a computer-readable medium. Background Art

[0002] Currently, traditional item search and recommendation technologies are no longer able to adequately meet users' growing needs. When browsing recommendation systems, users often face an information overload. They hope to find the products they need quickly and save costs through well-designed item descriptions. A common approach is to manually analyze item content, such as titles and attributes, as well as item reviews, to generate item descriptions.

[0003] However, when using the above method to generate item copy, the following technical problems often occur: most items have little review information or the existing review information is of low value, and it is impossible to effectively generate high-quality item copy based on the review information. Summary of the Invention

[0004] The content of this disclosure is used to briefly introduce concepts that will be described in detail in the detailed description section below. The content of this disclosure is not intended to identify key features or essential features of the claimed technical solution, nor is it intended to limit the scope of the claimed technical solution.

[0005] Some embodiments of the present disclosure propose an article copy generation network training method, an article copy generation network training device, an apparatus, a device and a computer-readable medium to solve one or more of the technical problems mentioned in the above background technology section.

[0006] In a first aspect, some embodiments of the present disclosure provide a method for training an item copy generation network, the method comprising: obtaining item description information of each item in an item set, wherein the item description information comprises: item title information, item attribute information, and at least one comment information of the item; performing data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set; using each item description information in the processed item description information set and the pre-written item copy corresponding to each item description information as training samples to train an initial first item copy generation network to obtain a trained first item copy generation network; using the title information, attribute information, and item copy corresponding to each item description information in the processed item description information set as training samples for an initial second item copy generation network, and using a knowledge distillation method to train the initial second item copy generation network based on the trained first item copy generation network to obtain a trained second item copy generation network.

[0007] Optionally, the above-mentioned data preprocessing is performed on the item description information set corresponding to the above-mentioned item set to obtain a processed item description information set, including: determining the number of comment information for each item description information in the above-mentioned item description information set; removing the item description information whose number of comment information is less than a predetermined threshold from the above-mentioned item description information set to obtain the removed item description information set; removing the comment information whose comment content meets a predetermined condition from each item description information in the above-mentioned removed item description information set to generate processed item description information, to obtain the above-mentioned processed item description information set.

[0008] Optionally, the above-mentioned first item copy generation network after training is trained and the knowledge distillation method is used to train the above-mentioned initial second item copy generation network to obtain the trained second item copy generation network, including: using the KL distance set between the conditional probability corresponding to the first target vector outputted from the above-mentioned trained first item copy generation network and the conditional probability set corresponding to the second target vector set outputted from the above-mentioned second item copy generation network as training constraints, training the above-mentioned initial second item copy generation network to obtain the above-mentioned trained second item copy generation network.

[0009] Optionally, the loss function of the initial second item copy generation network is generated based on the KL distance formula, a loss function that characterizes the correlation between the item copy generated by the first item copy generation network and the item's attribute information, and a loss function that characterizes the correlation between the item copy generated by the second item copy generation network and the item's attribute information.

[0010] In a second aspect, some embodiments of the present disclosure provide a method for generating an item copy, the method comprising: obtaining title information and attribute information of a target item; inputting the above-mentioned title information and the above-mentioned attribute information into a trained second item copy generation network to obtain an item copy corresponding to the above-mentioned target item, wherein the above-mentioned trained second item copy generation network is trained based on the trained first item copy generation network by using a knowledge distillation method to train the initial second item copy generation network.

[0011] Optionally, the above-mentioned inputting the above-mentioned title information and the above-mentioned attribute information into the trained second item copy generation network to obtain the item copy corresponding to the above-mentioned target item includes: performing word vector conversion on the above-mentioned title information and the above-mentioned attribute information to obtain a first vector corresponding to the above-mentioned title information and a second vector corresponding to the above-mentioned attribute information; encoding the above-mentioned first vector and the above-mentioned second vector to obtain a third vector set corresponding to the above-mentioned title information and a fourth vector set corresponding to the above-mentioned attribute information; for each third vector in the third vector set and the fourth vector corresponding to the above-mentioned third vector, linearly combining the above-mentioned third vector and the above-mentioned fourth vector to obtain a fifth vector; decoding the obtained fifth vector set to obtain the item copy corresponding to the above-mentioned target item.

[0012] Optionally, the encoding of the first vector and the second vector to obtain the third vector set corresponding to the title information and the fourth vector set corresponding to the attribute information includes: inputting the first vector and the third vector into a pre-trained encoding network respectively to obtain the third vector set and the fourth vector set, wherein the encoding network includes at least one encoding layer.

[0013] Optionally, for each third vector in the third vector set and the fourth vector corresponding to the third vector, a linear combination of the third vector and the fourth vector is performed to obtain a fifth vector, including: multiplying the third vector by a value η to obtain a first multiplication result, wherein the value η is a value between 0 and 1; multiplying the fourth vector by a value 1-η to obtain a second multiplication result; and adding the first multiplication result to the second multiplication result to obtain the fifth vector.

[0014] Optionally, decoding the fifth vector to obtain the item text corresponding to the target item includes: inputting the fifth vector into a pre-trained decoding network with a copy mechanism to obtain the item text corresponding to the target item.

[0015] In a third aspect, some embodiments of the present disclosure provide an item copy generation network training device, the device comprising: an acquisition unit, configured to acquire item description information of each item in an item set, wherein the item description information comprises: item title information, item attribute information and at least one comment information of the item; a preprocessing unit, configured to perform data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set; a first training unit, configured to use each item description information in the processed item description information set and the pre-written item copy corresponding to each item description information as training samples to train an initial first item copy generation network to obtain a trained first item copy generation network; a second training unit, configured to use the title information, attribute information and the item copy corresponding to each item description information in the processed item description information set as training samples for an initial second item copy generation network, and train the initial second item copy generation network based on the trained first item copy generation network using a knowledge distillation method to obtain a trained second item copy generation network.

[0016] Optionally, the preprocessing unit is further configured to: determine the number of comment information for each item description information in the above-mentioned item description information set; remove the item description information whose number of comment information is less than a predetermined threshold from the above-mentioned item description information set to obtain the removed item description information set; remove the comment information whose comment content meets the predetermined conditions from each item description information in the above-mentioned removed item description information set to generate processed item description information, and obtain the above-mentioned processed item description information set.

[0017] Optionally, the second training unit is further configured to: use the KL distance set between the conditional probability corresponding to the first target vector outputted by the trained first item copy generation network and the conditional probability set corresponding to the second target vector set outputted by the second item copy generation network as training constraints, train the initial second item copy generation network, and obtain the trained second item copy generation network.

[0018] Optionally, the loss function of the initial second item copy generation network is generated based on the KL distance formula, a loss function that characterizes the correlation between the item copy generated by the first item copy generation network and the item's attribute information, and a loss function that characterizes the correlation between the item copy generated by the second item copy generation network and the item's attribute information.

[0019] In a fourth aspect, some embodiments of the present disclosure provide an article copy generation device, comprising: an acquisition unit configured to acquire title information and attribute information of a target article; an input unit configured to input the above-mentioned title information and the above-mentioned attribute information into a trained second article copy generation network to obtain an article copy corresponding to the above-mentioned target article, wherein the above-mentioned trained second article copy generation network is trained based on the trained first article copy generation network by using a knowledge distillation method to train an initial second article copy generation network.

[0020] Optionally, the input unit is further configured to: perform word vector conversion on the above-mentioned title information and the above-mentioned attribute information to obtain a first vector corresponding to the above-mentioned title information and a second vector corresponding to the above-mentioned attribute information; encode the above-mentioned first vector and the above-mentioned second vector to obtain a third vector set corresponding to the above-mentioned title information and a fourth vector set corresponding to the above-mentioned attribute information; for each third vector in the third vector set and the fourth vector corresponding to the above-mentioned third vector, linearly combine the above-mentioned third vector and the above-mentioned fourth vector to obtain a fifth vector; decode the obtained fifth vector set to obtain the item text corresponding to the above-mentioned target item.

[0021] Optionally, the input unit is further configured to: input the above-mentioned first vector and the above-mentioned third vector into a pre-trained coding network respectively to obtain the above-mentioned third vector set and the above-mentioned fourth vector set, wherein the above-mentioned coding network includes at least one coding layer.

[0022] Optionally, the input unit is further configured to: multiply the above-mentioned third vector by the numerical value η to obtain a first multiplication result, wherein the above-mentioned numerical value η is a value between 0-1; multiply the above-mentioned fourth vector by the numerical value 1-η to obtain a second multiplication result; add the above-mentioned first multiplication result and the above-mentioned second multiplication result to obtain the above-mentioned fifth vector.

[0023] Optionally, the input unit is further configured to: input the fifth vector into a pre-trained decoding network with a copy mechanism to obtain the item text corresponding to the target item.

[0024] In a fifth aspect, some embodiments of the present disclosure provide an electronic device comprising: one or more processors; a storage device on which one or more programs are stored, and when the one or more programs are executed by one or more processors, the one or more processors implement any method as in the first aspect or the second aspect.

[0025] In a sixth aspect, some embodiments of the present disclosure provide a computer-readable medium having a computer program stored thereon, wherein when the program is executed by a processor, the method of any one of the first aspect or the second aspect is implemented.

[0026] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: Through the item copy generation network training method of some embodiments of the present disclosure, the trained first item copy generation network can guide the training of the initial second item copy generation network to generate item copy, so that the trained second item copy generation network can accurately and effectively generate item copy based on the item title information and item attribute information. Specifically, most items have few review information or the existing review information is of low value, and it is not possible to effectively generate high-quality item copy based on the review information. Based on this, the item copy generation network training method of some embodiments of the present disclosure first obtains item description information for each item in the item set, wherein the item description information includes: the item title information, the item attribute information, and at least one item review information. Then, data preprocessing is performed on the item description information set corresponding to the above item set to obtain a processed item description information set. Here, the data preprocessing of the above item description information set is used to remove meaningless review information to avoid affecting the training accuracy of the second item copy generation network. Furthermore, each item description in the processed item description information set and the pre-written item copy corresponding to each item description are used as training samples to train the initial first item copy generation network, thereby obtaining a trained first item copy generation network. Here, the trained first item copy generation network can generate high-quality item copies based on the input item description information. Finally, the title information, attribute information, and item copy corresponding to each item description in the processed item description information set are used as training samples for the initial second item copy generation network. Based on the trained first item copy generation network, the initial second item copy generation network is trained using a knowledge distillation method, thereby obtaining a trained second item copy generation network. Using the trained first item copy generation network to guide the training of the initial second item copy generation network allows the second item copy generation network to learn certain characteristic information used by the trained first item copy generation network to generate high-quality item copies, without relying on at least one review of the input item. This effectively addresses the issue of most items having few or low-value reviews, making it difficult to effectively generate high-quality item descriptions based on these reviews. Consequently, the aforementioned item description generation network training method enables the trained second item description generation network to accurately and effectively generate high-quality item descriptions based on item title information and item attribute information. BRIEF DESCRIPTION OF THE DRAWINGS

[0027] The above and other features, advantages, and aspects of the various embodiments of the present disclosure will become more apparent with reference to the following detailed description in conjunction with the accompanying drawings. Throughout the drawings, the same or similar reference numerals represent the same or similar elements. It should be understood that the drawings are schematic and that components and elements are not necessarily drawn to scale.

[0028] Figure 1 is a schematic diagram of an application scenario diagram of the network training method for generating article texts in some embodiments of the present disclosure;

[0029] Figure 2 is a flowchart of some embodiments of the network training method for generating article copywriting according to the present disclosure;

[0030] Figure 3 is a schematic diagram of an application scenario diagram of the article copy generation method in some embodiments of the present disclosure;

[0031] Figure 4 is a flowchart of some embodiments of the article copy generation method according to the present disclosure;

[0032] Figure 5 It is a schematic diagram of the structure of some embodiments of the network training device for generating article copywriting according to the present disclosure;

[0033] Figure 6 is a schematic structural diagram of some embodiments of the article copy generation device according to the present disclosure;

[0034] Figure 7 It is a structural diagram of an electronic device suitable for implementing some embodiments of the present disclosure. DETAILED DESCRIPTION

[0035] Embodiments of the present disclosure will be described in more detail below with reference to the accompanying drawings. Although certain embodiments of the present disclosure are shown in the accompanying drawings, it should be understood that the present disclosure can be implemented in various forms and should not be construed as being limited to the embodiments described herein. On the contrary, these embodiments are provided to provide a more thorough and complete understanding of the present disclosure. It should be understood that the drawings and embodiments of the present disclosure are for illustrative purposes only and are not intended to limit the scope of protection of the present disclosure.

[0036] It should also be noted that, for ease of description, only the parts related to the invention are shown in the drawings. In the absence of conflict, the embodiments and features in the embodiments of the present disclosure may be combined with each other.

[0037] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.

[0038] It should be noted that the modifications of "one" and "plurality" mentioned in the present disclosure are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, they should be understood as "one or more".

[0039] The names of the messages or information exchanged between multiple devices in the embodiments of the present disclosure are only used for illustrative purposes and are not used to limit the scope of these messages or information.

[0040] The present disclosure will be described in detail below with reference to the accompanying drawings and in conjunction with embodiments.

[0041] Figure 1 It is a schematic diagram of an application scenario diagram of the network training method for generating article copywriting in some embodiments of the present disclosure.

[0042] like Figure 1As shown, electronic device 101 can first obtain item description information for each item in item set 102. Item description information includes: item title information, item attribute information, and at least one item review. In this application scenario, item set 102 includes: a first item 1021, a second item 1022, and a third item 1023. First item 1021 corresponds to item description information 103. Second item 1022 corresponds to item description information 104. Third item 1023 corresponds to item description information 105. Item description information 103 includes: title information 1031, attribute information 1033, and at least one review 1032. The at least one review 1032 includes: a first review, a second review, and a third review. Item description information 104 includes: title information 1041, attribute information 1043, and at least one review 1042. The at least one review 1042 includes: a fourth review and a fifth review. The item description information 105 includes title information 1051, attribute information 1053, and at least one comment information 1052. The at least one comment information 1052 includes a sixth comment information, a seventh comment information, and an eighth comment information. Then, data preprocessing is performed on the item description information set corresponding to the item set 102 to obtain a processed item description information set. In this application scenario, the processed item description information set includes item description information 103 and item description information 105. Furthermore, each item description information in the processed item description information set and the pre-written item text corresponding to each item description information are used as training samples to train the initial first item text generation network 108, obtaining a trained first item text generation network 109. In this application scenario, the training sample set of the initial first item text generation network 108 may include training samples consisting of item description information 105 and item text 106, and training samples consisting of item description information 103 and item text 107. Finally, the title information, attribute information, and corresponding item text of each item description in the processed item description information set are used as training samples for the initial second item text generation network 110. Based on the trained first item text generation network 109, the knowledge distillation method is used to train the initial second item text generation network 110, resulting in a trained second item text generation network 111. In this application scenario, the training sample set for the initial second item text generation network 110 includes: training samples consisting of item text 106, attribute information from item description information 105, and title information from item description information 105; and training samples consisting of item text 107, attribute information from item description information 103, and title information from item description information 103.

[0043] It should be noted that the electronic device 101 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0044] It should be understood that Figure 1 The number of electronic devices in the embodiment is merely illustrative. Any number of electronic devices may be provided according to implementation requirements.

[0045] Continue to refer Figure 2 , shows a process 200 of some embodiments of the article copy generation network training method according to the present disclosure. The article copy generation network training method includes the following steps:

[0046] Step 201: Obtain item description information of each item in the item set.

[0047] In some embodiments, the execution subject of the article copy generation network training method (for example Figure 1 The electronic device 101 shown can obtain item description information for each item in the item collection via a wired or wireless connection. The item description information includes: item title information, item attribute information, and at least one item review. The item title information can be a short sentence describing the item's content. The item attribute information can include, but is not limited to, at least one of the following: item function information, item appearance and color information, item material information, and item style information.

[0048] As an example, the aforementioned items may be shoes.

[0049] The title information of the item can be: "Special price, *** official flagship, *** co-branded women's shoes retro canvas shoes, women's shoes, Baogai shoes, white and red".

[0050] The attribute information of an item can be:

[0051] “Function: breathable, wear-resistant;

[0052] Style: leisure;

[0053] Color: white, black, blue,

[0054] Upper material: textile".

[0055] Item review information can be:

[0056] "The shoes are very breathable and the colors are very beautiful,"

[0057] "The shoes are relatively new in style and affordable in price."

[0058] "The shoes are of good quality, but the delivery time is longer."

[0059] It should be noted that the above-mentioned wireless connection methods may include but are not limited to 3G / 4G / 5G connection, WiFi connection, Bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection, and other wireless connection methods currently known or to be developed in the future.

[0060] Step 202 : performing data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set.

[0061] In some embodiments, the execution entity may perform data pre-processing on the item description information set corresponding to the item set to obtain a processed item description information set.

[0062] In some optional implementations of some embodiments, the data preprocessing of the item description information set corresponding to the item set to obtain the processed item description information set may include the following steps:

[0063] The first step is to determine the number of comments for each item description in the item description information set. For example, the execution entity may determine the number of comments for each item description in the item description information set by querying a database storing comment information.

[0064] In the second step, the item description information having the number of review information less than a predetermined threshold is removed from the item description information set to obtain the item description information set after removal. As an example, the predetermined threshold may be a value of "3".

[0065] In the fourth step, the review information whose review content meets a predetermined condition is removed from each item description information in the removed item description information set to generate processed item description information, thereby obtaining the processed item description information set. The review content meeting the predetermined condition may be content that meets a predetermined template or content that is not of much reference value.

[0066] Step 203: Train the initial first item copy generation network to obtain a trained first item copy generation network.

[0067] In some embodiments, the execution entity trains the initial first item description generation network using each item description in the processed item description information set as training data in the training sample, and uses the pre-written item description corresponding to each item description as the annotation of the training data, thereby obtaining a trained first item description generation network. It should be noted that the training process of the initial first item description generation network is a relatively common training step and will not be further described here.

[0068] Step 204: Use a knowledge distillation method to train the initial second item copy generation network to obtain a trained second item copy generation network.

[0069] In some embodiments, the execution entity may use the title information and attribute information of each item description information in the processed item description information set as training data in the training sample and the item text corresponding to each item description information as the annotation of the training data in the training sample. Based on the trained first item text generation network, the initial second item text generation network may be trained using a knowledge distillation method to obtain a trained second item text generation network. The first item text generation network may be the Teacher network in the Teacher-Student network. Correspondingly, the second item text generation network may be the Student network in the Teacher-Student network. The knowledge distillation method may utilize transfer knowledge to obtain a small model that is more suitable for reasoning through a trained large model.

[0070] It should be noted that since the training samples of the trained first item copy generation network include the review information of the item, the trained first item copy generation network can learn to generate high-quality item copy through the title information, attribute information and at least one review information of the item description information. However, in the recommendation system, most items have a small amount of review information, resulting in the item copy generated by the trained first item copy generation network being of insufficient quality. Furthermore, the first item copy generation network is used as the Teacher network and the second item copy generation network is used as the Student network. The training samples of the second item copy generation network include the title information and attribute information of the item description information, but do not include the review information of the item. The trained first item copy generation network guides the second item copy generation network to train, so that the second item copy generation network learns the knowledge of the trained first item copy generation network to generate high-quality item copy.

[0071] In some optional implementations of some embodiments, the KL distance set between the conditional probability corresponding to the first target vector outputted by the trained first item copy generation network and the conditional probability set corresponding to the second target vector set outputted by the second item copy generation network is used as a training constraint, and the initial second item copy generation network is trained to obtain the trained second item copy generation network. The conditional probability corresponding to the first target vector can be: p(y t |H item ). Among them, y t It can be the tth word of the generated item copy. item It can be the first target vector corresponding to the item. The conditional probability corresponding to the second target vector can be p(y t |E′ R ). E′ R It can be the second target vector corresponding to the item. The KL distance can be calculated using the following formula:

[0072]

[0073] Among them, θ can be a parameter, The parameter can be θ to fish for KL distance.

[0074] As an example, for the first article copywriting generation network after training, first, at least one comment information, attribute information and title information of the target object are converted into word vectors to obtain a vector corresponding to the at least one comment information, a vector corresponding to the attribute information and a vector corresponding to the title information. The vector corresponding to the at least one comment information, the vector corresponding to the attribute information and the vector corresponding to the title information are respectively input into a coding network including multiple layers of coding layers to obtain a set of encoded vectors corresponding to the at least one comment information, a set of encoded vectors corresponding to the attribute information and a set of encoded vectors corresponding to the title information. The coding network is a network in which multiple layers of coding layers are connected in series. Each coding layer in the coding network corresponds to a coding output vector. As an example, the coding network can be a coding network of a Transformer model. The coding network of the Transformer model includes multiple layers of coding layers.

[0075] Then, the vector with the highest weight is selected from the encoded vector set corresponding to the at least one comment information as a first target vector. The first target vector is input into a pre-trained forward neural network to obtain a first output vector.

[0076] Next, based on the coding layer corresponding to each vector in the encoded vector set corresponding to the attribute information and each vector in the encoded vector set corresponding to the title information, feature fusion is performed on the encoded vector set corresponding to the attribute information and the encoded vector set corresponding to the title information to obtain a fused vector set. The fused vector set is input to an activation function to obtain a second output vector set. The activation function may be a GELU (Gaussian Error Linerar Units) activation function.

[0077] Finally, the vector with the highest weight is selected from the first and second output vectors as the third output vector. The third output vector is added to the first output vector to obtain a first summed vector. The first summed vector is normalized to obtain a fourth output vector. The fourth output vector is input into a pre-trained feedforward neural network to obtain a fifth output vector. The fifth output vector is added to the fourth output vector to obtain a second summed vector. The second summed vector is normalized to obtain a sixth output vector, which serves as the first target vector outputted by the trained first item copy generation network.

[0078] For the trained first item copy generation network, the second target vector set outputted by the second item copy generation network may correspond to the fifth vector set in the above-mentioned item copy generation method.

[0079] In some optional implementations of some embodiments, the loss function of the initial second item copy generation network is generated based on the KL distance formula, a loss function that characterizes the correlation between the item copy generated by the first item copy generation network and the item's attribute information, and a loss function that characterizes the correlation between the item copy generated by the second item copy generation network and the item's attribute information. As an example, the loss function of the initial second item copy generation network is the following formula:

[0080]

[0081] in, The parameter θ can be the loss function of the initial second item copy generation network. α can be a tuning parameter with a value range of [0, 1]. It can be a loss function with a parameter θ, which characterizes the correlation between the item copy generated by the second item copy generation network and the attribute information of the item. It can be a loss function with a parameter θfishing, which characterizes the correlation between the item copy generated by the above-mentioned first item copy generation network and the attribute information of the item.

[0082] in, It can be the following formula:

[0083]

[0084] Among them, E′ R It can be the second target vector corresponding to the target item. t It can be the tth word of the generated target item’s copy. t |E′ R ) can represent the output of the decoding network. θ can be a parameter.

[0085] It can be the following formula:

[0086]

[0087] Among them, y t It can be the tth word in the generated target item copy. |S| can be the number of words in the generated item copy. ＜t The set of words from the first to the t-1th word of the generated target item copy can be represented. T can be the first vector corresponding to the title information. A can be the second vector corresponding to the attribute information. R fuse (y ＜t ) can be the relevance scores of each of the 1st to t-1th words.

[0088] R fuse (y ＜t ) is generated by the following formula:

[0089] R fuse (y * )=βR Coh (y * )+(1-β)R RG (y * ),

[0090] Where β can be a tuning parameter with a value range of [0, 1]. * It can be y ＜t . R Coh (y * ) can represent the generated y * Overlap with pre-written copy. RG (y * ) can represent y * ROUGE score.

[0091] R Coh (y * ) can be the following formula:

[0092]

[0093] Among them, |y * | can be the number of words. f(·) can be a word frequency function.

[0094] It should be noted that if y * If the words in exist in the pre-written article copy as a label, then If y * If the words in the text exist in the pre-written article copy, then

[0095] The above-mentioned various embodiments of the present disclosure have the following beneficial effects: Through the item copy generation network training method of some embodiments of the present disclosure, the trained first item copy generation network can guide the training of the initial second item copy generation network to generate item copy, so that the trained second item copy generation network can accurately and effectively generate item copy based on the item title information and item attribute information. Specifically, most items have few review information or the existing review information is of low value, and it is not possible to effectively generate high-quality item copy based on the review information. Based on this, the item copy generation network training method of some embodiments of the present disclosure first obtains item description information for each item in the item set, wherein the item description information includes: the item title information, the item attribute information, and at least one item review information. Then, data preprocessing is performed on the item description information set corresponding to the above item set to obtain a processed item description information set. Here, the data preprocessing of the above item description information set is used to remove meaningless review information to avoid affecting the training accuracy of the second item copy generation network. Furthermore, each item description in the processed item description information set and the pre-written item copy corresponding to each item description are used as training samples to train the initial first item copy generation network, thereby obtaining a trained first item copy generation network. Here, the trained first item copy generation network can generate high-quality item copies based on the input item description information. Finally, the title information, attribute information, and item copy corresponding to each item description in the processed item description information set are used as training samples for the initial second item copy generation network. Based on the trained first item copy generation network, the initial second item copy generation network is trained using a knowledge distillation method, thereby obtaining a trained second item copy generation network. Using the trained first item copy generation network to guide the training of the initial second item copy generation network allows the second item copy generation network to learn certain characteristic information used by the trained first item copy generation network to generate high-quality item copies, without relying on at least one review of the input item. This effectively addresses the issue of most items having few or low-value reviews, making it difficult to effectively generate high-quality item descriptions based on these reviews. Consequently, the aforementioned item description generation network training method enables the trained second item description generation network to accurately and effectively generate high-quality item descriptions based on item title information and item attribute information.

[0096] Figure 3 It is a schematic diagram of an application scenario diagram of the article copy generation method in some embodiments of the present disclosure.

[0097] like Figure 3As shown, electronic device 301 can obtain title information 3031 and attribute information 3032 of target item 302. The title information 3031 and attribute information 3032 are then input into a trained second item description generation network 304 to generate an item description 305 corresponding to target item 302. The trained second item description generation network 304 is trained using a knowledge distillation method based on the trained first item description generation network. In this application scenario, target item 302 may be "shoes." The title information 3031 in the item description information 303 may be: "Title Information: Special Price, *** Official Flagship, *** Co-branded Women's Shoes, Retro Canvas Shoes, *** Women's Shoes, Baogai Shoes, Women's White and Red." The attribute information 3032 in the item description information 303 may be: "Attribute Information: Function: Breathable, Wear-Resistant; Style: Casual; Color: White, Red, Black, Blue; Upper Material: Fabric." The text 305 of the above-mentioned item can be: "Shoe text: *** joint collaboration model, combining classic elements with current trends, the logo design embellished on the tongue, personalized and fashionable, with a high street style, taking you to play on the streets easily, soft and comfortable, and enhancing the wearing experience."

[0098] It should be noted that the electronic device 301 can be hardware or software. When the electronic device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or it can be implemented as a single server or a single terminal device. When the electronic device is embodied as software, it can be installed in the hardware devices listed above. It can be implemented as multiple software or software modules for providing distributed services, for example, or it can be implemented as a single software or software module. No specific limitation is made here.

[0099] It should be understood that Figure 3 The number of electronic devices in the embodiment is merely illustrative. Any number of electronic devices may be provided according to implementation requirements.

[0100] Continue to refer Figure 4 , shows a process 400 of some embodiments of the article copy generation method according to the present disclosure. The article copy generation method includes the following steps:

[0101] Step 401: Obtain the title information and attribute information of the target item.

[0102] In some embodiments, the execution subject of the article copy generation method (for example Figure 3 The electronic device 301 shown can obtain the item description information of each item in the item set through a wired connection or a wireless connection.

[0103] Step 402: Input the title information and the attribute information into the trained second item copy generation network to obtain the item copy corresponding to the target item.

[0104] In some embodiments, the execution entity may input the title information and attribute information into a trained second item description generation network to obtain an item description corresponding to the target item. The trained second item description generation network is trained based on the trained first item description generation network using a knowledge distillation method to train the initial second item description generation network.

[0105] In some optional implementations of some embodiments, inputting the title information and the attribute information into the trained second item description generation network to obtain the item description corresponding to the target item may include the following steps:

[0106] The first step is to perform word embedding conversion on the title information and the attribute information to obtain a first vector corresponding to the title information and a second vector corresponding to the attribute information. As an example, the execution entity may first perform word segmentation on the title information and the attribute information to obtain a word set corresponding to the title information and a word set corresponding to the attribute information. Then, word embedding processing is performed on the word set corresponding to the title information and the word set corresponding to the attribute information to obtain the first vector and the second vector.

[0107] In the second step, the first vector and the second vector are encoded to obtain a third vector set corresponding to the title information and a fourth vector set corresponding to the attribute information.

[0108] In the third step, for each third vector in the third vector set and the fourth vector corresponding to the third vector, linearly combine the third vector and the fourth vector to obtain a fifth vector.

[0109] The fourth step is to decode the fifth vector set obtained to obtain the item text corresponding to the target item.

[0110] Optionally, the first vector and the third vector are respectively input into a pre-trained encoding network to obtain the third vector set and the fourth vector set. The encoding network includes at least one encoding layer. It should be noted that the encoding network is a network in which multiple encoding layers are connected in series. Each encoding layer in the encoding network corresponds to an encoding output vector.

[0111] Optionally, for each third vector in the third vector set and the fourth vector corresponding to the third vector, performing a linear combination of the third vector and the fourth vector to obtain the fifth vector may include the following steps:

[0112] In the first step, the third vector is multiplied by a value η to obtain a first multiplication result, where the value η is a value between 0 and 1.

[0113] In the second step, the fourth vector is multiplied by the value 1-η to obtain a second multiplication result.

[0114] In the third step, the first multiplication result and the second multiplication result are added to obtain the fifth vector.

[0115] Optionally, the fifth vector set is input into a pre-trained decoding network with a copy mechanism to obtain the item text corresponding to the target item.

[0116] The aforementioned embodiments of the present disclosure have the following beneficial effects: The item description generation methods of some embodiments of the present disclosure can first obtain the title and attribute information of a target item. Then, these title and attribute information are input into a trained second item description generation network, enabling accurate and efficient generation of high-quality item descriptions corresponding to the target item. The trained second item description generation network is trained using a knowledge distillation method based on the trained first item description generation network.

[0117] Continue to refer Figure 5 As an implementation of the above methods, the present disclosure provides some embodiments of a network training device for generating article texts. These device embodiments are similar to Figure 2 Corresponding to the above method embodiments, the device can be specifically applied to various electronic devices.

[0118] like Figure 5As shown, in some embodiments, the item text generation network training device 500 includes: an acquisition unit 501, a preprocessing unit 502, a first training unit 503, and a second training unit 504. The acquisition unit 501 is configured to acquire item description information for each item in an item set, wherein the item description information includes the item's title information, item attribute information, and at least one item review information. The preprocessing unit 502 is configured to perform data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set. The first training unit 503 is configured to train an initial first item text generation network using each item description in the processed item description information set and the pre-written item text corresponding to each item description as training samples to obtain a trained first item text generation network. The second training unit 504 is configured to use the title information, attribute information and item text corresponding to each item description information in the processed item description information set as training samples for an initial second item text generation network, and train the initial second item text generation network using a knowledge distillation method based on the trained first item text generation network to obtain a trained second item text generation network.

[0119] In some optional implementations of some embodiments, the preprocessing unit 502 of the item copy generation network training device 500 can be further configured to: determine the number of comment information for each item description information in the above-mentioned item description information set; remove the item description information with the number of comment information less than a predetermined threshold from the above-mentioned item description information set to obtain the removed item description information set; remove the comment information whose comment content meets the predetermined conditions from each item description information in the above-mentioned removed item description information set to generate processed item description information, and obtain the above-mentioned processed item description information set.

[0120] In some optional implementations of some embodiments, the second training unit 504 of the item copy generation network training device 500 can be further configured to: use the KL distance set between the conditional probability corresponding to the first target vector outputted internally by the trained first item copy generation network and the conditional probability set corresponding to the second target vector set outputted internally by the second item copy generation network as training constraints, train the initial second item copy generation network, and obtain the trained second item copy generation network.

[0121] It is understood that the units described in the device 500 are similar to those in the reference Figure 2Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 500 and the units included therein, and will not be repeated here.

[0122] Continue to refer Figure 6 As an implementation of the above methods, the present disclosure provides some embodiments of an article copy generation device. These device embodiments are similar to Figure 4 Corresponding to the above method embodiments, the device can be specifically applied to various electronic devices.

[0123] like Figure 6 As shown, in some embodiments, an item description generation apparatus 600 includes an acquisition unit 601 and an input unit 602. The acquisition unit 601 is configured to acquire title information and attribute information of a target item. The input unit 602 inputs the title information and attribute information into a trained second item description generation network to obtain an item description corresponding to the target item. The trained second item description generation network is trained based on the trained first item description generation network using a knowledge distillation method to train an initial second item description generation network.

[0124] In some optional implementations of some embodiments, the input unit 602 of the article copy generation device 600 can be further configured to: perform word vector conversion on the above-mentioned title information and the above-mentioned attribute information to obtain a first vector corresponding to the above-mentioned title information and a second vector corresponding to the above-mentioned attribute information; encode the above-mentioned first vector and the above-mentioned second vector to obtain a third vector set corresponding to the above-mentioned title information and a fourth vector set corresponding to the above-mentioned attribute information; for each third vector in the third vector set and the fourth vector corresponding to the above-mentioned third vector, linearly combine the above-mentioned third vector and the above-mentioned fourth vector to obtain a fifth vector; decode the obtained fifth vector set to obtain the article copy corresponding to the above-mentioned target article.

[0125] In some optional implementations of some embodiments, the input unit 602 of the article copy generation device 600 can be further configured to: input the above-mentioned first vector and the above-mentioned third vector into a pre-trained encoding network respectively to obtain the above-mentioned third vector set and the above-mentioned fourth vector set, wherein the above-mentioned encoding network includes at least one encoding layer.

[0126] In some optional implementations of some embodiments, the input unit 602 of the article copy generation device 600 can be further configured to: multiply the above-mentioned third vector by the numerical value η to obtain a first multiplication result, wherein the above-mentioned numerical value η is a value between 0 and 1; multiply the above-mentioned fourth vector by the numerical value 1-η to obtain a second multiplication result; and add the above-mentioned first multiplication result and the above-mentioned second multiplication result to obtain the above-mentioned fifth vector.

[0127] In some optional implementations of some embodiments, the input unit 602 of the item copy generation device 600 can be further configured to: input the above-mentioned fifth vector set into a pre-trained decoding network with a copy mechanism to obtain the item copy corresponding to the above-mentioned target item.

[0128] It is understood that the units described in the device 600 are similar to those in the reference Figure 4 Therefore, the operations, features and beneficial effects described above for the method are also applicable to the device 600 and the units included therein, and will not be repeated here.

[0129] Reference below Figure 7 , which shows an electronic device (eg, Figure 1 or Figure 3 Schematic diagram of the structure of the electronic device 700. Figure 7 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0130] like Figure 7 As shown, the electronic device 700 may include a processing device (e.g., a central processing unit, a graphics processing unit, etc.) 701, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 702 or a program loaded from a storage device 708 into a random access memory (RAM) 703. Various programs and data required for the operation of the electronic device 700 are also stored in the RAM 703. The processing device 701, the ROM 702, and the RAM 703 are connected to each other via a bus 704. An input / output (I / O) interface 705 is also connected to the bus 704.

[0131] Typically, the following devices may be connected to the I / O interface 705: an input device 706 including, for example, a touch screen, a touchpad, a keyboard, a mouse, a camera, a microphone, an accelerometer, a gyroscope, etc.; an output device 707 including, for example, a liquid crystal display (LCD), a speaker, a vibrator, etc.; a storage device 708 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 709. The communication device 709 may allow the electronic device 700 to communicate with other devices wirelessly or by wire to exchange data. Although Figure 7 The electronic device 700 is shown with various devices, but it should be understood that it is not required to implement or possess all of the devices shown. More or fewer devices may be implemented or possessed instead. Figure 7 Each block shown in the figure may represent one device, or may represent multiple devices as needed.

[0132] In particular, according to some embodiments of the present disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of the present disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program comprising program code for executing the method shown in the flowchart. In some such embodiments, the computer program can be downloaded and installed from a network via the communication device 709, or installed from the storage device 708, or installed from the ROM 702. When the computer program is executed by the processing device 701, the above-mentioned functions defined in the method of some embodiments of the present disclosure are performed.

[0133] It should be noted that in some embodiments of the present disclosure, the computer-readable medium mentioned above may be a computer-readable signal medium or a computer-readable storage medium, or any combination of the two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, 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 disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In some embodiments of the present disclosure, the computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or device. In some embodiments of the present disclosure, the computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to wires, optical cables, RF (radio frequency), etc., or any suitable combination thereof.

[0134] In some embodiments, the client and server can communicate using any currently known or future developed network protocol, such as HTTP (HyperText Transfer Protocol), and can be interconnected with any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include a local area network ("LAN"), a wide area network ("WAN"), an internet (e.g., the Internet), and a peer-to-peer network (e.g., an ad hoc peer-to-peer network), as well as any currently known or future developed network.

[0135] The computer-readable medium may be included in the apparatus, or may exist independently and not incorporated into the electronic device. The computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: obtain item description information for each item in the item set, wherein the item description information includes: item title information, item attribute information, and at least one item review information; perform data preprocessing on the item description information set corresponding to the item set to obtain a processed item description information set; train an initial first item description generation network using each item description in the processed item description information set and the pre-written item copy corresponding to each item description information as training samples to obtain a trained first item copy generation network; and train an initial second item copy generation network using the title information, attribute information, and the item copy corresponding to each item description information in the processed item description information set as training samples for an initial second item copy generation network, and train the initial second item copy generation network using the trained first item copy generation network using a knowledge distillation method to obtain a trained second item copy generation network. Obtain title information and attribute information of the target item; input the above title information and the above attribute information into the trained second item copy generation network to obtain the item copy corresponding to the above target item, wherein the above trained second item copy generation network is based on the trained first item copy generation network and uses the knowledge distillation method to train the initial second item copy generation network.

[0136] Computer program code for performing the operations of some embodiments of the present disclosure may be written in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., through the Internet using an Internet service provider).

[0137] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the module, program segment, or a part of code contains one or more executable instructions for realizing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in a different order than that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram and / or flowchart, and the combination of the boxes in the block diagram and / or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0138] The units described in some embodiments of the present disclosure may be implemented in software or hardware. The units described may also be provided in a processor. For example, they may be described as follows: a processor includes an acquisition unit, a preprocessing unit, a first training unit, and a second training unit. The names of these units do not, in some cases, limit the units themselves. For example, the acquisition unit may also be described as a "unit for acquiring item description information for each item in an item set."

[0139] The functions described above herein may be performed, at least in part, by one or more hardware logic components. For example, and without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chip (SOCs), complex programmable logic devices (CPLDs), and the like.

[0140] The above description is only an illustration of some preferred embodiments of the present disclosure and the technical principles used. Those skilled in the art should understand that the scope of the invention involved in the embodiments of the present disclosure is not limited to the technical solutions formed by the specific combination of the above-mentioned technical features, but should also cover other technical solutions formed by any combination of the above-mentioned technical features or their equivalent features without departing from the above-mentioned inventive concept. For example, the above-mentioned features are replaced with (but not limited to) technical features with similar functions disclosed in the embodiments of the present disclosure.

Claims

1. A method for training a network for generating article text, comprising: Obtaining item description information for each item in the item collection, wherein the item description information includes: item title information, item attribute information, and at least one item review information; Determining the number of comments for each item description in the item description information set; removing item description information having a number of review information less than a predetermined threshold from the item description information set, to obtain a post-removal item description information set; removing the review information whose review content meets a predetermined condition from each item description information in the removed item description information set to generate processed item description information, thereby obtaining the processed item description information set; Using each item description in the processed item description information set and the pre-written item copy corresponding to each item description as training samples, training an initial first item copy generation network to obtain a trained first item copy generation network; The title information, attribute information, and item copy corresponding to each item description information in the processed item description information set are used as training samples for an initial second item copy generation network. Based on the trained first item copy generation network, the initial second item copy generation network is trained using a knowledge distillation method to obtain a trained second item copy generation network.

2. The method according to claim 1, wherein The method of training the initial second item copy generation network using a knowledge distillation method based on the trained first item copy generation network to obtain a trained second item copy generation network includes: The KL distance set between the conditional probability corresponding to the first target vector outputted by the trained first item copy generation network and the conditional probability set corresponding to the second target vector set outputted by the second item copy generation network is used as a training constraint condition to train the initial second item copy generation network to obtain the trained second item copy generation network.

3. The method according to claim 1, wherein The loss function of the initial second item copy generation network is generated based on the KL distance formula, a loss function that characterizes the correlation between the item copy generated by the first item copy generation network and the item's attribute information, and a loss function that characterizes the correlation between the item copy generated by the second item copy generation network and the item's attribute information.

4. A method for generating article copy, comprising: Get the title information and attribute information of the target item; The title information and the attribute information are input into the trained second item copy generation network to obtain the item copy corresponding to the target item, wherein the trained second item copy generation network is trained based on the trained first item copy generation network using the knowledge distillation method to train the initial second item copy generation network, and the trained second item copy generation network and the trained first item copy generation network are obtained according to any one of the methods described in claims 1-3.

5. The method according to claim 4, wherein Inputting the title information and the attribute information into the trained second item copy generation network to obtain the item copy corresponding to the target item includes: Performing word vector conversion on the title information and the attribute information to obtain a first vector corresponding to the title information and a second vector corresponding to the attribute information; Encoding the first vector and the second vector to obtain a third vector set corresponding to the title information and a fourth vector set corresponding to the attribute information; For each third vector in the third vector set and a fourth vector corresponding to the third vector, linearly combine the third vector and the fourth vector to obtain a fifth vector; The obtained fifth vector set is decoded to obtain the item copy corresponding to the target item.

6. The method according to claim 5, wherein: The encoding of the first vector and the second vector to obtain a third vector set corresponding to the title information and a fourth vector set corresponding to the attribute information includes: The first vector and the third vector are respectively input into a pre-trained encoding network to obtain the third vector set and the fourth vector set, wherein the encoding network includes at least one encoding layer.

7. The method according to claim 5, wherein: For each third vector in the third vector set and a fourth vector corresponding to the third vector, performing a linear combination of the third vector and the fourth vector to obtain a fifth vector includes: Multiplying the third vector by a value η to obtain a first multiplication result, wherein the value η is a value between 0 and 1; multiplying the fourth vector by the value 1-n to obtain a second multiplication result; The first multiplication result and the second multiplication result are added to obtain the fifth vector.

8. The method according to claim 5, wherein The decoding of the obtained fifth vector set to obtain the item text corresponding to the target item includes: The fifth vector set is input into a pre-trained decoding network with a copy mechanism to obtain the item text corresponding to the target item.

9. A network training device for generating article text, comprising: an acquisition unit configured to acquire item description information of each item in the item collection, wherein the item description information includes: item title information, item attribute information, and at least one comment information of the item; a pre-processing unit configured to determine the number of comments for each item description information in the item description information set; remove item description information having a number of comments less than a predetermined threshold from the item description information set to obtain a post-removal item description information set; and remove comment information having comment content that meets a predetermined condition from each item description information in the post-removal item description information set to generate processed item description information, thereby obtaining the processed item description information set; a first training unit configured to train an initial first item description generation network using each item description in the processed item description information set and a pre-written item description corresponding to each item description as a training sample, thereby obtaining a trained first item description generation network; The second training unit is configured to use the title information, attribute information and item text corresponding to each item description information in the processed item description information set as training samples for an initial second item text generation network, and train the initial second item text generation network using a knowledge distillation method based on the trained first item text generation network to obtain a trained second item text generation network.

10. An article copy generation device, comprising: an acquiring unit configured to acquire title information and attribute information of a target item; An input unit is configured to input the title information and the attribute information into a trained second item copy generation network to obtain an item copy corresponding to the target item, wherein the trained second item copy generation network is trained based on the trained first item copy generation network using a knowledge distillation method to train an initial second item copy generation network, and the trained second item copy generation network and the trained first item copy generation network are obtained according to any one of the methods described in claims 1-3.

11. An electronic device comprising: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors are enabled to implement the method according to any one of claims 1 to 8.

12. A computer-readable medium having a computer program stored thereon, wherein: When the program is executed by a processor, the method according to any one of claims 1 to 8 is implemented.

Images