Methods, apparatus, equipment, media, and programs for generating item attribute information
By generating item image feature information and text prompt information, the attribute value information associated with the item image feature information is filtered out, which solves the problem of low accuracy of item attribute values caused by single-format prompt information and realizes more accurate generation of item attribute values.
Patent Information
- Application Number
- CN202211112134.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2022-09-13
- Publication Date
- 2026-11-17
- Estimated Expiration
- 2042-09-13
AI Technical Summary
In existing technologies, when generating item attribute values using a single type of prompt information, there are fewer learnable features, resulting in lower accuracy of the generated item attribute values.
By generating item image feature information and text prompt information, the attribute value information that meets the preset conditions and is associated with the item image feature information is selected as the target attribute value information, and finally the actual attribute value of the item is determined.
It improves the accuracy of generating item attribute values. By comprehensively learning image and text features, it enhances feature learning capabilities and ensures that the generated item attribute information is more accurate.
Smart Images

Figure CN117743623B_ABST
Abstract
Description
Technical Field
[0001] The embodiments disclosed herein relate to the field of artificial intelligence technology, specifically to methods, apparatus, devices, media, and program products for generating item attribute information. Background Technology
[0002] When an item has many attributes, it is usually necessary to determine the attribute values. Currently, the common method for determining attribute values is to generate the attribute values for the corresponding item attributes through a simple prompt message.
[0003] However, the inventors discovered that when using the above method to generate item attribute values, the following technical problems often arise:
[0004] Generating attribute values for corresponding items using only a single type of prompt information results in fewer learnable features, leading to lower accuracy in the generated item attribute values.
[0005] The information disclosed in this background section is only intended to enhance the understanding of the background of the inventive concept, and therefore may contain information that does not constitute prior art known to those skilled in the art. Summary of the Invention
[0006] The summary portion of this disclosure is intended to provide a brief overview of the concepts, which will be described in detail in the detailed description portion. This summary portion is not intended to identify key or essential features of the claimed technical solutions, nor is it intended to limit the scope of the claimed technical solutions.
[0007] Some embodiments of this disclosure provide methods, apparatus, electronic devices, computer-readable media, and program products for generating item attribute information to address one or more of the technical problems mentioned in the background section above.
[0008] In a first aspect, some embodiments of this disclosure provide a method for generating item attribute information. The method includes: generating item image feature information corresponding to an image of a target item based on pre-generated image prompt information for a target item; determining an attribute value information set corresponding to a preset set of item attribute values for an attribute name of the target item, wherein the attribute value information is generated based on item text prompt information, and the target item attribute name is the attribute name of the target item; selecting attribute value information from the attribute value information set whose association with the item image feature information satisfies a preset condition, and using this as target attribute value information; and selecting a preset item attribute value from the preset set of item attribute values that corresponds to the target attribute value information, and using this as the actual item attribute value of the target item.
[0009] Optionally, the attribute value information in the aforementioned attribute value information set is an attribute value weight matrix, which is generated through the following steps: determining the target item attribute name and a preset item attribute value set for the target item attribute name; determining item text prompt information, wherein the item text prompt information includes: attribute name prompt information and attribute value prompt information set; determining attribute name feature information corresponding to the target item attribute name based on the target item attribute name and the attribute name prompt information; for each preset item attribute value in the aforementioned preset item attribute value set, generating attribute value feature information corresponding to the preset item attribute value based on the preset item attribute value and each attribute value prompt information in the aforementioned attribute value prompt information set, thereby obtaining an attribute value feature information set corresponding to the preset item attribute value; and for each preset item attribute value, determining the attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value.
[0010] Optionally, determining the attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value includes: performing feature fusion processing on the attribute name feature information and each attribute value feature information in the attribute value feature information set to generate attribute value feature fusion information, thereby obtaining an attribute value feature fusion information set; and performing weighted processing on each attribute value feature fusion information included in the attribute value feature fusion information set according to a preset weighting coefficient to obtain the attribute value weight matrix corresponding to the preset item attribute value.
[0011] Optionally, the above-mentioned determination of the attribute value information set corresponding to the preset set of item attribute values for the target item attribute name includes: obtaining pre-determined attribute name feature information and the attribute value feature information set corresponding to each preset item attribute value; determining the attribute name weight matrix and the attribute value weight matrix based on the above-mentioned item image feature information; for each preset item attribute value corresponding to the attribute value feature information set, performing weighted feature fusion processing on the above-mentioned attribute name feature information and each attribute value feature information in the above-mentioned attribute value feature information set based on the above-mentioned attribute name weight matrix and the above-mentioned attribute value weight matrix to generate attribute value feature fusion information, thereby obtaining the attribute value feature fusion information set corresponding to the above-mentioned preset item attribute value.
[0012] Optionally, the above-mentioned determination of the attribute value information set corresponding to the preset item attribute value set for the target item attribute name further includes: for each preset item attribute value, performing the following attribute value information generation steps: based on the above-mentioned item image feature information, determining the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set to obtain a weight matrix set, wherein the above-mentioned attribute value feature fusion information set corresponds to the above-mentioned preset item attribute value; based on the above-mentioned weight matrix set, weighting each attribute value feature fusion information in the above-mentioned attribute value feature fusion information set to obtain the attribute value information corresponding to the above-mentioned preset item attribute value.
[0013] Optionally, the aforementioned image prompt information is an image prompt matrix; and the aforementioned generation of item image feature information corresponding to the target item image based on the pre-generated image prompt information for the target item includes: extracting features from each image block in the target item image to generate an image block feature matrix, thereby obtaining an image block feature matrix set; determining the image position encoding matrix corresponding to each image block in the target item image, thereby obtaining an image position encoding matrix set; performing matrix fusion on the image block feature matrix set and the image position encoding matrix set to obtain a fused matrix; and generating the item image feature information based on the fused matrix, the embeddable matrix, and the image prompt matrix.
[0014] Optionally, generating the item image feature information based on the fused matrix, the embeddable matrix, and the image cue matrix includes: performing matrix fusion processing on any one of the fused matrix and the embeddable matrix with the image cue matrix to obtain the item image feature information.
[0015] Secondly, some embodiments of this disclosure provide an item attribute information generation apparatus, the apparatus comprising: a generation unit configured to generate item image feature information corresponding to an image of a target item based on pre-generated image prompt information for a target item; a determination unit configured to determine an attribute value information set corresponding to a preset item attribute value set for an attribute name of a target item, wherein the attribute value information is generated based on item text prompt information, and the attribute name of the target item is the attribute name of the target item; a first filtering unit configured to filter attribute value information from the attribute value information set that satisfies a preset condition for the association between the attribute value information and the image feature information of the target item as target attribute value information; and a second filtering unit configured to filter preset item attribute values corresponding to the target attribute value information from the preset item attribute value set as actual item attribute values of the target item.
[0016] Optionally, the determining unit is further configured to: determine the target item attribute name and a preset set of item attribute values for the target item attribute name; determine item text prompt information, wherein the item text prompt information includes: a set of attribute name prompt information and a set of attribute value prompt information; determine attribute name feature information corresponding to the target item attribute name based on the target item attribute name and the attribute name prompt information; for each preset item attribute value in the preset item attribute value set, generate attribute value feature information corresponding to the preset item attribute value based on each attribute value prompt information in the preset item attribute value and the attribute value prompt information set, thereby obtaining a set of attribute value feature information corresponding to the preset item attribute value; and for each preset item attribute value, determine an attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the set of attribute value feature information corresponding to the preset item attribute value.
[0017] Optionally, the determining unit is further configured to: perform feature fusion processing on the attribute name feature information and each attribute value feature information in the attribute value feature information set to generate attribute value feature fusion information and obtain an attribute value feature fusion information set; and perform weighted processing on each attribute value feature fusion information included in the attribute value feature fusion information set according to a preset weighting coefficient to obtain an attribute value weight matrix corresponding to the preset item attribute value.
[0018] Optionally, the determining unit is further configured to: acquire predetermined attribute name feature information and a set of attribute value feature information corresponding to each preset item attribute value; determine an attribute name weight matrix and an attribute value weight matrix based on the item image feature information; for each set of attribute value feature information corresponding to each preset item attribute value, perform weighted feature fusion processing on the attribute name feature information and each attribute value feature information in the set of attribute value feature information based on the attribute name weight matrix and the attribute value weight matrix to generate attribute value feature fusion information, thereby obtaining a set of attribute value feature fusion information corresponding to the preset item attribute value.
[0019] Optionally, the determining unit is further configured to: for each preset item attribute value, perform the following attribute value information generation steps: based on the item image feature information, determine the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set, to obtain a weight matrix set, wherein the attribute value feature fusion information set corresponds to the preset item attribute value; based on the weight matrix set, perform weighted processing on each attribute value feature fusion information in the attribute value feature fusion information set to obtain the attribute value information corresponding to the preset item attribute value.
[0020] Optionally, the generation unit is further configured to: extract features from each image block in the target item image to generate an image block feature matrix, thereby obtaining an image block feature matrix set; determine the image position encoding matrix corresponding to each image block in the target item image, thereby obtaining an image position encoding matrix set; perform matrix fusion on the image block feature matrix set and the image position encoding matrix set to obtain a fused matrix; and generate the item image feature information based on the fused matrix, the embeddable matrix, and the image cue matrix.
[0021] Optionally, the generation unit is further configured to perform matrix fusion processing on either the fused matrix or the embeddable matrix and the image cue matrix to obtain the item image feature information.
[0022] Thirdly, some embodiments of this disclosure provide an electronic device, including: one or more processors; and a storage device having one or more programs stored thereon, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the method described in any implementation of the first aspect above.
[0023] Fourthly, some embodiments of this disclosure provide a computer-readable medium having a computer program stored thereon, wherein the program, when executed by a processor, implements the method described in any of the implementations of the first aspect above.
[0024] Fifthly, some embodiments of this disclosure provide a computer program product, including a computer program that, when executed by a processor, implements the method described in any of the implementations of the first aspect above.
[0025] The above embodiments of this disclosure have the following beneficial effects: the method for generating item attribute information according to some embodiments of this disclosure can improve the accuracy of the generated item attribute values. Specifically, the reason why the accuracy of the generated item attribute values is low due to the method of generating related item attribute values is that the attribute values of the corresponding item attributes are generated through a single form of prompt information, resulting in fewer learnable features. Based on this, the method for generating item attribute information according to some embodiments of this disclosure first generates item image feature information corresponding to the target item image based on pre-generated image prompt information for the target item. Thus, the item image feature information corresponding to the target item image can be determined through the image prompt information. Second, a set of attribute value information corresponding to a preset set of item attribute values for the target item attribute name is determined, wherein the attribute value information is generated based on item text prompt information, and the target item attribute name is the attribute name of the target item. Thus, the text features corresponding to the target item can be determined based on the name of the target item. Then, attribute value information that satisfies the preset conditions of the association relationship between the attribute value information and the item image feature information is selected from the set of attribute value information as target attribute value information. Therefore, by filtering attribute value information that meets the preset conditions for correlation with the feature information of the target item, it is convenient to determine the item attribute value based on the filtered attribute value information. Finally, preset item attribute values corresponding to the target attribute value information are filtered from the preset item attribute value set and used as the actual item attribute values of the target item. Thus, the attribute that best matches the target item can be determined. Because item image feature information corresponding to the target item image is generated based on the pre-generated image prompt information for the target item, the item image feature information corresponding to the target item image can be determined through the image prompt information, which enhances the feature learning ability. Also, because the set of attribute value information corresponding to the preset item attribute value set for the target item attribute name is determined, where the attribute value information is generated based on the item text prompt information, the attribute value information of each item attribute value can be determined through the item text prompt information, which enhances the feature learning ability. Therefore, by providing feature learning for both images and text simultaneously, more forms of feature learning can be provided, enhancing the feature learning ability and thus generating item attribute information more accurately. Therefore, the item attribute information generation method of some embodiments of this disclosure can improve the accuracy of the generated item attribute information. Attached Figure Description
[0026] The above and other features, advantages, and aspects of the embodiments of this disclosure will become more apparent from the accompanying drawings and the following detailed description. Throughout the drawings, the same or similar reference numerals denote the same or similar elements. It should be understood that the drawings are schematic, and elements are not necessarily drawn to scale.
[0027] Figure 1 This is a schematic diagram illustrating an application scenario of some embodiments of the item attribute information generation method disclosed herein;
[0028] Figure 2 This is a flowchart of some embodiments of the method for generating item attribute information according to this disclosure;
[0029] Figure 3 These are flowcharts of some other embodiments of the method for generating item attribute information according to this disclosure;
[0030] Figure 4 These are schematic diagrams illustrating the structure of some embodiments of the item attribute information generation apparatus according to this disclosure;
[0031] Figure 5 This is a schematic diagram of the structure of an electronic device suitable for implementing some embodiments of the present disclosure. Detailed Implementation
[0032] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.
[0033] It should also be noted that, for ease of description, only the parts relevant to the invention are shown in the accompanying drawings. Unless otherwise specified, the embodiments and features described in this disclosure can be combined with each other.
[0034] It should be noted that the concepts of "first" and "second" mentioned in this disclosure are used only to distinguish different devices, modules or units, and are not used to limit the order of functions performed by these devices, modules or units or their interdependencies.
[0035] It should be noted that the terms "a" and "a plurality of" used in this disclosure are illustrative rather than restrictive, and those skilled in the art should understand that, unless otherwise expressly indicated in the context, they should be understood as "one or more".
[0036] The names of messages or information exchanged between multiple devices in the embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of such messages or information.
[0037] This disclosure will now be described in detail with reference to the accompanying drawings and embodiments.
[0038] Figure 1 This is a schematic diagram illustrating an application scenario of the method for generating item attribute information according to some embodiments of this disclosure.
[0039] exist Figure 1 In the application scenario, firstly, the computing device 101 can generate item image feature information 104 corresponding to the target item image 103 based on pre-generated image prompt information 102 for the target item. For example, the target item image 103 can be an image of capri pants. Secondly, an attribute value information set 106 corresponding to a preset item attribute value set 105 for the target item attribute name is determined. For example, the preset item attribute value set 105 can include: "capri pants", "7 / 8 pants", and "9 / 10 pants". The attribute value information is generated based on the item text prompt information, and the target item attribute name is the attribute name of the target item. Then, attribute value information that satisfies the preset conditions for association with the item image feature information 104 is selected from the attribute value information set 106 and used as target attribute value information 107. Finally, preset item attribute values corresponding to the target attribute value information 107 are selected from the preset item attribute value set 105 and used as the actual item attribute value 108 of the target item. For example, the actual item attribute value 108 can be "capri pants".
[0040] It should be noted that the aforementioned computing device 101 can be either hardware or software. When the computing device is hardware, it can be implemented as a distributed cluster consisting of multiple servers or terminal devices, or as a single server or a single terminal device. When the computing device is software, it can be installed within the hardware devices listed above. It can be implemented as, for example, multiple software programs or software modules used to provide distributed services, or as a single software program or software module. No specific limitations are made here.
[0041] It should be understood that Figure 1 The number of computing devices shown is merely illustrative. Any number of computing devices can be used depending on implementation needs.
[0042] Continue to refer to Figure 2 The diagram illustrates a flow 200 of some embodiments of a method for generating item attribute information according to the present disclosure. This method for generating item attribute information includes the following steps:
[0043] Step 201: Generate item image feature information corresponding to the target item image based on the pre-generated image prompt information for the target item.
[0044] In some embodiments, the executor of the item attribute information generation method (e.g., Figure 1The computing device 101 shown can generate item image feature information corresponding to the target item image based on pre-generated image prompts for the target item. The target item can be an item whose attribute value needs to be determined. For example, the target item can be pants whose length type needs to be determined. The image prompts can be prompts obtained by initializing a pre-trained model. For example, the image prompts can be a matrix obtained by initializing a pre-trained model. The pre-trained model can be a CLIP (Contrastive Language-Image Pre-Training) model. The target item image can be an image of an item whose attribute needs to be determined. For example, the target item image can be an item image input by the user. The item image feature information can be information representing the global image features of the target item image. For example, the item image feature information can be a feature matrix. In practice, the executing entity can input the image prompts and the target item image into an image encoder to obtain the item image feature information. For example, the image encoder can be an image encoder in a CLIP model.
[0045] Optionally, the aforementioned image cue information can be an image cue matrix. This image cue matrix can be a matrix generated through cue learning on object images.
[0046] In some optional implementations of certain embodiments, the execution entity may generate item image feature information corresponding to the target item image by performing the following steps based on pre-generated image prompt information for the target item:
[0047] The first step involves extracting features from each image patch in the target item image to generate an image patch feature matrix, resulting in an image patch feature matrix set. In practice, the execution entity can use a pre-trained image classification model to perform convolution extraction on each image patch in the target item image to generate an image patch feature matrix, thus obtaining an image patch feature matrix set. The image classification model can be a model used for image classification. For example, the pre-trained image classification model can be a ViT (Visual Transformer) model.
[0048] The second step is to determine the image location encoding matrix for each image block in the target item image, thus obtaining a set of image location encoding matrices. In practice, the execution entity can use trigonometric functions to map the position of each image block to generate an image location encoding matrix, thereby obtaining a set of image location encoding matrices.
[0049] The third step involves fusing the aforementioned image block feature matrix set and the aforementioned image location encoding matrix set to obtain a fused matrix. As an example, firstly, the executing entity can sum the image block feature matrix and the corresponding image location encoding matrix for each image block to obtain a merged matrix. Then, the executing entity can fuse the merged matrices corresponding to each image block to obtain a fused matrix. As another example, the executing entity can also directly sum the aforementioned image block feature matrix set and the aforementioned image location encoding matrix set to obtain a fused matrix. As yet another example, the executing entity can also directly concatenate the aforementioned image block feature matrix set and the aforementioned image location encoding matrix set to obtain a fused matrix.
[0050] The fourth step involves generating the item image feature information based on the fused matrix, the embeddable matrix, and the image cue matrix. The embeddable matrix can be a matrix used to conveniently obtain global image features. For example, the embeddable matrix can be a class token in a ViT model. As an example, firstly, the executing entity can perform matrix fusion of the fused matrix, the embeddable matrix, and the image cue matrix to obtain a matrix sequence. Then, the executing entity can input this matrix sequence into the Transformer model to obtain the item image feature information.
[0051] Therefore, by adding an embeddable matrix, global image features can be easily obtained, thereby improving the accuracy of the obtained object image feature information.
[0052] In some optional implementations of certain embodiments, the execution entity can perform matrix fusion processing with either the fused matrix or the embeddable matrix and the image cue matrix to obtain the item image feature information. In practice, firstly, the execution entity can perform weighted processing with either the fused matrix or the embeddable matrix and the image cue matrix according to pre-set weighting coefficients to obtain a weighted matrix. For example, the execution entity can average the product of the fused matrix and pre-set coefficients corresponding to the fused matrix with the product of the image cue matrix and pre-set coefficients corresponding to the image cue matrix to obtain the weighted matrix. Then, the execution entity can input the weighted matrix into the Transformer model to obtain the item image feature information.
[0053] Step 202: Determine the set of attribute feature information corresponding to the preset set of item attribute values for the target item attribute name.
[0054] In some embodiments, the execution entity may determine a set of attribute value information corresponding to a preset set of item attribute values for a target item attribute name. The attribute value information may be generated based on item text prompt information. The target item attribute name may be the attribute name of the target item. For example, the target item attribute name may be "trouser length type". The preset item attribute value set may include multiple preset item attribute values. The preset item attribute value may be a pre-set name representing the specific type of the corresponding item attribute name. For example, the preset item attribute value may be "five-quarter pants". The preset item attribute value set may include "five-quarter pants", "seven-quarter pants", and "nine-quarter pants". The attribute value information set may include multiple attribute feature information. Attribute value information corresponds one-to-one with preset item attribute values. The attribute value information may be information representing the text features of the preset item attribute values. For example, the attribute value information may be a feature matrix. The item text prompt information may be prompt information obtained by initializing a pre-trained model. For example, the item text prompt information may be a matrix obtained by initializing a pre-trained model. In practice, the aforementioned executing entity can input the preset item attribute values and the corresponding item text prompts into the text encoder to obtain the attribute value information corresponding to the preset item attribute values. For example, the aforementioned text encoder can be the text encoder in the CLIP model.
[0055] In some optional implementations of certain embodiments, the attribute value information in the aforementioned attribute value information set can be an attribute value weight matrix.
[0056] Optionally, the entity that generates the aforementioned attribute value weight matrix can be... Figure 1 The computing device described can also be other computing devices. Based on this, the attribute value weight matrix can be generated through the following steps:
[0057] The first step is to determine the target item attribute name and the set of preset item attribute values for the target item attribute name.
[0058] The second step is to determine the item text prompts, which include a set of attribute name prompts and a set of attribute value prompts. The attribute name prompts can be learned from the item's attribute names. The attribute value prompts in the attribute value set can be obtained by initializing a pre-trained model. For example, both the attribute name and attribute value prompts can be matrices obtained by initializing a pre-trained model. Different attribute value prompts can be obtained by different initializations of the pre-trained model. It should be noted that different target item attribute names can use the same item text prompts. Different target item attribute names can also use different item text prompts.
[0059] The third step is to determine the attribute name feature information corresponding to the target item attribute name based on the aforementioned target item attribute name and the aforementioned attribute name prompt information. In practice, the aforementioned target item attribute name and the aforementioned attribute name prompt information can be input into a text encoder to obtain the attribute name feature information corresponding to the aforementioned target item attribute name.
[0060] Fourth, for each preset item attribute value in the aforementioned set of preset item attribute values, based on the preset item attribute value and each attribute value prompt in the aforementioned set of attribute value prompts, attribute value feature information corresponding to the preset item attribute value is generated, resulting in an attribute value feature information set corresponding to the preset item attribute value. In practice, each attribute value prompt and the aforementioned preset item attribute value can be sequentially input into a text encoder to generate attribute value feature information corresponding to the aforementioned attribute value prompts, thus obtaining an attribute value feature information set corresponding to the aforementioned preset item attribute value.
[0061] Fifth, for each preset item attribute value, determine the attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value. As an example, firstly, the attribute value feature information in the attribute value feature information set can be weighted according to a pre-set weighting coefficient to obtain weighted attribute value feature information. Secondly, the weighted attribute value feature information can be averaged with the attribute name feature information to obtain the attribute value weight matrix.
[0062] Therefore, by dividing text information into attribute names and attribute values, and then determining the characteristics of the attribute names and attribute values respectively, inefficient processing of long text input sequences can be avoided, and the efficiency of generating item text prompts can be improved.
[0063] In some optional implementations of certain embodiments, the execution entity may determine the attribute value weight matrix corresponding to the preset item attribute value by performing the following steps: based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value.
[0064] The first step is to perform feature fusion processing on the attribute name feature information and each attribute value feature information in the attribute value feature information set to generate attribute value feature fusion information, thus obtaining an attribute value feature fusion information set. As an example, the attribute name feature information and each attribute value feature information in the attribute value feature information set can be averaged to generate attribute value feature fusion information, thus obtaining an attribute value feature fusion information set. As another example, the attribute name feature information and each attribute value feature information in the attribute value feature information set can also be weighted according to a pre-set weighting coefficient to obtain attribute value feature fusion information.
[0065] The second step is to perform weighted processing on the above attribute value feature fusion information set, including the fusion information of each attribute value feature, according to the pre-set weighting coefficients, to obtain the attribute value weight matrix corresponding to the above-preset item attribute value.
[0066] Therefore, by using different attribute value prompts to determine multiple attribute value feature information, an integrated attribute value weight matrix is obtained through weighted processing, thereby improving the accuracy of the obtained attribute value weight matrix.
[0067] Step 203: Select attribute value information from the attribute value information set that meets the preset conditions for the correlation between the attribute value information and the feature information of the item image, and use it as the target attribute value information.
[0068] In some embodiments, the executing entity can filter attribute value information from the attribute value information set that satisfies a preset condition regarding its correlation with the item image feature information, and use this as the target attribute value information. The preset condition may be that the attribute value information has the highest correlation with the item image feature information. In practice, firstly, the executing entity can determine the similarity of each attribute value information to the item image feature information, obtaining a similarity set corresponding to the attribute value information set. Then, the executing entity can determine the attribute value information with the highest similarity as the target attribute value information. As an example, the executing entity can determine the similarity of each attribute value information to the item image feature information through a linear transformation.
[0069] Step 204: Select the preset item attribute values that correspond to the target attribute value information from the preset item attribute value set, and use them as the actual item attribute values of the target item.
[0070] In some embodiments, the execution entity may select preset item attribute values corresponding to the target attribute value information from the preset item attribute value set and use them as the actual item attribute values of the target item.
[0071] The above embodiments of this disclosure have the following beneficial effects: the method for generating item attribute information according to some embodiments of this disclosure can improve the accuracy of the generated item attribute values. Specifically, the reason why the accuracy of the generated item attribute values is low due to the method of generating related item attribute values is that the attribute values of the corresponding item attributes are generated through a single form of prompt information, resulting in fewer learnable features. Based on this, the method for generating item attribute information according to some embodiments of this disclosure first generates item image feature information corresponding to the target item image based on pre-generated image prompt information for the target item. Thus, the item image feature information corresponding to the target item image can be determined through the image prompt information. Second, a set of attribute value information corresponding to a preset set of item attribute values for the target item attribute name is determined, wherein the attribute value information is generated based on item text prompt information, and the target item attribute name is the attribute name of the target item. Thus, the text features corresponding to the target item can be determined based on the name of the target item. Then, attribute value information that satisfies the preset conditions of the association relationship between the attribute value information and the item image feature information is selected from the set of attribute value information as target attribute value information. Therefore, by filtering attribute value information that meets the preset conditions for correlation with the feature information of the target item, it is convenient to determine the item attribute value based on the filtered attribute value information. Finally, preset item attribute values corresponding to the target attribute value information are filtered from the preset item attribute value set and used as the actual item attribute values of the target item. Thus, the attribute that best matches the target item can be determined. Because item image feature information corresponding to the target item image is generated based on the pre-generated image prompt information for the target item, the item image feature information corresponding to the target item image can be determined through the image prompt information, which enhances the feature learning ability. Also, because the set of attribute value information corresponding to the preset item attribute value set for the target item attribute name is determined, where the attribute value information is generated based on the item text prompt information, the attribute value information of each item attribute value can be determined through the item text prompt information, which enhances the feature learning ability. Therefore, by providing feature learning for both images and text simultaneously, more forms of feature learning can be provided, enhancing the feature learning ability and thus generating item attribute information more accurately. Therefore, the item attribute information generation method of some embodiments of this disclosure can improve the accuracy of the generated item attribute information.
[0072] Further reference Figure 3The diagram illustrates a flow 300 of another embodiment of the method for generating item attribute information according to the present disclosure. This method for generating item attribute information includes the following steps:
[0073] Step 301: Generate item image feature information corresponding to the target item image based on the pre-generated image prompt information for the target item.
[0074] In some embodiments, the specific implementation of step 301 above and its resulting technical effects can be referred to Figure 2 Step 201 in the corresponding embodiment will not be repeated here.
[0075] Step 302: Obtain the predetermined attribute name feature information and the set of attribute value feature information corresponding to each preset item attribute value.
[0076] In some embodiments, the execution entity may obtain predetermined attribute name feature information and a set of attribute value feature information corresponding to each preset item attribute value. As an example, the execution entity may include a storage unit. The storage unit stores predetermined attribute name feature information and a set of attribute value feature information corresponding to each preset item attribute value. The execution entity can obtain the predetermined attribute name feature information and the set of attribute value feature information corresponding to each preset item attribute value from the storage unit.
[0077] Optionally, the above-mentioned attribute name feature information and the above-mentioned attribute value feature information set can be generated through the following steps:
[0078] The first step is to determine the target item attribute name and the set of preset item attribute values for the target item attribute name.
[0079] The second step is to determine the item text prompt information, which includes a set of attribute name prompt information and attribute value prompt information.
[0080] The third step is to determine the attribute name feature information corresponding to the above target item attribute name based on the above target item attribute name and the above attribute name prompt information.
[0081] Fourth step: For each preset item attribute value in the above preset item attribute value set, generate attribute value feature information corresponding to the above preset item attribute value based on the above preset item attribute value and each attribute value prompt information in the above attribute value prompt information set, and obtain the attribute value feature information set corresponding to the above preset item attribute value.
[0082] In some embodiments, the specific implementation of steps one through four above and the resulting technical effects can be referred to Figure 2 The corresponding implementation examples will not be described in detail here.
[0083] Step 303: Based on the above-mentioned item image feature information, determine the attribute name weight matrix and the attribute value weight matrix.
[0084] In some embodiments, the executing entity can determine the attribute name weight matrix and the attribute value weight matrix based on the item image feature information. In practice, firstly, the executing entity can multiply the pre-set matrix with the item image feature information to obtain the attribute name matrix and the attribute value matrix. Then, the executing entity can normalize the obtained attribute name matrix and attribute value matrix using a softmax logistic regression transformation to obtain the attribute name weight matrix and the attribute value weight matrix.
[0085] Step 304: For each preset item attribute value corresponding to the attribute value feature information set, according to the attribute name weight matrix and the attribute value weight matrix, the attribute name feature information and each attribute value feature information in the attribute value feature information set are weighted feature fusion to generate attribute value feature fusion information, and thus obtain the attribute value feature fusion information set corresponding to the preset item attribute value.
[0086] In some embodiments, the execution entity may, for each preset item attribute value corresponding to an attribute value feature information set, perform weighted feature fusion processing on the attribute name feature information and each attribute value feature information in the attribute value feature information set according to the attribute name weight matrix and the attribute value weight matrix, to generate attribute value feature fusion information, thereby obtaining an attribute value feature fusion information set corresponding to the preset item attribute value. In practice, for each preset item attribute value corresponding to an attribute value feature information set, the execution entity may average the product of the attribute name feature information and the attribute name weight matrix with the product of each attribute value feature information and the attribute value weight matrix to obtain attribute value feature fusion information.
[0087] Step 305: For each preset item attribute value, perform the following attribute value information generation steps:
[0088] Step 3051: Based on the item image feature information, determine the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set to obtain a weight matrix set, wherein the attribute value feature fusion information set corresponds to the preset item attribute value.
[0089] In some embodiments, the execution entity can determine the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set based on the item image feature information, thereby obtaining a weight matrix set, wherein the attribute value feature fusion information set corresponds to the preset item attribute value. In practice, firstly, the execution entity can multiply the item image feature information with a pre-set matrix to obtain the weight matrix corresponding to each attribute value feature fusion information. The number of elements in the pre-set matrix is the same as the number of elements in the item image feature information. Then, the execution entity can normalize the obtained matrices using a softmax logistic regression transformation to obtain the weight matrix set of the attribute value feature fusion information set corresponding to the preset item attribute value.
[0090] Step 3052: Based on the weight matrix set, the attribute value feature fusion information in the attribute value feature fusion information set is weighted to obtain the attribute value information of the corresponding preset item attribute value.
[0091] In some embodiments, the execution entity can perform weighted processing on each attribute value feature fusion information in the attribute value feature fusion information set according to the weight matrix set, to obtain the attribute value information corresponding to the preset item attribute value. In practice, the execution entity can determine the product of each attribute value feature fusion information in the attribute value feature fusion information set with the corresponding weight matrix, and then the execution entity can average the determined product to obtain the attribute value information corresponding to the preset item attribute value.
[0092] Step 306: Select attribute value information from the attribute value information set that meets the preset conditions for the correlation between the attribute value information and the feature information of the item image, and use it as the target attribute value information.
[0093] Step 307: Select the preset item attribute values that correspond to the target attribute value information from the preset item attribute value set, and use them as the actual item attribute values of the target item.
[0094] In some embodiments, the specific implementation of steps 306-307 above and the resulting technical effects can be referred to Figure 2 Steps 203-204 in the corresponding embodiments will not be repeated here.
[0095] from Figure 3 It can be seen from this that, with Figure 2 Compared to the description of some corresponding embodiments, Figure 3In some corresponding embodiments, the process 300 of the method for generating item attribute information includes determining an attribute name weight matrix and an attribute value weight matrix, as well as a weight matrix for the fusion information of each attribute value feature, based on the item image feature information. Then, attribute value feature fusion information and attribute value information are obtained based on the determined weight matrices. Therefore, determining the weight matrix based on the item image feature information, and subsequently obtaining the attribute value feature fusion information and attribute value information, can improve the matching degree between the selected attribute value information and the aforementioned item image feature information. This further improves the accuracy of the generated item attribute information.
[0096] Further reference Figure 4 As an implementation of the methods shown in the above figures, this disclosure provides some embodiments of an item attribute information generation device, which are similar to... Figure 2 Corresponding to the method embodiments shown, the device can be specifically applied to various electronic devices.
[0097] like Figure 4 As shown, some embodiments of the SMS unsubscribe device 400 include: a generation unit 401, a determination unit 402, a first filtering unit 403, and a second filtering unit 404. The generation unit 401 is configured to generate item image feature information corresponding to the image of the target item based on pre-generated image prompt information for the target item. The determination unit 402 is configured to determine a set of attribute value information corresponding to a preset set of item attribute values for the attribute name of the target item, wherein the attribute value information is generated based on item text prompt information, and the attribute name of the target item is the attribute name of the target item. The first filtering unit 403 is configured to filter attribute value information from the set of attribute value information that satisfies a preset condition regarding the association between the attribute value information and the item image feature information, as target attribute value information. The second filtering unit 404 is configured to filter preset item attribute values corresponding to the target attribute value information from the preset set of item attribute values, as the actual item attribute values of the target item.
[0098] In some optional implementations of embodiments, the determining unit 402 may be further configured to: determine the target item attribute name and a preset set of item attribute values for the target item attribute name; determine item text prompt information, wherein the item text prompt information includes: a set of attribute name prompt information and a set of attribute value prompt information; determine attribute name feature information corresponding to the target item attribute name based on the target item attribute name and the attribute name prompt information; for each preset item attribute value in the preset item attribute value set, generate attribute value feature information corresponding to the preset item attribute value based on each attribute value prompt information in the preset item attribute value and the attribute value prompt information set, thereby obtaining a set of attribute value feature information corresponding to the preset item attribute value; and for each preset item attribute value, determine an attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the set of attribute value feature information corresponding to the preset item attribute value.
[0099] In some optional implementations of the embodiments, the determining unit 402 may be further configured to: perform feature fusion processing on the attribute name feature information and each attribute value feature information in the attribute value feature information set to generate attribute value feature fusion information and obtain an attribute value feature fusion information set; and perform weighted processing on each attribute value feature fusion information included in the attribute value feature fusion information set according to a preset weighting coefficient to obtain an attribute value weight matrix corresponding to the preset item attribute value.
[0100] In some optional implementations of the embodiments, the determining unit 402 may be further configured to: acquire predetermined attribute name feature information and a set of attribute value feature information corresponding to each preset item attribute value; determine an attribute name weight matrix and an attribute value weight matrix based on the item image feature information; for each set of attribute value feature information corresponding to each preset item attribute value, perform weighted feature fusion processing on the attribute name feature information and each attribute value feature information in the set of attribute value feature information based on the attribute name weight matrix and the attribute value weight matrix to generate attribute value feature fusion information, thereby obtaining a set of attribute value feature fusion information corresponding to the preset item attribute value.
[0101] In some optional implementations of embodiments, the determining unit 402 may be further configured to: for each preset item attribute value, perform the following attribute value information generation steps: based on the item image feature information, determine the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set to obtain a weight matrix set, wherein the attribute value feature fusion information set corresponds to the preset item attribute value; based on the weight matrix set, perform weighted processing on each attribute value feature fusion information in the attribute value feature fusion information set to obtain the attribute value information corresponding to the preset item attribute value.
[0102] In some optional implementations of the embodiments, the generation unit 401 may be further configured to: extract features from each image block in the target item image to generate an image block feature matrix, thereby obtaining an image block feature matrix set; determine the image position encoding matrix corresponding to each image block in the target item image, thereby obtaining an image position encoding matrix set; perform matrix fusion on the image block feature matrix set and the image position encoding matrix set to obtain a fused matrix; and generate the item image feature information based on the fused matrix, the embeddable matrix, and the image cue matrix.
[0103] In some optional implementations of the embodiments, the generation unit 401 may be further configured to perform matrix fusion processing on any one of the fused matrix and the embeddable matrix with the image cue matrix to obtain the item image feature information.
[0104] It is understandable that the units described in the device 400 are related to the reference. Figure 2 The steps in the described method correspond accordingly. Therefore, the operations, features, and beneficial effects described above for the method also apply to the device 400 and the units contained therein, and will not be repeated here.
[0105] The following is for reference. Figure 5 It illustrates electronic devices suitable for implementing some embodiments of the present disclosure (such as...). Figure 1 The diagram shows the structure of the computing device 101)500. Figure 5 The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of the embodiments of this disclosure.
[0106] like Figure 5As shown, the electronic device 500 may include a processing unit (e.g., a central processing unit, a graphics processing unit, etc.) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage device 508 into a random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the electronic device 500. The processing unit 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.
[0107] Typically, the following devices can be connected to I / O interface 505: input devices 506 including, for example, touchscreens, touchpads, keyboards, mice, cameras, microphones, accelerometers, gyroscopes, etc.; output devices 507 including, for example, liquid crystal displays (LCDs), speakers, vibrators, etc.; and communication devices 509. Communication device 509 allows electronic device 500 to communicate wirelessly or wiredly with other devices to exchange data. Although Figure 5 An electronic device 500 with various devices is shown; however, 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 alternatively. Figure 5 Each box shown can represent a device or multiple devices as needed.
[0108] In particular, according to some embodiments of this disclosure, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, some embodiments of this disclosure include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the flowcharts. In such embodiments, the computer program can be downloaded and installed from a network via a communication device 509, or installed from a storage device 508, or installed from a ROM 502. When the computer program is executed by the processing device 501, it performs the functions defined in the methods of some embodiments of this disclosure.
[0109] It should be noted that, in some embodiments of this disclosure, the computer-readable medium may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In some embodiments of this disclosure, a 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, apparatus, or device. In some embodiments of this disclosure, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium can be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium can be transmitted using any suitable medium, including but not limited to: wires, optical fibers, RF (radio frequency), etc., or any suitable combination thereof.
[0110] In some implementations, clients and servers can communicate using any currently known or future-developed network protocol such as HTTP (Hypertext Transfer Protocol) and can interconnect with digital data communication (e.g., communication networks) of any form or medium. Examples of communication networks include local area networks (“LANs”), wide area networks (“WANs”), the Internet (e.g., the Internet of Things), and peer-to-peer networks (e.g., ad hoc peer-to-peer networks), as well as any currently known or future-developed networks.
[0111] The aforementioned computer-readable medium may be included in the aforementioned electronic device; or it may exist independently and not assembled into the electronic device. The aforementioned computer-readable medium carries one or more programs that, when executed by the electronic device, cause the electronic device to: generate image feature information corresponding to the image of the target item based on pre-generated image prompt information for the target item; determine a set of attribute value information corresponding to a preset set of item attribute values for the attribute name of the target item, wherein the attribute value information is generated based on the item text prompt information, and the attribute name of the target item is the attribute name of the target item; filter attribute value information from the set of attribute value information whose association with the image feature information satisfies a preset condition, and use it as target attribute value information; and filter preset item attribute values corresponding to the target attribute value information from the preset set of item attribute values, and use them as the actual item attribute values of the target item.
[0112] Computer program code for performing operations of some embodiments of this disclosure can be written in one or more programming languages or a combination thereof, including object-oriented programming languages such as Java, Smalltalk, and C++, and conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, as a standalone 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 remote computers, the remote computer can be connected to the user's computer via any type of network—including a local area network (LAN) or a wide area network (WAN)—or can be connected to an external computer (e.g., via the Internet using an Internet service provider).
[0113] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.
[0114] The units described in some embodiments of this disclosure can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor may be described as including a generation unit, a determination unit, a first filtering unit, and a second filtering unit. The names of these units do not necessarily limit the unit itself; for example, a determination unit may also be described as "a unit that determines a set of attribute value information corresponding to a preset set of item attribute values for a target item attribute name."
[0115] The functions described above in this document can be performed, at least in part, by one or more hardware logic components. For example, exemplary types of hardware logic components that can be used, without limitation, include: Field Programmable Gate Arrays (FPGAs), Application-Specific Integrated Circuits (ASICs), Application Standard Products (ASSPs), System-on-Chip (SoCs), Complex Programmable Logic Devices (CPLDs), and so on.
[0116] Some embodiments of this disclosure also provide a computer program product, including a computer program that, when executed by a processor, implements any of the above-described methods for generating item attribute information.
[0117] The above description is merely a selection of preferred embodiments of this disclosure and an explanation of the technical principles employed. Those skilled in the art should understand that the scope of the invention involved in the embodiments of this disclosure is not limited to technical solutions formed by specific combinations of the above-described technical features, but should also cover other technical solutions formed by arbitrary combinations of the above-described technical features or their equivalents without departing from the above-described inventive concept. For example, technical solutions formed by substituting the above-described features with (but not limited to) technical features with similar functions disclosed in the embodiments of this disclosure.
Claims
1. A method for generating item attribute information, comprising: Based on pre-generated image prompts for the target item, generate item image feature information corresponding to the target item image; Determine the set of attribute value information corresponding to the preset set of item attribute values for the target item attribute name, wherein the attribute value information is generated based on the item text prompt information, and the target item attribute name is the attribute name of the target item; Attribute value information that satisfies the preset conditions in relation to the feature information of the item image is selected from the attribute value information set and used as the target attribute value information; Select the preset item attribute values that correspond to the target attribute value information from the preset item attribute value set, and use them as the actual item attribute values of the target item; The attribute value information in the attribute value information set is an attribute value weight matrix, which is generated through the following steps: Determine the target item attribute name and the preset set of item attribute values for the target item attribute name; Determine the item text prompt information, wherein the item text prompt information includes: a set of attribute name prompt information and attribute value prompt information; Based on the target item attribute name and the attribute name prompt information, determine the attribute name feature information corresponding to the target item attribute name; For each preset item attribute value in the preset item attribute value set, attribute value feature information corresponding to the preset item attribute value is generated based on the preset item attribute value and each attribute value prompt information in the attribute value prompt information set, thereby obtaining the attribute value feature information set corresponding to the preset item attribute value; For each preset item attribute value, an attribute value weight matrix is determined based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value.
2. The method of claim 1, wherein, The step of determining the attribute value weight matrix corresponding to the preset item attribute value based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value includes: The attribute name feature information and each attribute value feature information in the attribute value feature information set are subjected to feature fusion processing to generate attribute value feature fusion information, thus obtaining an attribute value feature fusion information set. Based on the pre-set weighting coefficients, the attribute value feature fusion information included in the attribute value feature fusion information set is weighted to obtain the attribute value weight matrix corresponding to the preset item attribute value.
3. The method of claim 1, wherein, The determination of the attribute value information set corresponding to the preset set of item attribute values for the target item attribute name includes: Obtain the pre-determined attribute name feature information, as well as the set of attribute value feature information corresponding to each preset item attribute value; Based on the feature information of the item image, determine the attribute name weight matrix and the attribute value weight matrix; For each preset item attribute value corresponding to the attribute value feature information set, according to the attribute name weight matrix and the attribute value weight matrix, the attribute name feature information and each attribute value feature information in the attribute value feature information set are subjected to weighted feature fusion processing to generate attribute value feature fusion information, thereby obtaining the attribute value feature fusion information set corresponding to the preset item attribute value.
4. The method according to claim 3, wherein, The step of determining the set of attribute value information corresponding to the preset set of item attribute values for the target item attribute name also includes: For each preset item attribute value, perform the following attribute value information generation steps: Based on the item image feature information, determine the weight matrix corresponding to each attribute value feature fusion information in the attribute value feature fusion information set to obtain a weight matrix set, wherein the attribute value feature fusion information set corresponds to the preset item attribute value; Based on the weight matrix set, the attribute value feature fusion information in the attribute value feature fusion information set is weighted to obtain the attribute value information corresponding to the preset item attribute value.
5. The method according to claim 1, wherein, The image prompt information is an image prompt matrix; as well as The step of generating item image feature information corresponding to the target item image based on pre-generated image prompt information for the target item includes: Feature extraction is performed on each image patch in the target item image to generate an image patch feature matrix, resulting in an image patch feature matrix set; Determine the image location encoding matrix corresponding to each image block in the target item image to obtain an image location encoding matrix set; The image block feature matrix set and the image location encoding matrix set are fused to obtain a fused matrix; The item image feature information is generated based on the fused matrix, the embeddable matrix, and the image cue matrix, wherein the embeddable matrix is a matrix used to obtain global image features.
6. The method according to claim 5, wherein, The step of generating the item image feature information based on the fused matrix, the embeddable matrix, and the image cue matrix includes: The fused matrix and any one of the embeddable matrix are fused with the image cue matrix to obtain the item image feature information.
7. An item attribute information generation device, comprising: The generation unit is configured to generate item image feature information corresponding to the target item image based on pre-generated image prompt information for the target item; The determining unit is configured to determine a set of attribute value information corresponding to a preset set of item attribute values for a target item attribute name, wherein the attribute value information is generated based on item text prompt information, and the target item attribute name is the attribute name of the target item; The first filtering unit is configured to filter attribute value information from the attribute value information set that has a correlation with the feature information of the item image that meets a preset condition, and use it as target attribute value information; The second filtering unit is configured to filter out preset item attribute values that correspond to the target attribute value information from the preset item attribute value set, and use them as the actual item attribute values of the target item. The attribute value information in the attribute value information set is an attribute value weight matrix, which is generated through the following steps: Determine the target item attribute name and the preset set of item attribute values for the target item attribute name; Determine the item text prompt information, wherein the item text prompt information includes: a set of attribute name prompt information and attribute value prompt information; Based on the target item attribute name and the attribute name prompt information, determine the attribute name feature information corresponding to the target item attribute name; For each preset item attribute value in the preset item attribute value set, attribute value feature information corresponding to the preset item attribute value is generated based on the preset item attribute value and each attribute value prompt information in the attribute value prompt information set, thereby obtaining the attribute value feature information set corresponding to the preset item attribute value; For each preset item attribute value, an attribute value weight matrix is determined based on the attribute name feature information and the attribute value feature information set corresponding to the preset item attribute value.
8. An electronic device, comprising: One or more processors; Storage device, on which one or more programs are stored, When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-6.
9. A computer-readable medium having a computer program stored thereon, wherein, When the program is executed by the processor, it implements the method as described in any one of claims 1-6.
10. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-6.
Citation Information
Patent Citations
A method and device for generating article description information
CN109584013A
Method and device for determining attribute of article
CN114511380A
