Method and system for generating attribute information, and computer storage medium

Abstract

The embodiment of the present application discloses a method and system for generating attribute information, and a computer storage medium. The attribute information generation system obtains a first main image of an object to be published, and obtains characteristic information of the object to be published based on the first main image; uses the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain attribute information of the repaired second candidate object; and generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object.

Description

Technical Field

[0001] The present invention relates to the field of Internet technology, and in particular to a method and system for generating attribute information, and a computer storage medium. Background Art

[0002] With the rapid development of e-commerce business, the number of commodities operated by various e-commerce platforms has reached billions. A large number of commodities are released through various e-commerce platforms. When releasing commodities, it is necessary to fill in various attribute information or detailed information of the commodities.

[0003] In related technologies, merchants are usually required to manually fill in attribute information during the product release process. However, due to the large amount of attribute information involved in the product, relying on manual generation of attribute information will affect the efficiency of product release, is prone to errors, and cannot guarantee the accuracy of the attribute information. Summary of the Invention

[0004] The embodiments of the present application provide a method and system for generating attribute information, and a computer storage medium, which can improve the efficiency and accuracy of filling in commodity attribute information.

[0005] The technical solution of the embodiment of the present application is implemented as follows:

[0006] In a first aspect, an embodiment of the present application provides a method for generating attribute information, the method comprising:

[0007] Obtaining a first main image of an object to be published, and obtaining feature information of the object to be published based on the first main image;

[0008] Obtaining a first candidate object corresponding to the object to be published from a pre-stored object information library using the characteristic information, and performing expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object;

[0009] Performing attribute repair processing according to the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object;

[0010] The target attribute information of the object to be published is generated according to the restored attribute information of the second candidate object.

[0011] In a second aspect, an embodiment of the present application provides a system for generating attribute information, the system comprising an acquisition unit, a processing unit, and a generation unit.

[0012] The acquisition unit is configured to acquire a first main image of the object to be published, and obtain characteristic information of the object to be published based on the first main image; and to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library using the characteristic information, and to perform expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object;

[0013] The processing unit is configured to perform attribute repair processing based on the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object;

[0014] The generating unit is configured to generate target attribute information of the to-be-published object according to the restored attribute information of the second candidate object.

[0015] In a third aspect, an embodiment of the present application provides a system for generating attribute information, which further includes a processor and a memory storing instructions executable by the processor. When the instructions are executed by the processor, the method for generating attribute information as described above is implemented.

[0016] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a program stored thereon, which is applied to a system for generating attribute information, and is characterized in that when the program is executed by a processor, the above-mentioned method for generating attribute information is implemented.

[0017] An embodiment of the present application provides a method and system for generating attribute information, and a computer storage medium. The attribute information generation system obtains a first main image of an object to be published, and obtains characteristic information of the object to be published based on the first main image; uses the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain attribute information of the repaired second candidate object; and generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object. That is to say, in an embodiment of the present application, characteristic information of the object to be published can be first obtained based on the first main image of the object to be published, and then the first candidate object corresponding to the object to be published can be obtained in the pre-stored object information library using the characteristic information, and then the second candidate object can be determined. Since the second candidate object is determined based on the first candidate object, and the second candidate object includes the first candidate object, after the attribute repair processing is performed using the second candidate object with a larger range, the obtained attribute information can be used as accurate attribute information, and the target attribute information of the object to be published can be generated using the attribute information of the repaired second candidate object, thereby realizing the automatic generation of attribute information of the object to be published and improving the efficiency and accuracy of filling in the attribute information. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] Figure 1 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 1 ;

[0019] Figure 2 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 2 ;

[0020] Figure 3 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 3 ;

[0021] Figure 4 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 4 ;

[0022] Figure 5 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 5 ;

[0023] Figure 6 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 6 ;

[0024] Figure 7 A schematic diagram of the attribute information proposed in an embodiment of the present application;

[0025] Figure 8 Schematic diagram of the composition structure of the attribute information generation system proposed in the embodiment of this application Figure 1 ;

[0026] Figure 9 Schematic diagram of the composition structure of the attribute information generation system proposed in the embodiment of this application Figure 2 . DETAILED DESCRIPTION

[0027] The following will be combined with the accompanying drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. It should be understood that the specific embodiments described herein are only used to explain the related applications and are not intended to limit the applications. It should also be noted that for ease of description, only the portions relevant to the related applications are shown in the drawings.

[0028] With the rapid development of e-commerce, the number of commodities sold on various e-commerce platforms has also increased exponentially, reaching billions. A large number of commodities are released through the merchant-side product platforms of various platforms. Traditional product releases require merchants to fill in the product title, main image, attributes, business details, etc. The entire process takes a long time. For example, the attributes of many commodities are at least 20, and information appliances can have 50 or even more attributes. Filling in this information takes a lot of time, of which the time spent on attribute filling accounts for the largest proportion. At the same time, when merchants manually fill in this information, it is inevitable that errors will be made. Therefore, how to efficiently, quickly, and accurately fill in the attribute information of the commodities to be released is currently very urgent.

[0029] In related technologies, merchants are usually required to manually fill in attribute information during the product release process. However, due to the large amount of attribute information involved in the product, relying on manual generation of attribute information will affect the efficiency of product release, is prone to errors, and cannot guarantee the accuracy of the attribute information.

[0030] In order to solve the problems existing in the existing filling of attribute information, the embodiment of the present application provides a method and system for generating attribute information, and a computer storage medium. Specifically, the attribute information generation system obtains a first main image of the object to be published, and obtains characteristic information of the object to be published based on the first main image; uses the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain the attribute information of the repaired second candidate object; generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object, thereby improving the efficiency and accuracy of filling in the commodity attribute information.

[0031] The technical solutions in the embodiments of the present application will be described clearly and completely below in conjunction with the drawings in the embodiments of the present application.

[0032] Example 1

[0033] This application embodiment provides a method for generating attribute information. Figure 1 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 1 ,like Figure 1 As shown, in an embodiment of the present application, the method for generating attribute information may include the following steps:

[0034] Step 101: Obtain a first main image of an object to be published, and obtain feature information of the object to be published based on the first main image.

[0035] In an embodiment of the present application, the attribute information generation system may first obtain a first main image of the object to be published, and obtain feature information of the object to be published based on the first main image.

[0036] It should be noted that, in the embodiment of the present application, the object to be published refers to a commodity to be published, and attribute information needs to be generated for it; for example, the object to be published may be a top.

[0037] Furthermore, in the embodiments of the present application, the first main image refers to the main image of the object to be published. The first main image can intuitively display the appearance and structure of the object to be published. For example, if the object to be published is a top, the first main image can display the collar type, sleeve length, material, and other information of the top. It is understood that the first main image can contain one or more images of the object to be published.

[0038] Specifically, in an embodiment of the present application, after obtaining the first main image of the object to be published, the first main image can be firstly processed for image information pulling to obtain image information of the first main image, and then feature extraction processing can be performed on the image information to obtain feature information.

[0039] It can be understood that, in the embodiment of the present application, the feature information of the object to be published is feature information extracted from the first main image, which can represent the features of the object to be published.

[0040] Furthermore, in an embodiment of the present application, there are multiple ways to obtain feature information, such as scale-invariant features transform (SIFT), accelerated robust features (SURF), and convolutional neural networks (CNN).

[0041] It should be noted that, in an embodiment of the present application, the feature information of the object to be published is obtained from the first main image based on the above-mentioned SIFT, SURF, CNN and other methods. Therefore, the feature information of the object to be published may be the feature vector or deep semantic information of the first main image; for example, when the feature information of the object to be published is obtained from the first main image based on SIFT or SURF, the feature information of the object to be published may be the feature vector of the first main image; and when the feature information of the object to be published is obtained from the first main image based on CNN, the feature information of the object to be published may be the deep semantic information of the first main image.

[0042] Step 102: Obtain a first candidate object corresponding to the object to be published from a pre-stored object information library using the feature information, and perform expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object.

[0043] In an embodiment of the present application, after obtaining the first main image of the object to be published and obtaining the characteristic information of the object to be published based on the first main image, the attribute information generation system can use the characteristic information to obtain the first candidate object corresponding to the object to be published from the pre-stored object information library, and expand the first candidate object to obtain the second candidate object; wherein the second candidate object includes the first candidate object.

[0044] It should be noted that, in the embodiments of the present application, the pre-stored object information database refers to a database containing characteristic information of a large number of pre-stored objects, and the pre-stored objects include the first candidate object corresponding to the object to be published. Exemplarily, the pre-stored object information database contains characteristic information of a large number of pre-stored objects, and these pre-stored objects include the first candidate object for the object to be published.

[0045] Furthermore, in an embodiment of the present application, the first candidate object is of the same category as the object to be published, and the first candidate object may include multiple candidates. For example, if the object to be published is a sweatshirt, the first candidate object may be a sweatshirt of the same brand, model, or appearance as the sweatshirt, and the first candidate object may include multiple sweatshirts similar to the object to be published.

[0046] It is understood that in the embodiments of the present application, expansion processing refers to the process of obtaining a second candidate object based on a first candidate object. In other words, the candidate object is expanded based on the first candidate object to obtain the second candidate object, so that the number of candidate objects contained in the second candidate object is greater than that of the first candidate object, that is, the second candidate object includes the first candidate object. For example, if the first candidate objects are A, B, C, D, then the second candidate objects can be A, B, C, D, E, F, G.

[0047] Furthermore, in an embodiment of the present application, the second object to be selected is of the same category as the first object to be selected. Exemplarily, the first object to be selected includes multiple sweatshirts similar to the object to be published, and further, the second object to be selected includes the first object to be selected and sweatshirts similar to the first object to be selected.

[0048] It should be noted that, in the embodiments of the present application, since the attribute information of the first candidate object may contain erroneous attribute information, after the first candidate object is determined based on the object to be published, the attribute information of the first candidate cannot be directly determined as the target attribute information of the object to be published; in order to ensure the accuracy of the target attribute information of the object to be published, the present application performs extended processing based on the first candidate object to obtain the second candidate object, and then performs corresponding processing based on the second candidate object with a larger scope to obtain the target attribute information of the object to be published.

[0049] Step 103: Perform attribute repair processing based on the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object.

[0050] In an embodiment of the present application, the attribute information generation system obtains the first candidate object corresponding to the object to be published from the pre-stored object information library using feature information, and performs expansion processing on the first candidate object. After obtaining the second candidate object, the attribute repair processing can be performed based on the attribute information of the second candidate object to obtain the attribute information of the repaired second candidate object.

[0051] It should be noted that, in an embodiment of the present application, the attribute repair processing is used to repair the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object; and the repaired attribute information of the second candidate object can also be used as the attribute information of the first candidate object.

[0052] Furthermore, in an embodiment of the present application, the attribute restoration process includes word segmentation, similarity calculation, and clustering. Specifically, first, the attribute information of the second candidate object is subjected to word segmentation to obtain segmented attribute information; then, similarity calculation is performed on the segmented attribute information to obtain first similarities between different second candidate objects; finally, the second candidate objects are clustered based on the obtained first similarities, the clustering result is used as the third candidate object, and the restored attribute information of the second candidate object is obtained based on the attribute information of the third candidate object.

[0053] Step 104: Generate target attribute information of the object to be published based on the restored attribute information of the second candidate object.

[0054] In an embodiment of the present application, after performing attribute repair processing based on the attribute information of the second candidate object and obtaining the repaired attribute information of the second candidate object, the attribute information generation system can generate target attribute information of the object to be published based on the repaired attribute information of the second candidate object.

[0055] It can be understood that in the embodiments of the present application, the target attribute information refers to the attribute information of the object to be published, that is, the attribute information of the object to be published can be directly generated based on the attribute information of the repaired second selected object, thereby realizing the automatic generation of the attribute information of the object to be published without manual filling, thereby improving the efficiency and accuracy of filling in the product attribute information.

[0056] Furthermore, in an embodiment of the present application, since the attribute information of the repaired second object to be selected can also be used as the attribute information of the first object to be selected, when generating the target attribute information of the object to be published based on the attribute information of the repaired second object to be selected, corresponding processing can also be performed in combination with the first object to be selected; specifically, the second similarity between the first main image and the second main image of the first object to be selected can be calculated first, and then the weight parameter of the attribute information of the repaired second object to be selected can be determined based on the second similarity, and finally the target attribute information can be generated based on the attribute information and weight parameter of the first object to be selected.

[0057] Furthermore, in an embodiment of the present application, the attribute information of the repaired second candidate object is used as the attribute information of the first candidate object, and then the attribute information of the first candidate object can be stored as alternative attribute information. When the attribute information generation system detects another object to be published that is the same as the object to be published, it can directly call the stored alternative attribute information to automatically generate the target attribute information of the other object to be published; therefore, by determining the attribute information of the repaired second candidate object as the attribute information of the first candidate object, the attribute information of the first candidate object can be repeatedly used to generate attribute information subsequently, which can improve the efficiency of generating attribute information.

[0058] Figure 2 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 2 ,like Figure 2 As shown, the attribute information generation system performs attribute repair processing based on the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object, that is, step 103 may include the following steps:

[0059] Step 103a: perform word segmentation processing on the attribute information of the second candidate object to obtain attribute information after word segmentation.

[0060] In an embodiment of the present application, the attribute information generation system performs attribute repair processing based on the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object. Specifically, the attribute information generation system can first perform word segmentation processing on the attribute information of the second candidate object to obtain the attribute information after word segmentation.

[0061] It should be noted that in the embodiments of the present application, the attribute information may include a title. The title is a special type of attribute information. Therefore, the attribute information of the second candidate object includes the title of the second candidate object. Furthermore, performing word segmentation processing on the attribute information of the second candidate object specifically refers to performing word segmentation processing on the title of the second candidate object, so that the obtained attribute information after word segmentation includes the title after word segmentation and other attribute information.

[0062] Furthermore, in the embodiments of the present application, word segmentation processing can be achieved through a variety of methods, such as jieba, Han Language Processing (HanLP), Hidden Markov Model (HMM) + Conditional Random Field (CRF), Long Short-Term Memory Artificial Neural Network (LSTM) + CRF and other methods.

[0063] Step 103b: perform similarity calculation based on the attribute information after word segmentation to obtain first similarities between different second candidate objects.

[0064] In an embodiment of the present application, the attribute information generation system performs word segmentation processing on the attribute information of the second candidate object. After obtaining the segmented attribute information, it can perform similarity calculation processing based on the segmented attribute information to obtain the first similarity between different second candidate objects.

[0065] It should be noted that in the embodiments of the present application, similarity calculation processing is performed on the attribute information after word segmentation. Specifically, since the attribute information after word segmentation includes the title after word segmentation and other attribute information, similarity calculation processing is performed on the attribute information after word segmentation, that is, similarity calculation processing is performed on the title after word segmentation and other attribute information.

[0066] Furthermore, in an embodiment of the present application, since the number of objects to be selected contained in the second objects to be selected is multiple, when performing similarity calculation on the second objects to be selected, the similarity between every two objects to be selected in the second objects to be selected is calculated, and the similarity result obtained is the similarity between different second objects to be selected, which is the first similarity.

[0067] Furthermore, in an embodiment of the present application, the similarity calculation processing method can adopt the Locality Sensitive Hashing (LSH) algorithm in the Approximate Nearest Neighbor (ANN) method, and can be combined with the computing engine (SPARK) to optimize the iterative workload.

[0068] Step 103c: cluster the second candidate objects according to the first similarity to obtain third candidate objects.

[0069] In an embodiment of the present application, after performing similarity calculation based on the attribute information after word segmentation and obtaining the first similarity between different second candidate objects, the attribute information generation system can cluster the second candidate objects based on the first similarity to obtain the third candidate object.

[0070] It should be noted that in an embodiment of the present application, the third candidate object is the clustering result obtained after clustering the second candidate object; the third candidate objects are all candidates of the same category, and the attribute information of the first candidate object can be determined based on the attribute information of the third candidate object.

[0071] Furthermore, in the embodiment of the present application, a label propagation algorithm (LPA) or a connectivity graph algorithm may be used to implement clustering.

[0072] Step 103d: Determine the attribute information of the third candidate object as the restored attribute information of the second candidate object.

[0073] In an embodiment of the present application, after clustering the second candidate objects according to the first similarity to obtain the third candidate object, the attribute information generation system may determine the attribute information of the third candidate object as the restored attribute information of the second candidate object.

[0074] It can be understood that in an embodiment of the present application, since the third candidate object obtained after clustering processing is a candidate object belonging to the same category, the attribute information of the third candidate object can be used as the optimal choice for the attribute information of the repaired second candidate object, so that the attribute information of the third candidate object can be determined as the attribute information of the repaired second candidate object.

[0075] Furthermore, in an embodiment of the present application, the attribute information generation system performs clustering processing on the second candidate objects according to the first similarity to obtain the third candidate object, which may include the following steps:

[0076] Step 201: Determine a second candidate object whose first similarity is greater than or equal to a preset similarity threshold as a third candidate object.

[0077] In an embodiment of the present application, the attribute information generation system clusters the second candidate object according to the first similarity to obtain the third candidate object. Specifically, the attribute information generation system can determine the second candidate object whose first similarity is greater than or equal to the preset similarity threshold as the third candidate object.

[0078] It can be understood that in the embodiment of the present application, the role of the preset similarity threshold is to screen out the third candidate object; therefore, when the second candidate object is clustered according to the first similarity, the second candidate object whose first similarity is greater than or equal to the preset similarity threshold is determined as the third candidate object.

[0079] Figure 3 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 3 ,like Figure 3 As shown, in an embodiment of the present application, the attribute information generation system may use the feature information to obtain the first candidate object corresponding to the object to be published from the pre-stored object information library, which may include the following steps:

[0080] Step 301: Perform dimensionality reduction processing on the feature information to obtain a dense vector corresponding to the feature information.

[0081] In an embodiment of the present application, the attribute information generation system uses feature information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and can first perform dimensionality reduction processing on the feature information to obtain a dense vector corresponding to the feature information.

[0082] It should be noted that in the embodiment of the present application, the feature information obtained from the first main image belongs to a high-dimensional sparse vector, and it is difficult to perform effective feature space division from the high-dimensional sparse vector, and feature selection is very inefficient; therefore, it is necessary to first perform dimensionality reduction processing on the feature information and convert the feature information into a dense vector, which can better represent the overall feature information.

[0083] Furthermore, in an embodiment of the present application, an image classification algorithm or an image recognition algorithm can be used for dimensionality reduction processing, and a dense vector of feature information can be obtained by setting a fully connected layer in the algorithm; for example, the algorithm uses a convolutional neural network (CNN) to obtain a dense vector of feature information by setting a fully connected layer in the CNN structure.

[0084] It should be noted that in an embodiment of the present application, before performing dimensionality reduction processing on the feature information, it is also necessary to train the preset model for implementing the dimensionality reduction processing, so as to optimize the preset model to obtain better processing efficiency; for example, the preset model is CNN, and a fully connected layer is set in the CNN structure, and the preset model is trained using a preset training data set to obtain a trained model, and finally the trained model is used to obtain a dense vector of the feature information.

[0085] Step 302: Search the pre-stored object information database according to the dense vector to obtain the first candidate object.

[0086] In an embodiment of the present application, after performing dimensionality reduction processing on the feature information and obtaining a dense vector corresponding to the feature information, the attribute information generation system can search the pre-stored object information library according to the dense vector to obtain the first candidate object.

[0087] It should be noted that, in the embodiment of the present application, the function of the retrieval process is to obtain the first candidate object; that is, the dense vector is used to search in the pre-stored object information library, so as to obtain the first candidate object.

[0088] Furthermore, in an embodiment of the present application, retrieval processing can be implemented through methods such as Facebook AI Similarity Search (Faiss), Approximate Nearest Neighbors Oh Yeah (Annoy), LSH, k-dimensional tree (kd-tree) or ball-tree.

[0089] For example, in an embodiment of the present application, Faiss is used to implement retrieval processing. Specifically, Faiss has multiple indexing methods, including indexFlatL2, indexFlatIP, IndexLSH, IndexIVFPQ, etc.

[0090] Figure 4 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 4 ,like Figure 4 As shown, in an embodiment of the present application, the method for generating attribute information may further include the following steps:

[0091] Step 401: Obtain a first main image, perform attribute extraction processing on the first main image using a preset image classification model, and obtain attribute information of the object to be published.

[0092] In an embodiment of the present application, the attribute information generation system may further obtain a first main image, perform attribute extraction processing on the first main image using a preset image classification model, and obtain attribute information of the object to be published.

[0093] It should be noted that in the embodiments of the present application, attribute information can be obtained from multiple channels, and the attribute information obtained through different acquisition channels is also different. For example, when the object to be published is a coat, attribute information can be obtained from the first main image of the object to be published, the care label, the brand name, etc. Attribute information such as collar type, sleeve length, and sleeve type can be obtained from the first main image, while attribute information such as material and fabric can be obtained from the care label. Therefore, after obtaining the first main image of the object to be published, some attribute information can also be obtained directly from the first main image.

[0094] It should be noted that, in an embodiment of the present application, a preset image classification model is used to obtain attribute information from the first main image; the preset image classification model can adopt ResNet152.

[0095] Furthermore, in an embodiment of the present application, before using the preset image classification model to perform attribute extraction processing on the first main image, the preset image classification model needs to be trained, wherein the training data includes a large amount of pre-stored attribute information; for example, the training data may include a large amount of attribute information of collar types, so that the initial image classification model is trained using the training data to obtain the preset image classification model, so that the preset image classification model can effectively extract the attribute information about the collar type in the first main image.

[0096] Step 402: Generate target attribute information according to the attribute information of the object to be published.

[0097] In an embodiment of the present application, the attribute information generation system obtains the first main image, uses a preset image classification model to perform attribute extraction processing on the first main image, and obtains the attribute information of the object to be published. Then, the target attribute information can be generated based on the attribute information of the object to be published.

[0098] It can be understood that in the embodiments of the present application, the attribute information that can be obtained from the first main image is attribute information that can be intuitively reflected. Therefore, the attribute information of the object to be published obtained using the preset image classification model is only part of the attribute information of the object to be published. The attribute information of the object to be published can be supplemented with the target attribute information of the object to be published based on the attribute information of the first candidate object determined above.

[0099] Figure 5 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 5 ,like Figure 5As shown, in an embodiment of the present application, the attribute information generation system generates target attribute information of the object to be published based on the restored attribute information of the second candidate object, that is, step 104 may include the following steps:

[0100] Step 104a: Calculate a second similarity between the first main image and the second main image of the first candidate object.

[0101] In an embodiment of the present application, the attribute information generation system generates target attribute information of the object to be published based on the restored attribute information of the second candidate object. Specifically, the second similarity between the first main image and the second main image of the first candidate object can be calculated first.

[0102] It can be understood that in the embodiment of the present application, the second similarity refers to the similarity between the first main image and the second main image of the first candidate object; the second similarity represents the degree of similarity between each candidate object in the first candidate object and the object to be published.

[0103] Step 104b: Determine a weight parameter of the restored attribute information of the second candidate object according to the second similarity.

[0104] In an embodiment of the present application, after calculating the second similarity between the first main image and the second main image of the first candidate object, the attribute information generation system can determine the weight parameter of the restored attribute information of the second candidate object according to the second similarity.

[0105] It can be understood that in the embodiment of the present application, since the second similarity represents the degree of similarity between each candidate object in the first candidate object and the object to be published, and the attribute information of the repaired second candidate object can be used as the attribute information of the first candidate object; therefore, the weight parameter of the attribute information of the repaired second candidate object can be determined according to the second similarity, so as to obtain more accurate target attribute information of the object to be published based on the weight parameter and the attribute information of the repaired second candidate object.

[0106] Step 104c: Generate target attribute information based on the restored attribute information and weight parameters of the second candidate object.

[0107] In an embodiment of the present application, after determining the weight parameter of the attribute information of the first candidate object according to the second similarity, the attribute information generation system can generate target attribute information according to the restored attribute information of the second candidate object and the weight parameter.

[0108] It should be noted that, in an embodiment of the present application, a specific method of generating target attribute information based on the attribute information and weight parameters of the repaired second candidate object is to sum the attribute information of the repaired second candidate object in combination with the weight parameters. For example, a regression model optimization method can be used to generate target attribute information based on the attribute information and weight parameters of the repaired second candidate object.

[0109] Figure 6 Schematic diagram of the implementation process of the method for generating attribute information proposed in this embodiment of the application Figure 6 ,like Figure 6 As shown, in an embodiment of the present application, the attribute information generation system may obtain feature information of the object to be published according to the first main image, including the following steps:

[0110] Step 501: extract image information of the first main image to obtain image information of the first main image.

[0111] In an embodiment of the present application, the attribute information generation system obtains characteristic information of the object to be published based on the first main image. Specifically, the attribute information generation system can first pull image information of the first main image to obtain image information of the first main image.

[0112] It should be noted that, in the embodiment of the present application, the picture information pulling process refers to the rapid pulling of picture information through information flow, thereby obtaining the picture information of the first main picture.

[0113] Furthermore, in an embodiment of the present application, image information pulling and processing can be implemented in a variety of ways, such as Apache Kafka, Message Queue (MQ), Flume, etc.

[0114] Step 502: Perform feature extraction processing on the image information to obtain feature information.

[0115] In an embodiment of the present application, after the attribute information generation system performs image information pulling processing on the first main image to obtain the image information of the first main image, it can perform feature extraction processing on the image information to obtain feature information.

[0116] It can be understood that, in the embodiment of the present application, the feature extraction process is used to obtain feature information of the object to be published from the image information.

[0117] Furthermore, in the embodiments of the present application, feature extraction processing can be implemented through various methods, such as scale-invariant feature transform (SIFT), histogram of oriented gradients (HOG), deep learning, etc.

[0118] An embodiment of the present application provides a method for generating attribute information. The attribute information generation system obtains a first main image of an object to be published, and obtains feature information of the object to be published based on the first main image; uses the feature information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain attribute information of the repaired second candidate object; and generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object. That is to say, in an embodiment of the present application, characteristic information of the object to be published can be first obtained based on the first main image of the object to be published, and then the first candidate object corresponding to the object to be published can be obtained in the pre-stored object information library using the characteristic information, and then the second candidate object can be determined. Since the second candidate object is determined based on the first candidate object, and the second candidate object includes the first candidate object, after the attribute information of the second candidate object is repaired, the obtained attribute information can be used as the accurate attribute information of the first candidate object. Furthermore, the target attribute information of the object to be published is generated by the repaired attribute information of the second candidate object, thereby realizing the automatic generation of attribute information of the object to be published and improving the efficiency and accuracy of filling in the attribute information.

[0119] Example 2

[0120] Based on the above embodiment, illustratively, in another embodiment of the present application, Figure 7 This is a schematic diagram of the attribute information proposed in the embodiment of the present application, such as Figure 7 As shown, the attribute information of different types of stock keeping units (SKUs) is also different. The attribute information may include extended attributes and specification attributes. Among them, when the SKU is an electronic product such as a mobile phone, notebook, gaming notebook, or desktop computer, its specification attributes may include model, video memory capacity, video memory type, camera, etc.; extended attributes may include resolution, type, system, processor, graphics card model, etc.; and when the SKU is a non-standard product such as a top, skirt, or sweatshirt, its specification attribute may be a packing list, and extended attributes may include craftsmanship, collar type, sleeve length, applicable scenarios, thickness, etc.

[0121] Furthermore, in an embodiment of the present application, the attribute information generation system may be used to automatically generate attribute information of the object to be published in the SKU and standard product unit (SPU) dimensions.

[0122] Illustratively, in an embodiment of the present application, when generating characteristic information of an object to be published, a main image of the object to be published may be obtained first, and then the characteristic information of the object to be published may be obtained based on the main image.

[0123] Furthermore, in an embodiment of the present application, a preset model is set and trained or optimized, thereby using the optimized preset model to convert feature information of the object to be published into a low-dimensional dense vector. The preset model can be an image classification algorithm or an image recognition algorithm, and a fully connected layer needs to be set in the algorithm to convert the feature information into a low-dimensional dense vector using the preset model.

[0124] Furthermore, in an embodiment of the present application, since the pre-stored object information library contains feature information of a large amount of pre-stored objects, after obtaining the dense vector of the object to be published, retrieval processing can be performed in the pre-stored object information library based on the dense vector, and multiple pre-stored objects that are most similar to the object to be published can be obtained. These pre-stored objects are the first candidate objects.

[0125] Furthermore, in the embodiments of the present application, although the first candidate object has been determined, the accuracy of the attribute information of the first candidate object cannot be guaranteed. Therefore, corresponding processing is required based on the determination of the first candidate object to obtain accurate attribute information of the first candidate object; specifically, the first candidate object is first expanded to obtain the second candidate object, and then attribute repair processing is performed based on the attribute information of the second candidate object to obtain the target attribute information of the object to be published; wherein, the second candidate object includes the first candidate object.

[0126] Furthermore, in an embodiment of the present application, the attribute information of the repaired second candidate object can be obtained through the knowledge graph, and the process can be to perform word segmentation processing on the attribute information of the second candidate object. Specifically, the word segmentation processing is mainly performed on the title information in the attribute information of the second candidate object, and then after the title information is segmented, the similarity calculation processing is performed on the segmented title and other attribute information, so as to obtain a first similarity that can represent the degree of similarity between different second candidate objects; and then the second candidate object is clustered using the first similarity. The clustering result obtained after clustering is the third candidate object. The third candidate object can be represented as an entity, which represents the same type of goods. That is, the third candidate object may be the same type of goods sold by different selling objects, and then the attribute information of the repaired second candidate object can be determined based on the attribute information of the third candidate object; wherein, LPA or connectivity graph algorithm can be used to implement clustering processing.

[0127] Furthermore, in an embodiment of the present application, in addition to the above process, a preset image classification model can be used to extract attribute information from the first main image, so that the target attribute information of the object to be published can be jointly determined by using the attribute information of the object to be published extracted from the first main image and the attribute information of the repaired second selected object.

[0128] Furthermore, in the embodiments of the present application, the accuracy of the attribute information determined above can be ensured by introducing a regular expression engine.

[0129] Furthermore, in an embodiment of the present application, after determining the attribute information of the repaired second candidate object, the second similarity between the first main image of the object to be published and the second main image of the first candidate object can also be calculated, and the second similarity can be used to determine the weight parameter of the attribute information of the repaired second candidate object. By combining the weight parameter to sum the attribute information of the repaired second candidate object, further attribute optimization can be achieved to determine the target attribute information of the object to be published.

[0130] To sum up, when merchants need to publish products, in order to ensure that the merchants publish products in compliance with regulations and improve the quality and efficiency of product release, this application realizes multimodal product release through image processing technology, image retrieval technology, and knowledge graph technology. In addition, the attribute information generation method proposed in the embodiment of this application has been verified and the effect is positive, and it can realize the automatic generation of product attribute information.

[0131] An embodiment of the present application provides a method for generating attribute information. The attribute information generation system obtains a first main image of an object to be published, and obtains feature information of the object to be published based on the first main image; uses the feature information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain attribute information of the repaired second candidate object; and generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object. That is to say, in an embodiment of the present application, characteristic information of the object to be published can be first obtained based on the first main image of the object to be published, and then the first candidate object corresponding to the object to be published can be obtained in the pre-stored object information library using the characteristic information, and then the second candidate object can be determined. Since the second candidate object is determined based on the first candidate object, and the second candidate object includes the first candidate object, after the attribute repair processing is performed using the second candidate object with a larger range, the obtained attribute information can be used as accurate attribute information, and the target attribute information of the object to be published can be generated using the attribute information of the repaired second candidate object, thereby realizing the automatic generation of attribute information of the object to be published and improving the efficiency and accuracy of filling in the attribute information.

[0132] Example 3

[0133] This application embodiment provides a method for generating attribute information. Figure 8 Schematic diagram of the composition structure of the attribute information generation system proposed in the embodiment of this application Figure 1 ,like Figure 8 As shown, the attribute information generation system 10 proposed in the embodiment of the present application includes: an acquisition unit 11, a processing unit 12, a generation unit 13, and an extraction unit 14.

[0134] The acquisition unit 11 is configured to acquire a first main image of the object to be published, and obtain characteristic information of the object to be published based on the first main image; and to use the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and to perform expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object.

[0135] The processing unit 12 is configured to perform attribute restoration processing based on the attribute information of the second candidate object to obtain restored attribute information of the second candidate object.

[0136] The generating unit 13 is configured to generate target attribute information of the to-be-published object according to the restored attribute information of the second candidate object.

[0137] Furthermore, the processing unit 12 is specifically used to perform word segmentation processing on the attribute information of the second candidate object to obtain the attribute information after word segmentation; perform similarity calculation processing based on the attribute information after word segmentation to obtain the first similarity between different second candidate objects; perform clustering processing on the second candidate objects based on the first similarity to obtain the third candidate object; and determine the attribute information of the third candidate object as the attribute information of the repaired second candidate object.

[0138] Furthermore, the processing unit 12 is specifically configured to determine the second candidate object whose first similarity is greater than or equal to a preset similarity threshold as the third candidate object.

[0139] Furthermore, the acquisition unit 11 is specifically configured to perform dimensionality reduction processing on the feature information to obtain a dense vector corresponding to the feature information; and perform a search process on the pre-stored object information library according to the dense vector to obtain the first candidate object.

[0140] The extraction unit 14 is configured to obtain the first main image, perform attribute extraction processing on the first main image using a preset image classification model, and obtain attribute information of the object to be published.

[0141] Furthermore, the generating unit 13 is further configured to generate the target attribute information according to the attribute information of the object to be published.

[0142] Furthermore, the generation unit 13 is specifically used to calculate a second similarity between the first main image and the second main image of the first candidate object; determine a weight parameter of the attribute information of the repaired second candidate object based on the second similarity; and generate the target attribute information based on the attribute information of the repaired second candidate object and the weight parameter.

[0143] Furthermore, the acquisition unit 11 is specifically used to perform image information pulling processing on the first main image to obtain the image information of the first main image; and perform feature extraction processing on the image information to obtain the feature information.

[0144] Figure 9 Schematic diagram of the composition structure of the attribute information generation system proposed in the embodiment of this application Figure 2 ,like Figure 9 As shown, the attribute information generation system proposed in this application also includes a processor 15, a memory 16 storing instructions executable by the processor 15, a communication interface 17, and a bus 18 for connecting the processor 15, the memory 16 and the communication interface 17.

[0145] In an embodiment of the present application, the processor 15 may be at least one of an application-specific integrated circuit (ASIC), a digital signal processor (DSP), a digital signal processing device (DSPD), a programmable logic device (PLD), a field programmable gate array (FPGA), a central processing unit (CPU), a controller, a microcontroller, and a microprocessor. It is understandable that for different devices, the electronic device used to implement the above-mentioned processor function may also be other, and the embodiment of the present application is not specifically limited. A memory 16 may also be included, which may be connected to the processor 15, wherein the memory 16 is used to store executable program code, the program code including computer operating instructions, and the memory 16 may include a high-speed RAM memory, and may also include a non-volatile memory, for example, at least two disk memories.

[0146] In the embodiment of the present application, the bus 18 is used to connect the communication interface 17, the processor 15, and the memory 16, as well as to facilitate mutual communication between these devices.

[0147] In the embodiment of the present application, the memory 16 is used to store instructions and data.

[0148] Furthermore, in an embodiment of the present application, the above-mentioned processor 15 is used to obtain a first main image of the object to be published, and obtain characteristic information of the object to be published based on the first main image; use the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and perform expansion processing on the first candidate object to obtain a second candidate object; wherein, the second candidate object includes the first candidate object; perform attribute repair processing based on the attribute information of the second candidate object to obtain the attribute information of the repaired second candidate object; and generate the target attribute information of the object to be published based on the attribute information of the repaired second candidate object.

[0149] In practical applications, the memory 16 may be a volatile memory, such as a random-access memory (RAM); or a non-volatile memory, such as a read-only memory (ROM), a flash memory, a hard disk drive (HDD) or a solid-state drive (SSD); or a combination of the above types of memory, and provide instructions and data to the processor 15.

[0150] In addition, the functional modules in this embodiment may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The above-mentioned integrated units may be implemented in the form of hardware or software functional modules.

[0151] If the integrated unit is implemented in the form of a software functional module and is not sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this embodiment, or the part that contributes to the existing technology, or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium and includes a number of instructions for causing a computer device (which can be a personal computer, server, or network device, etc.) or a processor to execute all or part of the steps of the method of this embodiment. The aforementioned storage medium includes various media that can store program code, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0152] An embodiment of the present application provides a system for generating attribute information, wherein the system obtains a first main image of an object to be published, and obtains characteristic information of the object to be published based on the first main image; uses the characteristic information to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library, and performs expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object; performs attribute repair processing based on the attribute information of the second candidate object to obtain attribute information of the repaired second candidate object; and generates target attribute information of the object to be published based on the attribute information of the repaired second candidate object. That is to say, in an embodiment of the present application, characteristic information of the object to be published can be first obtained based on the first main image of the object to be published, and then the first candidate object corresponding to the object to be published can be obtained in the pre-stored object information library using the characteristic information, and then the second candidate object can be determined. Since the second candidate object is determined based on the first candidate object, and the second candidate object includes the first candidate object, after the attribute repair processing is performed using the second candidate object with a larger range, the obtained attribute information can be used as accurate attribute information, and the target attribute information of the object to be published can be generated using the attribute information of the repaired second candidate object, thereby realizing the automatic generation of attribute information of the object to be published and improving the efficiency and accuracy of filling in the attribute information.

[0153] An embodiment of the present application provides a first computer-readable storage medium having a program stored thereon. When the program is executed by a first processor, the method of the first and second embodiments is implemented.

[0154] Specifically, the program instructions corresponding to the method for generating attribute information in this embodiment may be stored on a storage medium such as a CD, a hard disk, or a USB flash drive. When the program instructions corresponding to the method for generating attribute information in the storage medium are read or executed by an electronic device, the following steps are included:

[0155] Obtaining a first main image of an object to be published, and obtaining feature information of the object to be published based on the first main image;

[0156] Obtaining a first candidate object corresponding to the object to be published from a pre-stored object information library using the characteristic information, and performing expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes the first candidate object;

[0157] Performing attribute repair processing according to the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object;

[0158] The target attribute information of the object to be published is generated according to the restored attribute information of the second candidate object.

[0159] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may adopt the form of hardware embodiments, software embodiments, or embodiments combining software and hardware. Furthermore, the present application may adopt the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to magnetic disk storage and optical storage, etc.) containing computer-usable program code.

[0160] The present application is described with reference to the implementation flow charts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of the present application. It should be understood that each process and / or box in the flow charts and / or block diagrams, as well as the combination of processes and / or boxes in the flow charts and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the steps in the implementation flow charts. Figure 1 a process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0161] These computer program instructions may also be stored in a computer readable memory that can direct a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer readable memory produce an article of manufacture comprising an instruction device, which is implemented in the implementation flow diagram. Figure 1 a process or multiple processes and / or boxes Figure 1 The function specified in one or more boxes.

[0162] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process described in the flowchart. Figure 1 a process or multiple processes and / or boxes Figure 1 A step that specifies a function in one or more boxes.

[0163] The above description is merely a preferred embodiment of the present application and is not intended to limit the scope of protection of the present application.

Claims

1. A method for generating attribute information, characterized in that: The method comprises: Obtaining a first main image of an object to be published, and obtaining feature information of the object to be published based on the first main image; Obtaining a first candidate object corresponding to the object to be published from a pre-stored object information database using the characteristic information, and performing expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes a greater number of candidate objects than the first candidate object, and the second candidate object includes the first candidate object; Performing attribute repair processing according to the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object; the attribute repair processing includes word segmentation processing, similarity calculation processing and clustering processing; generating target attribute information of the object to be published according to the restored attribute information of the second candidate object; The performing of attribute repair processing according to the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object includes: performing word segmentation processing on the attribute information of the second candidate object to obtain attribute information after word segmentation; Performing similarity calculation based on the attribute information after word segmentation to obtain first similarities between different second candidate objects; performing clustering processing on the second candidate objects according to the first similarity to obtain a third candidate object; The attribute information of the third candidate object is determined as the attribute information of the restored second candidate object.

2. The method according to claim 1, characterized in that The clustering of the second candidate objects according to the first similarity to obtain a third candidate object includes: The second candidate object whose first similarity is greater than or equal to a preset similarity threshold is determined as the third candidate object.

3. The method according to claim 1, characterized in that The obtaining a first candidate object corresponding to the object to be published from a pre-stored object information library using the characteristic information includes: Performing dimensionality reduction processing on the feature information to obtain a dense vector corresponding to the feature information; The pre-stored object information database is searched according to the dense vector to obtain the first candidate object.

4. The method according to claim 1, wherein After obtaining the first main image of the object to be published, the method further includes: Using a preset image classification model to perform attribute extraction processing on the first main image to obtain attribute information of the object to be published; The target attribute information is generated according to the attribute information of the object to be published.

5. The method according to any one of claims 1 to 2, characterized in that Generating the target attribute information of the object to be published according to the restored attribute information of the second candidate object includes: Calculating a second similarity between the first main image and a second main image of the first candidate object; determining a weight parameter of the restored attribute information of the second candidate object according to the second similarity; The target attribute information is generated according to the restored attribute information of the second candidate object and the weight parameter.

6. The method according to claim 1, characterized in that The obtaining characteristic information of the object to be published according to the first main image includes: Performing image information pulling processing on the first main image to obtain image information of the first main image; Perform feature extraction processing on the image information to obtain the feature information.

7. A system for generating attribute information, characterized in that: The attribute information generation system includes an acquisition unit, a processing unit and a generation unit. The acquisition unit is configured to acquire a first main image of the object to be published, and obtain characteristic information of the object to be published based on the first main image; and to obtain a first candidate object corresponding to the object to be published from a pre-stored object information library using the characteristic information, and to perform expansion processing on the first candidate object to obtain a second candidate object; wherein the second candidate object includes a greater number of candidate objects than the first candidate object, and the second candidate object includes the first candidate object; The processing unit is configured to perform attribute repair processing based on the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object; the attribute repair processing includes word segmentation processing, similarity calculation processing, and clustering processing; The generating unit is configured to generate target attribute information of the object to be published based on the restored attribute information of the second candidate object; The performing of attribute repair processing according to the attribute information of the second candidate object to obtain the repaired attribute information of the second candidate object includes: performing word segmentation processing on the attribute information of the second candidate object to obtain attribute information after word segmentation; Performing similarity calculation based on the attribute information after word segmentation to obtain first similarities between different second candidate objects; performing clustering processing on the second candidate objects according to the first similarity to obtain a third candidate object; The attribute information of the third candidate object is determined as the attribute information of the restored second candidate object.

8. A system for generating attribute information, characterized in that: The attribute information generation system further includes a processor and a memory storing instructions executable by the processor. When the instructions are executed by the processor, the attribute information generation method according to any one of claims 1 to 6 is implemented.

9. A computer-readable storage medium having a program stored thereon, applied to a system for generating attribute information, characterized in that: When the program is executed by a processor, the method for generating attribute information according to any one of claims 1 to 6 is implemented.

Images