A system and method for mining new keywords for products in e-commerce

Abstract

The present invention discloses a system and method for mining new keywords for goods in e-commerce. In this method, if a hot word performs one or more of the following operations, such as collection operation, order operation, adding to shopping cart operation, and consultation operation, on each item corresponding to the same item within a preset time after being searched by more than a preset number of consumers, it can be judged that the hot word is related to the item, that is, the purpose of the consumer searching for the hot word is to find the item. Thus, the server can automatically associate the hot word, which is a product label word that is not widely used by stores but has a large number of user searches, with the target item that the consumer wants to search for, and add the new keyword obtained based on the hot word to the keyword library of the item. When other consumers search for this word again in the future, the item items corresponding to the item can be directly returned to the consumer without manually judging which items the word is related to, thereby improving the consistency of judging the hot word and the item.

Description

Technical Field

[0001] The present invention relates to the technical field of e-commerce, and in particular to a system and method for mining new keywords for commodities used in e-commerce. Background Art

[0002] Electronic commerce platforms, or e-commerce platforms, provide consumers with search functions. Consumers can enter search terms in the search box of the e-commerce platform's application or website. After the matching algorithm used by the e-commerce platform matches the product corresponding to the search term, the server will return the matched product entry to the current page and present it to the consumer, so that the consumer can view the corresponding product and perform subsequent collections, add to shopping carts, place orders, or initiate consultations with a store's customer service.

[0003] The inventor discovered that the search terms entered by consumers are the names of products in their cognition, and each store on the e-commerce platform will give corresponding titles to the products according to the product names in their cognition. When the product keywords (i.e., product names) in the consumer's cognition are different from the product names in the e-commerce platform, the matching algorithm will not be able to match the corresponding product entries according to the product titles. In other words, in order to allow consumers to search for a certain product more efficiently, new keywords for the product should be discovered in a timely manner and updated to achieve the unity of cognition between consumers and store staff.

[0004] After searching, patent document 1 discloses a keyword definition method. The technical contribution of this solution lies in obtaining hot search words and when it is determined that the hot search words are related to a product, using the hot words to update the product keywords, thereby obtaining new keywords that are more timely for the product, making product queries more timely.

[0005] However, when determining whether a hot search word is relevant to a product, Patent Document 1 requires manual positive or negative instructions. Specifically, a judgment instruction input by a user is received to determine whether a new keyword is relevant to the product. When the user inputs a positive judgment instruction, the new keyword is determined to be relevant to the product. When the user inputs a negative judgment instruction, the new keyword is determined to be irrelevant to the product.

[0006] Obviously, the above-mentioned solution for judging whether a hot search word is related to a product can only be applied to the staff of a store to judge the products of their own store. If the e-commerce platform wants to recruit a group of staff to judge all the products of the thousands of stores on the platform, it will lead to high labor costs. In addition, since these staff are not practitioners in the field of the product, they may not have enough understanding of the specific product. At the same time, everyone has different perceptions of the same product. Therefore, when judging whether a hot word is related to the product, it will lead to inconsistent judgment standards and the risk of misjudgment.

[0007] Patent document 1, Chinese patent, publication number, CN106547762A, patent name, a keyword definition method and system, publication date, 2017-03-29. Summary of the invention

[0008] The purpose of the present invention is to solve at least one of the technical problems existing in the prior art, and to provide a system and method for mining new keywords for products for e-commerce, which can automatically determine whether hot words are related to products through data analysis of the e-commerce platform, and if relevant, use the hot words as new keywords and add them to the keyword library of the products, thereby improving the consistency of the judgment results.

[0009] In a first aspect, the present invention provides a method for mining new keywords for commodities for e-commerce, the method comprising:

[0010] Continuously obtain the search terms entered by consumers in the search box at a preset frequency;

[0011] When it is detected that the number of times a search word is input is greater than a preset threshold and the search word does not belong to an existing keyword, the search word is used as a hot word;

[0012] Determine whether the hot word is related to the product;

[0013] When the hot word is related to the product, the hot word is added as a new keyword of the product to the keyword library of the product, so as to search for the product item corresponding to the product according to the new keyword;

[0014] The step of judging whether the hot word is related to the product specifically includes:

[0015] After a hot word is input in the search box and within a preset time, when the number of consumers who perform a target operation exceeds a preset number, it is determined that the hot word is related to the product; the target operation includes one or more of the following operations: a collection operation, an order operation, an add-to-shopping cart operation, and a consultation operation for each product item corresponding to the same product, wherein each product item in the same product set is regarded as each product item corresponding to the same product.

[0016] In a second aspect, the present invention provides a system for mining new keywords for commodities in e-commerce, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0017] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which implements the steps of the above method when executed by a processor.

[0018] The system and method for mining new keywords for products in e-commerce provided by the present invention have at least the following beneficial effects:

[0019] If a hot word is searched by more than a preset number of consumers within a preset time period and one or more of the following operations are performed on the product items corresponding to the same product, such as collection, ordering, adding to shopping cart, and consultation, it can be determined that the hot word is related to the product, that is, the purpose of the consumer's search for the hot word is to find the product. Therefore, the server can automatically associate the product label words that are not widely used by stores but are searched by users with the target products that consumers want to search for, and add the new keywords obtained based on the hot word to the keyword library of the product. When other consumers search for this word in the future, the product items corresponding to this product can be directly returned to the consumer without the need to manually determine which products the word is related to, thereby improving the consistency of determining the relevance of the hot word and the product.

[0020] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] The present invention is further described below in conjunction with the accompanying drawings and embodiments;

[0022] Figure 1 The present invention is a flowchart of a method for mining new keywords for commodities for e-commerce provided in one embodiment.

[0023] Figure 2 The present invention is a flowchart of a method for mining new keywords for commodities for e-commerce provided in one embodiment.

[0024] Figure 3 The present invention is a flowchart of a method for mining new keywords for commodities for e-commerce provided in one embodiment.

[0025] Figure 4 The present invention is a flowchart of a method for mining new keywords for commodities for e-commerce provided in one embodiment.

[0026] Figure 5 The present invention is a flowchart of a method for mining new keywords for commodities for e-commerce provided in one embodiment. DETAILED DESCRIPTION

[0027] This section will describe in detail the specific embodiments of the present invention. The preferred embodiments of the present invention are shown in the accompanying drawings. The purpose of the accompanying drawings is to supplement the description of the text part of the specification with graphics, so that people can intuitively and vividly understand each technical feature and the overall technical solution of the present invention, but it cannot be understood as a limitation on the scope of protection of the present invention.

[0028] In order to facilitate those skilled in the art to understand the present invention, the following is an in-depth description of the background technology of the present invention made by the inventor:

[0029] Generally speaking, the search terms entered by consumers are the names of products in their cognition, and each store on the e-commerce platform will give the corresponding titles to the products according to the names of the products in their cognition. Store staff will generally give multiple product names to a product as the title text of the product item. For example, for a certain Daddy Shoes product, the title text given by Store A is: Daddy Shoes, Forrest Gump Shoes, Jogging Shoes; the title text given by Store B is: Daddy Shoes, Tuchao Shoes, Big Toe Shoes; the title text given by Store C is: Daddy Shoes, Silly and Cute Big Toe Shoes. When consumers enter Daddy Shoes for search, all product items with Daddy Shoes in the title text may be returned to the current page. However, based on the sorting strategy of the e-commerce platform, such as low price, high sales, high ratings, and the latest listing time, different product items will be sorted differently on the current page. For example, if the search for Daddy Shoes hits 10,000 product items, the worst case for Store B is that its product may be ranked last, that is, consumers must keep turning pages until the last position of the page to see the product items of Store B.

[0030] The matching algorithm used by e-commerce platforms in the prior art is as follows: the search term entered in the search box is directly matched as a whole with the title text of each product item. If the match fails, the search term is re-segmented. For example, if the search term is "stupid shoes" and the search term fails to match, the search term is re-segmented into "stupid" and "shoes" and then matched again. Thus, the product item corresponding to the "stupid and cute big-headed shoes" can be matched, but other product items with the words "stupid" and "shoes" in the title text may also be introduced. Alternatively, another method to help consumers search for products faster is the search term recommendation function, that is, when the user enters "stupid shoes" in the search box, the term "stupid and cute big-headed shoes" will be directly displayed below the search box to help consumers improve search efficiency.

[0031] These recommended search terms are extracted by the platform from the title texts of a large number of product items. The platform will automatically extract keywords that appear in large quantities in the title texts of product items. For example, when the word "stupid big-headed shoes" appears in more than 300 title texts, the platform will trigger automatic extraction to facilitate subsequent recommendations. For keywords that appear in small quantities in the title texts of product items, the extraction of keywords will not be triggered based on cost savings. However, when many products are first recognized by consumers, the names of the products recognized by consumers may be different from the names widely used on e-commerce platforms. For example, taking dad shoes as an example, the name widely used on a certain e-commerce platform is dad shoes, and further is Forrest Gump shoes, jogging shoes, and a small number of names are stupid big-headed shoes (for example, among 2,000 items of dad shoes, only 5 items have the word "stupid big-headed shoes" in the title text, 1,500 items have the word "dad shoes" in the title, 300 items have the word "Forrest Gump shoes", and 195 items have the word "jogging shoes"). When many consumers learn that the name is stupid shoes from the introduction of a public figure and do not know the name of dad shoes, Many consumers enter "stupid shoes" when searching on e-commerce platforms. At this time, since stupid and cute big-toed shoes do not meet the criteria for inclusion of recommended words, consumers cannot get recommended words below the search box and can only search on the search results page. The search results page may introduce many product entries with the words "stupid" and "shoes" in the titles. The worst case for the product entry of stupid and cute big-toed shoes is that it may be ranked last. After inefficiently finding the page with the product entry of stupid and cute big-toed shoes in the title, consumers can further search on the e-commerce platform for a large number of product entries of daddy shoes for purchase through the daddy shoes recorded in the title of the entry.

[0032] It can be found that in the above scenario, the technical inspiration provided by the solution of patent document 1 can only realize the statistics of hot words, that is, which hot words appear. Based on business practices, hot words are the content that e-commerce platforms need to pay attention to. However, before the staff of a large number of stores add the names of stupid shoes or stupid cute big-headed shoes to the product entries, because the product name used by consumers when searching is stupid shoes instead of stupid cute big-headed shoes, that is, it is not a complete or accurate product name that has been included in the store on the e-commerce platform, the search result page cannot accurately return the product entry of stupid cute big-headed shoes to consumers, resulting in low search efficiency. At the same time, since the platform cannot obtain which product the consumer wants to buy when searching for the new search term stupid shoes, the platform cannot provide a search term recommendation function. If the solution of patent document 1 is used to update the new keywords of the product to obtain new keywords, it is necessary to introduce a large number of personnel to manually judge which product the hot word stupid shoes is related to, as described in the background technology, so as to update the keyword library of the product to obtain new keywords for the product. This leads to high labor costs, and because these staff members are not practitioners in the field of the product, they may not have sufficient understanding of the specific product. At the same time, each person has a different understanding of the same product. Therefore, when judging whether a hot word is related to the product, the judgment criteria will be inconsistent, and there is a risk of misjudgment.

[0033] How to identify whether the purpose of consumers entering search terms is to inquire about what products to buy, that is, which products a hot word is related to, when the input search terms cannot be completely matched with the title texts of the product items on the e-commerce platform and the search recommendation terms corresponding to the search terms are not generated, is a technical difficulty that the present invention needs to overcome.

[0034] The following will explain in detail through the accompanying drawings and embodiments a system and method for mining new keywords for products for e-commerce provided by the present invention that can solve the above technical difficulties.

[0035] like Figure 1 As shown, in one embodiment, a method for mining new keywords for commodities for e-commerce is provided, the method is run in a network composed of a server and a terminal, and the method comprises the following steps:

[0036] Step S202: The server continuously obtains the search terms entered by the consumer in the search box at a preset frequency.

[0037] Consumers conduct data transmission between the application or browser on their terminals and the server to browse the product entries provided by the server. Each product entry is displayed using a page, which includes a product display area and a toolbar. The product display area sequentially includes a product image, a product title, and a product details introduction (including text and images) from top to bottom. On the toolbar, there are a favorite button, an add to cart button, a consult customer service button, and a purchase button. Due to screen size limitations, it is impossible to directly display all the content in the product display area. Therefore, users can slide the screen, operate the mouse wheel, or drag the scroll bar on one side of the page downwards to move more content into the product display area for display. It can be understood that when first entering the page corresponding to a certain product entry, the scroll bar is at the starting point of the scroll bar track. When sliding the screen, operating the mouse wheel, or dragging the scroll bar on one side of the page downwards, the scroll bar will gradually move towards the end point of the scroll bar track to display the content below in the product display area.

[0038] Among them, a search box will be displayed on the home page of the application or e-commerce website. When the user types text in the search box and triggers the search operation by pressing the Enter key or clicking the magnifying glass button, it is considered that the consumer has entered a search term in the search box, and the server will return the corresponding page to the application or browser in response to the input search term.

[0039] Step S204, when the server detects that the number of times the search term is input is greater than the preset threshold and the search term does not belong to the existing keywords, the search term is taken as a hot word.

[0040] For example, the search term is "stupid shoes", the preset threshold is 2000, the server will count the number of times each search term is input by consumers. When it is counted that the number of times the term "stupid shoes" is input is greater than 2000 and "stupid shoes" does not belong to the existing keywords, "stupid shoes" is taken as a hot word and added to the hot word set. Therefore, existing keywords can be filtered, that is, the keywords already in the product keyword library can be filtered out, and new keywords can be mined more efficiently, which is different from Patent Document 1.

[0041] Step S206, the server determines whether the hot word is related to the product.

[0042] For the hot words added to the hot word set, the server will determine whether the hot word is related to the product, specifically including:

[0043] When the server detects that the number of consumers who perform the target operation exceeds a preset number after a hot word is entered in the search box and within a preset time period, the server determines that the hot word is related to the product; the target operation includes one or more of the following operations: a collection operation, an order operation, an add to shopping cart operation, and a consultation operation for each product item corresponding to the same product, wherein each product item in the same product set is regarded as each product item corresponding to the same product.

[0044] For example, the preset duration is 30 minutes, and the preset number is 2000. When the server counts that 2000 consumers have performed one or more of the following operations on the product items in the product set "Dad Shoes" within 30 minutes after searching for the hot word "Stupid Shoes", such as collecting, ordering, adding to shopping cart, and consulting, it can be determined that the word "Stupid Shoes" is related to the product "Dad Shoes", that is, the purpose of consumers searching for the search word "Stupid Shoes" is to find the product "Dad Shoes".

[0045] Wherein, the method further comprises:

[0046] Step S702: the server automatically extracts the product tag words corresponding to the product entry according to the title text of the product entry.

[0047] For example, for a certain dad shoe product, the title text assigned by store A is: Dad Shoes, Forrest Gump Shoes, Jogging Shoes, the product entry is defined as 11, and the extracted product label words include "Dad Shoes", "Forrest Gump Shoes", and "Jogging Shoes"; the title text assigned by store B is: Dad Shoes, Rustic Shoes, Big Toe Shoes, the product entry is defined as 12, and the extracted product label words include "Dad Shoes", "Trendy Shoes", and "Big Toe Shoes"; the title text assigned by store C is: Dad Shoes, Stupid and Cute Big Toe Shoes, the product entry is defined as 13, and the extracted product label words include "Jogging Shoes" and "Stupid and Cute Big Toe Shoes".

[0048] Step S704: the server performs pairwise matching on two product entries as a group, and adds two product entries having at least one identical target product tag word to the same product set. The target product tag word refers to a product tag word whose search volume is greater than a target threshold.

[0049] If the target threshold is 3000, the search volume for "dad shoes" is 5000, and the search volume for "jogging shoes" is 3100, then item items 11, 12, and 13 are added to the same item set, and the item set is defined as item 100. If the target threshold is 3300, only item items 11 and 12 are added to the same item set. The purpose of setting the target threshold is to avoid misjudgment caused by associating some search terms that are not highly associated with the item "dad shoes" with the item, because sometimes consumers may use the same word to represent two different items when searching for them. For example, "speaker" is likely to be used as a search term by consumers when searching for speakers and pants, but more consumers will use the more standardized search term "bell-bottom pants" for pants, so the number of times "speaker" is used when searching for pants is less than the number of times when purchasing speakers. Therefore, store staff generally set "bell-bottom pants" as the title for bell-bottom pants and "speaker" as the title for speakers. Even though using the search term "horn" to search for "bell-bottom pants" is the habit of only a small number of consumers, while using "horn" to search for "speaker" is the habit of most consumers, setting target thresholds can regulate the use of title keywords and search terms by consumers and store staff.

[0050] Step S706: The server uses the commodity items in the same commodity set as commodity items corresponding to the same commodity.

[0051] In this example, the commodity set of commodity 100 includes commodity items 11 , 12 , and 13 , that is, commodity items 11 , 12 , and 13 are the same commodity, that is, commodity 100 .

[0052] Through step S702 to step S704, some commodity items are automatically divided into corresponding commodity sets, that is, it is automatically determined what commodity the commodity item is.

[0053] In step S708, the server removes duplicate product tag words of each product item in the same product set and sorts them according to the search volume, and adds the product tag words in the preset ranking as product keywords to the product keyword library.

[0054] For example, the top five product tag words are used as keywords for the corresponding products.

[0055] Step S710: When any keyword in the product keyword library is entered in the search box, all product items in the product set are displayed in the search result page.

[0056] Through steps S702 to S710, product items represented by search terms or keywords widely used by consumers or stores can be automatically divided into corresponding product sets and when any keyword in the product keyword library is entered in the search box, all product items in the product set are displayed on the search results page, thereby improving the efficiency of consumers' search for products.

[0057] Step S208: When the hot word is related to the product, the server adds the hot word as a new keyword of the product to the keyword library of the product, so as to search for the product entry corresponding to the product according to the new keyword.

[0058] However, in steps S702 to S710, for product tag words that are not widely used by stores, they still cannot be quickly searched even when the user search volume is large. In this regard, steps S202 to S208 can be combined to overcome this deficiency. Specifically, if the hot word "stupid shoes" is searched by 2,000 consumers within 30 minutes after being searched by 2,000 consumers, one or more of the following operations are performed on the product items in the product set "dad shoes", such as collection, ordering, adding to shopping cart, and consultation, it can be determined that the word "stupid shoes" is related to the product "dad shoes", that is, the purpose of consumers searching for the search word "stupid shoes" is to find the product "dad shoes". Therefore, the server can automatically associate the product tag word "stupid shoes" which is not widely used by stores but is a hot word with a large search volume by users with the target product that consumers want to search for, namely "dad shoes", and add the new keyword "stupid shoes" to the keyword library of the product "dad shoes". When other consumers search for the word "stupid shoes" in the future, all product entries in the product collection of "dad shoes" can be directly returned to the consumer. There is no need to manually judge which products the word "stupid shoes" is related to, which improves the consistency of judging the relevance of hot words and products.

[0059] In one embodiment, the target operation includes a first sub-operation and a second sub-operation performed after the first sub-operation;

[0060] After a hot word is input into the search box and within a preset time, the method further includes:

[0061] The input hot word is used as the search word with serial number 1. After the hot word is input, the search result page with serial number 1 is returned and the corresponding product entries are displayed.

[0062] Similarly, after inputting the search term with serial number N, the search result page with serial number N is returned and the corresponding product entries are displayed; N is an integer greater than 2.

[0063] In one example, after entering the hot word "stupid shoes" and before triggering the second sub-operation, the consumer entered 2 search terms in sequence, and browsed the corresponding search result pages returned through the terminal each time a search term was entered. Using the entered hot word as the search term numbered 1, the subsequent 2 search terms are numbered 2 and 3 respectively.

[0064] Among them, the first sub-operation includes:

[0065] The search term numbered N appears in the text of the product entry on the search result page numbered N - 1 that has been browsed by the consumer.

[0066] For example, for a certain pair of dad shoes product, the title text given by store A is: Dad Shoes,阿甘 Shoes, Jogging Shoes, and the product entry is defined as 11. The extracted product label words include "dad shoes", "阿甘 shoes", and "jogging shoes"; the title text given by store B is: Dad Shoes, Ugly Trendy Shoes, Big Head Shoes, and the product entry is defined as 12. The extracted product label words include "dad shoes", "ugly trendy shoes", and "big head shoes"; the title text given by store C is: Dad Shoes, Stupid and Cute Big Head Shoes, and the product entry is defined as 13. The extracted product label words include "jogging shoes" and "stupid and cute big head shoes".

[0067] For example, after entering the hot word (stupid shoes) and before triggering the second sub-operation, the consumer entered 2 search terms in sequence and browsed them. Specifically, on the search result page numbered 1, that is, the search result page of the hot word "stupid shoes", the consumer did not find the product they wanted by rummaging through this page. After continuing to rummage, they saw product entry 13. The consumer entered the page of product entry 13 for browsing and found that it was the product they wanted. After seeing the word "jogging shoes" recorded in it, they returned to the search result page numbered 1 and entered "jogging shoes" in the search box for further search, so that the search term numbered 2, "jogging shoes", appears in the title text of product entry 13 on the search result page numbered 1 that has been browsed by the consumer. After entering the search term numbered 2, "jogging shoes", the consumer rummaged through the search result page numbered 2 for the product they wanted and found product entry 11. They further guessed that this product might also be called "dad shoes", so they returned to the search result page numbered 2 and entered "dad shoes" in the search box for further search, so that the search term numbered 3, "jogging shoes", appears in the title text of product entry 11 on the search result page numbered 2 that has been browsed by the consumer.

[0068] It is understandable that after inputting the hot word and before the second sub-operation, if each search word with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1, it means that the two consecutive searches are related, that is, from searching the hot word to searching the search word with sequence number N, the consumer's purpose is to point to the same product. Therefore, when the first sub-operation is detected, that is, when it is determined that the search word with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1, it can be determined that from searching the hot word to searching the search word with sequence number N, the consumer's purpose is to point to the same product, and it can be determined that the consumer's purpose of searching the hot word is to find the product corresponding to the second sub-operation.

[0069] The second sub-operation includes one or more of the following operations:

[0070] The collection operations, ordering operations, adding to shopping cart operations, and consultation operations for each product item corresponding to the same product.

[0071] After entering the search term "dad shoes" with serial number 3, consumers will find a large number of product items on the search page with serial number 3 that are the products they want, that is, dad shoes, so they will enter the corresponding interface to select their favorite styles for collection, ordering, adding to shopping cart, consultation, etc.

[0072] It is understandable that since the search terms used each time in the process of gradually retrieving the product "Daddy Shoes" from the hot word "Stupid Shoes" are dependent on the search results provided by the previous search term, it means that the purpose of the consumer's initial input of the hot word "Stupid Shoes" is to find the product "Daddy Shoes". Further, if a large number of consumers (i.e., consumers exceeding the preset number) search for Daddy Shoes according to the above search process after searching for the hot word "Stupid Shoes" and perform a second sub-operation such as purchasing or adding the same product (i.e., Daddy Shoes) to the shopping cart, it means that for a large number of consumers, the purpose of searching for the hot word "Stupid Shoes" is to find the product. That is, since the process from hot word search to the second sub-operation such as placing an order ultimately points to the same product, the relevance of the product that the consumer wants to buy to the hot word can be guaranteed. If the hot word is not added to the keyword library of the product, it will cause the consumer's search process to require at least 3 searches as in the above example, and the search efficiency is low, and a large number of consumers cannot find the products they want to find in one search. In this embodiment, the hot word is added to the keyword library of the corresponding product. When subsequent consumers search through the word, the product entry of the corresponding product can be directly returned to the consumer, thereby improving search efficiency.

[0073] like Figure 2As shown, in one embodiment, the following process is used to determine whether the search term with a sequence number of N appears in the text of the commodity item that has been browsed by the consumer on the search result page with a sequence number of N-1:

[0074] Step S302: Acquire the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 as the target text.

[0075] Step S304, matching the search term with sequence number N in the target text.

[0076] Step S306: when the search term with serial number N is matched in the target text, it is determined that the search term with serial number N appears in the text of the product entry on the search result page with serial number N-1 that has been browsed by the consumer.

[0077] like Figure 3 As shown, in one example, the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 is obtained as the target text, specifically including:

[0078] Step S402: the terminal continuously monitors the position of the scroll bar in the page of the product item opened by the consumer.

[0079] Specifically, the position of the scroll bar can be monitored through the following JavaScript code.

[0080] / / Get the ratio of the scroll bar position to the total scroll height

[0081] function getScrollPercentage(){

[0082] var scrollTop = window.scrollY; / / Current scroll position

[0083] var totalHeight = document.documentElement.scrollHeight; / / Total height of the document

[0084] var windowHeight = window.innerHeight; / / window viewport height

[0085] / / Calculate the scrollable height (total height minus viewport height)

[0086] var scrollableHeight=totalHeight-windowHeight;

[0087] / / Calculate the percentage of the scroll bar position to the scrollable height

[0088] var scrollPercentage=(scrollTop / scrollableHeight)*100;

[0089] return scrollPercentage;

[0090] }

[0091] Wherein, scrollPercentage is the position of the scroll bar, that is, the calculated percentage of the distance between the current position of the scroll bar and the starting point to the scrollable height, for example, may be 60%.

[0092] Step S404: based on the maximum position reached by the scroll bar, the terminal obtains text elements contained in the page from the starting position of the scroll bar to the maximum position.

[0093] Assuming that the maximum position is 60%, the following JavaScript code can be used to obtain the text elements contained in the page from the starting position of the scroll bar to the corresponding maximum position.

[0094]

[0095]

[0096] Step S406: the terminal obtains the text of the product entry on the search result page with sequence number N-1 that has been browsed by the consumer from the text element as the target text.

[0097] During the website page loading process, these codes will be loaded and executed along with the rest of the web page, executed on the terminal web page or in the application embedded with WebView, so that only the text content browsed by the consumer in a product entry page is automatically obtained, without matching all the content in the page with the search term with sequence number N, thus reducing the matching resource overhead. On the other hand, the task of obtaining the target text is performed by the terminal, which can reduce the resource overhead of the server.

[0098] It should be noted that some web pages in the prior art can provide the function of matching text within a web page, for example, whether the word "shoes" appears in the current web page. However, this solution is not applicable to the present invention. Because if a consumer sees the Nth search word on the page of a certain product item, he must exit the current page before entering the Nth search word in the search box. Only then can the system know the Nth search word entered by the user. Since the page of the Nth search word appears before the user enters the search word in the search box, in this embodiment, it is necessary to specifically use the text of the product item that has been browsed by the consumer on the search result page numbered N-1 as the target text and extract these target texts to facilitate matching of the Nth search word.

[0099] It is understandable that, in another example, the terminal may also capture a screenshot of the screen, and then use an OCR (Optical Character Recognition) tool to recognize the text in the image to obtain the target text.

[0100] like Figure 4 As shown, in another example, the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 is obtained as the target text, specifically including:

[0101] Step S502: The terminal continuously monitors the position of the scroll bar in the page of the product item opened by the consumer.

[0102] Specifically, the code for monitoring the scroll bar position may refer to the implementation of step S402.

[0103] Step S504: When the terminal detects an unloading trigger signal of the current page, the terminal sends the maximum position reached by the scroll bar to the server.

[0104] Specifically, the beforeunload event in JavaScript can be used to detect the unloading trigger signal of the current page. The beforeunload event is triggered when the browser is about to unload the page. This is usually because the user closes the tab, clicks the back button, or navigates to another page.

[0105] Step S506: Based on the maximum position that the scroll bar has reached, the server obtains text elements contained in the page from the starting position of the scroll bar to the maximum position.

[0106] Specifically, the code for obtaining the text elements contained in the page corresponding to the scroll bar starting position to the maximum position may refer to the implementation of step S404.

[0107] Step S508: The server obtains the text of the product entry in the search result page with sequence number N-1 that has been browsed by the consumer from the text element as the target text.

[0108] In this example, the consumer searches for products on the e-commerce platform through the browser on the terminal, and the browser sends the maximum position that the scroll bar has reached to the server, and the server obtains the text elements contained in the page corresponding to the scroll bar starting position to the maximum position. The consumer's search process for hot words on the e-commerce platform can also be recorded without using a special application, which improves the scope of application of this method. And only the text content that the consumer has browsed in a product entry page is automatically obtained, without matching all the content in the page with the search word with sequence number N, which reduces the matching resource overhead.

[0109] It should be noted that some web pages in the prior art can provide the function of matching text within a web page, for example, whether the word "shoes" appears in the current web page. However, this solution is not applicable to the present invention. Because if a consumer sees the Nth search word on the page of a certain product item, he must exit the current page before entering the Nth search word in the search box. Only then can the system know the Nth search word entered by the user. Since the page of the Nth search word appears before the user enters the search word in the search box, in this embodiment, it is necessary to specifically use the text of the product item that has been browsed by the consumer on the search result page numbered N-1 as the target text and extract these target texts to facilitate matching of the Nth search word.

[0110] It is understandable that, in another example, the server may also capture a screenshot of the screen, and then use an OCR (optical character recognition) tool to recognize the text in the image to obtain the target text.

[0111] In one embodiment, the method further comprises:

[0112] When the terminal detects that the consumer selects the jump text in the search result page numbered N-1 and triggers the jump search mode, the jump text is input into the search box as the search term numbered N and it is determined that the search term numbered N appears in the text of the product entry that has been browsed by the consumer on the search result page numbered N-1.

[0113] In one example, a consumer logs into the website of an e-commerce platform on a browser of a terminal. In the search result page of the hot word "stupid shoes" with serial number 1, the consumer enters the page of a certain product item. In this page, the consumer selects the search word "jogging shoes" with serial number 2 by pressing and holding the left button of the mouse and releases the left button. Then, an operation pop-up window pops up on the right side of the selected search word on the webpage. A "jump control" is set above the operation pop-up window. When the consumer clicks the "jump control" with the mouse, it is considered that the jump search mode is triggered, which triggers the browser to return to the search result page and type the search word "jogging shoes" with serial number 2 into the search box and trigger the search, so that the user can view the search results for jogging shoes. It can be seen that when the jump search mode is triggered, the search word with serial number 2 must appear in the text of the product item that has been browsed by the consumer on the search result page with serial number 1. At this time, there is no need to perform the above steps S302 to S306, which reduces the system computer resource overhead. At the same time, it also facilitates users to search for the next search term, realizing the search for the search term in one step without exiting the current page, thereby improving user search efficiency.

[0114] It is understandable that after inputting the hot word and before the second sub-operation, if each search word with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1, it means that the two adjacent searches are related, that is, from the search for the hot word to the search for the search word with sequence number N, the consumer's purpose is to point to the same product. Therefore, when the first sub-operation is detected, that is, when it is determined that the search word with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1, it can be determined that from the search for the hot word to the search for the search word with sequence number N, the consumer's purpose is to point to the same product. Therefore, if the terminal detects that the consumer selects the jump text and triggers the jump search mode in the search result page with sequence number N-1 after searching the hot word and before the second sub-operation, it can be determined that the consumer's purpose of searching the hot word is to find the product corresponding to the second sub-operation.

[0115] In another example, a consumer can use an e-commerce platform through an application on a mobile phone. At this time, after selecting a word "jogging shoes" on the page by long pressing the finger, shake the phone to trigger the jump search mode. In this case, there is no need to perform the above steps S302 to S306, which reduces the system computer resource overhead. At the same time, it is also convenient for users to search for the next search word, realizing the search of the search word in one step without exiting the current page, thereby improving the user's search efficiency.

[0116] like Figure 5 As shown, in one embodiment, the method further includes:

[0117] Step S602: When the server detects that the number of co-occurrences of the search term with sequence number N and the adjacent search term with sequence number N-1 is greater than a preset value, the server adds the two as a related word combination to the recommendation word library.

[0118] It is understandable that the N in the search term with sequence number N and the adjacent search term with sequence number N-1 can be any integer greater than 1. For example, in the search term sequence input by a consumer, the search term "jogging shoes" with sequence number 3 and the adjacent search term "dad shoes" with sequence number 2, and in the search term sequence input by another consumer, the search term "dad shoes" with sequence number 4 and the adjacent search term "jogging shoes" with sequence number 3, it can be considered that the number of times "dad shoes" and "jogging shoes" appear together is 2, where "dad shoes" and "jogging shoes" are arranged in no order, which can also be described as the number of times "jogging shoes" and "dad shoes" appear together is 2. If the preset value is 1000, then when the number of times "jogging shoes" and "dad shoes" appear together is greater than 1000 times, it is considered that the two words "jogging shoes" and "dad shoes" are associated, and "jogging shoes" and "dad shoes" are added to the recommended word library as an associated word combination.

[0119] Step S604: when the number of co-occurrences of a search word in the associated word combination as the search word with sequence number N and the adjacent new search word with sequence number N-1 or N+1 is greater than a preset value, the server adds the new search word to the associated word combination.

[0120] For example, in the search term sequence input by one consumer, the search term "jogging shoes" with sequence number 3 and the adjacent new search term "Forrest Gump shoes" with sequence number 4, and in the search term sequence input by another consumer, the search term "jogging shoes" with sequence number 4 and the adjacent new search term "Forrest Gump shoes" with sequence number 3, it is considered that the number of times "jogging shoes" and "Forrest Gump shoes" appear together is 2 times. If the preset value is 1000, then when the number of times "jogging shoes" and "Forrest Gump shoes" appear together is greater than 1000 times, it is considered that the two words "jogging shoes" and "Forrest Gump shoes" are associated, and the server adds the new search term "Forrest Gump shoes" to the associated word combination.

[0121] Step S606: When a search word in a certain associated word combination in the recommended word library is retrieved and entered by the consumer in the search box, the server displays other search words in the associated word combination on the search result page.

[0122] For example, when a user enters "Forrest Gump shoes" in the search box, the search results are returned, and other search terms in the associated word combination, such as "jogging shoes" and "dad shoes", are also displayed on the search results page to help consumers improve search efficiency. The display location can be below the search box or other locations.

[0123] On the other hand, in one embodiment, a system for mining new keywords for products in e-commerce is provided, comprising a memory, a processor and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the above method.

[0124] The system embodiment and method embodiment of the present invention are based on the same inventive concept and will not be described in detail here.

[0125] In a third aspect, the present invention provides a computer-readable storage medium having a computer program / instruction stored thereon, which implements the steps of the above method when executed by a processor.

[0126] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory may include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory may include random access memory (RAM) or external cache memory. As an illustration and not limitation, RAM is available in many forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRA), memory bus (Rambus) direct RAM (RDRA), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

[0127] The technical features of the above embodiments may be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

Claims

1. A method for mining new keywords for products in e-commerce, characterized in that: The method comprises: Continuously obtain the search terms entered by consumers in the search box at a preset frequency; When it is detected that the number of times a search word is input is greater than a preset threshold and the search word does not belong to an existing keyword, the search word is used as a hot word; Determine whether the hot word is related to the product; When the hot word is related to the product, the hot word is added as a new keyword of the product to the keyword library of the product, so as to search for the product item corresponding to the product according to the new keyword; The step of judging whether the hot word is related to the product specifically includes: After a hot word is input in the search box and within a preset time, when the number of consumers who perform a target operation exceeds a preset number, it is determined that the hot word is related to the product; the target operation includes one or more of the following operations: a collection operation, an order operation, an add-to-cart operation, and a consultation operation for each product item corresponding to the same product, wherein each product item in the same product set is regarded as each product item corresponding to the same product; The target operation includes a first sub-operation and a second sub-operation executed after the first sub-operation; After a hot word is input into the search box and within a preset time, the method further includes: The input hot word is used as the search word with serial number 1. After the hot word is input, the search result page with serial number 1 is returned and the corresponding product items are displayed; Similarly, after inputting the search term with the serial number N, the search result page with the serial number N is returned and the corresponding product items are displayed; N is an integer greater than 2; The first sub-operation includes: The search term with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1; The second sub-operation includes one or more of the following operations: Collect, order, add to cart, and consult for each item of the same product; Among them, the return result corresponding to the search box is the page corresponding to the product entry, and the multiple search terms numbered 1 to N in the first sub-operation are all search terms formed when a user uses the same application or the same e-commerce website once.

2. A method for mining new keywords for products in e-commerce according to claim 1, characterized in that: The following process is used to determine whether the search term with sequence number N appears in the text of the product entry that has been browsed by the consumer on the search result page with sequence number N-1: Obtain the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 as the target text; Match the search term with sequence number N in the target text; When the search term with serial number N is matched in the target text, it is determined that the search term with serial number N appears in the text of the commodity entry on the search result page with serial number N-1 that has been browsed by the consumer.

3. A method for mining new keywords for products in e-commerce according to claim 2, characterized in that: The step of obtaining the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 as the target text specifically includes: The terminal continuously monitors the position of the scroll bar in the page of the product item opened by the consumer; The terminal obtains text elements contained in a page from a starting position of the scroll bar to a corresponding maximum position based on the maximum position reached by the scroll bar; The terminal obtains the text of the commodity item that has been browsed by the consumer on the search result page with sequence number N-1 from the text element as the target text.

4. A method for mining new keywords for commodities in e-commerce according to claim 2, characterized in that: The step of obtaining the text of the product item that has been browsed by the consumer on the search result page with sequence number N-1 as the target text specifically includes: The terminal continuously monitors the position of the scroll bar in the page of the product item opened by the consumer; When the terminal detects the unloading trigger signal of the current page, it sends the maximum position that the scroll bar has reached to the server; The maximum position that the server scroll bar has reached, and the text elements contained in the page corresponding to the maximum position from the scroll bar starting point; The server obtains the text of the commodity entry in the search result page with sequence number N-1 that has been browsed by the consumer from the text element as the target text.

5. The method for mining new keywords for commodities in e-commerce according to claim 1, characterized in that: The method further comprises: When it is detected that the consumer selects the jump text in the search result page numbered N-1 and triggers the jump search mode, the jump text is entered into the search box as the search term numbered N and it is determined that the search term numbered N appears in the text of the product entry that has been browsed by the consumer on the search result page numbered N-1.

6. A method for mining new keywords for products in e-commerce according to claim 2, characterized in that: The method further comprises: When it is detected that the number of co-occurrences of the search term with sequence number N and the adjacent search term with sequence number N-1 is greater than a preset value, the two are added to the recommended word library as a related word combination; When the number of co-occurrences of a search word in the associated word combination as the search word with sequence number N and the adjacent new search word with sequence number N-1 or N+1 is greater than a preset value, the new search word is added to the associated word combination; When a search word in a certain associated word combination entered by a consumer in a recommended word library is retrieved in a search box, other search words in the associated word combination are displayed on a search page or a search result page.

7. A method for mining new keywords for products in e-commerce according to claim 1, characterized in that: The method further comprises: Automatically extract the product tag words corresponding to the product entry according to the title text of the product entry; Two product entries are matched in pairs, and two product entries with at least one identical target product tag word are added to the same product set; Treat each product item in the same product set as each product item corresponding to the same product; The server removes duplicate product tag words of each product item in the same product set and sorts them according to the search volume, and adds the product tag words in the preset ranking as product keywords to the product keyword library; When any keyword in the product keyword library is entered in the search box, all product items in the product collection are displayed on the search result page.

8. A system for mining new keywords for commodities in e-commerce, comprising a memory, a processor and a computer program stored in the memory, characterized in that: The processor executes the computer program to implement the steps of the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program / instruction stored thereon, characterized in that: When the computer program / instructions are executed by a processor, the steps of the method according to any one of claims 1 to 7 are implemented.

Images