Commodity query method and device
By segmenting the query text of e-commerce platforms and constructing a matching term pool, combined with fine-grained data within the products, the display of the list page is dynamically adjusted, solving the problem of poor user experience in e-commerce platforms and achieving the effect of "what you see is what you search for".
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- 阿里巴巴(中国)网络技术有限公司
- Filing Date
- 2025-12-03
- Publication Date
- 2026-05-01
AI Technical Summary
In e-commerce platforms, existing search relevance models rely on text matching mechanisms, which leads to SKUs or attributes that do not correspond to the main image/main title being recalled and displayed, resulting in a serious mismatch at the user perception level and a poor user experience.
By segmenting the query text and building a pool of terms to be matched, and combining this with a pre-built set of fine-grained data within the products for scanning and matching, the display of the list page is dynamically adjusted to ensure that what users see is what they search for.
It improved the user experience, reduced discrepancies between images and text, misleading titles, and anchor text failures, and significantly improved user conversion efficiency and the accuracy of relevance ranking.
Smart Images
Figure CN121958640A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of data processing technology, specifically to a method and apparatus for querying goods. Background Technology
[0002] Currently, e-commerce platforms commonly exhibit the phenomenon of multi-SKU clustering, where a single product (offer ID) contains multiple SKUs, each differing significantly in price, specifications, and attributes (such as color, size, and packaging). While this design improves operational efficiency for merchants, it also introduces serious biases in judging search relevance.
[0003] Current search relevance models mainly rely on text matching mechanisms. When the relevance match hits a SKU or attribute that is not the main image or main title, it will still be displayed as the main image and title, which will lead to a serious mismatch in user perception and a poor user experience. Summary of the Invention
[0004] Based on this, this application provides a product query method and apparatus to enable querying and displaying matching products, thereby improving user experience.
[0005] According to one aspect of this application, a product query method is proposed, comprising: segmenting the query text of the product to obtain word segmentation results, and constructing a pool of terms to be matched based on the word segmentation results; performing product search based on the word segmentation results and performing relevance filtering, and selecting products whose relevance meets preset conditions as matching products; when the title of the matching product does not completely match the word segmentation results, scanning and matching with a pre-constructed fine-grained data set of the product internal data based on the pool of terms to be matched according to matching judgment rules, to obtain a target matching result, wherein the target matching result includes matching SKU or matching product attributes; and dynamically rendering the list page according to the target matching result.
[0006] According to some embodiments, the method further includes: when the title of the matched product completely matches the word segmentation result, displaying the product UI data and price data of the matched product normally.
[0007] According to some embodiments, the query text of the product is segmented to obtain word segmentation results, and a pool of terms to be matched is constructed based on the word segmentation results, including: extracting keywords from the word segmentation results to obtain a keyword set; performing semantic relationship expansion on the keywords in the keyword set to obtain a semantic equivalence set for each keyword; and forming a pool of terms to be matched based on multiple keywords in the keyword set and their semantic equivalence sets.
[0008] According to some embodiments, keywords are extracted from the word segmentation results to obtain a keyword set, including: matching the word segmentation results based on a pre-built candidate pool to obtain keywords, thereby forming a keyword set.
[0009] According to some embodiments, when the title of the matched product does not completely match the word segmentation result, based on the matching judgment rules, a scan and matching is performed with a pre-built fine-grained data set of the product internal data to obtain a target matching result. This includes: when the title of the matched product does not completely match the word segmentation result, matching the term pool to be matched with the product SKUs in the pre-built fine-grained data set of the product internal data to obtain a first matching result; if the first matching result includes a successful match, the matched SKU is taken as the target matching result; if the first matching result includes a failed match, matching the term pool to be matched with the product attributes in the pre-built fine-grained data set of the product internal data to obtain a second matching result; if the second matching result includes a successful match, the matched product attribute is taken as the target matching result.
[0010] According to some embodiments, the list page is dynamically rendered based on the target matching result, including: if the target matching result includes matching product attributes, the relevant data of the matching product attributes is used as the data to be displayed; on the list page, the data to be displayed is displayed in front, thus completing the dynamic rendering of the list page.
[0011] According to some embodiments, the list page is dynamically rendered based on the target matching result, including: if the target matching result includes a matching SKU, determining the target SKU based on sales data; using the relevant data of the target SKU as the data to be displayed; and displaying the data to be displayed in front of the list page to complete the dynamic rendering of the list page.
[0012] According to some embodiments, on the list page, the data to be displayed is displayed in front to complete the dynamic rendering of the list page, including: replacing the product UI data and / or price data displayed on the list page according to the data to be displayed, and completing the dynamic rendering of the list page, wherein the product UI data includes product title, product image and / or default options on the details page.
[0013] According to some embodiments, when the target matching result includes a matching SKU, determining the target SKU based on sales data includes: when the target matching result includes a matching SKU and the matching SKU is a multi-dimensional SKU, extracting the dimensions of terms in the term pool to be matched from the matching SKU to obtain multiple dimensions; sorting the multiple dimensions based on sales data; and determining the target SKU based on the sorting result.
[0014] According to some embodiments, on the list page, the data to be displayed is displayed in front to complete the dynamic rendering of the list page, including: selecting a preset number of dimensions as the target dimensions from multiple dimensions according to the sorting result; extracting the images of the data to be displayed in the target dimensions as the images to be displayed; concatenating the titles of the data to be displayed from multiple dimensions according to the sorting result as the titles to be displayed; and replacing the images and titles displayed on the list page with the images and titles to be displayed.
[0015] According to some embodiments, when the target matching result includes a matching SKU, the target SKU is determined based on sales data, including: when the target SKU lacks a key description, selecting the next SKU as the new target SKU.
[0016] According to some embodiments, when the second matching result includes a successful match, the matched product attribute is used as the target matching result, including: when the target matching result includes a matched SKU and the matched SKU is a multi-dimensional SKU, cross-validation is performed on the multi-dimensional SKU based on the terms in the term pool to be matched.
[0017] According to one aspect of this application, a product query device includes: a term construction unit, configured to segment the query text of the product to obtain word segmentation results, and construct a term pool to be matched based on the word segmentation results; a first matching unit, configured to perform product search based on the word segmentation results, and perform relevance filtering, selecting products whose relevance meets preset conditions as matched products; a second matching unit, configured to, when the title of the matched product does not completely match the word segmentation results, scan and match the term pool to be matched with a pre-constructed fine-grained data set of the product to obtain a target matching result, wherein the target matching result includes matching SKU or matching product attributes; and a dynamic rendering unit, configured to dynamically render the list page based on the target matching result.
[0018] According to one aspect of this application, an electronic device is provided, comprising: one or more processors; a storage device for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the method as described above.
[0019] According to one aspect of this application, a computer-readable medium is provided that stores a computer program or instructions thereon, which, when executed by a processor, implement the method as described above.
[0020] Through the embodiments provided in this application, the query text for products is segmented to obtain word segmentation results. Based on the word segmentation results, a high-recall pool of terms to be matched is constructed. Based on the word segmentation results, product search is performed to obtain matching products. If the title of the matching product does not completely match the word segmentation results, the current query and display do not match. In this case, based on the pool of terms to be matched, further precise matching is performed with the fine-grained information inside the product to obtain the target matching result. Based on the target matching result, the list page is driven to realize the corresponding dynamic rendering, so as to realize the query and display of matching products, ensuring that what the user sees is what they search for, and improving the user experience. Attached Figure Description
[0021] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application.
[0022] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings, without exceeding the scope of protection claimed by this application.
[0023] Figure 1 A flowchart illustrating the product query method provided in this application embodiment; Figure 2 The flowchart illustrates how to segment the query text for products provided in this application embodiment to obtain word segmentation results, and how to construct a pool of terms to be matched based on the word segmentation results. Figure 3 The flowchart provided in this application embodiment shows that when the title of the matched product does not completely match the word segmentation result, the target matching result is obtained by scanning and matching the pre-built fine-grained data set inside the product based on the matching judgment rules and the term pool to be matched. Figure 4 A flowchart illustrating the dynamic rendering of a list page based on target matching results, provided in this embodiment of the application; Figure 5 A flowchart illustrating the dynamic rendering of a list page based on target matching results, provided in this embodiment of the application; Figure 6 A flowchart for determining the target SKU based on sales data when the target matching result includes a matching SKU, provided as an embodiment of this application; Figure 7 A block diagram of a product query device provided in an embodiment of this application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation
[0024] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of this application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without creative effort are within the scope of protection of this application.
[0025] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.
[0026] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.
[0027] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.
[0028] It should be understood that although the terms first, second, third, etc., may be used herein to describe various components, these components should not be limited by these terms. These terms are used to distinguish one component from another. Therefore, the first component discussed below may be referred to as the second component without departing from the teachings of this application. As used herein, the term "and / or" includes all combinations of any one and more of the associated listed items.
[0029] Current search relevance models primarily rely on text matching mechanisms. In implementation, they comprehensively evaluate the relevance of the query to the following information: product title, SKU (Stock Keeping Unit) name and description, attributes and attribute values (e.g., "Material: Cotton", "Target Audience: Children"), merchant tags, and multi-dimensional workflow tags. When the relevance match hits an SKU or attribute that is not the main image / title, the display will still show the main image and title, leading to a severe mismatch in user perception. Specifically: 1) Irrelevant display image. This means that the main product image shows a style that does not match the SKU (such as the red version), but because another SKU (such as the blue version) matches the query, the product will be recalled and judged as highly relevant. However, since the main image is still displayed, users will find that the image and text do not match after clicking.
[0030] 2) The title does not reflect the matching relationship. That is, the main title of the product does not contain the query keyword, and the semantics cannot reflect the relevance, but it is recalled because of the text matching of a certain SKU, which misleads the user's expectations.
[0031] 3) Displayed price ≠ matching SKU price. The price displayed on the product listing page is for the lower-priced SKU, while the actual price of the SKU matching the query is higher, creating a "low-price inducement" feeling for users when they enter the details page.
[0032] 4) List to OD anchoring failure. This means that after a user clicks from the list page to enter the product detail page (OD), the page does not automatically anchor to the SKU that matches the query, requiring manual searching and increasing decision-making costs.
[0033] Based on this, this application proposes a method and apparatus for querying goods.
[0034] For specific implementation details, please refer to the following examples.
[0035] Figure 1 A flowchart illustrating the product query method provided in this application embodiment. Figure 1 As shown, the method includes steps S110-S140.
[0036] In step S110, the query text of the product is segmented to obtain word segmentation results, and a pool of terms to be matched is constructed based on the word segmentation results.
[0037] The query text for a product is the query text that the user enters into the search system of the target e-commerce platform.
[0038] To improve the search system's ability to understand users' true intentions, the original query text is deeply analyzed to identify keywords (terms) that express core needs, and a pool of terms to be matched with rich semantic generalization capabilities is constructed.
[0039] In the deep analysis process, the query text is first segmented to obtain word segmentation results. Based on the word segmentation results, multi-dimensional intent features are extracted and expanded to obtain a pool of terms to be matched.
[0040] In step S120, a product search is performed based on the word segmentation results, and relevance filtering is conducted. Products whose relevance meets the preset conditions are selected as matching products.
[0041] The query text after word segmentation is compared with the corresponding products on the platform to determine the text relevance and obtain the matching products.
[0042] It is understood that this step can be understood as relevance matching before the improvements of this application: using a text matching mechanism, comprehensively evaluate the relevance of word segmentation results with the product's main title, SKU name and description, attributes and attribute values (such as "material: cotton", "applicable audience: children"), merchant tags and multi-dimensional tags of the workflow, and extract products whose relevance meets the preset conditions as matching products.
[0043] The preset conditions can be set according to the circumstances. This application does not impose any restrictions on this.
[0044] In step S130, if the title of the matched product does not completely match the word segmentation result, the matching judgment rule is used to scan and match the pre-built fine-grained data set of the product based on the pool of terms to be matched, so as to obtain the target matching result. The target matching result includes matching SKU or matching product attributes.
[0045] If the word segmentation results of the product title and the query text do not match perfectly, the target matching result is obtained by matching each word in the word pool to be matched with the pre-built fine-grained data set of the product, according to the matching judgment rules.
[0046] Understandably, the pre-built fine-grained data set of products includes product-level dump data, containing fine-grained information such as titles, prices, inventory, attributes, and images for each SKU.
[0047] According to the example embodiment, step S130 is performed using the Ha3 Summary plug-in matching engine.
[0048] In practice, based on the query or query rewrite results in the term pool to be matched, rule matching is performed with each SKU in the fine-grained data set within the product. The output is the SKU ID of the matched SKU or the matching source field (e.g., SKU title, CPV (category-prop-value, product category-attribute-attribute value)). The most relevant SKU or the most relevant attribute is automatically identified based on the matching results. The most relevant SKU is recorded as the matched SKU, and the most relevant attribute is recorded as the matched product attribute, thus obtaining the target matching result.
[0049] Furthermore, the matching determination rules include: a match is considered successful if any expanded term (original word or synonym / hypothesis) in the term pool to be matched is found, without requiring complete consistency, thus satisfying the diversity of user searches.
[0050] In step S140, the list page is dynamically rendered based on the target matching result.
[0051] Based on the target matching results, the corresponding dynamic rendering strategy of the list page is triggered, ultimately achieving the following: the display elements such as main image, title, price, and tags are all optimized and adjusted around the "actual matching point" to ensure that what the user sees is what they search for.
[0052] This application constructs a closed-loop search query system that integrates semantic matching and tracing with dynamic front-end display. By analyzing the matching path between query segmentation granularity and fine-grained information within the product (SKU title, attribute value), it ultimately drives the dynamic adjustment of the main image, title, price, and OD anchoring, significantly improving the user's perception of search results consistency and conversion efficiency, ensuring that what the user sees is what they search for.
[0053] Through the embodiments provided in this application, the query text for products is segmented to obtain word segmentation results. Based on the word segmentation results, a high-recall pool of terms to be matched is constructed. Based on the word segmentation results, product search is performed to obtain matching products. If the title of the matching product does not completely match the word segmentation results, the current query and display do not match. In this case, based on the pool of terms to be matched, further precise matching is performed with the fine-grained information inside the product to obtain the target matching result. Based on the target matching result, the list page is driven to realize the corresponding dynamic rendering, so as to realize the query and display of matching products, ensuring that what the user sees is what they search for, and improving the user experience.
[0054] According to some embodiments, the method further includes step S150.
[0055] In step S150, if the title of the matched product completely matches the word segmentation result, the product UI data and price data of the matched product are displayed normally.
[0056] Specifically, if the product title and the word segmentation results of the query text match perfectly, the product's UI data (including the main image and title) and price data will be displayed normally on the list page, and the process will end.
[0057] According to some embodiments, refer to Figure 2 In step S110, the query text of the product is segmented to obtain the word segmentation results, and a pool of terms to be matched is constructed based on the word segmentation results. This can be achieved through steps S210-S230.
[0058] In step S210, keywords are extracted from the word segmentation results to obtain a keyword set.
[0059] In this step, after the query text is segmented into words, the core words with a clear filtering intent are identified based on the segmentation results and used as keywords. Multiple keywords form a keyword set.
[0060] According to the example implementation, multi-dimensional intent features such as color, style, design, and model are extracted based on the word segmentation results. Core words with clear filtering intent are extracted, such as style words ("trendy", "fashion"), color words ("red", "light blue"), function words ("pure cotton", "linen"), and IP words ("Nezha", "My Little Pony").
[0061] In step S220, the semantic relationships of the keywords in the keyword set are expanded to obtain a semantically equivalent set for each keyword.
[0062] Semantic relation expansion is performed on the keywords in the keyword set. That is, for each keyword, its synonyms, near-synonyms, hyponyms, and common expression variations are associated to form a semantically equivalent set.
[0063] In practical implementation, it can be achieved based on a pre-built matching knowledge base. Specifically, the matching knowledge base is a pre-set knowledge base of synonyms and hyponyms corresponding to keywords. It can be implemented based on a data generation model or other methods, which are not limited in this application.
[0064] In step S230, a pool of terms to be matched is formed based on multiple keywords in the keyword set and their semantic equivalent sets.
[0065] A pool of terms to be matched is formed by using multiple keywords in the keyword set and their semantically equivalent sets.
[0066] This application embodiment starts from the user's search intent, deeply analyzes the semantic roles of different terms in the query text to obtain keywords, and expands them through synonyms and hyponyms to obtain a pool of terms to be matched.
[0067] According to some embodiments, in step S210, keywords are extracted from the word segmentation results to obtain a keyword set, which can be specifically implemented through step S211.
[0068] In step S211, based on the pre-built candidate pool, the word segmentation results are matched to obtain keywords, thereby forming a keyword set.
[0069] In practice, the query text for products is segmented using term weight. After term weight segmentation, words of different parts of speech will have different tags, such as color, material, style, IP, etc., which are used as the segmentation results. Based on this, the segmentation results are matched with a pre-built candidate pool to extract keywords from the candidate pool and form a keyword set.
[0070] The pre-built candidate pool includes commonly used keywords of various parts of speech.
[0071] According to some embodiments, refer to Figure 3 In step S130, if the title of the matched product does not completely match the word segmentation result, the matching judgment rule is used to scan and match the pre-built fine-grained data set of the product based on the pool of terms to be matched, so as to obtain the target matching result. This can be achieved through steps S310-S340.
[0072] In step S310, if the title of the matched product does not completely match the word segmentation result, the term pool to be matched is matched with the product SKU in the pre-built fine-grained data set of the product to obtain the first matching result.
[0073] If the word segmentation results of the product title and the query text do not match perfectly, each term in the term pool to be matched is matched with the SKU of the product in the fine-grained data set inside the product. If a match is successful, the first matching result is recorded as a successful match; if a match is unsuccessful, the first matching result is recorded as a failed match.
[0074] In step S320, if the first matching result includes a successful match, the matched SKU is taken as the target matching result.
[0075] In step S330, if the first matching result includes a matching failure, the term pool to be matched is matched with the product attributes in the pre-built fine-grained data set of the product to obtain a second matching result.
[0076] If the first matching result includes a match failure, that is, no corresponding SKU is matched, then the term in the term pool to be matched is matched with the product attributes in the fine-grained data set inside the product. If the match is successful, the second matching result is recorded as a successful match; if the match is unsuccessful, the second matching result is recorded as a failed match.
[0077] In step S340, if the second matching result includes a successful match, the matched product attribute is taken as the target matching result.
[0078] If the second matching result includes a successful match, that is, a match for the corresponding attribute, then the matched product attribute will be used as the target matching result for subsequent display of the attribute title.
[0079] Furthermore, if the second matching result includes a match failure, i.e. no matching of the corresponding attribute, the main image and title of the product will be displayed normally, and the process will end.
[0080] This application specifies the specific steps for matching and sets specific rules for matching in different tiers to ensure that what the user sees is what they search for as much as possible.
[0081] According to some embodiments, refer to Figure 4 In step S140, the list page is dynamically rendered based on the target matching result, which can be implemented through steps S410-S420.
[0082] In step S410, if the target matching result includes matching product attributes, the relevant data of the matching product attributes is used as the data to be displayed.
[0083] If the target matching result includes matching product attributes, that is, only matching product attributes and attribute values, then the relevant data of the matching product attributes will be used as the data to be displayed.
[0084] In step S420, the data to be displayed is displayed in front of the list page, completing the dynamic rendering of the list page.
[0085] Dynamic rendering includes dynamic rendering strategies such as attribute exposure.
[0086] According to some embodiments, dynamic rendering includes at least: displaying the matching product attributes in the first position on the list page.
[0087] According to the example implementation, if the query text is "pure cotton socks", and a certain term "pure cotton" in the term pool can only be matched by the CPV attribute (such as "material: pure cotton"), then the attribute value will be displayed in the first position on the list page.
[0088] According to some embodiments, dynamic rendering may also include: highlighting matching product attributes in the product tag area.
[0089] According to the example implementation, if the query text is "children's picture books" and the target audience is "children", then "suitable for children" will be highlighted in the product tag area.
[0090] According to some embodiments, refer to Figure 5 In step S140, the list page is dynamically rendered based on the target matching result, which can be implemented through steps S510-S530.
[0091] In step S510, if the target matching result includes a matching SKU, the target SKU is determined based on the sales data.
[0092] If the target matching result includes matching SKUs, that is, only the SKU copy is matched, then the target SKU that the user is most likely to be interested in is determined based on sales data.
[0093] Specifically, when the matched SKU is a single-dimensional SKU, the matched SKUs are sorted according to sales data, and the target SKU is determined in descending order of sales. When the matched SKU is a multi-dimensional SKU, it is determined whether the missing terms in the title are concentrated in one dimension, and the target SKU is determined according to the dimension.
[0094] In step S520, the relevant data of the target SKU is used as the data to be displayed.
[0095] In step S530, the data to be displayed is displayed in front of the list page to complete the dynamic rendering of the list page.
[0096] Dynamic rendering includes strategies such as image replacement, title replacement, attribute exposure, price synchronization, and OD anchoring.
[0097] Prioritize displaying relevant data for the target SKU, including but not limited to changing images, titles, price synchronization, and OD anchoring, to ensure that what users see is what they search for.
[0098] According to some embodiments, in step S420 or step S530, the data to be displayed is displayed in front of the list page to complete the dynamic rendering of the list page, which can be specifically implemented through step S101.
[0099] In step S101, the product UI data and / or price data displayed on the list page are replaced according to the data to be displayed, and the dynamic rendering of the list page is completed. The product UI data includes the product title, product image and / or default options on the details page.
[0100] In this embodiment, dynamic rendering includes at least: 1) Image replacement strategy: When a non-main image SKU is matched, the image of the SKU most relevant to the query is displayed as the main image of the list. 2) Title change strategy: Based on the original product title, insert keywords or phrases that are strongly related to the query (from SKU, CPV) to create a more targeted display title. 3) Attribute Transparency and Exposure: Key attributes that match the match (such as "waterproof" or "full-length") will be displayed as tags on the list page to enhance information transparency.
[0101] According to some embodiments, if the target matching result includes matching product attributes, dynamic rendering includes at least: attribute exposure, that is, displaying the matching product attributes in the first order on the list page.
[0102] According to some embodiments, if the target matching result includes a matching SKU, and the matching SKU is a single-dimensional SKU, then dynamic rendering includes at least: 1) Image strategy: Prioritize displaying images that match the SKU (if no image is available, retain the original main image); 2) Title Strategy: Prioritize displaying the complete description of the SKU; if there is remaining space, add the main product title. 3) Pricing Strategy: Display the actual price of the matching SKU; 4) OD Anchoring: Clicking this will take you to the details page, where the corresponding SKU will be selected by default.
[0103] According to some embodiments, refer to Figure 6 In step S510, if the target matching result includes a matching SKU, the target SKU is determined based on the sales data, which can be achieved through steps S610-S630.
[0104] In step S610, if the target matching result includes a matching SKU and the matching SKU is a multi-dimensional SKU, the dimensions of the terms in the term pool to be matched are extracted from the matching SKU to obtain multiple dimensions.
[0105] Specifically, the specification dimension to which each matching term belongs in the term pool to be matched is identified and extracted, resulting in multiple dimensions.
[0106] In step S620, the sales data is sorted according to multiple dimensions.
[0107] In step S630, the target SKU is determined based on the sorting results.
[0108] The sales data is sorted according to multiple dimensions, and the combined SKUs are determined in descending order of sales volume, which serve as the target SKUs.
[0109] According to some embodiments, in step S530, the data to be displayed is displayed in front of the list page to complete the dynamic rendering of the list page, which can be specifically implemented through steps S531-S534.
[0110] In step S531, based on the sorting results, a preset number of dimensions are selected from multiple dimensions as the target dimensions.
[0111] According to the example implementation, the first matching dimension among multiple belonging dimensions is selected as the target belonging dimension.
[0112] In step S532, the images of the data to be displayed in the dimension to which the target belongs are extracted and used as the images to be displayed.
[0113] According to the example embodiment, the image strategy in the dynamic rendering strategy includes: extracting the SKU image corresponding to the first matching dimension as the image to be displayed, so as to give priority to display.
[0114] In step S533, based on the sorting result, the multi-dimensional titles of the data to be displayed are concatenated to form the title to be displayed.
[0115] According to the example implementation, the title strategy in the dynamic rendering strategy includes: concatenating the specification text of the multi-dimensional title (e.g., "Red + Thickened") in the order of matching dimensions as the title to be displayed.
[0116] In step S534, the image and title displayed on the list page are replaced according to the image to be displayed and the title to be displayed.
[0117] Prioritize displaying the images and titles to be shown, and replace the content in the list page.
[0118] According to some embodiments, in step S510, if the target matching result includes a matching SKU, the target SKU is determined based on sales data, which can be specifically implemented through step S511.
[0119] In step S511, if the target SKU lacks a key description, the next SKU is selected as the new target SKU.
[0120] In other words, if a certain SKU is missing other key term descriptions, the SKU is skipped, and the next better match is selected as the target SKU in order of sales volume, thereby preventing mismatches.
[0121] According to some embodiments, in step S340, if the second matching result includes a successful match, the matched product attribute is taken as the target matching result, which can be specifically implemented through step S341.
[0122] In step S341, if the target matching result includes a matching SKU and the matching SKU is a multi-dimensional SKU, cross-validation is performed on the multi-dimensional SKU based on the terms in the term pool to be matched.
[0123] When the matched SKU is a multi-dimensional SKU, that is, multiple SKUs match the same term, other terms in the query (i.e. terms in the term pool to be matched) are introduced to cross-validate all candidate SKUs, thereby preventing false matches.
[0124] In one specific embodiment, the product query method provided in this application is configured on an online trading platform. After calculation, it effectively improves the experience bottlenecks such as "mismatch between images and text, misleading titles, and anchor failure" after going live, significantly enhances the user experience, and drives a 2.36% (significant) increase in valuable GMV (total merchandise transaction volume) and a 0.93% increase in GPM (gross merchandise sales per thousand impressions), achieving an experience upgrade from "calculable" to "visible" relevance ranking.
[0125] The following describes an apparatus embodiment of this application, which can be used to perform the method embodiment of this application. For details not disclosed in the apparatus embodiment of this application, please refer to the method embodiment of this application.
[0126] Figure 7 A block diagram of a product query device according to an exemplary embodiment is shown.
[0127] Figure 7 The apparatus shown can perform the product query method described above according to the embodiments of this application.
[0128] like Figure 7 As shown, the product query device may include: See Figure 7 Referring to the preceding description, the term construction unit 710 is used to segment the query text of the product to obtain the word segmentation results, and to construct a pool of terms to be matched based on the word segmentation results.
[0129] The first matching unit 720 is used to search for products based on word segmentation results and perform relevance filtering, selecting products whose relevance meets preset conditions as matching products.
[0130] The second matching unit 730 is used to perform scanning and matching with a pre-built fine-grained data set of the product based on the matching judgment rules and the term pool to be matched when the title of the matched product does not completely match the word segmentation result, so as to obtain the target matching result. The target matching result includes matching SKU or matching product attributes.
[0131] The dynamic rendering unit 740 is used to dynamically render the list page based on the target matching results.
[0132] The device performs functions similar to those described above; other functions are described in the preceding descriptions and will not be repeated here.
[0133] This application discloses an electronic device, including: a processor; and a memory storing a computer program, which, when executed by the processor, causes the processor to execute the above-described instruction generation method.
[0134] For example, refer to Figure 8 , Figure 8 The illustrated electronic device 800 includes a processor 801 and a memory 803. The processor 801 and the memory 803 are connected, for example, via a bus 802. Optionally, the electronic device 800 may also include a transceiver 804. It should be noted that in practical applications, the transceiver 804 is not limited to one type, and the structure of this electronic device 800 does not constitute a limitation on the embodiments of this application.
[0135] Processor 801 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It can implement or execute the various exemplary logic blocks, modules, and circuits described in this application. Processor 801 may also be a combination that implements computational functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, etc.
[0136] Bus 802 may include a pathway for transmitting information between the aforementioned components. Bus 802 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus, etc. Bus 802 can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.
[0137] The memory 803 may be a ROM (Read Only Memory) or other type of static storage device capable of storing static information and instructions, RAM (Random Access Memory) or other type of dynamic storage device capable of storing information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), CD-ROM (Compact Disc Read Only Memory) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital universal optical discs, Blu-ray discs, etc.), magnetic disk storage media or other magnetic storage devices, or any other storage medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto.
[0138] The memory 803 stores application code that executes the scheme of this application, and its execution is controlled by the processor 801. The processor 801 executes the application code stored in the memory 803 to implement the content shown in the foregoing method embodiments.
[0139] Figure 8 The electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.
[0140] This application discloses a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, causes the processor to execute an instruction generation method.
[0141] It should be understood that although the steps in the flowcharts of the accompanying figures are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some steps in the flowcharts of the accompanying figures may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily completed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be performed alternately or in turn with other steps or at least some of the sub-steps or stages of other steps.
[0142] The above are only some embodiments of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.
Claims
1. A method for querying products, characterized in that, include: The query text of the product is segmented to obtain word segmentation results, and a pool of terms to be matched is constructed based on the word segmentation results; Based on the word segmentation results, a product search is performed, and relevance filtering is conducted. Products whose relevance meets preset conditions are selected as matching products. If the title of the matched product does not completely match the word segmentation result, the matching judgment rule is used to scan and match the pre-built fine-grained data set of the product based on the pool of terms to be matched, so as to obtain the target matching result, wherein the target matching result includes matching SKU or matching product attribute; Based on the target matching results, the list page is dynamically rendered.
2. The method according to claim 1, characterized in that, The method further includes: If the title of the matched product completely matches the word segmentation result, the product UI data and price data of the matched product will be displayed normally.
3. The method according to claim 1, characterized in that, The query text for the product is segmented to obtain word segmentation results, and a pool of terms to be matched is constructed based on the word segmentation results, including: Keywords are extracted from the word segmentation results to obtain a keyword set; Semantic relation expansion is performed on the keywords in the keyword set to obtain a semantically equivalent set for each keyword; A pool of terms to be matched is formed based on multiple keywords in the keyword set and their semantic equivalent sets.
4. The method according to claim 3, characterized in that, Keywords are extracted from the word segmentation results to obtain a keyword set, including: Based on a pre-built candidate pool, the word segmentation results are matched to obtain the keywords, thereby forming the keyword set.
5. The method according to claim 1, characterized in that, If the title of the matched product does not completely match the word segmentation result, based on the matching judgment rules and the pool of terms to be matched, a scan and matching process is performed with a pre-built fine-grained data set within the product to obtain the target matching result, including: If the title of the matched product does not completely match the word segmentation result, the pool of terms to be matched is matched with the product SKUs in the pre-built fine-grained data set of the product to obtain the first matching result; If the first matching result includes a successful match, the matched SKU obtained from the match will be used as the target matching result; If the first matching result includes a matching failure, the term pool to be matched is matched with the product attributes in the pre-built fine-grained data set of the product to obtain a second matching result; If the second matching result includes a successful match, the matched product attribute is taken as the target matching result.
6. The method according to claim 5, characterized in that, Based on the target matching results, the list page is dynamically rendered, including: If the target matching result includes matching product attributes, the relevant data of the matching product attributes will be used as the data to be displayed. On the list page, the data to be displayed is displayed in front of the view, thus completing the dynamic rendering of the list page.
7. The method according to claim 5, characterized in that, Based on the target matching results, the list page is dynamically rendered, including: If the target matching result includes a matching SKU, the target SKU is determined based on sales data; The relevant data of the target SKU will be used as the data to be displayed. On the list page, the data to be displayed is displayed in front of the view, thus completing the dynamic rendering of the list page.
8. The method according to claim 6 or 7, characterized in that, On the list page, the data to be displayed is displayed in front of the viewer to complete the dynamic rendering of the list page, including: Based on the data to be displayed, replace the product UI data and / or price data displayed on the list page to complete the dynamic rendering of the list page. The product UI data includes the product title, product image, and / or default options on the details page.
9. The method according to claim 7, characterized in that, If the target matching result includes a matching SKU, the target SKU is determined based on sales data, including: When the target matching result includes a matching SKU and the matching SKU is a multi-dimensional SKU, the dimensions to which the terms in the term pool to be matched are extracted from the matching SKU to obtain multiple dimensions; Based on the sales data, sort the multiple related dimensions; The target SKU is determined based on the sorting results.
10. The method according to claim 9, characterized in that, On the list page, the data to be displayed is displayed in front of the viewer to complete the dynamic rendering of the list page, including: Based on the sorting results, select the first preset number of belonging dimensions from the multiple belonging dimensions as the target belonging dimension; Extract the image of the data to be displayed in the dimension to which the target belongs, and use it as the image to be displayed; Based on the sorting results, the multi-dimensional titles of the data to be displayed are concatenated to form the title to be displayed; Replace the image and title displayed on the list page with the image and title to be displayed.
11. The method according to claim 7, characterized in that, If the target matching result includes a matching SKU, the target SKU is determined based on sales data, including: If the target SKU lacks a key description, select the next SKU as the new target SKU.
12. The method according to claim 5, characterized in that, If the second matching result includes a successful match, the matched product attributes are used as the target matching result, including: If the target matching result includes a matching SKU and the matching SKU is a multi-dimensional SKU, cross-validation is performed on the multi-dimensional SKU based on the terms in the term pool to be matched.
13. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-12.
14. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed by a processor, they implement the method as described in any one of claims 1-12.