Article information processing method and device and storage medium

Abstract

The invention provides an article information processing method and device and a storage medium, and the method comprises the steps: obtaining the basic information of a target article in response to a copywriting generation instruction for the target article; according to the basic information, a candidate article set matched with the target article is retrieved from a preset article library, and the preset document library comprises association relationships between multiple preset articles and corresponding matched documents; and performing small sample learning according to the matching copywriting corresponding to each candidate article in the candidate article set, and generating a target copywriting matched with the target article. The matching copywriting is automatically generated based on the basic information of the article, the diversity of the article copywriting is improved, the manpower and time cost is reduced, and the user experience is improved.

Description

Technical Field

[0001] The present application relates to the field of information processing technology, and in particular to an item information processing method, device and storage medium. Background Art

[0002] Product digitization is to obtain the original product data through visual photography, field extraction, etc., and design, edit, and output these original data according to business needs to form digital product pictures and texts, and directly display them to consumers through screens (such as mobile phones, TVs, etc.), finally completing the entire process of information transmission and transactions. It also includes the process of making a complete product portrait (based on business labeling) according to one's own business needs. With the increase in the share of e-commerce and the advancement of new retail, product digitization is the only way for modern brand companies to manage digitalization, and it is the hub of the IT system (information system, intelligent system) of retail companies.

[0003] For the traditional links of digitalization of goods under the new retail industry, in scenarios such as digital promotion of goods, multi-channel sharing of goods, and operation of goods, it is always impossible to skip the manually designed creative marketing copy. The daily update and sharing of goods has an insurmountable ceiling due to the time-consuming manual work; the copy currently shared to the community in the industry mainly comes from the creativity of shopping guides with different experience backgrounds, such as brands, social platforms, and shopping guides. The generation of marketing copy depends on the manual creative design and input of shopping guides. Manual creative work takes a long time, and the quality of marketing copy is ultimately uneven, resulting in poor efficiency in product promotion and sharing. Summary of the invention

[0004] The main purpose of the embodiments of the present application is to provide an item information processing method, device and storage medium, which realizes the automatic generation of matching copy based on the basic information of the item, which not only improves the diversity of item copy but also reduces manpower and time costs and improves user experience.

[0005] In a first aspect, an embodiment of the present application provides an item information processing method, comprising: in response to a text generation instruction for a target item, obtaining basic information of the target item; based on the basic information, retrieving a set of candidate items matching the target item from a preset item library, wherein the preset text library includes associations between multiple preset items and corresponding matching texts; performing small sample learning based on the matching texts corresponding to each candidate item in the candidate item set to generate a target text matching the target item.

[0006] In one embodiment, the step of obtaining basic information of the target item in response to a text generation instruction for the target item includes: obtaining identification information of the target item entered by a user in response to a text generation instruction for the target item, and extracting the basic information of the target item from a database based on the identification information.

[0007] In one embodiment, the step of obtaining basic information of the target object in response to a text generation instruction for the target object includes: obtaining image information of the target object uploaded by a user in response to a text generation instruction for the target object; identifying the image information to obtain attribute information of the target object, wherein the basic information includes the attribute information.

[0008] In one embodiment, retrieving a set of candidate items matching the target item in a preset item library based on the basic information includes: extracting a first feature of the target item based on the basic information; and retrieving a set of candidate items matching the target item in a preset item library based on the first feature.

[0009] In one embodiment, the preset item library includes the second features of the multiple preset items; and searching the preset item library for a set of candidate items that match the target item based on the first feature includes: determining a first similarity between the first feature and the second feature of each of the preset items; sorting the multiple preset items from large to small according to the corresponding first similarities, and determining the multiple preset items with the first similarity sorted in a front preset order as the set of candidate items that match the target item.

[0010] In one embodiment, the basic information includes first user information applicable to the target item; the preset item library includes second user information applicable to the candidate items; before performing small sample learning based on the matching text corresponding to each candidate item in the candidate item set to generate a target text matching the target item, it also includes: determining a second similarity between the first user information and the second user information of each candidate item; sorting each candidate item in the candidate item set from large to small according to the second similarity to obtain a final sorting result of the candidate item set.

[0011] In one embodiment, the candidate item set includes an association relationship between multiple candidate items and corresponding matching texts; the small sample learning is performed according to the matching texts corresponding to each candidate item in the candidate item set to generate a target text matching the target item, including: determining the multiple candidate items and the corresponding matching texts as a small sample set, training a preset large language model, and inputting a preset prompt word matching the target item into the preset large language model, so that the large language model outputs the target text matching the target item.

[0012] In one embodiment, before retrieving a set of candidate items matching the target item in a preset item library based on the basic information, the method further includes: obtaining a plurality of preset items and matching texts corresponding to each of the preset items, and determining an association relationship between the plurality of preset items and the corresponding matching texts; and respectively extracting second features corresponding to the plurality of preset items, and establishing the preset item library.

[0013] In one embodiment, responding to the copywriting generation instruction for the target item includes: responding to a user's sharing operation on the target item, generating the copywriting generation instruction for the target item.

[0014] In a second aspect, an embodiment of the present application provides a method for processing product information, the method comprising: in response to a copy generation instruction for a target product, obtaining basic information of the target product; based on the basic information, retrieving a set of candidate products matching the target product from a preset product library, wherein the preset copy library includes associations between multiple preset products and corresponding matching copies; performing small sample learning based on the matching copies corresponding to each candidate product in the candidate product set to generate a target copy matching the target product.

[0015] In a third aspect, an embodiment of the present application provides an item information processing device, including:

[0016] An acquisition module, configured to acquire basic information of a target item in response to a copywriting generation instruction for the target item;

[0017] A retrieval module, configured to retrieve a set of candidate items matching the target item from a preset item library according to the basic information, wherein the preset document library includes associations between a plurality of preset items and corresponding matching documents;

[0018] The generation module is used to perform small sample learning based on the matching copy corresponding to each candidate item in the candidate item set to generate a target copy matching the target item.

[0019] In one embodiment, the acquisition module is used to respond to a copywriting generation instruction for a target object, acquire identification information of the target object entered by a user, and extract basic information of the target object from a database according to the identification information.

[0020] In one embodiment, the acquisition module is used to obtain image information of the target object uploaded by the user in response to a text generation instruction for the target object; identify the image information to obtain attribute information of the target object, and the basic information includes the attribute information.

[0021] In one embodiment, the retrieval module is used to extract a first feature of the target item based on the basic information; and retrieve a set of candidate items matching the target item in a preset item library based on the first feature.

[0022] In one embodiment, the preset item library includes second features of the plurality of preset items; the retrieval module is used to determine a first similarity between the first feature and the second feature of each of the preset items; the plurality of preset items are sorted from large to small according to the corresponding first similarities, and the plurality of preset items whose first similarities are sorted in a front preset order are determined as a set of candidate items matching the target item.

[0023] In one embodiment, the basic information includes first user information applicable to the target item; the preset item library includes second user information applicable to the candidate items; the device also includes: a sorting module, which is used to determine the second similarity between the first user information and the second user information of each candidate item before performing small sample learning based on the matching text corresponding to each candidate item in the candidate item set to generate a target text matching the target item; and sort each candidate item in the candidate item set from large to small according to the second similarity to obtain a final sorting result of the candidate item set.

[0024] In one embodiment, the candidate item set includes associations between multiple candidate items and corresponding matching texts; the generation module is used to determine the multiple candidate items and the corresponding matching texts as a small sample set, train a preset large language model, and input a preset prompt word matching the target item into the preset large language model, so that the large language model outputs the target text matching the target item.

[0025] In one embodiment, it also includes: an establishment module, which is used to obtain a plurality of preset items and matching documents corresponding to each of the preset items before retrieving a set of candidate items matching the target item in the preset item library based on the basic information, and determine the association relationship between the plurality of preset items and the corresponding matching documents; respectively extract the second features corresponding to the plurality of preset items and establish the preset item library.

[0026] In one embodiment, the device further includes: a response module, configured to generate a text generation instruction for the target item in response to a user's sharing operation on the target item.

[0027] In a fourth aspect, an embodiment of the present application provides an electronic device, including:

[0028] at least one processor; and

[0029] a memory communicatively coupled to the at least one processor;

[0030] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the electronic device to execute the method described in any one of the above aspects.

[0031] In a fifth aspect, an embodiment of the present application provides a cloud device, including:

[0032] at least one processor; and

[0033] a memory communicatively coupled to the at least one processor;

[0034] The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the cloud device to execute the method described in any one of the above aspects.

[0035] In a sixth aspect, an embodiment of the present application provides a computer-readable storage medium, wherein the computer-readable storage medium stores computer-executable instructions. When a processor executes the computer-executable instructions, the method described in any one of the above aspects is implemented.

[0036] In a seventh aspect, an embodiment of the present application provides a computer program product, including a computer program, which implements the method described in any of the above aspects when executed by a processor.

[0037] The item information processing method, device and storage medium provided in the embodiments of the present application first obtain the basic information of the target item, such as the item title, category, material and other information, when it is necessary to generate promotional copy for the target item. Then, based on the basic information of the item, a search is performed in a preset item library. The preset item library is pre-configured with multiple items and matching copy corresponding to each item. A set of candidate items similar to the target item can be retrieved from the preset item library based on the basic information of the target item. Then, the candidate items and their corresponding matching copy are used as samples for small sample learning. Through small sample learning, the copy style corresponding to the candidate item set can be learned, and then promotional copy that conforms to the style of the target item can be automatically generated. This not only improves the diversity of item copy, but also reduces manpower and time costs, and improves user experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] The drawings herein are incorporated into the specification and constitute a part of the specification, illustrate embodiments consistent with the present application, and are used together with the specification to explain the principles of the present application. Obviously, the drawings described below are some embodiments of the present invention, and for ordinary technicians in this field, other drawings can be obtained based on these drawings without creative labor.

[0039] Figure 1 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application;

[0040] Figure 2 A schematic diagram of an application scenario of an item information processing solution provided in an embodiment of the present application;

[0041] Figure 3 An interactive schematic diagram of an item information processing system provided in an embodiment of the present application;

[0042] Figure 4 A flowchart of an item information processing method provided in an embodiment of the present application;

[0043] Figure 5 A flowchart of an item information processing method provided in an embodiment of the present application;

[0044] Figure 6 A schematic diagram of a specific application scenario flow of an item information processing solution provided in an embodiment of the present application;

[0045] Figure 7 A flowchart of a method for processing commodity information provided in an embodiment of the present application;

[0046] Figure 8 A schematic diagram of the structure of an item information processing device provided in an embodiment of the present application;

[0047] Fig. 9 A schematic diagram of the structure of a cloud device provided in an embodiment of the present application.

[0048] The above drawings have shown clear embodiments of the present application, which will be described in more detail later. These drawings and text descriptions are not intended to limit the scope of the present application in any way, but to illustrate the concept of the present application to those skilled in the art by referring to specific embodiments. DETAILED DESCRIPTION

[0049] Here, exemplary embodiments are described in detail, and examples thereof are shown in the accompanying drawings. When the following description refers to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The implementations described in the following exemplary embodiments do not represent all implementations consistent with the present application.

[0050] The term "and / or" in this article is used to describe the association relationship of associated objects, specifically indicating that there may be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone.

[0051] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0052] In order to clearly describe the technical solution of the embodiment of the present application, the terms involved in the present application are first defined:

[0053] NLP: Natural Language Processing.

[0054] LLM: Large Language Model. This model is based on a deep learning model and can be used for various natural language processing (NLP) tasks. This deep model uses the Transformer architecture, carries a large number of hyperparameters, and requires a large amount of data sets for training. Therefore, LLM is a large-scale deep model.

[0055] Few-shot Learning: Few-shot learning, that is, using fewer samples to complete parameter training based on the backbone network, can still have excellent task execution performance. Few-shot learning in the LLM model no longer uses traditional training of small samples in deep models. By building and optimizing Prompt, the LLM model can effectively learn small-sample knowledge.

[0056] Prompt: The prompt word of the LLM model, that is, the input text of the LLM model. The text in Prompt will be encoded into text features by the encoder and enter the language model.

[0057] ID: Identity document, ID card identification number, account number, unique code, exclusive number.

[0058] BERT: Bidirectional Encoder Representations from Transformers, a bidirectional encoder model built on Transformer, uses a masked language model (MLM) to pre-train the bidirectional Transformer architecture, and uses unsupervised training to represent the deep context in all layers. It is often used as a pre-training model for various NLP downstream tasks.

[0059] StructBert: An improved BERT model.

[0060] GLM: General Language Model, is a general language model based on autoregressive blank filling, which can be fine-tuned for downstream tasks on various natural language understanding and generation tasks.

[0061] Text generation: an important and challenging task in the field of natural language processing. Text generation is to generate text based on structured input data, including machine translation, generative text summarization, intelligent question answering, etc.

[0062] Image recognition: Extract features and reweight features of input images based on deep models to identify individual classification categories in the image.

[0063] like Figure 1 As shown, this embodiment provides an electronic device 1, including: at least one processor 11 and a memory 12, Figure 1A processor is taken as an example. The processor 11 and the memory 12 are connected via a bus 10. The memory 12 stores instructions that can be executed by the processor 11, and the instructions are executed by the processor 11, so that the electronic device 1 can execute all or part of the process of the method in the following embodiment to realize automatic generation of matching copy based on the basic information of the item, which not only improves the diversity of the item copy, but also reduces the manpower and time cost, and improves the user experience.

[0064] In one embodiment, the electronic device 1 may be a mobile phone, a tablet computer, a laptop computer, a desktop computer, or a large computing system composed of multiple computers.

[0065] Figure 2 Schematic diagram of an application scenario system 200 of an item information processing solution provided in an embodiment of the present application. Figure 2 As shown, the system includes: a server 210 and a terminal 220, wherein:

[0066] The server 210 may be a data platform that provides item information processing services, such as an e-commerce shopping platform. In actual scenarios, an e-commerce shopping platform may have multiple servers 210. Figure 2 In the figure, one server 210 is taken as an example.

[0067] The terminal 220 may be a computer, mobile phone, tablet or other device used by a user to log in to an e-commerce shopping platform. There may also be multiple terminals 220. Figure 2 Two terminals 220 are taken as an example for illustration.

[0068] The terminal 220 and the server 210 can transmit information via the Internet, so that the terminal 220 can access the data on the server 210. The terminal 220 and / or the server 210 can be implemented by the electronic device 1.

[0069] The item information processing solution of the embodiment of the present application can be deployed on the server 210, can also be deployed on the terminal 220, or can be partially deployed on the server 210 and partially deployed on the terminal 220. In actual scenarios, it can be selected based on actual needs, and this embodiment does not limit it.

[0070] When the item information processing solution is fully or partially deployed on the server 210 , a calling interface may be opened to the terminal 220 to provide algorithm support to the terminal 220 .

[0071] The method provided in the embodiment of the present application can be implemented by executing the corresponding software code by the electronic device 1, and can be implemented by exchanging data with the server. The electronic device 1 can be a local terminal device. When the method is run on the server, the method can be implemented and executed based on the cloud interaction system, wherein the cloud interaction system includes a server and a client device.

[0072] In a possible implementation, the method provided by the embodiment of the present invention provides a graphical user interface through a terminal device, wherein the terminal device can be the local terminal device mentioned above, or can be the client device in the cloud interaction system mentioned above.

[0073] The item information processing method of the embodiment of the present application can be applied to any scenario where item copywriting needs to be generated.

[0074] Taking the e-commerce platform scenario as an example, with the development of the Internet, shopping through e-commerce platforms has become an important part of people's entertainment life. Product digitization is to obtain original product data through visual shooting, field extraction, etc., and according to business needs, these original data are designed, compiled, and output to form digital product pictures and texts, and directly displayed to consumers through screens (such as mobile phones, TVs, etc.), finally completing the entire process of information transmission and transactions. It also includes the process of making a complete product portrait (based on business labeling) according to its own business needs. With the increase in the share of e-commerce and the advancement of new retail, product digitization is the only way for modern brand companies to manage digitalization, and it is the hub of retail companies' IT systems (information systems, intelligent systems).

[0075] For the traditional links of product digitization under the new retail industry, it is always impossible to skip manually designed creative marketing copy in scenarios such as product digitalization, multi-channel product sharing, and product operation. Due to the high time-consuming manual work, the volume of products updated and shared daily has an insurmountable ceiling. At present, the copy shared to the community in the industry mainly comes from the creativity of shopping guides with different experience backgrounds, such as brands, social platforms, and shopping guides. The generation of marketing copy depends on the manual creative design and input of shopping guides. Therefore, not only does the manual creative work take a long time, but the quality of the final marketing copy is also uneven, resulting in poor efficiency in product promotion and sharing.

[0076] In related technologies, product descriptions or creative advertising texts can be generated based on keywords. However, different application scenarios for creative advertising text generation have different specialties. The focus is on generating short copy on advertising images, and the copy style is not easy to migrate. When copywriting with different copy styles is required, a large number of closed set data sets are required for fine-tuning, which is not conducive to the richness of copy generation. Generating a sentence of product function or style description based on keywords has limited text richness, and it is difficult to avoid training with large-scale data sets.

[0077] In order to solve the above problems, an embodiment of the present application provides an item information processing solution. When it is necessary to generate promotional copy for a target item, the basic information of the target item is first obtained, such as the item title, category, material and other information. Then, based on the basic information of the item, a search is performed in a preset item library. The preset item library is pre-configured with multiple items and matching copy corresponding to each item. A set of candidate items similar to the target item can be retrieved from the preset item library based on the basic information of the target item. Then, the candidate items and their corresponding matching copy are used as samples for small sample learning. Through small sample learning, the copy style corresponding to the candidate item set can be learned, and then promotional copy that conforms to the style of the target item can be automatically generated. This not only improves the diversity of item copy, but also reduces manpower and time costs, thereby improving user experience.

[0078] like Figure 3 As shown, it is an interactive schematic diagram of an item information processing system 300 provided by an embodiment of the present application. Taking the e-commerce scenario as an example, when a user needs to generate a promotional copy for a target product A, the user can log in to the page where the target product A is located in the e-commerce platform through the terminal 220 to make a request, triggering a copy generation instruction for the target product A, and the copy generation instruction is transmitted to the server 210 of the e-commerce platform. The server 210 can retrieve a set of candidate products similar to the target product in the preset product library based on the basic information of the target product A, and input each candidate product in the candidate product set and its corresponding matching copy to the large language model, so that the large language model uses each candidate product and its corresponding matching copy as a sample for small sample learning, learns the copy style corresponding to the candidate product set, and then automatically generates a promotional copy that conforms to the style of the target product A, and returns the promotional copy to the terminal 220 through the server 210, so that the user can view the promotional copy.

[0079] Some embodiments of the present application are described in detail below in conjunction with the accompanying drawings. In the case where there is no conflict between the embodiments, the following embodiments and the features in the embodiments can be combined with each other. In addition, the step sequence in the following method embodiments is only an example and not a strict limitation.

[0080] Please see Figure 4 , which is an item information processing method of an embodiment of the present application, the method can be Figure 1 The electronic device 1 shown is used to perform and can be applied to Figure 2-Figure 3 In the application scenario of item information processing shown in , the matching copy is automatically generated based on the basic information of the item, which not only improves the diversity of the item copy, but also reduces the manpower and time costs and improves the user experience. In this embodiment, the terminal 220 is used as an execution end as an example, and the method includes the following steps:

[0081] Step 401: In response to a copywriting generation instruction for a target item, basic information of the target item is obtained.

[0082] In this step, the target item can be an item selected by the user, such as a product on sale on an e-commerce platform. Taking the e-commerce scenario as an example, when a user needs to generate a promotional copy for a target product A, the user can log in to the page where the target product A is located on the e-commerce platform through terminal 220 to make a request, trigger a copy generation instruction for the target product A, and obtain the basic information of the product A in response to the generation instruction. The basic information here is used to characterize the basic characteristics of the target item. The basic information includes but is not limited to the category, title, and attribute information of the target item, wherein the attribute information includes but is not limited to the material, color, size, and other information of the target item.

[0083] In one embodiment, responding to the copywriting generation instruction for the target item in step 401 may specifically include: generating the copywriting generation instruction for the target item in response to the user's sharing operation on the target item.

[0084] In this embodiment, the user can trigger the copy generation instruction when sharing the target item through the interactive interface. Taking the e-commerce scenario as an example, when the user wants to share the information of product A with other users or other online platforms, the user can enter the sharing operation of product A through the interactive interface, which automatically triggers the copy generation instruction for product A and enters the copy generation process. Improve the interactive experience.

[0085] In one embodiment, step 401 may specifically include: in response to a copywriting generation instruction for a target item, obtaining identification information of the target item entered by a user, and extracting basic information of the target item from a database according to the identification information.

[0086] In this embodiment, the user can enter the identification information of the target item through the interactive interface. The identification information is used to uniquely identify the target item, such as the identity ID of the target item, category keywords, etc. Taking the e-commerce scenario as an example, when the user needs to generate a promotional copy for a target product A, the ID of product A can be entered through the interactive interface. The system searches the product library based on the ID of product A to obtain the basic information of product A, such as obtaining the title, category, and main image of product A from the details page of product A. For the main image of the product, features can also be extracted and classified based on the deep image recognition model to obtain the "material", "color", "category" and other attribute information of product A, and store it in text form. Improve the interactive experience.

[0087] In one embodiment, step 401 may further specifically include: in response to a copywriting generation instruction for a target item, obtaining image information of the target item uploaded by a user, recognizing the image information, and obtaining attribute information of the target item, wherein the basic information includes the attribute information.

[0088] In this embodiment, the user can actively upload the physical image of the target item. For example, the user uses a mobile phone to take a photo of a piece of clothing, obtains the image information of the clothing, and then uploads the image information of the clothing, triggering the copywriting instruction for the clothing. For the real picture of the product, features can be extracted and classified based on the deep image recognition model to obtain the attribute information of the "material", "color", "category" and so on of the clothing, and the attribute information can be stored in text form. Improve the interactive experience.

[0089] Step 402: According to the basic information, a set of candidate items matching the target item is retrieved from a preset item library, wherein the preset document library includes associations between a plurality of preset items and corresponding matching documents.

[0090] In this step, the preset item library pre-configures multiple preset items and matching texts corresponding to each preset item. Matching texts of corresponding styles can be pre-configured for preset items of different styles to form item-text pairs. Assuming that the preset item is a skirt suitable for young ladies, the corresponding matching text may be "This skirt can be tied at the waist, and the hem naturally droops into pleats, which not only modifies the lower body line, but also adds a girlish touch, making you more lovely and charming in summer."

[0091] Taking the e-commerce scenario as an example, multiple preset product copy pairs can be obtained from the product and its corresponding description copy data set of the e-commerce platform to form a preset item library. When the user needs to generate a description copy for the target product A, the preset item library is searched based on the basic information of the product A, and a candidate item set matching the target product A is retrieved from the preset item library. Here, the candidate item set can be a set of products similar to the target product A.

[0092] In one embodiment, step 402 may specifically include: extracting a first feature of the target item based on the basic information, and searching a set of candidate items matching the target item in a preset item library based on the first feature.

[0093] In this embodiment, the first feature refers to the feature information of the target item, which is used to characterize the characteristics of the target item. For example, if the basic information of the target item is stored in the form of text, the first feature extracted based on the basic information can be the text feature of the target item. For the text of the basic information of the target commodity A obtained in step 401, its text feature can be extracted based on the StructBert model as the first feature, and the feature dimension can be 1024. Then, based on the first feature, a set of candidate items matching the target commodity A is accurately retrieved in the preset item library.

[0094] In one embodiment, in step 402, according to the first feature, searching for a set of candidate items matching the target item in the preset item library includes: determining a first similarity between the first feature and a second feature of each preset item. Sorting the plurality of preset items from large to small according to the corresponding first similarities, and determining the plurality of preset items with the first similarities ranked in a first preset order as the set of candidate items matching the target item.

[0095] In this embodiment, the preset item library includes the second features of multiple preset items. The second feature refers to the feature information corresponding to the preset item, which is used to characterize the characteristics of the corresponding preset item. A feature bank corresponding to the preset item can be pre-constructed, and the second features of each preset item can be obtained by pre-extracting text features from the information of multiple preset items. For example, the information of the preset item is stored in the form of text, and the text features of the information of each preset item can be extracted based on the StructBert model as the corresponding second feature. The second features of each preset item form a feature bank. In the process of retrieving the candidate item set, the first similarity can be calculated by comparing the first feature of the target item with the second feature of each preset item in the feature bank. Here, the first similarity can be the cosine similarity between the first feature and the corresponding second feature. Then, the multiple preset items in the preset item library are sorted from large to small according to the first similarity, and the preset items with greater first similarity to the target item are sorted higher, and the multiple preset items sorted in the front preset order are determined as the candidate item set matching the target item, for example, the five preset items sorted in Top5 (top five) are selected as the candidate item set matching the target item. A list of candidate items with high similarity to the target item is achieved based on feature recall.

[0096] In one embodiment, before step 402, the method may further include: obtaining a plurality of preset items and a matching copy corresponding to each preset item, determining the association between the plurality of preset items and the corresponding matching copy, respectively extracting the second features corresponding to the plurality of preset items, and establishing a preset item library.

[0097] In this embodiment, the preset item library can be a database based on Faiss (an open source vector database, which is an efficient similarity search and clustering engine for dense vectors). Before the search in step 402, a feature bank corresponding to the preset items can be pre-constructed in the preset item library, and the second feature of each preset item is obtained by pre-extracting text features from the information of multiple preset items to form a feature bank. The information of the preset items includes but is not limited to the title, category, material, color, size and other information of the preset items.

[0098] Taking the e-commerce scenario as an example, the preset item copy pair can be a preset product data set. For the preset product description copy data set, based on user type classification, related products and matching multi-style copy pairs, it is assumed that about 2 million "product-copy" pairs are formed, and then the text features of the product title in each "product-copy" pair are extracted based on the StructBert model to form a feature bank of the "product-copy" pairs, and the product text feature library can be stored and maintained offline.

[0099] In one embodiment, after step 402 and before step 403, the method may further include: determining a second similarity between the first user information and the second user information of each candidate item, and sorting the candidate items in the candidate item set from large to small according to the second similarity to obtain a final sorting result of the candidate item set.

[0100] In this embodiment, the basic information of the target item includes the first user information applicable to the target item. The preset item library may include the second user information applicable to the candidate item. The user information applicable to the item may be the user category or user characteristics of the user who directly wears or uses the item. For example, if item B is a skirt, the user information applicable to item B may be information such as young ladies. After selecting a set of candidate items similar to the target item, the candidate item set may be further finely sorted, and the second similarity between the first user information applicable to the target item and the second user information applicable to each candidate item is calculated respectively. The greater the second similarity, the greater the similarity between the user group applicable to the target item and the corresponding candidate item, and the greater the style similarity between the target item and the candidate item. The candidate items in the candidate item set may be sorted from large to small according to the second similarity to obtain the final sorting result of the candidate item set, so as to achieve fine sorting of the candidate items according to the applicable user type distribution characteristics and improve the subsequent calculation accuracy.

[0101] Step 403: Perform small sample learning based on the matching text corresponding to each candidate item in the candidate item set to generate a target text matching the target item.

[0102] In this step, candidate items and their corresponding matching texts are used as samples for small sample learning. Through small sample learning, the text style corresponding to the candidate item set can be learned, and then the target text that matches the style of the target item is automatically generated, forming an automatic generation link for item texts, which not only improves the diversity of item texts, but also greatly reduces manpower and time costs. The target text can be used as the promotional text of the target item and shared on different user platforms.

[0103] In one embodiment, step 403 may specifically include: determining multiple candidate items and corresponding matching texts as a small sample set, training a preset large language model, and inputting preset prompt words matching the target item into the preset large language model, so that the large language model outputs a target text matching the target item.

[0104] In this embodiment, the candidate item set includes the association relationship between multiple candidate items and corresponding matching texts. Taking the e-commerce scenario as an example, assuming that the target item is product A, the candidate item set screened from the preset item library in step 402 can be the "product-text" pairs ranked in the top 5. The large language model (LLM) refers to a large deep learning model framework trained with a large amount of text data. It can be used for multiple natural language processing tasks, and can generate natural language text or understand the meaning of language text. Small sample learning, that is, using fewer samples to complete parameter training based on the backbone network, can still have excellent task execution performance. In the small sample learning in the LLM model, the traditional training of small samples in the deep model is no longer performed. By constructing and optimizing Prompt (prompt words), the LLM model can effectively learn small sample knowledge. The top 5 "product-copy" pairs can be used as a small sample set to train the preset large language model, so that the preset large language model can perform small sample learning. For example, based on the small sample set, the few-shot large language model ChatGLM-6B (an open source conversational language model based on the GLM architecture that supports Chinese and English bilingual question and answer) is used to learn text style knowledge. The input prompt words of the LLM model are optimized to achieve the expected copy format and special needs. For example, the prompt words can indicate the knowledge base, the tone type of the copy, the copy typesetting format, the number of words, the copy theme, the content and other information. Through small sample learning, the large language model can learn the copy style of the top 5 "product-copy" pairs, combined with the preset prompt words that match the target product A, and then automatically generate promotional copy that conforms to the style of product A and meets the expectations of the prompt words.

[0105] The above-mentioned item information processing method, when it is necessary to generate promotional copy for the target item, first obtains the basic information of the target item, such as the item title, category, material and other information, and then searches in a preset item library based on the basic information of the item. The preset item library is pre-configured with multiple items and matching copy corresponding to each item. Based on the basic information of the target item, a set of candidate items similar to the target item can be retrieved from the preset item library, and then the candidate items and their corresponding matching copy are used as samples for small sample learning. Through small sample learning, the copy style corresponding to the candidate item set can be learned, and then promotional copy that conforms to the style of the target item can be automatically generated, which not only improves the diversity of item copy but also reduces manpower and time costs, and improves user experience.

[0106] The embodiments of the present application start from two aspects. On the one hand, based on the basic copywriting generation capability of the LLM model, the richness and creativity of the copywriting are controlled by optimizing the prompt. On the other hand, based on the product description copywriting data set, for different types of products, different tones and styles of copywriting type test examples are recalled, and small sample learning is completed for the LLM model to generate product promotion copywriting of corresponding styles, solving the problems of copywriting style limitations and text richness existing in related technologies. By constructing an automated generation link for product promotion copywriting, not only can product promotion copywriting of different styles and text formats be generated, but also the manpower and time costs are reduced, and the ability to be easily implemented in multiple scenarios of product promotion is achieved.

[0107] like Figure 5 As shown, an item information processing method provided by an embodiment of the present application can be Figure 1 The electronic device 1 shown is used to perform and can be applied to Figure 2-Figure 3 In the application scenario of item information processing shown in , the automatic generation of matching copy based on the basic information of the item is realized, which not only improves the diversity of the item copy, but also reduces the manpower and time costs and improves the user experience. This embodiment takes the terminal 220 as an example of the execution end. Compared with the above embodiment, this embodiment takes the generation of marketing copy for goods in the e-commerce scenario as an example. The method includes the following steps:

[0108] Step 501: For each target product that enters the system and needs to generate marketing copy, basic information of the target product is obtained. The basic information may include product title, category, etc., and actual product photos (physical product photos) may be obtained through user upload, database retrieval, etc. For the actual product photos, image recognition can be performed based on a deep image recognition model, and the features of the actual product photos can be extracted and classified to obtain attribute information such as "material", "color", "category", etc. of the target product, and stored in text form.

[0109] Step 502: word embedding is performed on the text of the basic information of the target product obtained in step 501, and its text features are extracted based on the StructBert model. The feature dimension may be 1024. Then, step 503 is entered.

[0110] Step 503: Enter the feature bank based on the text features in step 502, perform feature retrieval in the feature bank based on the text features of the target product, sort the target product from large to small based on the first similarity between the target product and the candidate products, such as based on Faiss retrieval, rough sorting and fine sorting features, recall the list of candidate products with high similarity to the product, select the top 5 candidate products with the first similarity to form a candidate product set, the candidate product set can be presented in a table, and can include the index of each candidate product and the corresponding matching copy, and the feature similarity can be the cosine similarity between features. Then, according to the distribution of the first user type applicable to the target product and the second user type applicable to the candidate product, the top 5 candidate products can be finely sorted to form a final candidate product set.

[0111] Before step 503, step S0 may also be included: constructing a "product-copy" feature vector bank:

[0112] S01: Obtain a product description copy dataset, which contains about 2 million "product-copy" pairs, and associate products with multi-style copy pairs based on user type classification.

[0113] S02: Perform word embedding injection processing based on the product description copy dataset, and extract the product title text features of each "product-copy" pair in about 2 million "product-copy" pairs based on StructBert to form a preset product text feature library, namely the feature bank, which is stored and maintained offline.

[0114] Step 504: The final candidate product set obtained in step 503 is used as a small sample set, and the LLM model is trained based on the small sample set for small sample learning. Specifically, the input prompt words of the LLM model are optimized to achieve the format and special requirements of the expected copy. Based on the small sample set of the copy, the Few-shot large language model ChatGLM-6B is trained to learn text style knowledge and generate marketing copy that matches the target product.

[0115] like Figure 6 As shown in FIG. 1 , a schematic diagram of a specific application scenario process of an embodiment of the present application is shown. In this embodiment, a skirt in an e-commerce scenario is used as a target product as an example. The physical image of the target product for which marketing copywriting needs to be generated is obtained through user upload or database retrieval. Figure 6For an image named "Umbrella Skirt Women", image recognition can be performed on the image to obtain the following product attributes of the skirt: "Pattern - solid color", "Silhouette - A-type", "Waist type - high waist", "Skirt length - long skirt", "Applicable people - youth / ladies". Then, feature retrieval can be performed based on the product attributes to filter out a set of candidate products. Suppose the "Product - Copy" pair of a candidate product 1 is as follows: "Title: Miss Chipmunk White Skirt Women's Summer Medium-length Slim Lace-up Umbrella Skirt Small Fresh Skirt.

[0116] Copywriting: This skirt makes the waist look very slim and hides the belly. The A-shaped skirt modifies the lower body line. The skirt naturally hangs down into a small wave shape, with a sense of cuteness, and a little playfulness in elegance. "

[0117] The "product-copy" pair content of candidate product 1 is used as a small sample to train the LLM model, conduct small sample learning, and realize copy style transfer. Combined with Prompt optimization, the LLM model can generate marketing copy that matches the "Umbrella Skirt Girl" skirt, as follows:

[0118] "This skirt is simple yet elegant. The upper body is A-line, with a belt around the waist to tighten the waist. The skirt naturally droops into pleats, which not only modifies the lower body line but also adds a girlish touch, making you more lovely and charming in summer."

[0119] You can optimize the prompt and configure the text format, such as adding icons and other content at the end of a sentence to improve the affinity of the text.

[0120] In the embodiment of the present application, taking the e-commerce scenario as an example, based on the strong text generation capability of the ChatGLM-6B model, the basic format and highly relevant vocabulary of the copy are controlled by optimizing the prompt. For each product input prompt, the product title, product label, and attributes are cleaned, and the material, shape and other features of the product are image-recognized to improve the knowledge accuracy of text generation. According to the product attribute characteristics, suitable style copy samples are retrieved to learn the ChatGLM model with a small sample to generate marketing copy with different text and tone styles. It solves the marketing text generation problems such as copy style limitations and text richness existing in the relevant technology.

[0121] In order to improve the richness of copywriting and product matching, an end-to-end product copywriting generation system is built. Once the target product is known, product information extraction, feature retrieval, and training and reasoning of the generation model are carried out, including image recognition, text embedding, feature extraction, vector similarity matching, vector retrieval and sorting, text-to-text model generation, etc., and finally the corresponding product marketing copywriting is generated. In addition, the output copywriting can be entered into the algorithm and manual review process respectively to ensure the accuracy and reliability of the final copywriting generation.

[0122] Please see Figure 7 , which is a commodity information processing method of an embodiment of the present application, the method can be Figure 1 The electronic device 1 shown is used to perform and can be applied to Figure 2-Figure 3 In the application scenario of item information processing shown in , the matching copy is automatically generated based on the basic information of the item, which not only improves the diversity of the item copy, but also reduces the manpower and time costs and improves the user experience. This embodiment takes the terminal 220 as an example of the execution end. Compared with the above embodiment, this embodiment takes the generation of matching copy for items in the e-commerce scenario as an example. The method includes the following steps:

[0123] Step 701: In response to a copywriting generation instruction for a target product, basic information of the target product is obtained.

[0124] Step 702: According to the basic information, a set of candidate commodities matching the target commodity is retrieved from the preset commodity library, wherein the preset copy library includes associations between a plurality of preset commodities and corresponding matching copies.

[0125] Step 703: Perform small sample learning based on the matching copy corresponding to each candidate product in the candidate product set to generate a target copy that matches the target product.

[0126] For details of each step of the above method, please refer to the relevant description of the above-mentioned embodiments, which will not be repeated here.

[0127] Please see Figure 8 , which is an item information processing device 800 of an embodiment of the present application, the device can be applied to Figure 1 The electronic device 1 shown can be applied to Figure 2-Figure 3 In the application scenario of item information processing shown in , the matching copy is automatically generated based on the basic information of the item, which not only improves the diversity of the item copy, but also reduces the manpower and time costs and improves the user experience. The device includes: an acquisition module 801, a retrieval module 803 and a generation module 803, and the functional principles of each module are as follows:

[0128] The acquisition module 801 is used to acquire basic information of the target item in response to a copywriting generation instruction for the target item.

[0129] The search module 803 is used to search for a set of candidate items matching the target item in the preset item library according to the basic information, wherein the preset document library includes associations between a plurality of preset items and corresponding matching documents.

[0130] The generation module 803 is used to perform small sample learning based on the matching text corresponding to each candidate item in the candidate item set to generate a target text matching the target item.

[0131] In one embodiment, the acquisition module 801 is used to respond to a copywriting generation instruction for a target item, acquire identification information of the target item entered by a user, and extract basic information of the target item from a database according to the identification information.

[0132] In one embodiment, the acquisition module 801 is used to respond to the copywriting generation instruction for the target object, acquire the image information of the target object uploaded by the user, identify the image information, and obtain the attribute information of the target object, and the basic information includes the attribute information.

[0133] In one embodiment, the retrieval module 803 is used to extract a first feature of the target item based on the basic information and retrieve a set of candidate items matching the target item from a preset item library based on the first feature.

[0134] In one embodiment, the preset item library includes second features of multiple preset items. The retrieval module 803 is used to determine the first similarity between the first feature and the second feature of each preset item. The multiple preset items are sorted from large to small according to the corresponding first similarities, and the multiple preset items with the first similarities sorted in a front preset order are determined as a candidate item set matching the target item.

[0135] In one embodiment, the basic information includes first user information applicable to the target item. The preset item library includes second user information applicable to the candidate items. The device also includes: a sorting module, which is used to perform small sample learning based on the matching copy corresponding to each candidate item in the candidate item set, and before generating a target copy matching the target item, determine the second similarity between the first user information and the second user information of each candidate item. Sort each candidate item in the candidate item set from large to small according to the second similarity to obtain a final sorting result of the candidate item set.

[0136] In one embodiment, the candidate item set includes associations between multiple candidate items and corresponding matching texts. The generation module 803 is used to determine the multiple candidate items and the corresponding matching texts as a small sample set, train a preset large language model, and input the preset prompt words matching the target item into the preset large language model, so that the large language model outputs a target text matching the target item.

[0137] In one embodiment, it further includes: a building module, which is used to obtain a plurality of preset items and matching documents corresponding to each preset item before searching for a set of candidate items matching the target item in the preset item library according to the basic information, and determine the association relationship between the plurality of preset items and the corresponding matching documents. The second features corresponding to the plurality of preset items are respectively extracted to build a preset item library.

[0138] In one embodiment, the device further includes: a response module, configured to generate a text generation instruction for the target item in response to a user's sharing operation on the target item.

[0139] For a detailed description of the above-mentioned item information processing device 800, please refer to the description of the relevant method steps in the above-mentioned embodiment. Its implementation principle and technical effect are similar, and will not be repeated here in this embodiment.

[0140] Fig. 9 The following is a schematic diagram of the structure of a cloud device 90 provided in an exemplary embodiment of the present application. The cloud device 90 can be used to run the method provided in any of the above embodiments. Fig. 9 As shown, the cloud device 90 may include: a memory 904 and at least one processor 905, Fig. 9 A processor is taken as an example.

[0141] The memory 904 is used to store computer programs and can be configured to store various other data to support operations on the cloud device 90. The memory 904 can be an object storage service (OSS).

[0142] Memory 904 can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read-only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic storage, flash memory, magnetic disk or optical disk.

[0143] The processor 905 is coupled to the memory 904 and is used to execute the computer program in the memory 904 to implement the solution provided by any of the above method embodiments. The specific functions and technical effects that can be achieved are not repeated here.

[0144] Furthermore, if Fig. 9 The cloud device also includes: a firewall 901, a load balancer 902, a communication component 906, a power supply component 903 and other components. Fig. 9 Only some components are shown schematically, which does not mean that the cloud device only includes Fig. 9 Components shown.

[0145] In one embodiment, the above Fig. 9 The communication component 906 in is configured to facilitate wired or wireless communication between the device where the communication component 906 is located and other devices. The device where the communication component 906 is located can access a wireless network based on a communication standard, such as WiFi, 2G, 3G, 4G, LTE (Long Term Evolution, Long Term Evolution, referred to as LTE), 5G and other mobile communication networks, or a combination thereof. In an exemplary embodiment, the communication component 906 receives a broadcast signal or broadcast-related information from an external broadcast management system via a broadcast channel. In an exemplary embodiment, the communication component 906 also includes a near field communication (Near Field Communication, referred to as NFC) module to facilitate short-range communication. For example, the NFC module can be based on Radio Frequency Identification (Radio Frequency Identification, referred to as RFID) technology, Infrared Data Association (Infrared Data Association, referred to as IrDA) technology, Ultra Wide Band (UWB) technology, Bluetooth (BT) technology and other technologies.

[0146] In one embodiment, the above Fig. 9 The power supply component 903 provides power to various components of the device where the power supply component 903 is located. The power supply component 903 may include a power management system, one or more power supplies, and other components associated with generating, managing, and distributing power to the device where the power supply component is located.

[0147] An embodiment of the present application further provides a computer-readable storage medium, in which computer-executable instructions are stored. When a processor executes the computer-executable instructions, the method of any of the aforementioned embodiments is implemented.

[0148] An embodiment of the present application also provides a computer program product, including a computer program, which implements the method of any of the aforementioned embodiments when executed by a processor.

[0149] In the several embodiments provided in this application, it should be understood that the disclosed devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic, for example, the division of modules is only a logical function division, and there may be other division methods in actual implementation, such as multiple modules can be combined or integrated into another system, or some features can be ignored or not executed.

[0150] The above-mentioned integrated module implemented in the form of a software function module can be stored in a computer-readable storage medium. The above-mentioned software function module is stored in a storage medium, including a number of instructions for enabling a computer device (which can be a personal computer, a server, or a network device, etc.) or a processor to perform some steps of the methods of various embodiments of the present application.

[0151] It should be understood that the above-mentioned processor can be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), etc. The general-purpose processor can be a microprocessor or the processor can also be any conventional processor, etc. The steps of the method disclosed in the application can be directly embodied as being executed by a hardware processor, or can be executed by a combination of hardware and software modules in the processor. The memory may include a high-speed RAM (Random Access Memory) memory, and may also include non-volatile storage NVM (Nonvolatile memory, NVM for short), such as at least one disk storage, and can also be a USB flash drive, a mobile hard disk, a read-only memory, a disk or an optical disk, etc.

[0152] The above storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as static random access memory (SRAM), electrically erasable programmable read only memory (EEPROM), erasable programmable read-only memory (EPROM), programmable read-only memory (PROM), read-only memory (ROM), magnetic memory, flash memory, magnetic disk or optical disk. The storage medium can be any available medium that can be accessed by a general or special computer.

[0153] An exemplary storage medium is coupled to a processor so that the processor can read information from the storage medium and write information to the storage medium. Of course, the storage medium can also be a component of the processor. The processor and the storage medium can be located in an application specific integrated circuit (ASIC). Of course, the processor and the storage medium can also exist as discrete components in an electronic device or a main control device.

[0154] It should be noted that, in this article, the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, an element defined by the sentence "comprises a ..." does not exclude the existence of other identical elements in the process, method, article or device including the element.

[0155] The serial numbers of the above-mentioned embodiments of the present application are for description only and do not represent the advantages or disadvantages of the embodiments.

[0156] Through the description of the above implementation methods, those skilled in the art can clearly understand that the above-mentioned embodiment methods can be implemented by means of software plus a necessary general hardware platform, and of course by hardware, but in many cases the former is a better implementation method. Based on such an understanding, the technical solution of the present application, or the part that contributes to the prior art, can be embodied in the form of a software product, which is stored in a storage medium (such as ROM / RAM, a disk, or an optical disk), and includes a number of instructions for a terminal device (which can be a mobile phone, a computer, a server, an air conditioner, or a network device, etc.) to execute the methods of each embodiment of the present application.

[0157] In the technical solution of this application, the collection, storage, use, processing, transmission, provision and disclosure of user data and other information involved are in compliance with the provisions of relevant laws and regulations and do not violate public order and good morals.

[0158] The above are only preferred embodiments of the present application, and are not intended to limit the patent scope of the present application. Any equivalent structure or equivalent process transformation made using the contents of the present application specification and drawings, or directly or indirectly applied in other related technical fields, are also included in the patent protection scope of the present application.

Claims

1. A method for processing item information, characterized in that: The method comprises: In response to a copywriting generation instruction for a target item, obtaining basic information of the target item; According to the basic information, a set of candidate items matching the target item is retrieved from a preset item library, wherein the preset document library includes associations between a plurality of preset items and corresponding matching documents; Small sample learning is performed based on the matching text corresponding to each candidate item in the candidate item set to generate a target text matching the target item.

2. The method according to claim 1, characterized in that The step of obtaining basic information of the target item in response to the copywriting generation instruction for the target item includes: In response to a copywriting generation instruction for a target item, identification information of the target item entered by a user is obtained, and basic information of the target item is extracted from a database according to the identification information.

3. The method according to claim 1, characterized in that The step of obtaining basic information of the target item in response to the copywriting generation instruction for the target item includes: In response to a copywriting generation instruction for a target object, acquiring image information of the target object uploaded by a user; The image information is identified to obtain attribute information of the target object, wherein the basic information includes the attribute information.

4. The method according to claim 1, characterized in that: The step of retrieving a set of candidate items matching the target item from a preset item library according to the basic information includes: Extracting a first feature of the target object according to the basic information; According to the first feature, a set of candidate items matching the target item is retrieved from a preset item library.

5. The method according to claim 4, characterized in that The preset item library includes the second features of the plurality of preset items; and searching the preset item library for a set of candidate items matching the target item according to the first features includes: Determining a first similarity between the first feature and a second feature of each of the preset objects; The plurality of preset objects are sorted from largest to smallest according to the corresponding first similarities, and the plurality of preset objects whose first similarities are sorted in a front preset order are determined as a set of candidate objects matching the target object.

6. The method according to claim 1 or 5, characterized in that: The basic information includes first user information applicable to the target item; the preset item library includes second user information applicable to the candidate item; before performing small sample learning based on the matching copy corresponding to each candidate item in the candidate item set to generate a target copy matching the target item, it also includes: determining a second similarity between the first user information and the second user information of each candidate item; The candidate items in the candidate item set are sorted from large to small according to the second similarity to obtain a final sorting result of the candidate item set.

7. The method according to claim 1, characterized in that The candidate item set includes an association relationship between a plurality of candidate items and corresponding matching texts; performing small sample learning according to the matching texts corresponding to each candidate item in the candidate item set to generate a target text matching the target item includes: The multiple candidate items and corresponding matching texts are determined as a small sample set, a preset large language model is trained, and a preset prompt word matching the target item is input into the preset large language model, so that the large language model outputs the target text matching the target item.

8. The method according to claim 1, characterized in that Before searching for a candidate item set matching the target item in a preset item library according to the basic information, the method further includes: Acquire a plurality of preset items and a matching copy corresponding to each of the preset items, and determine an association relationship between the plurality of preset items and the corresponding matching copy; The second features corresponding to the plurality of preset items are respectively extracted to establish the preset item library.

9. The method according to claim 1, characterized in that: The response to the copywriting generation instruction for the target item includes: In response to a user's sharing operation on a target item, a copywriting generation instruction for the target item is generated.

10. A commodity information processing method, characterized in that: The method comprises: In response to a copywriting generation instruction for a target product, obtaining basic information of the target product; According to the basic information, a set of candidate commodities matching the target commodity is retrieved from a preset commodity library, wherein the preset copy library includes associations between a plurality of preset commodities and corresponding matching copies; Small sample learning is performed based on the matching copy corresponding to each candidate product in the candidate product set to generate a target copy matching the target product.

11. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively coupled to the at least one processor; The memory stores instructions that can be executed by the at least one processor, and the instructions are executed by the at least one processor to enable the electronic device to perform the method described in any one of claims 1 to 10.

12. A cloud device, characterized in that: include: at least one processor; and a memory communicatively coupled to the at least one processor; The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the cloud device to execute the method described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer-executable instructions, and when the processor executes the computer-executable instructions, the method according to any one of claims 1 to 10 is implemented.

Images