Auxiliary classification method and device, storage medium and equipment

Abstract

The embodiment of the invention discloses an auxiliary classification method and device, a storage medium and equipment, and the method comprises the steps: determining a corresponding recommendation category for a merchant through a pre-trained large language model based on the commodity features corresponding to the to-be-uploaded commodity information and the category features corresponding to each commodity category before the merchant uploads the commodity information, and carrying out the classification of the commodity information. And the merchant can select the commodity category to which the to-be-uploaded commodity information belongs according to the recommendation category, so that the efficiency of selecting the commodity category to which the to-be-uploaded commodity information belongs by the merchant is improved, and the accuracy of the commodity category selected by the merchant is also improved. Therefore, the probability of entering a category system auditing process due to the fact that merchants select wrong commodity categories subsequently is reduced, and the operation efficiency of all parties can be effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of computer technology, and in particular to an auxiliary classification method, device, storage medium and equipment. Background Art

[0002] At present, e-commerce platforms have been widely used around the world. Merchants can upload product information to e-commerce platforms to sell corresponding products through e-commerce platforms, and users can also purchase the products they need through e-commerce platforms.

[0003] Generally, e-commerce platforms usually maintain a huge commodity category tree, which usually contains more than 10,000 category nodes. In other words, the number of commodity categories preset in the e-commerce platform is at least more than 10,000. When merchants upload commodity information to the e-commerce platform, they need to manually select a commodity category for the commodity information. This is not only inefficient, but once the selection is inappropriate, the e-commerce platform will also start the corresponding category system review process to correct the commodity category to which the commodity information belongs. The category system review process is very long, which will further reduce the operational efficiency of all parties.

[0004] Therefore, the embodiments of this specification provide an auxiliary classification method, apparatus, storage medium and device. Summary of the invention

[0005] The embodiments of this specification provide an auxiliary classification method, device, storage medium and electronic device to partially solve the problems existing in the above-mentioned prior art.

[0006] The embodiments of this specification adopt the following technical solutions:

[0007] This specification provides an auxiliary classification method, the method comprising:

[0008] Get the product information to be uploaded entered by the merchant;

[0009] Determine the commodity characteristics corresponding to the commodity information to be uploaded;

[0010] Inputting the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category;

[0011] The recommended categories are displayed, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

[0012] Optionally, the product information to be uploaded includes text information and / or image information of the product;

[0013] Determining the product features corresponding to the product information to be uploaded specifically includes:

[0014] When the product information to be uploaded includes text information of the product, the text information is input into a text encoder to obtain text features output by the text encoder as product features corresponding to the product information to be uploaded;

[0015] When the product information to be uploaded includes image information of the product, the image information is input into an image encoder to obtain image features output by the image encoder as product features corresponding to the product information to be uploaded.

[0016] Optionally, the method further comprises:

[0017] Obtaining the commodity category selected by the merchant from among the commodity categories as the pending category;

[0018] If the recommended category is different from the pending category, the pending category is input into the large language model as prompt information, so that the large language model outputs the reason why the pending category is not the product category to which the product information to be uploaded belongs based on the product features and the prompt information, and displays it.

[0019] Optionally, the large language model includes a representation network and an adaptation network;

[0020] Pre-training large language models, including:

[0021] Pre-training the representation network of the large language model to align the representation space of the representation network with the representation space of a feature extractor, wherein the feature extractor includes the text encoder and the image encoder;

[0022] According to the sample product information and the annotations corresponding to the sample product information, the adaptation network of the large language model is trained based on the pre-trained representation network.

[0023] Optionally, pre-training the representation network of the large language model specifically includes:

[0024] Inputting the sample product information into the feature extractor, obtaining the product feature corresponding to the sample product information output by the feature extractor as the first sample feature; and obtaining product description information corresponding to the sample product information according to the sample product information, inputting the product description information into the representation network, and obtaining the second sample feature output by the representation network;

[0025] The representation network is pre-trained with reducing the difference between the first sample feature and the second sample feature as a training goal.

[0026] Optionally, according to the sample product information and the annotations corresponding to the sample product information, based on the pre-trained representation network, the adaptation network of the large language model is trained, specifically including:

[0027] According to the sample product information, determine the product category to which the sample product information belongs as a label corresponding to the sample product information;

[0028] Generate sample pairs according to the sample product information and the annotations, the sample pairs including positive sample pairs and negative sample pairs, the positive sample pairs are composed of the sample product information and the annotations, and the negative sample pairs are composed of the sample product information and other product categories except the annotations;

[0029] Inputting the sample product information in the sample pair into the feature extractor to obtain the sample product features corresponding to the sample product information; and inputting the product category in the sample pair into the representation network to obtain the category features;

[0030] Inputting the sample product features and the category features into the adaptation network to obtain an adaptation result output by the adaptation network, wherein the adaptation result includes a result of whether the sample product information in the input sample pair belongs to the product category in the sample pair;

[0031] According to the adaptation result, comparative training is performed on the adaptation network.

[0032] Optionally, after displaying the recommended categories, the method further includes:

[0033] The product information to be uploaded and the product category described in the product information to be uploaded selected by the merchant are uploaded.

[0034] This specification provides an auxiliary classification device, the device comprising:

[0035] The acquisition module obtains the product information to be uploaded entered by the merchant;

[0036] A determination module, for determining product features corresponding to the product information to be uploaded;

[0037] An input module inputs the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category;

[0038] The recommendation module displays the recommended categories, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

[0039] This specification provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the auxiliary classification method described above is implemented.

[0040] The present specification provides an electronic device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the above-mentioned auxiliary classification method when executing the program.

[0041] At least one of the above technical solutions adopted in the embodiments of this specification can achieve the following beneficial effects:

[0042] The embodiments of this specification disclose an auxiliary classification method. Before a merchant uploads product information, this method determines corresponding recommended categories for the merchant through a pre-trained large language model based on product features corresponding to the product information to be uploaded and category features corresponding to each product category, and recommends them to the merchant. The merchant can select the product category to which the product information to be uploaded belongs based on the recommended category. This not only improves the merchant's efficiency in selecting the product category to which the product information to be uploaded belongs, but also improves the accuracy of the product category selected by the merchant, thereby reducing the probability of the merchant entering the category system review process due to the wrong product category selection, and can effectively improve the operational efficiency of all parties. BRIEF DESCRIPTION OF THE DRAWINGS

[0043] The drawings described herein are used to provide a further understanding of this specification and constitute a part of this specification. The illustrative embodiments and descriptions of this specification are used to explain this specification and do not constitute an improper limitation on this specification. In the drawings:

[0044] Figure 1 A flowchart of an auxiliary classification method provided in this manual;

[0045] Figure 2 A schematic diagram of the structure of a large language model provided in the embodiments of this specification;

[0046] Figure 3 A schematic diagram of an auxiliary classification device provided in an embodiment of this specification;

[0047] Figure 4 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this specification. DETAILED DESCRIPTION

[0048] In order to make the purpose, technical solutions and advantages of this specification more clear, the technical solutions of this specification will be clearly and completely described below in combination with the specific embodiments of this specification and the corresponding drawings. Obviously, the described embodiments are only part of the embodiments of this specification, not all of them. Based on the embodiments in this specification, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this specification.

[0049] The technical solutions provided by the embodiments of this specification are described in detail below in conjunction with the accompanying drawings.

[0050] Figure 1 A flowchart of an auxiliary classification method provided in this specification may include the following steps:

[0051] S100: Obtain the product information to be uploaded input by the merchant.

[0052] The embodiment of this specification aims to assist merchants in selecting appropriate product categories for the product information to be uploaded before the merchant uploads the product information, so as to minimize the probability of entering the category system review process later. Based on this, the following can be implanted on the platform used to provide merchants with product information upload: Figure 1 The auxiliary classification function shown is referred to as the category recommendation assistant below. The merchant enters the product information to be uploaded on the platform, and before uploading, the preset interface can be automatically called to call the category recommendation assistant function to assist the merchant in selecting the appropriate product category before the merchant uploads the product information. Therefore, the entity that implements the auxiliary classification method described in the embodiment of this specification can be the above-mentioned platform for providing merchants with product information to upload, and specifically can be a server for implementing the platform.

[0053] Specifically, when the merchant inputs the product information to be uploaded, the server can obtain the product information to be uploaded input by the merchant.

[0054] S102: Determine product features corresponding to the product information to be uploaded.

[0055] Specifically, the product information to be uploaded input by the merchant includes at least one of the text information of the product and the image information of the product. In step S102, if the product information to be uploaded includes the text information of the product, the text information can be input into a pre-trained text encoder to obtain the text features output by the text encoder. If the product information to be uploaded includes the image information of the product, the image information can be input into a pre-trained image encoder to obtain the image features output by the image encoder.

[0056] Furthermore, if the product information to be uploaded only contains the text information of the product, the corresponding text features can be directly used as the product features corresponding to the product information to be uploaded. If the product information to be uploaded only contains the image information of the product, the corresponding image features can be directly used as the product features corresponding to the product information to be uploaded. If the product information to be uploaded contains both text information and image information, the product features corresponding to the product information to be uploaded can be determined based on the text features and the image features, for example, the text features and the image features can be spliced ​​or merged to obtain the product features corresponding to the product information to be uploaded.

[0057] S104: Input the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category.

[0058] The embodiments of this specification utilize the powerful understanding and representation capabilities of the large language model to pre-input the category information (such as category description information) of each commodity category into the large language model to obtain the category features of each commodity category output by the large language model.

[0059] In step S104, the product features and the category features of each product category may be input into the large language model, so that the large language model selects the product category to which the product information to be uploaded belongs from each product category as the recommended category based on the product features and the category features of each product category.

[0060] S106: Displaying the recommended categories, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

[0061] Specifically, the commodity category selected by the merchant in each commodity category can be first obtained as the pending category. If the pending category selected by the merchant belongs to the recommended category output by the large language model in step S104, the commodity information to be uploaded and the pending category selected by the merchant can be directly uploaded. If the pending category selected by the merchant does not belong to the recommended category output by the large language model in step S104, the pending category can also be input into the large language model as prompt information, so that the large language model outputs the reasons why the pending category selected by the merchant may be wrong based on the commodity characteristics and prompt information, and displays it, so that the merchant can try to modify the pending category selected by him. However, regardless of whether the merchant modifies the pending category selected by him, the commodity information to be uploaded and the commodity category (i.e., the pending category) selected by the merchant to which the commodity information to be uploaded belongs need to be uploaded in the end.

[0062] Through the above method, through the category recommendation assistant, before the merchant uploads the product information, the pre-trained large language model is used to determine the corresponding recommended categories for the merchant based on the product features corresponding to the product information to be uploaded and the category features corresponding to each product category, and recommend them to the merchant. The merchant can select the product category to which the product information to be uploaded belongs according to the recommended category. This not only improves the efficiency of the merchant in selecting the product category to which the product information to be uploaded belongs, but also improves the accuracy of the product category selected by the merchant, thereby reducing the probability of the merchant entering the category system review process due to the wrong product category selection, which can effectively improve the operational efficiency of all parties.

[0063] In the embodiments of this specification, Figure 1 The large language model shown in FIG. 1 includes at least two parts: a representation network and an adaptation network. For example, Figure 2 As shown. The representation network is used to characterize the input information (such as product information or product category), and the adaptation network is used to further extract deep features and match them according to the features of the representation network or external input. Therefore, the training of the large language model also includes at least two parts: pre-training of the representation network and training of at least the adaptation network.

[0064] When pre-training the representation network, it is necessary to make the representation space of the representation network consistent with the representation space including Figure 1 The representation spaces of the feature extractors including the text encoder and the image encoder shown are aligned, that is, for the feature representation of the same thing, the representation of the representation network and the representation of the feature extractor are made as similar as possible. Specifically, the sample product information can be input into the feature extractor to obtain the product features corresponding to the sample product information output by the feature extractor as the first sample features; and, based on the sample product information, the product description information corresponding to the sample product information is obtained, and the product description information is input into the representation network to obtain the second sample features output by the representation network; the representation network is pre-trained with the training goal of reducing the difference between the first sample feature and the second sample feature. Since both the sample product information and the product description information corresponding to the sample product information are essentially information describing the sample product, the representation network is pre-trained with the goal of minimizing the difference between the first sample feature extracted from the sample product information by the feature extractor and the second sample feature extracted from the product description information by the representation network, so as to achieve the purpose of aligning the representation space of the representation network and the feature extractor.

[0065] The sample product information may also include at least one of text information and image information of the sample product. The product description information corresponding to the sample product information may be obtained by inputting the sample product information into the large language model, and allowing the large language model to output the corresponding product description information based on the understanding of the input sample product information.

[0066] When training the adaptation network, the adaptation network can be trained based on the sample product information and the annotations corresponding to the sample product information (i.e., the product category to which the sample product information actually belongs) and the pre-trained representation network. Specifically, according to the sample product information, determine the product category to which the sample product information actually belongs as the annotation corresponding to the sample product information; generate sample pairs according to the sample product information and the annotations, the sample pairs include positive sample pairs and negative sample pairs, the positive sample pairs are composed of the sample product information and the annotations, and the negative sample pairs are composed of the sample product information and other product categories except the annotations; input the sample product information in the sample pair into the feature extractor to obtain the sample product features corresponding to the sample product information; and input the product categories in the sample pair into the pre-trained representation network to obtain the category features; input the sample product features and the category features into the adaptation network to obtain the adaptation results output by the adaptation network, the adaptation results include the results of whether the sample product information in the input sample pair belongs to the product category in the sample pair; according to the adaptation results, perform comparative training on the adaptation network.

[0067] Among them, when constructing sample pairs, the above method can specifically construct the sample product information (including at least one of the text information and image information of the sample product) and the product category to which it actually belongs as a positive sample pair, and construct the sample product information and other product categories as a negative sample pair.

[0068] After training the representation network and adaptation network in the large language model, the product categories can be input into the trained representation network in an offline state to obtain the category features output by the representation network and store them. Therefore, when assisting merchants in classification online, the product information to be uploaded input by the merchant can be input into the feature extractor to obtain product features, and the category features of each product category that has been stored offline in advance can be directly extracted, and then the recommended categories can be obtained through matching through the adaptation network.

[0069] The above is an auxiliary classification method provided in an embodiment of this specification. Based on the same idea, this specification also provides corresponding devices, storage media and electronic devices.

[0070] Figure 3 A schematic diagram of an auxiliary classification device provided in an embodiment of this specification, the device includes:

[0071] The acquisition module 301 acquires the product information to be uploaded input by the merchant;

[0072] Determination module 302, determining the commodity characteristics corresponding to the commodity information to be uploaded;

[0073] An input module 303 inputs the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category;

[0074] The recommendation module 304 displays the recommended categories, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

[0075] Optionally, the product information to be uploaded includes text information and / or image information of the product;

[0076] The determination module 302 is specifically used to, when the product information to be uploaded includes text information of the product, input the text information into a text encoder to obtain text features output by the text encoder as product features corresponding to the product information to be uploaded; when the product information to be uploaded includes image information of the product, input the image information into an image encoder to obtain image features output by the image encoder as product features corresponding to the product information to be uploaded.

[0077] Optionally, the recommendation module 304 is also used to obtain the product category selected by the merchant from each product category as a pending category; if the recommended category is different from the pending category, the pending category is input into the large language model as prompt information, so that the large language model outputs the reason why the pending category is not the product category to which the product information to be uploaded belongs based on the product features and the prompt information, and displays it.

[0078] Optionally, the large language model includes a representation network and an adaptation network;

[0079] The device also includes:

[0080] The training module 305 is used to pre-train the representation network of the large language model so that the representation space of the representation network is aligned with the representation space of the feature extractor, wherein the feature extractor includes the text encoder and the image encoder; according to the sample product information and the annotations corresponding to the sample product information, the adaptation network of the large language model is trained based on the pre-trained representation network.

[0081] Optionally, the training module 305 is specifically used to input the sample product information into the feature extractor, obtain the product feature corresponding to the sample product information output by the feature extractor, as the first sample feature; and, based on the sample product information, obtain the product description information corresponding to the sample product information, input the product description information into the representation network, and obtain the second sample feature output by the representation network; and pre-train the representation network with the training goal of reducing the difference between the first sample feature and the second sample feature.

[0082] Optionally, the training module 305 is specifically used to determine, based on the sample product information, the product category to which the sample product information belongs, as a label corresponding to the sample product information; generate a sample pair based on the sample product information and the label, the sample pair including a positive sample pair and a negative sample pair, the positive sample pair consisting of the sample product information and the label, and the negative sample pair consisting of the sample product information and other product categories except the label; input the sample product information in the sample pair into the feature extractor to obtain the sample product feature corresponding to the sample product information; and, input the product category in the sample pair into the representation network to obtain the category feature; input the sample product feature and the category feature into the adaptation network to obtain the adaptation result output by the adaptation network, the adaptation result including the result of whether the sample product information in the input sample pair belongs to the product category in the sample pair; and perform comparative training on the adaptation network based on the adaptation result.

[0083] Optionally, the device further comprises:

[0084] The uploading module 306 is used to upload the product information to be uploaded and the product category to which the product information to be uploaded belongs selected by the merchant after the recommendation module 304 displays the recommended category.

[0085] This specification also provides a computer-readable storage medium, wherein the storage medium stores a computer program, and when the computer program is executed by a processor, it can be used to perform the above Figure 1 Provided auxiliary classification method.

[0086] based on Figure 1 Based on the Figure 1 The auxiliary classification method shown in the present specification also provides Figure 4 The structural diagram of the electronic device shown in FIG. Figure 4 At the hardware level, the electronic device includes a processor, an internal bus, a network interface, a memory, and a non-volatile memory, and may also include other hardware required for the business. The processor reads the corresponding computer program from the non-volatile memory into the memory and then runs it to achieve the above Figure 1 The auxiliary classification method described.

[0087] Of course, in addition to software implementation, this specification does not exclude other implementation methods, such as logic devices or a combination of software and hardware, etc., that is to say, the executor of the following processing flow is not limited to each logic unit, but can also be hardware or logic devices.

[0088] In the 1990s, improvements to a technology could be clearly distinguished as hardware improvements (for example, improvements to the circuit structure of diodes, transistors, switches, etc.) or software improvements (improvements to the method flow). However, with the development of technology, many improvements to the method flow today can be regarded as direct improvements to the hardware circuit structure. Designers almost always obtain the corresponding hardware circuit structure by programming the improved method flow into the hardware circuit. Therefore, it cannot be said that an improvement in a method flow cannot be implemented using a hardware entity module. For example, a programmable logic device (PLD) (such as a field programmable gate array (FPGA)) is such an integrated circuit whose logical function is determined by the user's programming of the device. Designers can "integrate" a digital system on a PLD by programming it themselves, without having to ask a chip manufacturer to design and produce a dedicated integrated circuit chip. Moreover, nowadays, instead of manually making integrated circuit chips, this kind of programming is mostly implemented by "logic compiler" software, which is similar to the software compiler used when developing and writing programs, and the original code before compilation must also be written in a specific programming language, which is called hardware description language (HDL). There is not only one HDL, but many kinds, such as ABEL (Advanced Boolean Expression Language), AHDL (Altera Hardware Description Language), Confluence, CUPL (Cornell University Programming Language), HDCal, JHDL (Java Hardware Description Language), Lava, Lola, MyHDL, PALASM, RHDL (Ruby Hardware Description Language), etc. The most commonly used ones are VHDL (Very-High-Speed ​​Integrated Circuit Hardware Description Language) and Verilog. Those skilled in the art should also know that it is only necessary to program the method flow slightly in the above-mentioned hardware description languages ​​and program it into the integrated circuit, and then it is easy to obtain the hardware circuit that implements the logic method flow.

[0089] The controller can be implemented in any appropriate manner, for example, the controller can take the form of a microprocessor or processor and a computer-readable medium storing a computer-readable program code (such as software or firmware) that can be executed by the (micro)processor, a logic gate, a switch, an application-specific integrated circuit (ASIC), a programmable logic controller, and an embedded microcontroller. Examples of controllers include, but are not limited to, the following microcontrollers: ARC 625D, Atmel AT91SAM, Microchip PIC18F26K20, and Silicone Labs C8051F320. The memory controller can also be implemented as part of the control logic of the memory. Those skilled in the art also know that in addition to implementing the controller in a purely computer-readable program code manner, the controller can be implemented in the form of a logic gate, a switch, an application-specific integrated circuit, a programmable logic controller, and an embedded microcontroller by logically programming the method steps. Therefore, this controller can be considered as a hardware component, and the devices included therein for implementing various functions can also be regarded as structures within the hardware component. Or even, the devices for implementing various functions can be regarded as both software modules for implementing the method and structures within the hardware component.

[0090] The systems, devices, modules or units described in the above embodiments may be implemented by computer chips or entities, or by products with certain functions. A typical implementation device is a computer. Specifically, the computer may be, for example, a personal computer, a laptop computer, a cellular phone, a camera phone, a smart phone, a personal digital assistant, a media player, a navigation device, an email device, a game console, a tablet computer, a wearable device, or a combination of any of these devices.

[0091] For the convenience of description, the above device is described in various units according to their functions. Of course, when implementing this specification, the functions of each unit can be implemented in the same or multiple software and / or hardware.

[0092] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems, or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) that contain computer-usable program code.

[0093] This specification is described with reference to the flowcharts and / or block diagrams of the methods, devices (systems), and computer program products according to the embodiments of this specification. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of the processes and / or boxes in the flowchart and / or block diagram, 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 processes in the flowchart and / or block diagram. 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.

[0094] These computer program instructions may also be stored in a computer-readable memory capable of directing 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 implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0095] 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. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0096] In a typical configuration, a computing device includes one or more processors (CPU), input / output interfaces, network interfaces, and memory.

[0097] The memory may include non-permanent storage in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0098] Computer readable media include permanent and non-permanent, removable and non-removable media that can be implemented by any method or technology to store information. Information can be computer readable instructions, data structures, program modules or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technology, compact disk read-only memory (CD-ROM), digital versatile disk (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer readable media does not include temporary computer readable media (transitory media), such as modulated data signals and carrier waves.

[0099] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

[0100] Those skilled in the art will appreciate that the embodiments of this specification may be provided as methods, systems or computer program products. Therefore, this specification may take the form of a complete hardware embodiment, a complete software embodiment or an embodiment combining software and hardware. Moreover, this specification may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0101] This specification may be described in the general context of computer-executable instructions executed by a computer, such as program modules. Generally, program modules include routines, programs, objects, components, data structures, etc. that perform specific tasks or implement specific abstract data types. This specification may also be practiced in distributed computing environments where tasks are performed by remote processing devices connected through a communication network. In a distributed computing environment, program modules may be located in local and remote computer storage media, including storage devices.

[0102] Each embodiment in this specification is described in a progressive manner, and the same or similar parts between the embodiments can be referred to each other, and each embodiment focuses on the differences from other embodiments. In particular, for the system embodiment, since it is basically similar to the method embodiment, the description is relatively simple, and the relevant parts can be referred to the partial description of the method embodiment.

[0103] The above description is only an embodiment of the present specification and is not intended to limit the present specification. For those skilled in the art, the present specification may have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present specification shall be included in the scope of the claims of the present specification.

Claims

1. An auxiliary classification method, the method comprising: Get the product information to be uploaded entered by the merchant; Determine the commodity characteristics corresponding to the commodity information to be uploaded; Inputting the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category; The recommended categories are displayed, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

2. The method according to claim 1, wherein the product information to be uploaded includes text information and / or image information of the product; Determining the product features corresponding to the product information to be uploaded specifically includes: When the product information to be uploaded includes text information of the product, the text information is input into a text encoder to obtain text features output by the text encoder as product features corresponding to the product information to be uploaded; When the product information to be uploaded includes image information of the product, the image information is input into an image encoder to obtain image features output by the image encoder as product features corresponding to the product information to be uploaded.

3. The method of claim 1, further comprising: Obtain the commodity category selected by the merchant from among the commodity categories as the pending category; If the recommended category is different from the pending category, the pending category is input into the large language model as prompt information, so that the large language model outputs the reason why the pending category is not the product category to which the product information to be uploaded belongs based on the product features and the prompt information, and displays it.

4. The method according to claim 2, wherein the large language model comprises a representation network and an adaptation network; Pre-training large language models, including: Pre-training the representation network of the large language model to align the representation space of the representation network with the representation space of a feature extractor, wherein the feature extractor includes the text encoder and the image encoder; According to the sample product information and the annotations corresponding to the sample product information, the adaptation network of the large language model is trained based on the pre-trained representation network.

5. The method according to claim 4, wherein pre-training the representation network of the large language model specifically comprises: Inputting the sample product information into the feature extractor, and obtaining the product feature corresponding to the sample product information output by the feature extractor as the first sample feature; And, according to the sample product information, obtain product description information corresponding to the sample product information, input the product description information into the representation network, and obtain a second sample feature output by the representation network; The representation network is pre-trained with reducing the difference between the first sample feature and the second sample feature as a training goal.

6. The method according to claim 4, training the adaptation network of the large language model based on the pre-trained representation network according to the sample product information and the annotations corresponding to the sample product information, specifically comprising: According to the sample product information, determine the product category to which the sample product information belongs as a label corresponding to the sample product information; Generate sample pairs according to the sample product information and the annotations, the sample pairs including positive sample pairs and negative sample pairs, the positive sample pairs are composed of the sample product information and the annotations, and the negative sample pairs are composed of the sample product information and other product categories except the annotations; Inputting the sample product information in the sample pair into the feature extractor to obtain the sample product features corresponding to the sample product information; And, inputting the commodity categories in the sample pairs into the representation network to obtain category features; Inputting the sample product features and the category features into the adaptation network to obtain an adaptation result output by the adaptation network, wherein the adaptation result includes a result of whether the sample product information in the input sample pair belongs to the product category in the sample pair; According to the adaptation result, comparative training is performed on the adaptation network.

7. The method according to any one of claims 1 to 6, after displaying the recommended categories, the method further comprises: The product information to be uploaded and the product category to which the product information to be uploaded belongs, selected by the merchant, are uploaded.

8. An auxiliary classification device, comprising: The acquisition module obtains the product information to be uploaded entered by the merchant; A determination module, for determining product features corresponding to the product information to be uploaded; An input module inputs the product features into a pre-trained large language model, so that the large language model determines the product category to which the product information to be uploaded belongs as a recommended category based on the product features and category features corresponding to each product category; The recommendation module displays the recommended categories, so that the merchant can select the commodity category to which the commodity information to be uploaded belongs from various commodity categories according to the displayed recommended categories.

9. A computer-readable storage medium storing a computer program, wherein the computer program, when executed by a processor, implements the method according to any one of claims 1 to 7.

10. An electronic device comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor implements the method according to any one of claims 1 to 7 when executing the program.

Images