Reverse article matching method and device

Abstract

The invention discloses a reverse article matching method and device, and relates to the technical field of artificial intelligence. A specific embodiment of the method comprises the following steps: executing data enhancement on initial attribute data of a standard reverse article in a standard data set according to a preset basic data set to obtain enhanced attribute data of the standard reverse article, determining feature data of the standard reverse article from the enhanced attribute data by using a pre-trained feature extraction model; obtaining manually marked attribute data of a to-be-processed reverse article, and determining feature data of the to-be-processed reverse article from the manually marked attribute data by using the feature extraction model; and determining the similarity between the feature data of the to-be-processed reverse article and the feature data of each standard reverse article, and determining the standard reverse article corresponding to the maximum similarity as a matched article of the to-be-processed reverse article. According to the embodiment, accurate matching of the to-be-processed reverse article and the standard reverse article can be realized.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to a reverse object matching method and device. Background Art

[0002] For service-side platforms that handle online business, items that flow from the warehouse to the user (forward direction) based on user orders in a forward supply chain scenario are generally referred to as forward items, while items that flow in the opposite direction (the opposite direction of the "forward direction" above, for example, from the user to the warehouse) through channels such as returns and exchanges are referred to as reverse items. For example, reverse items can include items that pass through multiple links from the user's hands to the service-side warehouse. They can also include items that flow from the user to a pickup point or distribution station (including items that have arrived at the pickup point or distribution station and items in transit), items that flow from the pickup point or distribution station to a sorting center (including items that have arrived at the sorting center and items in transit), and items that flow from a sorting center to the warehouse (including items that have arrived at the warehouse and items in transit). Reverse items can be caused by user-initiated returns or exchanges, or by item returns initiated by the service side (including warehouses, sorting centers, distribution stations, etc.) (for example, items that enter the sorting center in the forward direction from the warehouse and are then returned, and then flow from the sorting center to the warehouse in the reverse direction), or other reasons.

[0003] Since reverse items initially lack standard data such as item names and categories, the server-side platform needs to manually mark the reverse items (hereinafter referred to as reverse items to be processed) after obtaining them. However, these manually marked data are not standard data and are quite arbitrary and non-standard. As a result, it is impossible to accurately match the reverse items to be processed with the standard reverse items in the annotated data set that have undergone data standardization through other channels through these manually marked data. As a result, it is impossible to use the historical data of the standard reverse items to provide reasonable guidance for the subsequent business links of the reverse items to be processed. Summary of the Invention

[0004] In view of this, embodiments of the present invention provide a reverse article matching method and apparatus, which can achieve accurate matching between a reverse article to be processed and a standard reverse article.

[0005] To achieve the above objective, according to one aspect of the present invention, a reverse item matching method is provided.

[0006] The reverse article matching method of an embodiment of the present invention includes: performing data enhancement on initial attribute data of a standard reverse article in a standard data set based on a preset basic data set to obtain enhanced attribute data of the standard reverse article, and determining feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model; obtaining manually labeled attribute data of a reverse article to be processed, and determining feature data of the reverse article to be processed from the manually labeled attribute data using the feature extraction model; determining similarity between the feature data of the reverse article to be processed and feature data of each standard reverse article, and determining the standard reverse article corresponding to the maximum similarity as a matching article for the reverse article to be processed.

[0007] Optionally, the initial attribute data of any standard reverse article includes the item name of the standard reverse article, and the basic data set includes the item name and multi-level category data of each standard reverse article; and performing data enhancement on the initial attribute data of the standard reverse article in the standard data set based on the preset basic data set includes: using the item name of any standard reverse article in the standard data set to query the basic data set to obtain multi-level category data of the standard reverse article; and combining the queried multi-level category data with the initial attribute data of the standard reverse article to obtain enhanced attribute data of the standard reverse article.

[0008] Optionally, the method of determining the feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model includes: inputting the category data of each level in the enhanced attribute data of any standard reverse article into the feature extraction model respectively to obtain multiple category features of the standard reverse article; identifying at least one named entity of a preset entity type based on the name of the standard reverse article, and concatenating the identified named entity with the corresponding entity type into the named entity feature of the standard reverse article; and determining the weighted sum of the multiple category features and the named entity feature as the feature data of the standard reverse article.

[0009] Optionally, the manually marked attribute data includes manually marked multi-level category data and manually marked item names; and the use of the feature extraction model to determine the feature data of the reverse item to be processed from the manually marked attribute data includes: inputting the category data of each level in the manually marked attribute data of the reverse item to be processed into the feature extraction model respectively to obtain multiple category features of the reverse item to be processed; identifying the item name of the reverse item to be processed based on at least one named entity of the entity type, and splicing the identified named entity with the corresponding entity type into the named entity feature of the reverse item to be processed; and determining the weighted sum of the category features and the named entity features of the reverse item to be processed as the feature data of the reverse item to be processed.

[0010] Optionally, the standard data set includes historical supply quantity and historical value data of any standard reverse item; and the method further includes: establishing a quantity-value relationship model of the matching item based on the historical supply quantity and historical value data of the matching item of the reverse item to be processed; and using the quantity-value relationship model to determine the value data of the reverse item to be processed.

[0011] To achieve the above objective, according to another aspect of the present invention, a reverse object matching device is provided.

[0012] The reverse article matching device of an embodiment of the present invention may include: a first feature extraction unit, configured to perform data enhancement on initial attribute data of a standard reverse article in a standard data set based on a preset basic data set to obtain enhanced attribute data of the standard reverse article, and determine feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model; a second feature extraction unit, configured to obtain manually labeled attribute data of a reverse article to be processed, and determine feature data of the reverse article to be processed from the manually labeled attribute data using the feature extraction model; and a matching unit, configured to determine a similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article, and determine the standard reverse article corresponding to the maximum similarity as a matching article of the reverse article to be processed.

[0013] Optionally, the initial attribute data of any standard reverse article includes the item name of the standard reverse article, and the basic data set includes the item name and multi-level category data of each standard reverse article; and the first feature extraction unit is further used to: use the item name of any standard reverse article in the standard data set to query the basic data set to obtain the multi-level category data of the standard reverse article; and combine the queried multi-level category data with the initial attribute data of the standard reverse article to obtain enhanced attribute data of the standard reverse article.

[0014] Optionally, the manually marked attribute data includes manually marked multi-level category data and manually marked item names; and the first feature extraction unit is further used to: input the category data of each level in the enhanced attribute data of any standard reverse item into the feature extraction model respectively to obtain multiple category features of the standard reverse item; identify the item name of the standard reverse item based on at least one named entity of a preset entity type, and splice the identified named entity with the corresponding entity type into the named entity feature of the standard reverse item; and calculate the weighted sum of the multiple category features and the named entity features. The method further comprises: determining feature data of the standard reverse article; the second feature extraction unit is further used to: input the category data of each level in the manually marked attribute data of the reverse article to be processed into the feature extraction model respectively to obtain multiple category features of the reverse article to be processed; identifying the item name of the reverse article to be processed based on at least one named entity of the entity type, splicing the identified named entity with the corresponding entity type into the named entity feature of the reverse article to be processed; and determining the weighted sum of the category features and the named entity features of the reverse article to be processed as the feature data of the reverse article to be processed.

[0015] Optionally, the standard data set includes historical supply quantity and historical value data of any standard reverse item; and the device further includes: a quantity and price guidance unit, which is used to establish a quantity-value relationship model of the matching item based on the historical supply quantity and historical value data of the matching item of the reverse item to be processed, and use the quantity-value relationship model to determine the value data of the reverse item to be processed.

[0016] To achieve the above objective, according to another aspect of the present invention, an electronic device is provided.

[0017] An electronic device of the present invention includes: one or more processors; a storage device for storing one or more programs, and when the one or more programs are executed by the one or more processors, the one or more processors implement the reverse item matching method provided by the present invention.

[0018] To achieve the above objective, according to another aspect of the present invention, a computer-readable storage medium is provided.

[0019] A computer-readable storage medium of the present invention stores a computer program, which, when executed by a processor, implements the reverse article matching method provided by the present invention.

[0020] According to the technical solution of the present invention, the embodiments of the above invention have the following advantages or beneficial effects:

[0021] By performing data enhancement and feature extraction based on a machine learning model on the initial attribute data of each standard reverse item in the standard data set, feature data of the covered item names and multi-level category information of each standard reverse item are obtained; by performing feature extraction based on a machine learning model on the manually labeled data of the reverse items to be processed, feature data of the covered item names and multi-level category information of the reverse items to be processed are obtained, and then the similarity of the feature data between the reverse items to be processed and each standard reverse item is calculated. The standard reverse item that accurately matches the reverse item to be processed can be determined based on the maximum similarity. Subsequently, business guidance can be provided to the reverse items to be processed based on the quantity-value relationship model of the matching items, thereby solving the problem of difficulty in accurately matching the reverse items to be processed with the standard reverse items due to the non-standardization and semantic ambiguity of the manually labeled data.

[0022] The further effects of the above-mentioned non-conventional optional manner will be described below in conjunction with specific embodiments. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] The accompanying drawings are provided for a better understanding of the present invention and are not intended to limit the present invention.

[0024] Figure 1 1 is a schematic diagram of the main steps of the reverse item matching method according to an embodiment of the present invention;

[0025] Figure 2 2 is a schematic diagram of a technical implementation of a reverse item matching method according to an embodiment of the present invention;

[0026] Figure 3 Schematic diagram of components of a reverse object matching device according to an embodiment of the present invention;

[0027] Figure 4 is an exemplary system architecture diagram in which embodiments of the present invention may be applied;

[0028] Figure 5 Schematic diagram of the structure of an electronic device used to implement the reverse object matching method in an embodiment of the present invention. DETAILED DESCRIPTION

[0029] The following description of exemplary embodiments of the present invention is made in conjunction with the accompanying drawings, in which various details of the embodiments of the present invention are included to facilitate understanding. These details should be considered as merely exemplary. Therefore, it should be appreciated by those skilled in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for the sake of clarity and conciseness, descriptions of well-known functions and structures are omitted in the following description.

[0030] The following first describes the technical concept of the present invention. At present, for the server-side platform that undertakes online business, the items that flow from the warehouse to the user (forward direction) based on the user order in the forward supply chain scenario are generally called forward items, and the items that flow in the opposite direction (the opposite direction of the above "forward", such as from the user to the warehouse) through channels such as returns and exchanges are called reverse items. Exemplarily, reverse items may include: items that go from the hands of users through multiple links to the server-side warehouse, and may also include: items that flow from users to self-pickup points or distribution sites (including items that have arrived at self-pickup points or distribution sites and items in transit), items that flow from self-pickup points or distribution sites to sorting centers (including items that have arrived at sorting centers and items in transit), and items that flow from sorting centers to warehouses (including items that have arrived at warehouses and items in transit). The reason for the reverse item may be a return or exchange initiated by the user, or it may be an item return initiated by the service end (including warehouses, sorting centers, distribution stations, etc.) (for example, the item enters the sorting center in the forward direction from the warehouse and is returned, and then flows back from the sorting center to the warehouse in the reverse direction), or it may be other reasons.

[0031] After the server platform obtains a reverse commodity, it can be used as a reverse item to be processed. Since the reverse item to be processed may lack standard data such as the item name and category, the server platform needs to manually mark the relevant attribute data of the reverse item to be processed (hereinafter referred to as manually marked attribute data). For example, the manually marked attribute data may include: the item name, the various categories to which the item belongs, the state of the item, the item code, etc. Of course, manually marked attribute data is not standard data and is highly arbitrary. Subsequently, it is necessary to use the manually marked attribute data to help the reverse item to be processed to execute subsequent business processes. For example, it is necessary to determine the value data (such as price) of the reverse item to be processed for subsequent transactions. This requires using the manually marked attribute data to match the reverse item to reverse items that have completed business processes (e.g., completed transactions) and have been archived (hereinafter referred to as standard reverse items). The historical data of these reverse items can be used to support subsequent business operations for the reverse item to be processed. However, due to the arbitrary and non-standard nature of the manually marked attribute data, the accuracy of this matching is low and it cannot support the actual business progress.

[0032] To address the above issues, the present invention provides a method for accurately matching reverse items to be processed with standard reverse items based on manually labeled attribute data. In practical applications, the relevant data of standard reverse items (hereinafter referred to as initial attribute data) is stored in a standard dataset. This dataset can be any suitable data storage method, such as a database, database table, or file. The standard dataset is primarily used to support current business operations by using historical transaction information (such as daily prices and sales). It is generally compiled from order table data based on user orders. This data can include various positive and reverse items traded within a historical period. The relevant data in the standard dataset is reliable standard data.

[0033] For standard reverse items, initial attribute data may include: item name, supply quantity (such as sales volume), value data, and order identifiers. Since this initial attribute data in the standard dataset often comes from the order table, it lacks basic information such as the item's multi-level category and SKU (Stock Keeping Unit). This basic information is generally stored in the basic dataset. The basic dataset can store relatively comprehensive information such as the item name, SKU, multi-level category, and specifications for each item.

[0034] In the technical solution of the present invention, it is necessary to perform preprocessing of relevant data on the standard reverse items and the reverse items to be processed from two aspects, and use a pre-trained feature extraction model based on a machine learning algorithm to extract feature data that contains complete syntactic and semantic information and is suitable for performing similarity calculations, thereby improving the accuracy of item matching.

[0035] For standard reverse items, since the data in the standard dataset is incomplete, the technical solution of the present invention first uses the basic dataset to enhance the initial attribute data of the standard reverse items in the standard dataset to obtain enhanced attribute data for the standard reverse items. The data enhancement method can be determined according to actual needs, for example, by supplementing the initial attribute data with multi-level category data for the same standard reverse item. Subsequently, a pre-trained feature extraction model based on a machine learning algorithm can be used to extract feature data for the standard reverse items from the enhanced attribute data containing the item name and multi-level category data, which serves as the data basis for item matching. This feature extraction model can be a well-known model in the field of natural language processing, such as BERT, TRANSFORMER, or ELMO. Preferably, since item names have a certain degree of semantic ambiguity, NER (Named Entity Recognition) can be performed on the item names before inputting the enhanced attribute data into the feature extraction model to improve data quality. It should be understood that the above NER step is only an optional step.

[0036] Regarding feature extraction, the traditional solution is to concatenate related features and input them into the model together, obtaining the final features output by the model in one go. This method is relatively simple, but the model lacks hierarchy during calculation, resulting in poor feature extraction results. It also makes it impossible to perform certain manual interventions, such as distinguishing the importance of different fields. In view of the above problems, the embodiment of the present invention inputs the category data of each level of the standard reverse item into the model for calculation, and inputs the item name (or NER result) into the model for calculation. Finally, based on the manually preset weight values, the various calculation results of the model are integrated to obtain the final feature data of the standard reverse item, thereby solving the aforementioned problems of the traditional method.

[0037] For the reverse items to be processed, their manually marked attribute data contains key information such as item names and multi-level category data. At this time, a feature extraction method can be performed accordingly, in which model calculations are performed on each data separately and then the data is integrated. The item names can also be subjected to NER before being input into the model. Of course, the traditional feature extraction method of model calculation after data splicing can also be used, or NER can be omitted.

[0038] Finally, matching can be performed based on the characteristic data of the reverse items to be processed and the standard reverse items obtained through the above steps, thereby achieving a more accurate matching effect, which helps to provide guidance for subsequent business links of the reverse items to be processed based on the historical transaction records of the standard reverse items.

[0039] It should be pointed out that, in the absence of conflict, the embodiments of the present invention and the technical features therein may be combined with each other.

[0040] Figure 1 1 is a schematic diagram of the main steps of the reverse item matching method according to an embodiment of the present invention. Figure 2 Schematic diagram of the technical implementation of the reverse item matching method in an embodiment of the present invention. The above method can be executed by a specific server on the server platform.

[0041] like Figure 1 As shown, the reverse item matching method according to the embodiment of the present invention can be specifically performed according to the following steps:

[0042] Step S101: The server performs data enhancement on the initial attribute data of the standard reverse article in the standard data set according to the preset basic data set to obtain enhanced attribute data of the standard reverse article, and uses a pre-trained feature extraction model to determine the feature data of the standard reverse article from the enhanced attribute data.

[0043] Because the initial attribute data for each standard reverse item in the standard dataset is incomplete, in this step, the server can utilize the basic dataset containing the complete basic information of the item to perform data enhancement on the initial attribute data of the standard reverse item. For example, this can include the multi-level category data to which the standard reverse item belongs in the basic dataset into the initial attribute data. It will be appreciated that the method of data enhancement can be determined based on actual needs, and any other applicable fields can also be added to the initial attribute data as needed. If multi-level category data is required, the server first uses the item name of any standard reverse item in the standard dataset to query the basic dataset to obtain the multi-level category data for that standard reverse item. The server then combines the retrieved multi-level category data with the initial attribute data of that standard reverse item to generate the enhanced attribute data for that standard reverse item.

[0044] After obtaining the enhanced attribute data of the standard reverse item, the server can concatenate the various data contained in the enhanced attribute data and input the concatenated data into a feature extraction model to obtain the feature data of the standard reverse item. More preferably, the server can input the category data of each level in the enhanced attribute data of any standard reverse item into the feature extraction model to obtain multiple category features for that standard reverse item. For example, if a standard reverse item has a primary category, a secondary category, and a tertiary category, the primary category, the secondary category, and the tertiary category can be input into the feature extraction model to obtain three category features. Alternatively, the server can identify at least one named entity based on a preset entity type in the name of the standard reverse item and concatenate the identified named entity with the corresponding entity type to form the named entity feature of the standard reverse item. These entity types can be preset based on actual scenarios, such as brand, specification, ingredient, etc. The NER process can utilize applicable methods such as LAC (Lexical Analysis of Chinese). Finally, the server determines the weighted sum of multiple category features and named entity features as the feature data of this standard reverse item. The weights used to calculate the weighted sum can be flexibly set manually according to actual conditions to distinguish the different importance of different fields.

[0045] Step S102: The server obtains the manually marked attribute data of the reverse article to be processed, and uses a feature extraction model to determine the feature data of the reverse article to be processed from the manually marked attribute data.

[0046] This step can be performed before or after step S101, or the two steps can be performed simultaneously. In this step, the manual marking of the attribute data of the reverse article to be processed can be performed in a manner similar to step S101, or in a different manner.

[0047] As a preferred solution, the server can, on the one hand, input the category data at each level from the manually labeled attribute data of the reverse article to be processed into a feature extraction model to obtain multiple category features of the reverse article to be processed. On the other hand, the server can identify at least one named entity of the aforementioned entity types in the name of the reverse article to be processed and concatenate the identified named entity with the corresponding entity type to form the named entity features of the reverse article to be processed. Finally, the server determines the feature data of the reverse article to be processed as the weighted sum of the category features and the named entity features.

[0048] Step S103: The server determines the similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article, and determines the standard reverse article corresponding to the maximum similarity as a matching article of the reverse article to be processed.

[0049] In this step, the server calculates the similarity between the reverse object to be processed and each standard reverse object based on the feature data of the reverse object to be processed and the features of the standard reverse objects. The above calculation can be implemented using applicable methods such as cosine similarity, Pearson correlation coefficient, and Euclidean distance. Finally, the server determines the standard reverse object corresponding to the maximum similarity as the matching object of the reverse object to be processed.

[0050] After determining the matching items for the reverse item to be processed, the historical data of the matching items in the standard dataset can be used to assist in determining subsequent business parameters for the reverse item to be processed. For example, the standard dataset may include the historical supply quantity (e.g., sales volume) and historical value data (e.g., price) of any standard reverse item in each statistical period (e.g., one day). In this case, the server can establish a quantity-value relationship model for the matching items based on the historical supply quantity and historical value data of the matching items for the reverse item to be processed, i.e., a mathematical model of how sales volume changes with price. The server can then use this quantity-value relationship model in combination with relevant data in the manually labeled attribute data, such as the degree of recency, to determine the value data of the reverse item to be processed.

[0051] In the technical solution of the embodiment of the present invention, data enhancement and feature extraction based on a machine learning model are performed on the initial attribute data of each standard reverse item in the standard data set to obtain feature data of the covered item name and multi-level category information of each standard reverse item; feature extraction based on a machine learning model is performed on the manually labeled data of the reverse item to be processed to obtain feature data of the covered item name and multi-level category information of the reverse item to be processed. Thereafter, the similarity between the feature data of the reverse item to be processed and each standard reverse item is calculated. The standard reverse item that accurately matches the reverse item to be processed can be determined based on the maximum similarity. Subsequently, business guidance can be provided to the reverse item to be processed based on the quantity-value relationship model of the matching items, thereby solving the problem of difficulty in accurately matching the reverse item to be processed with the standard reverse item due to the non-standardization and semantic ambiguity of the manually labeled data.

[0052] It should be noted that the collection, collection, updating, analysis, processing, use, transmission, and storage of user personal information involved in the technical solution of the present invention all comply with the provisions of relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken with respect to user personal information to prevent unauthorized access to user personal information data and maintain the security of user personal information, network security, and national security.

[0053] For ease of description, the aforementioned method embodiments are described as a series of actions. However, those skilled in the art should be aware that the present invention is not limited to the order of the actions described, and certain steps can actually be performed in other orders or simultaneously. In addition, those skilled in the art should also be aware that the embodiments described in this specification are preferred embodiments, and the actions and modules involved are not necessarily required to implement the present invention.

[0054] In order to better implement the above solutions of the embodiments of the present invention, relevant devices for implementing the above solutions are also provided below.

[0055] See also Figure 3 As shown, the reverse object matching device 300 provided by the embodiment of the present invention may include: a first feature extraction unit 301 , a second feature extraction unit 302 and a matching unit 303 .

[0056] Among them, the first feature extraction unit 301 can be used to perform data enhancement on the initial attribute data of the standard reverse article in the standard data set based on a preset basic data set to obtain enhanced attribute data of the standard reverse article, and use a pre-trained feature extraction model to determine the feature data of the standard reverse article from the enhanced attribute data; the second feature extraction unit 302 can be used to obtain manually marked attribute data of the reverse article to be processed, and use the feature extraction model to determine the feature data of the reverse article to be processed from the manually marked attribute data; the matching unit 303 can be used to determine the similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article, and determine the standard reverse article corresponding to the maximum similarity as the matching article of the reverse article to be processed.

[0057] In an embodiment of the present invention, the initial attribute data of any standard reverse article includes the item name of the standard reverse article, and the basic data set includes the item name and multi-level category data of each standard reverse article; and the first feature extraction unit 301 can be further used to: use the item name of any standard reverse article in the standard data set to query the basic data set to obtain the multi-level category data of the standard reverse article; and combine the queried multi-level category data with the initial attribute data of the standard reverse article to obtain enhanced attribute data of the standard reverse article.

[0058] As a preferred solution, the manually marked attribute data includes manually marked multi-level category data and manually marked item names; and the first feature extraction unit 301 can be further used to: input the category data of each level in the enhanced attribute data of any standard reverse item into the feature extraction model respectively to obtain multiple category features of the standard reverse item; identify the item name of the standard reverse item based on at least one named entity of a preset entity type, and splice the identified named entity with the corresponding entity type into a named entity feature of the standard reverse item; and combine the multiple category features with the named entity features. The weighted sum is determined as the feature data of the standard reverse article; the second feature extraction unit 302 can be further used to: input the category data of each level in the manually marked attribute data of the reverse article to be processed into the feature extraction model respectively to obtain multiple category features of the reverse article to be processed; identify the item name of the reverse article to be processed based on at least one named entity of the entity type, and splice the identified named entity with the corresponding entity type to form the named entity feature of the reverse article to be processed; and determine the weighted sum of the category features and the named entity features of the reverse article to be processed as the feature data of the reverse article to be processed.

[0059] In addition, in an embodiment of the present invention, the standard data set includes the historical supply quantity and historical value data of any standard reverse item; and the device 300 may further include: a quantity and price guidance unit, which is used to establish a quantity-value relationship model of the matching item based on the historical supply quantity and historical value data of the matching item of the reverse item to be processed, and use the quantity-value relationship model to determine the value data of the reverse item to be processed.

[0060] According to the technical solution of the embodiment of the present invention, by performing data enhancement and feature extraction based on a machine learning model on the initial attribute data of each standard reverse article in the standard data set, feature data of the covered item name and multi-level category information of each standard reverse article is obtained; by performing feature extraction based on a machine learning model on the manually labeled data of the reverse article to be processed, feature data of the covered item name and multi-level category information of the reverse article to be processed is obtained, and then the similarity between the feature data of the reverse article to be processed and each standard reverse article is calculated. The standard reverse article that accurately matches the reverse article to be processed can be determined based on the maximum similarity. Subsequently, business guidance can be provided to the reverse article to be processed based on the quantity-value relationship model of the matching items, thereby solving the problem of difficulty in accurately matching the reverse article to be processed with the standard reverse article due to the non-standardization and semantic ambiguity of the manually labeled data.

[0061] Figure 4 An exemplary system architecture 400 is shown to which the reverse item matching method or reverse item matching apparatus according to an embodiment of the present invention may be applied.

[0062] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, and 403, a network 404, and a server 405 (this architecture is merely an example, and the components included in the specific architecture may be adjusted based on the specific application). Network 404 is used to provide a medium for communication links between terminal devices 401, 402, and 403 and server 405. Network 404 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0063] Users can use terminal devices 401, 402, 403 to interact with server 405 via network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, 403, such as feature extraction applications (only as an example).

[0064] The terminal devices 401 , 402 , and 403 may be various electronic devices having a display screen and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, and desktop computers.

[0065] Server 405 may be a server that provides various services, such as a machine learning model application server (for example only) that supports feature extraction applications operated by users using terminal devices 401, 402, and 403. The machine learning model application server may process received feature extraction requests and feedback processing results (for example, extracted feature data—for example only) to terminal devices 401, 402, and 403.

[0066] It should be noted that the reverse item matching method provided in the embodiment of the present invention is generally executed by the server 405 , and accordingly, the reverse item matching device is generally disposed in the server 405 .

[0067] It should be understood that Figure 4 The number of terminal devices, networks and servers in the embodiment is merely illustrative. Any number of terminal devices, networks and servers may be provided as required.

[0068] The present invention also provides an electronic device. The electronic device in an embodiment of the present invention includes: one or more processors; and a storage device configured to store one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the reverse article matching method provided by the present invention.

[0069] Reference below Figure 5 , which shows a schematic structural diagram of a computer system 500 suitable for implementing an electronic device according to an embodiment of the present invention. Figure 5 The electronic device shown is only an example and should not limit the functions and scope of use of the embodiments of the present invention.

[0070] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 502 or a program loaded from a storage unit 508 into a random access memory (RAM) 503. Various programs and data required for the operation of the computer system 500 are also stored in the RAM 503. The CPU 501, the ROM 502, and the RAM 503 are connected to each other via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0071] The following components are connected to the I / O interface 505: an input section 506 including a keyboard, a mouse, and the like; an output section 507 including devices such as a cathode ray tube (CRT), a liquid crystal display (LCD), and a speaker; a storage section 508 including a hard disk; and a communication section 509 including a network interface card such as a LAN card or a modem. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to the I / O interface 505 as needed. A removable medium 511, such as a magnetic disk, an optical disk, a magneto-optical disk, or a semiconductor memory, is installed in the drive 510 as needed, so that computer programs read therefrom can be installed in the storage section 508 as needed.

[0072] In particular, according to embodiments disclosed herein, the processes described in the main step diagrams above can be implemented as computer software programs. For example, embodiments of the present invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for executing the methods shown in the main step diagrams. In the above embodiments, the computer program can be downloaded and installed from a network via the communication section 509 and / or installed from removable media 511. When the computer program is executed by the central processing unit 501, the above-described functions defined in the system of the present invention are performed.

[0073] It should be noted that the computer-readable medium described in the present invention may be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or component, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to, an electrical connection having one or more conductors, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination thereof. In the present invention, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, device, or component. In the present invention, a computer-readable signal medium may include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium that can transmit, propagate, or transport a program for use by or in conjunction with an instruction execution system, apparatus, or device. Program code embodied on a computer-readable medium may be transmitted using any suitable medium, including but not limited to wireless, wireline, optical cable, RF, or any suitable combination thereof.

[0074] The flowcharts and block diagrams in the accompanying drawings illustrate the possible implementation architecture, functions and operations of the systems, methods and computer program products according to various embodiments of the present invention. In this regard, each box in the flowchart or block diagram can represent a module, program segment, or a part of code, and the above-mentioned module, program segment, or a part of code contains one or more executable instructions for implementing the specified logical function. It should also be noted that in some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two boxes represented in succession can actually be executed substantially in parallel, and they can sometimes be executed in the opposite order, depending on the functions involved. It should also be noted that each box in the block diagram or flowchart, and the combination of boxes in the block diagram or flowchart, can be implemented with a dedicated hardware-based system that performs the specified function or operation, or can be implemented with a combination of dedicated hardware and computer instructions.

[0075] The units described in the embodiments of the present invention may be implemented via software or hardware. The units described may also be provided within a processor. For example, a processor may be described as comprising a first feature extraction unit, a second feature extraction unit, and a matching unit. The names of these units do not, in some cases, limit the units themselves. For example, the first feature extraction unit may also be described as a "unit that provides feature data of a standard reverse object to the matching unit."

[0076] As another aspect, the present invention further provides a computer-readable medium, which may be included in the device described in the above embodiment, or may exist independently and not be incorporated into the device. The computer-readable medium carries one or more programs, and when executed by the device, causes the device to perform the following steps: performing data enhancement on initial attribute data of a standard reverse article in a standard data set based on a preset basic data set to obtain enhanced attribute data of the standard reverse article, and determining feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model; obtaining manually labeled attribute data of a reverse article to be processed, and determining feature data of the reverse article to be processed from the manually labeled attribute data using the feature extraction model; and determining similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article, and determining the standard reverse article corresponding to the maximum similarity as a matching item for the reverse article to be processed.

[0077] In the technical solution of the embodiment of the present invention, data enhancement and feature extraction based on a machine learning model are performed on the initial attribute data of each standard reverse item in the standard data set to obtain feature data of the covered item name and multi-level category information of each standard reverse item; feature extraction based on a machine learning model is performed on the manually labeled data of the reverse item to be processed to obtain feature data of the covered item name and multi-level category information of the reverse item to be processed. Thereafter, the similarity between the feature data of the reverse item to be processed and each standard reverse item is calculated. The standard reverse item that accurately matches the reverse item to be processed can be determined based on the maximum similarity. Subsequently, business guidance can be provided to the reverse item to be processed based on the quantity-value relationship model of the matching items, thereby solving the problem of difficulty in accurately matching the reverse item to be processed with the standard reverse item due to the non-standardization and semantic ambiguity of the manually labeled data.

[0078] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A reverse item matching method, characterized in that: include: performing data enhancement on initial attribute data of the standard reverse article in the standard dataset based on a preset basic dataset to obtain enhanced attribute data of the standard reverse article, and determining feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model; Acquire manually marked attribute data of the reverse article to be processed, and determine feature data of the reverse article to be processed from the manually marked attribute data using the feature extraction model; The similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article is determined, and the standard reverse article corresponding to the maximum similarity is determined as a matching article of the reverse article to be processed.

2. The method according to claim 1, characterized in that The initial attribute data of any standard reverse item includes the item name of the standard reverse item, and the basic data set includes the item name and multi-level category data of each standard reverse item; Furthermore, performing data enhancement on the initial attribute data of the standard reverse item in the standard data set based on the preset basic data set includes: Using the name of any standard reverse item in the standard data set to query the basic data set, to obtain multi-level category data of the standard reverse item; The retrieved multi-level category data is combined with the initial attribute data of the standard reverse item to form enhanced attribute data of the standard reverse item.

3. The method according to claim 2, characterized in that Determining the feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model includes: Inputting the category data of each level in the enhanced attribute data of any standard reverse article into the feature extraction model respectively to obtain multiple category features of the standard reverse article; Identifying the name of the standard reverse item based on at least one named entity of a preset entity type, and concatenating the identified named entity with the corresponding entity type into a named entity feature of the standard reverse item; A weighted sum of the multiple category features and the named entity features is determined as feature data of the standard reverse item.

4. The method according to claim 3, characterized in that The manually marked attribute data includes manually marked multi-level category data and manually marked item names; and determining the feature data of the reverse item to be processed from the manually marked attribute data using the feature extraction model includes: Inputting the category data of each level in the manually marked attribute data of the reverse article to be processed into the feature extraction model respectively to obtain multiple category features of the reverse article to be processed; Identifying the name of the reverse article to be processed based on at least one named entity of the entity type, and concatenating the identified named entity with the corresponding entity type into a named entity feature of the reverse article to be processed; The weighted sum of each category feature and the named entity feature of the reverse article to be processed is determined as the feature data of the reverse article to be processed.

5. The method according to any one of claims 1 to 4, characterized in that: The standard data set includes historical supply quantity and historical value data of any standard reverse item; and the method further comprises: Establishing a quantity-value relationship model of the matching items based on historical supply quantity and historical value data of the matching items of the reverse item to be processed; The quantity-value relationship model is used to determine the value data of the reverse item to be processed.

6. A reverse object matching device, characterized in that: include: a first feature extraction unit, configured to perform data enhancement on initial attribute data of the standard reverse article in the standard dataset based on a preset basic dataset to obtain enhanced attribute data of the standard reverse article, and determine feature data of the standard reverse article from the enhanced attribute data using a pre-trained feature extraction model; a second feature extraction unit, configured to obtain manually marked attribute data of the reverse article to be processed, and determine feature data of the reverse article to be processed from the manually marked attribute data using the feature extraction model; The matching unit is configured to determine the similarity between the feature data of the reverse article to be processed and the feature data of each standard reverse article, and determine the standard reverse article corresponding to the maximum similarity as a matching article for the reverse article to be processed.

7. The device according to claim 6, characterized in that The initial attribute data of any standard reverse article includes the name of the standard reverse article, and the basic data set includes the name of each standard reverse article and multi-level category data; and the first feature extraction unit is further configured to: The basic data set is queried using the name of any standard reverse item in the standard data set to obtain multi-level category data of the standard reverse item; and the queried multi-level category data is combined with the initial attribute data of the standard reverse item to obtain enhanced attribute data of the standard reverse item.

8. The device according to claim 7, characterized in that The manually marked attribute data includes manually marked multi-level category data and manually marked item names; and The first feature extraction unit is further configured to: input each level of category data in the enhanced attribute data of any standard reverse article into the feature extraction model to obtain multiple category features of the standard reverse article; identify at least one named entity of a preset entity type based on the name of the standard reverse article, and concatenate the identified named entity with the corresponding entity type to form a named entity feature of the standard reverse article; and determine a weighted sum of the multiple category features and the named entity feature as feature data of the standard reverse article; The second feature extraction unit is further configured to: input the category data of each level in the manually marked attribute data of the reverse article to be processed into the feature extraction model respectively, to obtain multiple category features of the reverse article to be processed; The article name of the reverse article to be processed is identified based on at least one named entity of the entity type, the identified named entity is concatenated with the corresponding entity type to form a named entity feature of the reverse article to be processed; and a weighted sum of each category feature of the reverse article to be processed and the named entity feature is determined as feature data of the reverse article to be processed.

9. The device according to any one of claims 6 to 8, characterized in that: The standard data set includes historical supply quantity and historical value data of any standard reverse item; and the device further includes: The quantity and price guidance unit is used to establish a quantity-value relationship model of the matching items according to the historical supply quantity and historical value data of the matching items of the reverse items to be processed, and use the quantity-value relationship model to determine the value data of the reverse items to be processed.

10. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs, When the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 5.

11. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 5 is implemented.

Images