TWM667558U - Retail system with composite tags for products - Google Patents

Abstract

A retail system with composite tags for products is provided. The retail system includes a retail system frontend that provides consumers the product information, a retail system backend that provides providers to set product texts, a database and a computing server. The computing server is used to generate composite tags for products. the computing server retrieves texts of a product from a product text library, segments the text for obtaining multiple keywords that form attribute tags of the product, and combines the multiple attribute tags so as to form multiple sets of candidate composite tags for the product. After duplicate attribute tags are excluded, multiple sets of composite tags for the product are generated. The retail system frontend shows multiple products and the multiple sets of composite tags with respect to each of the products.

Description

Retail system with product combination labeling

The invention relates to an e-commerce platform, in particular to a retail system for generating a product combination label by breaking up product text into words to obtain product labels and combining them to form a combination label.

In addition to shopping in physical stores, e-commerce platforms use Internet technology to provide consumers with a more convenient way to search for products of interest. E-commerce platforms can also accurately recommend products to consumers through various means, which can increase consumers' motivation to purchase products and increase their stickiness to the e-commerce platform.

The common way to recommend products to consumers is to set a tag that matches the product for each product, and then analyze the products that users browse to learn the user's preferences, so that products that match the user's preferences can be provided by matching the tags. Another method is to use an intelligent model, such as an embedding model, to execute an embedding vector algorithm to convert the product text into an embedding vector, and then perform a similarity comparison in a vector space, which can effectively recommend highly similar products to consumers.

Different from the conventional method of establishing product-related labels to recommend products, the disclosure proposes a retail system with product combination labels. The retail system proposed mainly includes a retail system front desk for providing consumers with information on multiple products, a retail system back desk for providing multiple vendors with individual texts for multiple products, and setting multiple products on the shelves, and includes a database, which is built-in or external to the retail system back desk and has database 1 and database 2.

Database 1 includes a product text library, a product attribute tag library, a retail dictionary, and black and white lists set by various brands for their various products. The product text library records the text of each product, the product attribute tag library records multiple attribute tags of each product, and the retail dictionary records specialized terms in the retail industry.

Database 2 includes a product/category page combination tag library, which records one or more sets of combination tags for each product.

The retail system is provided with a computing server, which is connected to database 1 and database 2, and can obtain the text of a product from the product text library in database 1, and obtain multiple keywords after segmenting the text according to the retail dictionary and using a segmentation model. These keywords will form multiple native attribute tags of the product. The native attribute tags are obtained, and one or more whitelist attribute tags, one or more extended attribute tags and one or more replacement attribute tags set by the brand are obtained, excluding one or more blacklist attribute tags set by the brand, and together form multiple attribute tags of the product. Afterwards, multiple attribute tags are combined to form multiple candidate combination tags for the product, wherein each candidate combination tag has two or more attribute tags. After excluding duplicate combination tags from the multiple candidate combination tags, multiple combination tags ranked first can be selected according to a sorting result to become multiple combination tags associated with the product, and the product/category page combination tag library in database 2 can be established accordingly. Afterwards, multiple products and multiple combination tags of multiple products can be presented through the retail system front end.

Furthermore, when the computing server obtains multiple native attribute tags from the text, it also obtains one or more whitelist attribute tags, one or more extended attribute tags and one or more replacement attribute tags. The native attribute tags, one or more whitelist attribute tags, one or more extended attribute tags and one or more replacement attribute tags remove one or more blacklist attribute tags to form the multiple attribute tags of the product together.

Furthermore, when forming multiple candidate combination labels, a combination label whitelist can be introduced from the black and white list to list one or more combination labels recorded in the combination label whitelist as the first-priority combination labels appearing in the foreground; in addition, a combination label blacklist can be introduced from the black and white list to exclude one or more combination labels recorded in the combination label blacklist from multiple candidate combination labels.

Furthermore, the ranking result is obtained by ranking each set of combination labels according to the web search volume, so as to select multiple sets of combination labels ranked first to form the second-ranked combination labels.

Then, the first-order combination tags or the second-order combination tags can be added to form multiple groups of combination tags for the related products.

Furthermore, by calculating the timeliness of any of the multiple sets of combination tags associated with the product, the multiple sets of combination tags associated with the product can be automatically updated according to the timeliness.

Furthermore, the computing server can periodically or according to a schedule check whether each product on the shelf and its text have changed. If the check shows a change, the combination label of the product will be regenerated or updated.

When regenerating or updating the combination label of a product, the validity of the black and white list can also be checked. When the validity of the black and white list expires, the combination label will be modified according to the updated black and white list.

The retail system also has a tag recommendation application program interface. Based on the product/category page combination tag library, the computing server displays the combination tags of recommended specific products on the retail system front end through the tag recommendation application program interface. Furthermore, a product description page is entered on the retail system front end, which displays the content of multiple products, one or more combination tags of each product, and the specific product recommended by the tag recommendation application program interface.

In one implementation, after excluding duplicate combination tags from the multiple sets of candidate combination tags, a natural language processing technique is used to calculate the similarity between each set of candidate combination tags and a title of a product, and then the multiple sets of candidate combination tags are sorted according to the multiple similarities to select multiple sets of combination tags for related products.

In order to further understand the features and technical contents of the present invention, please refer to the following detailed description and drawings of the present invention. However, the drawings provided are only used for reference and description and are not used to limit the present invention.

The following is a specific and concrete example to illustrate the implementation of this creation. The technical personnel in this field can understand the advantages and effects of this creation from the content disclosed in this manual. This creation can be implemented or applied through other different specific embodiments. The details in this manual can also be modified and changed in various ways based on different viewpoints and applications without deviating from the concept of this creation. In addition, the drawings of this creation are only for simple schematic illustrations and are not depicted according to actual size. Please note in advance. The following implementation will further explain the relevant technical content of this creation in detail, but the disclosed content is not used to limit the scope of protection of this creation.

It should be understood that, although the terms "first", "second", "third", etc. may be used herein to describe various components or signals, these components or signals should not be limited by these terms. These terms are mainly used to distinguish one component from another component, or one signal from another signal. In addition, the term "or" used herein may include any one or more combinations of the associated listed items depending on the actual situation.

The disclosure document proposes a retail system with product combination tags, wherein the retail system provides consumers with product information through web pages or applications through product combination tags, and can link to other related product pages through the combination tags, or product pages that the brand intends to allow consumers to link to, so as to achieve the purpose of recommending products with products. Please refer to the schematic diagrams of the product description page, product classification page, tag classification page, and even the search result-free page with combination tags shown in Figures 6 to 11 of the disclosure document.

A retail system that runs a method for generating a product combination label can be implemented as an e-commerce platform, allowing consumers to browse the products on the shelves through the Internet and to make purchases. The system architecture is shown in FIG1 . The retail system architecture includes a backend 11, a frontend 12, a database (database 1 13, database 2 15), a server 14, and an application programming interface 16 that provides each end with access to data, which is implemented by the collaboration of computer software and hardware.

The front desk 12 provides a retail system front desk 103, which can provide consumers with information about multiple products in the retail system through a web page or an application program executed on a user terminal device. The back desk 11 is similar to the retail system back desk 101, and is connected to the retail system front desk 103 through a specific channel, and can provide multiple vendors (such as brand owners, manufacturers, distributors, etc.) with individual texts of multiple products, and provide multiple products that are set up for consumers.

According to the embodiment, the database is connected (including external or built-in) to the retail system backend 101, and can be divided into database 1 13 and database 2 15 according to the function. Database 1 13 is provided with at least a product text library 105, a product attribute tag library 107, and may be provided with a retail dictionary 109 and a black and white list 110 set by the brand; database 2 is provided with a product/category page combination tag library 113.

Among them, the product text library 105 records the text of each product, and according to the retail industry's specialized terms and word-breaking models recorded in the retail dictionary 109, the product native attribute tags can be obtained to establish the product attribute tag library 107. The blacklist attribute tags 32, whitelist attribute tags 34, extended attribute tags 31 and replacement attribute tags 33 set by the brand will enter the blacklist 110 to optimize the native attribute tags obtained by the product attribute tag library 107, and the candidate combination tags of each product will be processed by the calculation server 111, and then the blacklist 110 set by the brand for each product or multiple products under its brand will be introduced, the whitelist combination tags will be placed on the top, and the blacklist combination tags will be filtered. After calculation, they will become the combination tags on each product page.

The server 14 of the retail system is actually the computing server 111, which is a computer system such as a processor with data processing capabilities, a memory and related peripherals, in which various server programs that provide terminal services are run, and a method for generating a product combination label and a method for recommending products are executed.

According to the embodiment, the computing server 111 is connected to the database 1 13 and the database 2 15 in sequence, so that the text of each product can be obtained from the product text library 105 in the database 1 13, and the black and white lists 110 set by each brand for its various products can be obtained, and multiple attribute tags of each product can also be obtained from the product attribute tag library 107. After that, the computing server 111 executes the product combination tag generation method to obtain multiple sets of combination tags for each product, and establishes the product/category page combination tag library 113 in the database 2 15.

According to the needs, the combined label and the vectorized value are first temporarily stored in the cache memory module 115, and then output to the retail system front end 103 through the label recommendation application program interface 117, and then access the product text library 105 through the retail system front end 103. In this way, by combining the product text, the combined label and the recommended product, the product page content provided to the consumer is formed.

Through the retail system described above, the use of product combination tags can effectively provide consumers with product categories of interest in the e-commerce platform, perform product recommendations, and effectively increase the search volume of search engines. The relevant process can be referred to in Figure 2.

Through the product combination tags, when consumers use the search engine 201 to search for products, they can effectively search for products of interest and enter the related brand homepage 203, from which they can link to the product classification page 205, which displays product information and also provides combination tags and related products of each product; further, after clicking on the classification page combination tag 207, they can enter the tag classification page 209, where in addition to the combination tags for the product, they can also obtain products recommended by the retail system.

When consumers use search engines 201 to search for products, they can enter the tag classification page 209 of related products based on the search words, which contains product combination tags 211 and products. Clicking on a product can enter a specific product page 213, or by using combination tags, the circulation rate between multiple products in a brand website can be increased. It is worth mentioning that whether it is a tag classification page 209 or a product page 213, consumers can get product recommendations 215 from the retail system when using combination tags.

Regarding the generation of the combination label, reference may be made to FIG3 which shows a flow chart of an embodiment of a method for generating a product combination label executed by the computing server 111 (through a processor therein running a related software program), and the components used therein may refer to the embodiment diagram shown in FIG1 .

At the beginning of the process, the computing server (FIG. 1, 111) obtains the product-related text from the product text library (FIG. 1, 105) of the database 1 (FIG. 1, 13) shown in FIG. 1 (step S301). The product text library records the title and text data provided by the brand (or manufacturer or distributor) for each product through the retail system backend 101.

Next, based on the retail product names and descriptions in the product text library (Figure 1, 105), word segmentation is performed according to the retail dictionary (and the word segmentation model can be used). After word segmentation, other peripheral information can be referenced to obtain multiple keywords (step S303). These keywords will form multiple attribute tags for the product, and accurately provide attribute tags that meet the product attributes and increase the browsing rate, forming native attribute tags (step S305), which can be recorded in the product attribute tag library (Figure 1, 107). Furthermore, when forming native attribute tags, extended attribute tags 31, blacklist attribute tags 32, replacement attribute tags 33 and whitelist attribute tags 34 established by brand owners' requirements or specific needs can be introduced. In addition to native attribute tags, extended attribute tags 31 and whitelist attribute tags 34 set by brand owners that are not native attribute tags derived from product texts are added to increase the attribute tags covered by the products. Replacement attribute tags 33 can also be set to replace specific terms in native attribute tags with other terms to form replacement attribute tags 33, and then blacklist attribute tags 32 are excluded from native attribute tags to establish a product attribute tag library (Figure 1, 107). Through the above steps, one or any combination of the extended attribute tag 31, the blacklist attribute tag 32, the replacement attribute tag 33 and the whitelist attribute tag 34 is introduced to form multiple attribute tags of the product. After arranging and combining the multiple attribute tags generated according to the product attributes, multiple groups of candidate composite tags of the product are formed (step S307), wherein each group of candidate composite tags has two or more attribute tags.

It is worth mentioning that the brand owner can set other whitelist attribute tags 34, replacement attribute tags 33 and extended attribute tags 31 of the product that do not include the original attribute tags derived from the text through the retail system backend (Figure 1, 101), and through the processing of the computing server (Figure 1, 111), the attribute tags covered by the product can be expanded. Furthermore, the whitelist attribute tags 34, replacement attribute tags 33 and extended attribute tags 31 may be irrelevant to the product on the current page, and can be used to guide consumers to obtain additional product information, not limited to the original attribute tags derived from the current product text library 105. For example, when it is necessary to promote a product through a holiday, such as Valentine's Day or Christmas, even if the product text does not calculate attribute tags related to these holidays, or the product is not directly related, but a specific meaning is given to the specific holiday, the product can be set with an additional attribute tag through the extended attribute tag 31 to achieve the purpose of marketing. The replacement attribute tag 33 replaces a specific attribute tag in the native attribute tag with an attribute tag not obtained from the native attribute tag. For example, the name derived from the product text of a specific lipstick is lipstick, but the brand owner wants to replace lipstick with lipstick, so lipstick is used as the replacement attribute tag of this product.

It is worth noting that in the proposed method for generating product combination labels, keywords are obtained by parsing product text (including product name and description text) and applying a word-breaking model, and are sorted according to the importance of the keywords. The importance can refer to the number of times the keywords appear in the text, or the importance can be judged based on external information such as retail dictionaries, so that the top keywords are the native attribute labels of the products, and the product attribute labels are adjusted or added through one or any combination of the extended attribute labels 31, replacement attribute labels 33, blacklist attribute labels 32 and whitelist attribute labels 34 set by the brand or specific needs. Furthermore, the system can also provide the brand with the ability to set a schedule for changing the attribute labels or give them timeliness, or set them to be permanently valid.

In the process of FIG. 3 , after forming a plurality of candidate combination tags (step S307 ), the combination tag whitelist 35 may be used to generate the first-rank combination tag (step S309 ), and the combination tags recorded in the combination tag blacklist 36 may be excluded from the candidate combination tags to form the adjusted final candidate combination tags (step S311 ).

Furthermore, referring to specific external information 37, such as the search volume for each combination tag in the search engine, and excluding duplicate candidate combination tags in multiple candidate combination tags, natural language processing (NLP) technology is further used to understand the semantics of consumers using various search words in the search engine to search, and the similarity between each group of candidate combination tags and the title of the product is calculated, and then multiple groups of combination tags of related products are selected according to the multiple similarities of the multiple groups of candidate combination tags. The final candidate combination tags obtained from the above steps determine the second-order combination tags (step S313), and the first-order combination tags plus the second-order combination tags become multiple groups of combination tags of related products (step S315).

Based on the first-order combination tag, or adding the second-order combination tag, multiple groups of combination tags of related products are formed. At this time, the combination tags can be reviewed manually or by computer software to exclude inappropriate attribute tags and combination tags, generate product combination tags (step S317), and the product/category page combination tag library (Figure 1, 113) can be established based on this. The product/category page combination tag library records one or more groups of combination tags for each product.

After obtaining multiple sets of combination tags of a product, the combination tags can be dynamically updated according to the actual situation. According to one of the embodiments, the timeliness of any of the multiple sets of combination tags of related products can be set through the computing server, so that the system will automatically update the multiple sets of combination tags of related products according to the timeliness of the combination tags. Please refer to the flowchart of the embodiment of the method for updating the product combination tags executed by the computing server (Figure 1, 111) in the retail system shown in Figure 4.

When the product combination label is obtained and the product/category page combination label library (Figure 1, 113) is established (step S401), the system checks whether each product on the shelf and its text have changed through the computing server on a regular basis or according to a schedule? (Step S403), if there is a change (yes), including the brand removing the product from the shelf, updating the product name or text, or other changes in settings, the product combination label generation method will be re-executed to generate or update the combination label, such as executing the process described in Figure 4 (step S407). In the process of regenerating or updating the combination label, a time limit check is further performed (step S405). If the time limit of a specific black and white list 110 expires, the combination label will be modified according to the updated black and white list 110. On the contrary, if the checked product, text or its attribute tags have not changed (No), no update is required and the system continues to check the changes of the product and its text regularly or according to the schedule (step S403).

FIG5 then shows a flow chart of an embodiment of a method for generating multiple sets of combination labels for each product to recommend products using the above method.

First, the computing server ( FIG. 1 , 111 ) obtains the product content in the product page (product page “product” 51), the search term when the user searches without results (search without results page “search term” 52), and the combination tag obtained from the product web page (tag classification page “combination tag” 53), and then executes the vector algorithm (step S501), and obtains the vectors of all products to create a product vector set (step S503), which can be first stored in a specific database, such as a product vector database. Therefore, the similarity can be calculated using the vectors of the product/search term/combination tag and all products, including calculating the similarity between the product and all products (step S505), calculating the similarity between the search term and all products (step S507), and calculating the similarity between the combination tag and all products (step S509), and then sorting according to the similarity (step S511), obtaining the order of similar products (step S513), and displaying it on the product page/search no result page/tag classification page, as described in the following embodiments, and also referring to the schematic diagrams of Figures 6, 7, 8, 9, and 10.

Therefore, when the retail system displays one of the products through the product page, it can display one or more other recommended products according to the order of similar products of the product, for example, display one or more products recommended to the consumer by the system on the product page (step S515). When the consumer enters the search result page, it can display one or more other recommended products according to the order of similarity with the search term, for example, display one or more products recommended to the consumer by the system on the search result page (step S517). When the retail system displays products through the tag classification page, it can display one or more other recommended products according to the order of similarity with the combined tag, for example, display one or more products recommended to the consumer by the system on the tag classification page (step S519).

Referring to the retail system architecture diagram shown in FIG1 , the computing server ( FIG1 , 111 ) displays a combination tag for recommending a certain product on the retail system front end ( FIG1 , 103 ) through a tag recommendation application program interface ( FIG1 , 117 ) based on the established product/category page combination tag library ( FIG1 , 113 ). One of the purposes is to recommend the product. For related examples, please refer to the user interface implementation diagram provided by the web page or application on the mobile device shown below.

FIG6 is a schematic diagram showing an implementation example of a product description page, wherein it is shown that after entering a brand's website, products can be searched by search words, and consumers enter the product description page 60 at the front desk of the retail system, wherein the main contents include a product photo 601, a product description 603, and a combination tag 605 generated by the above-mentioned product combination tag generation method, and a recommended product 607 recommended by the tag recommendation application program interface, which schematically shows recommended product 1, recommended product 2, recommended product 3, and recommended product 4.

Next, the consumer can click on one of the combination tags 605 on the product description page 60. The displayed web page content can refer to Figure 7, which shows a schematic diagram of an implementation example of a tag classification page, wherein a tag classification page 70 is displayed, including a combination tag 701 under this category, and multiple recommended products obtained by calculating the similarity based on this combination tag vector, such as recommended product 703, which schematically shows recommended product one, recommended product two, recommended product three and recommended product four, and can bring out related products in the product attribute tag library 107 that have the same attribute tags as those that constitute this combination tag, such as related products 705 shown in the schematic diagram, which schematically shows product one, product two, product three and product four, to recommend more product options, thereby realizing a recommendation method of finding products by products.

Next, FIG8 shows another schematic diagram of an embodiment of a tag classification page 80, which includes a combination tag 801 under the tag classification page, and according to the design of the page, recommended products 805 (showing recommended products one to four) and related products 803 and related products 807 (showing products one to eight) are repeatedly displayed.

FIG9 is a schematic diagram showing another embodiment of a product description page. The product description page 90 shown in the figure includes product content 901, in which the product and its description are displayed. The product description page 90 can also display a combination tag 903 related to the product after scrolling, and recommended products 905, which schematically display recommended product 1, recommended product 2, and recommended product 3.

However, when searching for products in the entire site using a search word or other search method, no results may be found. FIG10 is a schematic diagram of an implementation example of a search no results page, in which a search no results page 100 is schematically displayed. In addition to this page being represented as a search no results page 100, a popular search term 1001 and recommended products 1003 similar to the search term vector are also provided, and recommended product 1, recommended product 2, and recommended product 3 are schematically displayed.

Finally, FIG. 11 shows a schematic diagram of an implementation example of a product classification page, which shows a product classification page 110, including a combination tag 1101 and related products 1103 under this classification page, schematically showing product one and product two.

In summary, according to the retail system with product combination tags described in the above embodiments, the calculated combination tags of each product provide consumers with more accurate browsing of products of interest and increase the number of browsing cycles in the site. Furthermore, using vectors to calculate similar products provides a more flexible product recommendation space, rather than the recommended products obtained by traditional comparison tags. The traditional recommendation space is relatively inflexible. Moreover, the combination tags of products can also refer to the search volume obtained by the brand in the search engine (such as Google™ search engine) by naturally searching for the combination tags.

The above disclosed contents are only the preferred feasible embodiments of the present invention, and do not limit the scope of the patent application of the present invention. Therefore, all equivalent technical changes made by using the contents of the specification and drawings of the present invention are included in the scope of the patent application of the present invention.

11:Backstage 12:Frontstage 13:Database 1 14:Server 15:Database 2 16:Application Programming Interface 101:Retail System Backstage 103:Retail System Frontstage 105:Product Text Library 107:Product Attribute Tag Library 109:Retail Dictionary 110:Black and White List 111:Computation Server 113:Product/Category Page Combination Tag Library 115:Cache Memory Module 117:Tag Recommendation Application Programming Interface 201:Search Engine 203:Brand Homepage 205:Product Category Page 207:Category Page Combination Tag 209:Tag Category Page 211:Product Combination Tag 213:Product Page 215: Product recommendation 31: Extended attribute tag 32: Blacklist attribute tag 33: Replace attribute tag 34: Whitelist attribute tag 35: Combination tag whitelist 36: Combination tag blacklist 37: External information 38: Review 51: Product page "Product" 52: Search no results page "Search term" 53: Tag classification page "Combination tag" 60: Product description page 601: Product photo 603: Product description 605: Combination tag 607: Recommended products 70: Tag classification page 701: Combination tag 703: Recommended products 705: Related products 80: Tag classification page 801: Combination tag 803: Related products 805: Recommended products 807: Related products 90: Product description page 901: Product content 903: Combination tags 905: Recommended products 100: Search no results page 1001: Popular search terms 1003: Recommended products 110: Product classification page 1101: Combination tags 1103: Related products Steps S301~S317: Process of generating product combination tags Steps S401~S407: Product combination tag update process Steps S501~S519: Product recommendation process

FIG1 is a flowchart showing an example of a retail system architecture;

FIG2 shows an example of a process for performing product recommendation;

FIG3 is a flow chart showing an embodiment of a method for generating a product combination label;

FIG4 is a flow chart showing an embodiment of a method for updating a product combination label;

FIG5 is a flow chart showing an embodiment of a product recommendation method;

FIG6 is a schematic diagram showing an embodiment of a web page version of a product description page;

FIG. 7 is a schematic diagram showing an embodiment of a web page version of a tag classification page;

FIG8 is a schematic diagram showing an implementation example of the mobile device version of the tag classification page;

FIG9 is a schematic diagram showing an implementation example of a mobile device version of a product description page;

FIG. 10 is a schematic diagram showing an implementation example of a search result-free page for mobile devices; and

FIG11 is a schematic diagram showing an implementation example of a mobile device version of a product category page.

11: Backstage

12: Front Desk

13: Database 1

14: Server

15: Database 2

16: Application Programming Interface

101: Retail system backend

103: Retail system front desk

105: Product text library

107: Product attribute tag library

109: Retail Dictionary

110: Black and White List

111: Computing Server

113: Product/category page combination tag library

115: Cache memory module

117: Tag recommendation API

Claims (10)

A retail system, comprising: A retail system front desk, providing consumers with information on multiple products; A retail system back desk, connected to the retail system front desk, providing multiple vendors with individual texts for the multiple products, and setting up the multiple products for consumers; A database, built-in or external to the retail system back desk, the database includes database 1 and database 2, wherein: Database 1 includes a product text library, a product attribute label library, a retail dictionary, and a black and white list set by each brand for its various products, wherein the product text library records the text of each product, the product attribute label library records multiple attribute labels of each product, and the retail dictionary records specialized terms in the retail industry; and The database 2 is provided with a product/category page combination tag library, wherein the product/category page combination tag library records one or more combination tags for each product; and a computing server, connected to the database 1 and the database 2, obtains a text of a product from the product text library in the database 1, performs word segmentation on the text according to the retail dictionary and uses a word segmentation model to obtain multiple keywords, so that the multiple keywords form multiple attribute tags of the product, and then combines the multiple attribute tags to form multiple groups of candidate combination tags of the product, wherein each group of candidate combination tags has two or more , then exclude the multiple sets of candidate combination tags including duplicate combination tags, and select the multiple sets of combination tags ranked first according to a sorting result to become the multiple sets of combination tags associated with the product, so as to establish the product/category page combination tag library in the second database, and present the multiple products and the multiple sets of combination tags of the multiple products to consumers through the retail system front end of the retail system. A retail system as described in claim 1, wherein when the computing server derives multiple native attribute tags from the text, it also obtains one or more whitelist attribute tags, one or more extended attribute tags, and one or more replacement attribute tags. The native attribute tags, the one or more whitelist attribute tags, the one or more extended attribute tags, and the one or more replacement attribute tags, remove one or more blacklist attribute tags, and together form the multiple attribute tags of the product. A retail system as described in claim 1, wherein, when forming the multiple sets of candidate combination labels, a combination label white list is introduced from the black and white list, and one or more combination labels recorded in the combination label white list are listed as a first-priority combination label; and/or a combination label black list is introduced from the black and white list to exclude one or more combination labels recorded in the combination label black list from the multiple sets of candidate combination labels. A retail system as described in claim 3, wherein the ranking result is obtained by ranking each set of combination tags according to the online search volume, so as to select the multiple sets of combination tags ranked first to form a second-rank combination tag, wherein the first-rank combination tag, or the second-rank combination tag plus the multiple sets of combination tags associated with the product. A retail system as described in claim 1, wherein a timeliness is set for any of the multiple sets of combination tags associated with the product through the computing server, that is, the multiple sets of combination tags associated with the product are automatically updated according to the timeliness. A retail system as described in claim 1, wherein the computing server periodically or according to a schedule checks whether each product on the shelf and its text have changed. If the check shows a change, the combination label of the product is regenerated or updated. A retail system as described in claim 6, wherein when the combination label of the product is regenerated or updated, the validity of the black and white list is also checked. When the validity of the black and white list expires, the combination label will be modified according to the updated black and white list. A retail system as described in claim 1, wherein the retail system is further provided with a tag recommendation application program interface, and the computing server displays the combination tag of recommended specific products on the front end of the retail system through the tag recommendation application program interface based on the product/category page combination tag library. A retail system as described in claim 8, wherein a product description page is entered at the front end of the retail system, which displays the contents of the multiple products, one or more sets of combined tags for each product, and the specific product recommended by the tag recommendation application program interface. A retail system as described in any one of claim items 1 to 9, wherein after excluding duplicate combination tags from the multiple sets of candidate combination tags, a natural language processing technology is used to calculate the similarity between each set of candidate combination tags and a title of the product, and then the multiple sets of candidate combination tags are sorted according to the multiple similarities to select the multiple sets of combination tags associated with the product.

Images