Cross-platform e-commerce resource dynamic matching search method and system

Abstract

The invention relates to the technical field of resource matching, and discloses a cross-platform e-commerce resource dynamic matching search method, which comprises the following steps: carrying out real-time analysis on commodity description information from different e-commerce platforms, extracting a core attribute of a commodity and generating a unified semantic representation; when cross-platform data conflicts occur, a conflict resolution strategy is dynamically generated, and the conflict resolution strategy is executed by a local preprocessing module; acquiring data content after conflict resolution, selecting a target edge node based on the data content in combination with user geographical location information, and dynamically configuring a cache data set of the target edge node; and generating corresponding multi-dimensional matching features based on user intention information in a natural query language in combination with the commodity data in the configured cache, determining a candidate commodity set and a corresponding search result based on the multi-dimensional matching features, obtaining feedback information of the user on the search result, and updating the intention analysis model and the matching strategy based on the feedback information. According to the invention, efficient cross-platform resource matching can be realized.

Description

Technical Field

[0001] The present application relates to the technical field of resource matching, and in particular to a method and system for dynamic matching and searching of cross-platform e-commerce resources. Background Art

[0002] With the globalization of e-commerce, the heterogeneity of cross-platform product data (such as format differences and semantic ambiguity) and the need for real-time performance have become increasingly prominent. Traditional methods rely on manual rules or static mappings, making it difficult to dynamically adapt to multi-platform data structures, resulting in low matching efficiency and high latency. Furthermore, existing technologies for resolving data conflicts (such as price and inventory inconsistencies) often rely on fixed priorities or local optimization, lacking dynamic coordination mechanisms based on user behavior and real-time needs. Furthermore, the lack of collaborative capabilities between edge computing and central servers further exacerbates system load and response delays.

[0003] Current cross-platform e-commerce systems generally adopt independent modular designs, with separate functions (such as data standardization, conflict detection, and order synchronization) making closed-loop optimization difficult. As can be seen from the above, achieving efficient cross-platform resource matching remains a challenge. Summary of the Invention

[0004] In order to achieve efficient cross-platform resource matching, the present application provides a cross-platform e-commerce resource dynamic matching search method and system.

[0005] In the first aspect, the present application provides a cross-platform e-commerce resource dynamic matching search method, which adopts the following technical solutions: A cross-platform e-commerce resource dynamic matching search method, comprising: At the data access end, product descriptions from different e-commerce platforms are parsed in real time to extract core product attributes and generate a unified semantic representation. Product descriptions include text and images, and core attributes include name, specifications, and price. The semantic representation establishes semantic mapping relationships between cross-platform attributes based on a context-aware synonym replacement mechanism and attribute association rules. The standardized data is then transmitted to the central database through a distributed consistency synchronization mechanism. Based on the product attribute information in the standardized data and the user's historical behavior characteristics, a conflict resolution strategy is dynamically generated when a cross-platform data conflict occurs. The conflict resolution strategy is executed by a local pre-processing module deployed on the edge node, wherein the local pre-processing module quickly responds to sudden data and batch processes periodic data; After the conflict is resolved, the conflict-resolved data content is obtained, and in the computing architecture composed of the edge node and the central server, the target edge node is selected based on the data content and the user's geographic location information, and the cache data set of the target edge node is dynamically configured; Obtain natural query language input by the user, parse corresponding user intent information based on the natural query language, generate corresponding multi-dimensional matching features based on the user intent information combined with product data in the configured cache, determine a set of candidate products through approximate nearest neighbor retrieval based on the multi-dimensional matching features, output corresponding search results based on the candidate product set, obtain user feedback information on the search results, and update the intent parsing model and matching strategy based on the feedback information.

[0006] Optionally, when generating the unified semantic representation, the method further includes: Correlate and map product core attributes with preference tags in user historical behavior to extract corresponding cross-platform behavior patterns; Based on the cross-platform behavior pattern, corresponding dynamic attribute weights are generated, the priorities of different attributes in the semantic representation are calculated through the dynamic attribute weight generation module, and the priorities of different attributes in the semantic representation are dynamically updated through the attribute priority adjustment module.

[0007] Optionally, in the conflict resolution strategy, the method further includes: Obtaining the source platform of the conflicting data and the geographic location information of the target user, and extracting the correlation features between the source platform and the region based on the source platform and the geographic location information; A cross-region data consistency verification rule is dynamically generated based on the association feature, wherein the cross-region data consistency verification rule preferentially adopts the platform data with the highest matching degree with the user's region as a conflict resolution benchmark.

[0008] Optionally, when dynamically configuring the cache data set of the target edge node, the method further includes: Extracting cross-platform price fluctuation trends and user purchase periodicity characteristics of products, and prioritizing cache updates of products based on the cross-platform price fluctuation trends and user purchase periodicity characteristics; Based on the cache update priority, a short-cycle forced refresh strategy is executed for commodities with frequent price fluctuations, and a long-cycle passive update strategy is executed for commodities with stable prices.

[0009] Optionally, when generating the multi-dimensional matching feature, the method further includes: Retrieving user intent information and cross-platform attribute association rules, and generating corresponding dynamic attribute filtering conditions based on the user intent information and the cross-platform attribute association rules; Based on the dynamic attribute filtering conditions, the range of candidate products is gradually narrowed down through a hierarchical screening mechanism, with priority given to matching the core attributes explicitly mentioned in the user's intent.

[0010] Optionally, when updating the matching strategy based on the feedback information, the method further includes: Obtaining user operations on search results, where user operations include clicks, purchases, and favorites on search results, obtaining cross-platform attribute features corresponding to products, performing correlation analysis based on the user operations and the cross-platform attribute features, and generating user behavior-driven attribute correlation enhancement rules; Based on the attribute association enhancement rules, the synonym replacement mechanism and the attribute alignment logic are updated through the semantic mapping optimization mechanism.

[0011] Optionally, when parsing the natural query language, the method further includes: Acquire semantic mapping relationships between cross-platform attributes, and generate a multi-language compatible standardized query template based on the semantic mapping relationships between the natural query language and the cross-platform attributes; Based on the standardized query template, the attribute description habits of different platforms are automatically adapted through a semantic expansion mechanism to generate cross-platform compatible query features.

[0012] Secondly, this application provides a cross-platform e-commerce resource dynamic matching and search system, which adopts the following technical solutions: A cross-platform e-commerce resource dynamic matching and search system, comprising: The semantic mapping relationship establishment module parses product descriptions from different e-commerce platforms in real time at the data access end, extracts the core attributes of the products, and generates a unified semantic representation. Product descriptions include text and images, and core attributes include name, specifications, and price. The semantic representation uses a context-aware synonym replacement mechanism and attribute association rules to establish semantic mapping relationships between cross-platform attributes. The standardized data is then transmitted to the central database through a distributed consistency synchronization mechanism. A conflict resolution strategy generation module, which dynamically generates a conflict resolution strategy when a cross-platform data conflict occurs based on the product attribute information in the standardized data and the user's historical behavior characteristics. The conflict resolution strategy is executed by a local pre-processing module deployed on the edge node, wherein the local pre-processing module quickly responds to sudden data and batch processes periodic data; A target edge node selection module, after conflict resolution, obtains the conflict-resolved data content, selects a target edge node based on the data content combined with the user's geographic location information in a computing architecture jointly constructed by edge nodes and central servers, and dynamically configures the cache data set of the target edge node; An update module obtains natural query language input by a user, parses corresponding user intent information based on the natural query language, generates corresponding multi-dimensional matching features based on the user intent information combined with product data in a configured cache, determines a set of candidate products through approximate nearest neighbor retrieval based on the multi-dimensional matching features, outputs corresponding search results based on the set of candidate products, obtains user feedback on the search results, and uses the feedback information to update the intent parsing model and matching strategy.

[0013] In a third aspect, this application provides a cross-platform e-commerce resource dynamic matching and search system, which adopts the following technical solutions: A cross-platform e-commerce resource dynamic matching and search system includes a processor running a program of any one of the above-mentioned cross-platform e-commerce resource dynamic matching and search methods.

[0014] In a fourth aspect, the present application provides a storage medium, which adopts the following technical solution: A storage medium storing a program for the cross-platform e-commerce resource dynamic matching and search method described in any one of the above.

[0015] In summary, this application includes at least one of the following beneficial technical effects: Through dynamic semantic mapping and distributed consistency synchronization, efficient cross-platform product data standardization is achieved. On the data access side, the system combines a context-aware synonym replacement mechanism and attribute association rules to convert product descriptions (such as text and images) from heterogeneous platforms into a unified semantic representation, resolving matching barriers caused by differences in attribute definitions across platforms. Furthermore, based on the distributed consistency synchronization mechanism, only differential data is transmitted, rather than the full data set, significantly reducing storage pressure on the central database and data synchronization latency.

[0016] During the resource matching phase, this solution achieves efficient cross-platform matching through multi-dimensional feature fusion and dynamic strategy optimization. First, the system combines user intent analysis with product features in cached data to generate multi-dimensional matching criteria. A hierarchical screening mechanism prioritizes matching core attributes explicitly mentioned by the user (such as name and specifications), gradually narrowing the candidate product list and avoiding invalid matches. Second, based on correlation analysis between user feedback (clicks, purchases, and favorites) and cross-platform attribute features, the system dynamically generates attribute association enhancement rules and reversely optimizes semantic mapping logic (such as synonym replacement and attribute alignment), forming a closed-loop improvement loop. Finally, through multi-language compatible standardized query templates and a semantic expansion mechanism, the system automatically adapts to attribute description conventions across different platforms, eliminating the impact of language and format differences on matching accuracy. This reduces cross-platform resource matching response time to milliseconds while significantly improving matching accuracy, ultimately achieving efficient, low-latency dynamic cross-platform resource matching. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The present invention is a flowchart of a cross-platform e-commerce resource dynamic matching search method according to an exemplary embodiment.

[0018] Figure 2 The figure is a structural block diagram of a cross-platform e-commerce resource dynamic matching and search system according to an exemplary embodiment. DETAILED DESCRIPTION

[0019] Embodiments of the present application are described in detail below, examples of which are illustrated in the accompanying drawings.

[0020] Throughout this specification, reference to the terms "certain embodiments," "one embodiment," "some embodiments," "illustrative embodiments," "examples," "specific examples," or "some examples" means that a specific feature, structure, material, or characteristic described in conjunction with the embodiment or example is included in at least one embodiment or example of the present application. In this specification, the schematic representations of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in any one or more embodiments or examples.

[0021] The present application discloses a cross-platform e-commerce resource dynamic matching search method, referring to Figure 1 ,include: S100 parses product description information from different e-commerce platforms in real time at the data access end, extracts the core attributes of the products and generates a unified semantic representation. Product description information includes text and images, and core attributes include name, specifications, and price. The semantic representation establishes a semantic mapping relationship between cross-platform attributes based on a context-aware synonym replacement mechanism and attribute association rules, and transmits the standardized data to the central database through a distributed consistency synchronization mechanism.

[0022] S100 standardizes heterogeneous product data (such as text and images) from different e-commerce platforms, generates a unified semantic representation, and transmits it to a central database through a distributed consistency synchronization mechanism. The following is an analysis of the specific execution steps and functions of S100: Step 1: Obtain product description information (including text description, images, price, specifications, etc.) in real time from the API interfaces or crawler systems of multiple e-commerce platforms, and use natural language processing (NLP) technology to parse the text content and extract the core attributes of the product (such as name, specifications, and price).

[0023] At the same time, image recognition technology is used to extract features from product images (such as color and shape) to supplement the deficiencies of text descriptions. For example, a product description of "waterproof phone case" on one e-commerce platform and "IP68-rated dust and water-resistant protective cover" on another platform will be interpreted as having the same core semantic attributes. This process addresses the issue of different product descriptions on different e-commerce platforms (such as "waterproof" and "IP68-rated waterproof") and ensures the integrity of product information through multimodal (text + image) analysis, avoiding matching failures due to the lack of a single data source.

[0024] Step 2: Based on the semantic dictionary and user behavior data of the e-commerce platform, a context-aware synonym replacement mechanism is used to dynamically replace cross-platform synonyms (e.g., "phone case" → "PhoneCase"), and the replacement priority is adjusted according to the context (e.g., "waterproof phone case" may be replaced with "WaterproofPhoneCase" instead of "CaseforMobile").

[0025] At the same time, a cross-platform attribute mapping relationship is established through a preset rule library (for example, the "brand + model" combination uniquely identifies a product) and a machine learning model (for example, mapping "memory capacity" on one e-commerce platform to "RAMSize" on another e-commerce platform). Ultimately, a unified semantic vector representation is generated as the basis for standardized data. This process eliminates semantic ambiguity in cross-platform product descriptions (such as "waterproof" and "IP68 waterproof"), ensuring that product attributes on different platforms can be accurately matched. It also provides consistent attribute definitions for subsequent cross-platform resource matching (such as search and recommendation), reducing matching errors caused by semantic differences.

[0026] In step 3, the standardized data is transmitted to the central database using a distributed consistency synchronization mechanism (such as a consistent hashing algorithm). Only differential data (such as new products and price changes) is transmitted, rather than the full data set, to reduce network bandwidth consumption. For example, if the price of a product on an e-commerce platform is updated from 100 yuan to 120 yuan, the synchronization mechanism will only transmit the incremental data "price = 120 yuan" rather than the entire product description.

[0027] The distributed consistency synchronization mechanism significantly reduces the storage and bandwidth overhead of cross-platform data transmission through incremental synchronization and differential data compression (for example, traditional full synchronization requires the transmission of several GB of data, while incremental synchronization only requires the transmission of a few KB). At the same time, it ensures the consistency of the data status of the central database and each edge node, avoiding distorted matching results due to data delays or conflicts (such as users seeing old prices when searching).

[0028] In step 4, the standardized data is written to a central database (such as MySQL or Elasticsearch) and indexed to support subsequent fast retrieval. The database design must be compatible with multi-platform attribute structures (for example, mapping "store ratings" on one e-commerce platform and "user reviews" on another e-commerce platform to a unified "reputation score" field).

[0029] The above process provides unified data support for subsequent conflict resolution, cache configuration, and user intent analysis through centralized data management. It also achieves millisecond-level retrieval of cross-platform products through pre-built indexes and optimized query statements (such as Elasticsearch's inverted index) (for example, when a user enters "waterproof phone case", matching results can be returned within 100ms).

[0030] Through context-aware synonym replacement and attribute association rules, product descriptions on different e-commerce platforms are converted into a unified semantic representation, eliminating format, terminology and semantic differences, and providing a standardized basis for cross-platform matching. At the same time, a distributed consistency synchronization mechanism is used to reduce full data transmission, reduce the load on the central database and ensure data real-time performance. In addition, the standardized data provides a reliable premise for subsequent conflict resolution, cache configuration and user intent analysis, and supports high-concurrency real-time processing of millions of products through multimodal analysis and incremental synchronization mechanisms to meet the dynamic needs of e-commerce platforms.

[0031] S200, based on the product attribute information in the standardized data and the user's historical behavior characteristics, dynamically generates a conflict resolution strategy when a cross-platform data conflict occurs. The conflict resolution strategy is executed by the local preprocessing module deployed on the edge node, where the local preprocessing module quickly responds to bursty data and batch processes periodic data.

[0032] S200 dynamically generates conflict resolution strategies based on standardized product attribute information and user historical behavior characteristics when cross-platform data conflicts occur, and executes them through local pre-processing modules deployed on edge nodes. The following are its specific execution steps and function analysis: In step 1, the system first detects potential conflicts in real time by monitoring the differences in standardized product attribute data (such as price, inventory, and specifications) across different platforms. For example, if a product is priced at 100 yuan on one e-commerce platform and 120 yuan on another, the system will trigger a conflict signal.

[0033] At the same time, the system analyzes the context of the conflict by combining historical user behavior characteristics (such as search preferences, purchase history, and geographic location). For example, if a user is located in Beijing and tends to trust the logistics services of a particular e-commerce platform, the system may prioritize the pricing data of that platform as a benchmark. This process precisely locates the conflicting attributes (such as price and inventory) and the scope of impact, preventing user complaints or transaction failures caused by data inconsistencies. It also enhances the personalization of conflict resolution strategies and user experience through context-aware decision-making (such as the priority of the user's frequently visited platforms).

[0034] Step 2. Next, the system dynamically generates a resolution strategy based on the type of conflicting data (e.g., price, inventory), user characteristics (e.g., geographic location, platform preferences), and product attributes (e.g., price fluctuation trends). For example, for price conflicts involving high-frequency fluctuating products (e.g., promotional items), data from the most recently updated platform is prioritized; for stable products (e.g., daily necessities), a weighted average of data from multiple platforms is used. The logic for handling inventory conflicts involves prioritizing data from platforms with inventory and marking the source of the conflict. The generated strategy is encapsulated as executable instructions (e.g., "Use e-commerce platform prices" or "Merge inventory data") and passed to the local preprocessing module of the edge node. This process avoids the rigidity of single rules by flexibly addressing complex scenarios (e.g., price, inventory, and description conflicts), reduces the load on the central server, and improves response efficiency.

[0035] Third, in step 3, the local preprocessing module on the edge node implements conflict resolution strategies using two mechanisms: rapid response to sudden data changes and batch processing of periodic data. For sudden conflicts (such as sudden price changes caused by promotional events), the local module directly implements a pre-set policy (such as "force the latest price") without waiting for confirmation from the central server. For example, if the price of an item on an e-commerce platform suddenly drops to 50 yuan, the local module immediately overwrites the existing data and synchronizes it to the cache node.

[0036] For periodic conflicts (such as daily inventory updates), the local module performs batch data consistency checks at fixed intervals (e.g., hourly) and performs merge or overwrite operations. For example, it compares inventory data from one e-commerce platform with that from another. If the difference is less than a threshold (e.g., 5%), the data is merged into an average. If the difference is too large, it is flagged as an anomaly and triggers manual review. This mechanism reduces pressure on central servers by balancing real-time and stability (critical conflicts are resolved immediately, while periodic operations avoid system jitter) and leveraging the advantages of edge computing (reducing cross-network communication overhead and reducing latency to milliseconds).

[0037] Finally, in step 4, the local preprocessing module modifies conflicting data (such as adjusting prices and consolidating inventory) based on the generated policy and updates the product information in the edge node cache. Subsequently, the conflict-resolved data is transmitted to the central database through a distributed consistency synchronization mechanism (such as the consistent hashing algorithm) to ensure global data consistency. For example, if the local module uses the price of one e-commerce platform to overwrite the data of another e-commerce platform, the central database will be updated synchronously to avoid cross-platform information discrepancies. This process, through the collaboration between the edge node and the central database, ensures that all platform users see unified product information (such as price and inventory), avoiding user confusion or transaction failures caused by data delays. Furthermore, if an edge node detects an anomaly after executing the policy (such as a misjudged conflict), it can quickly roll back to the pre-conflict state through the log records in the central database, achieving fault tolerance and rollback capabilities.

[0038] By dynamically generating conflict resolution strategies (such as prioritizing trusted data sources based on user behavior characteristics) and relying on local pre-processing modules at edge nodes, we can achieve millisecond-level responses to cross-platform data conflicts (such as price and inventory differences), significantly reducing the load on central servers (for example, processing 100,000 conflicting data items per second during the Double 11 shopping festival). Combined with personalized strategies (such as prioritizing data from platforms frequently visited by users), we can reduce user complaints due to data inconsistencies. At the same time, through a combination of rapid sudden responses and periodic batch processing mechanisms, we can support high-concurrency, real-time conflict resolution for millions of products, ensuring the consistency of product information and user experience across platforms.

[0039] S300, after the conflict is resolved, obtain the data content after the conflict is resolved, select the target edge node based on the data content and the user's geographic location information in the computing architecture jointly constructed by the edge node and the central server, and dynamically configure the cache data set of the target edge node.

[0040] After cross-platform data conflicts are resolved, the target edge node is dynamically selected and its cached data set is configured based on the user's geographic location information and the data content after the conflict is resolved. The following are the specific execution steps and function analysis of S300: In step 1, the system first obtains the standardized product attribute data (such as price, inventory, specifications) and the execution result of the conflict resolution strategy (such as "adopt the price of one of the e-commerce platforms" or "merge inventory data") from the conflict resolution module (S200). The data after conflict resolution usually contains the core attributes of the product (such as brand, model, price, inventory status), a conflict resolution mark (such as "one of the e-commerce platform data takes precedence" or "merged data") and a timestamp (used to determine the cache timeliness). For example, if the price of a product on one e-commerce platform is 100 yuan and another e-commerce platform is 120 yuan, after the conflict is resolved, it may be marked as "the e-commerce platform data with a product price of 100 yuan takes precedence" and the update time is recorded as the current timestamp.

[0041] The above process ensures the freshness of cached data through timestamps, preventing users from obtaining old data due to cache expiration (such as price restoration after a promotion ends). At the same time, it provides a consistent data foundation for subsequent cache configuration and ensures data consistency between edge nodes and the central database.

[0042] In step 2, the user's geographic location (e.g., "Beijing user" or "Shanghai user") is obtained through their IP address, GPS location, or carrier base station information. The location tag is then refined (e.g., "Beijing user prefers a certain e-commerce platform") based on the user's historical behavior (e.g., frequently visited e-commerce platforms). Based on this, the system prioritizes the edge node closest to the user's geographic location based on proximity (e.g., "Beijing user → Beijing edge node"), while also considering platform preference adaptation and selecting the optimal node from multiple candidate nodes based on a load balancing strategy. For example, if there are three edge nodes in Beijing, the system will select the node with the lowest load and that supports direct data connection to the preferred e-commerce platform.

[0043] The above significantly reduces network latency by reducing cross-regional data transmission paths (for example, reducing the response from seconds for cross-regional transmission to milliseconds for local transmission), and ensures that users get data from frequently visited platforms first through platform preference adaptation (for example, "Beijing users see the real-time price of one of the e-commerce platforms"), thereby improving user experience.

[0044] In step 3, a caching strategy is dynamically generated based on product attributes (e.g., hot-selling items, promotional items) and user needs (e.g., frequently searched keywords). High-value data (e.g., "hot-selling mobile phones") is prioritized for caching at the target edge node, and cache capacity is allocated based on product popularity (e.g., hot-selling items occupy a larger portion, while less popular items are loaded on demand). For frequently changing attributes such as price and inventory, an incremental update mechanism is used (e.g., only synchronizing price changes); for static attributes (e.g., product descriptions), the cache is refreshed periodically (e.g., hourly). The system then writes conflict-resolved data (e.g., "Price on a certain e-commerce platform = 120 yuan") to the target edge node's cache and accelerates query efficiency by establishing a multidimensional index (e.g., "product ID + platform + location").

[0045] The above process significantly improves the cache hit rate (for example, the cache hit rate of popular products can exceed 90%) through dynamic priority allocation and cache update mechanism, while reducing the load on the central server (only processing unpopular data requests), and flexibly responding to traffic fluctuations (such as increasing the cache of popular products during promotions), supporting the high concurrent access needs of millions of products.

[0046] S400, obtain natural query language input by the user, parse the corresponding user intent information based on the natural query language, generate corresponding multi-dimensional matching features based on the user intent information and the product data in the configured cache, determine the candidate product set through approximate nearest neighbor retrieval based on the multi-dimensional matching features, output corresponding search results based on the candidate product set, obtain user feedback information on the search results, and update the intent parsing model and matching strategy based on the feedback information.

[0047] The intent information in the user's natural language query is analyzed, and multi-dimensional matching features are generated by combining cached data. The search results are output based on approximate nearest neighbor retrieval. Finally, the intent parsing model and matching strategy are dynamically optimized based on user feedback. The following is an analysis of the specific execution steps and functions of S400: Step 1: First, use natural language processing (NLP) technology to perform word segmentation, entity recognition, and semantic analysis on the natural language query input by the user (such as "Find a mobile phone with good waterproof performance") to extract key intents (such as "waterproof performance" and "mobile phone").

[0048] The extracted intents are then categorized into functional requirements (such as "waterproof"), price ranges (such as "under 2,000 yuan"), and brand preferences (such as "A"). These are then dynamically weighted based on the user's historical behavior (such as search history and purchase preferences). For example, if a user has searched for "waterproof phone" multiple times, the "waterproof" intent will be weighted higher than other attributes. This process uses structured intent information to accurately understand user needs, avoiding search result bias caused by natural language ambiguity. It also incorporates personalized weighting to improve search result matching (for example, prioritizing frequently purchased brands).

[0049] Step 2: Cache data from configured edge nodes (such as standardized attribute data for popular and promotional items) is retrieved and combined with user intent information to construct a multi-dimensional matching feature vector. For example, "waterproof" is mapped to standardized attributes such as "IP68 waterproof." If the user intent includes "less than 2,000 yuan," products priced above 2,000 yuan are filtered out. If the user prefers "A," the matching weight for product A is increased. The system then converts the multi-dimensional features into numerical vectors for subsequent retrieval.

[0050] The above process avoids the limitations of single-dimensional matching by covering multi-dimensional requirements such as function, price, and brand. At the same time, it reuses the cached data of S300, reduces the real-time query pressure on the central database, and significantly improves the response speed.

[0051] In step 3, based on the generated multi-dimensional feature vector, the system uses an efficient approximate nearest neighbor (ANN) algorithm (such as Faiss and HNSW) to quickly retrieve the most similar candidate products from the cached data. For example, given the feature vector of "IP68 waterproof, price under 2000 yuan, brand A," the system retrieves candidate products from the cache that meet the requirements (e.g., mobile phone B corresponding to brand A). The candidate products are then ranked based on their similarity scores, and invalid products (such as "out of stock" products) are filtered out based on real-time inventory and promotional information.

[0052] The above process meets the needs of high-concurrency scenarios through millisecond-level large-scale data retrieval (such as a million-level product library), while ensuring that search results are highly relevant to user intent through multi-dimensional feature vectors and similarity sorting (such as prioritizing the display of mobile phone A with strong waterproof performance).

[0053] In step 4, the system trains an intent parsing model (such as BERT or Transformer) based on user feedback data (such as clicked keywords) to optimize the accuracy and generalization of intent recognition. For example, if a user frequently associates "waterproof" with "IP68 rating," the model will reinforce this mapping. Simultaneously, the system dynamically adjusts the matching strategy based on this feedback data: increasing brand preference weights, optimizing ANN algorithm parameters (such as increasing the similarity threshold for price), and updating cached data priorities (such as increasing the cache hit rate for frequently clicked items). This process, through a closed feedback loop, enables the continuous evolution of the intent parsing model and matching strategy, adapting to long-term changes in user needs (such as a shift in preference from "waterproof" to "battery life") while reducing operational costs. Automatic strategy adjustments also improve search result accuracy and user satisfaction.

[0054] In step 5, the intent parsing model (such as BERT or Transformer) is trained based on user feedback data (such as clicked keywords) to optimize the accuracy and generalization of intent recognition. For example, if a user frequently associates "waterproof" with "IP68 rating," the model will reinforce this mapping. Simultaneously, the matching strategy is dynamically adjusted based on this feedback data: brand preference weighting is increased, ANN algorithm parameters are optimized (such as increasing the similarity threshold for price), and cached data priorities are updated (such as increasing the cache hit rate for frequently clicked items). This process, through a closed feedback loop, enables the continuous evolution of the intent parsing model and matching strategy, adapting to long-term changes in user needs (such as a shift in preference from "waterproof" to "battery life") while reducing operational costs. Automatic policy adjustments also improve search result accuracy and user satisfaction.

[0055] Through multimodal data parsing and semantic standardization, the S100 addresses cross-platform product information heterogeneity and builds a unified data foundation. Leveraging natural language processing (NLP) and image recognition technologies, the S100 extracts core product attributes (such as name, price, and specifications). It then uses context-aware synonym replacement (e.g., "waterproof" maps to "IP68 waterproof") and attribute association rules to eliminate semantic differences, ensuring accurate matching of product descriptions across platforms. Furthermore, the integration of distributed consistency synchronization mechanisms (such as incremental data transfer) significantly reduces the load on the central database and ensures real-time data availability. The S200 further utilizes local pre-processing modules on edge nodes to dynamically generate conflict resolution strategies (e.g., prioritizing data from frequently accessed platforms), achieving millisecond-level response times (e.g., for price and inventory discrepancies), and avoiding performance bottlenecks caused by centralized processing. This "edge-based rapid response + central-level collaboration" architecture supports high-concurrency, real-time processing of millions of products, providing efficient and consistent data support for cross-platform resource matching.

[0056] Efficient resource matching is achieved through multi-dimensional feature vectors and an approximate nearest neighbor (ANN) search algorithm, with continuous optimization leveraging user feedback. The S300 dynamically configures edge node caches based on user location and behavioral preferences (e.g., prioritizing popular products). Combining incremental updates with periodic batch processing significantly improves cache hit rates (e.g., over 90%) and reduces pressure on central servers. The S400 uses natural language processing (NLP) to analyze user intent (e.g., "waterproof phone"), combines cached data to generate multi-dimensional matching features (features, price, and brand), and employs an ANN algorithm to rapidly retrieve candidate products from a library of millions of products. The intent parsing model (e.g., BERT) and matching strategies (e.g., weighting) are dynamically adjusted based on user feedback (clicks, purchases). This closed-loop chain of "data standardization → conflict resolution → cache optimization → intent parsing → feedback" not only enables real-time and accurate cross-platform resource matching, but also continuously improves the user experience through personalized strategies and adaptive optimization, ultimately forming an efficient and flexible cross-platform e-commerce matching system.

[0057] In the embodiment of the present application, when generating a unified semantic representation, the method further includes: Step 1: First, extract user preference tags, such as "waterproof," "high cost-effective," and "Brand A," from historical user behavior data (such as search keywords, click history, purchase history, and product ratings). Next, semantically associate these fuzzy preference tags with the product's core attributes. For example, "waterproof" is mapped to the product's "waterproof rating" attribute (such as IP68 or IP67), and "high cost-effective" is mapped to the "price / performance ratio" attribute. Simultaneously, by analyzing user behavior data across different platforms, common needs and differentiated preferences across platforms are identified. For example, the "waterproof" attribute has seen a significant increase in popularity across multiple platforms, while "authenticity guarantee" is a core user demand on one e-commerce platform.

[0058] In step 2, during the dynamic weight calculation phase, the system combines cross-platform behavioral pattern data (e.g., "waterproof" accounts for 30%, "price" accounts for 20%) with the specific values ​​of product attributes (e.g., "IP68 waterproof" = 1.0, "less than 2,000 yuan = 0.8") to calculate the dynamic weight of each attribute. For example, if "waterproof" accounts for a high proportion of the behavioral pattern and the corresponding product attribute value is high, its weight will be increased to 0.5, while "price" may be assigned a weight of 0.3.

[0059] The priority adjustment module then dynamically adjusts weights based on real-time user feedback (e.g., click-through rate, conversion rate). For example, if the click-through rate of a product with the "waterproof" attribute increases significantly, its weight will be further increased. Simultaneously, the system sets threshold rules (e.g., "priority lock triggers when a certain attribute's weight exceeds 0.7") to prevent excessive weight skew, thereby achieving a balance between global needs (e.g., high-frequency attributes across platforms) and personalized needs (e.g., brand preferences for specific users).

[0060] In step 3, the system finally injects the dynamically calculated attribute weights into the semantic representation vector, replacing the original static weight configuration. For example, the original static weights of "brand = 0.2, waterproof = 0.4, price = 0.4" will be updated to "brand = 0.3, waterproof = 0.5, price = 0.3".

[0061] The above update process is synchronized to the cache of edge nodes, ensuring that high-priority attributes (such as "waterproof") are loaded first during searches. At the same time, the system optimizes search efficiency using the Approximate Nearest Neighbor (ANN) algorithm, prioritizing products with high-weighted attributes. For example, for a user-entered query such as "IP68 waterproof + less than 2,000 yuan + Brand A," the system will prioritize products with a high match for the "waterproof" attribute, thereby improving the relevance of search results. Furthermore, cache priority optimization reduces invalid searches for products with unmatched low-weighted attributes, reducing system load and improving response speed.

[0062] By dynamically associating user behavior with product attributes, adjusting attribute weights in real time, and optimizing semantic representation priorities, we achieve adaptive optimization of cross-platform resource matching. This approach accurately captures ambiguous user needs (e.g., translating "waterproof" into an IP68-rated waterproof attribute weight), dynamically responds to market changes (adjusting weights based on cross-platform behavior patterns and real-time click-through rates), and significantly reduces system load and improves matching efficiency in high-concurrency scenarios through cache prioritization and retrieval algorithm optimization.

[0063] In the embodiment of the present application, in the conflict resolution strategy, the method further includes: In step 1, when a data conflict occurs, the system first identifies the source platform of the conflicting data and the geographic location of the target user. Subsequently, the system analyzes historical data or pre-set rules to identify correlations between different platforms and user regions. For example, it may be possible to determine that users of e-commerce platform A are primarily concentrated in China, that users of e-commerce platform B prefer authentic products, and that users of overseas platforms are more concerned with shipping time. Furthermore, the system uses statistical analysis or machine learning models, taking into account regional characteristics, to quantify the compatibility between platforms and regions.

[0064] Step 2: After extracting the associated features between platforms and regions, the system dynamically adjusts the priority of the verification rules based on the degree of match. For example, if the user is located in China and platform a has the highest degree of match with Chinese users, the data from platform a will be prioritized as the verification benchmark. If the user is located elsewhere and platform b has a lower degree of match with the region, the priority of platform b's data will be lowered. For conflicting fields (such as price, inventory, and logistics information), the system generates differentiated verification rules. For example, for the price field, the platform data with the highest degree of match will be prioritized, and if there is a conflict, data from other platforms will be referenced. For the logistics information field, the platform data with a high degree of match with the user's region will be prioritized.

[0065] In step 3, based on the generated cross-regional validation rules, data from platforms with high matching scores is prioritized as a benchmark, and conflicting data from other platforms is corrected. The system then collects user feedback (such as click-through rates and complaint rates) after conflict resolution to evaluate the effectiveness of the dynamic rules. For example, if it finds that "e-commerce platform b's matching with the corresponding local user base has improved," the priority weight is adjusted to reflect this change. Through a continuous feedback loop, the system can dynamically adjust the association characteristics between platforms and regions, avoiding rigid rules (such as changing user preferences in emerging markets) and automatically adapting to new scenarios (such as new user groups in local markets) as cross-regional business expands.

[0066] By dynamically associating user geographic location with platform characteristics to generate verification rules, we can accurately adapt to regional user preferences (for example, Chinese users prioritize verifying local platform data), reduce invalid verification steps, and prioritize correcting highly matching data, thereby shortening response time. At the same time, through real-time feedback optimization and rule iteration, we can achieve adaptive expansion of cross-regional business scenarios and significantly reduce manual intervention and maintenance costs.

[0067] In an embodiment of the present application, when dynamically configuring the cache data set of the target edge node, the method further includes: Step 1: First, use data interfaces or automated tools to collect product price data from different platforms and analyze the frequency and magnitude of price fluctuations. For example, identify products with a 5% daily price reduction during promotional periods or products with stable prices during non-promotional periods. Combined with historical data, determine whether the product is experiencing periods of high volatility (such as Singles' Day and Black Friday) or stability (such as daily sales).

[0068] In addition, based on user historical behavior (such as purchase records and browsing cycles), the purchase cycle characteristics of products are analyzed. For example, high-frequency purchases (such as milk and toilet paper) are distinguished from low-frequency purchases (such as home appliances), and the classification is further refined based on product types.

[0069] In step 2, the system dynamically prioritizes cache updates based on product price fluctuations and purchase cycles. Products with frequent price fluctuations (such as promotional and seasonal items) are assigned high priority and require a short-term forced refresh strategy (e.g., hourly updates). Products with stable prices (e.g., daily necessities and basic items) are assigned low priority and can be updated passively over a longer period (e.g., daily or weekly updates). For products with cyclical purchases (e.g., monthly repurchases), the system further prioritizes them based on price fluctuations. For example, products with high price volatility and high purchase frequency (e.g., promotional milk) retain high priority, while products with low price volatility and low purchase frequency (e.g., air conditioning filters) are adjusted to low priority.

[0070] The system then deploys high-priority items to edge nodes close to the user area (such as front-end nodes in a multi-level cache architecture) with a short lifespan (e.g., cache expiration after 1 hour). Low-priority items are deployed to regional central nodes with a long lifespan (e.g., cache expiration after 24 hours). This process optimizes cache resource allocation, preventing user churn due to outdated data for high-priority items while reducing the cost of ineffective refreshes for low-priority items. Furthermore, high-priority items are prioritized by edge nodes, shortening user access latency.

[0071] In step 3, for high-priority items (such as promotional items), the system periodically pulls the latest price data from the source (e.g., syncing prices hourly via an API) and overwrites the edge node cache. If an abnormal price fluctuation (e.g., an illegal low price) is detected, a forced refresh is immediately triggered (e.g., by actively requesting an API to update the cache). For low-priority items (e.g., daily necessities), the system retrieves the latest data from the source only after the cache expires (e.g., updating it daily at midnight). Furthermore, based on the user's purchase cycle (e.g., the average user purchase cycle is 68 days), the cache is updated in advance (e.g., 60 days before the active period) to prevent outdated data from affecting conversion rates.

[0072] The purpose of the above process is to dynamically adapt to market changes (such as promptly responding to price callbacks after the end of a promotion), avoid user complaints caused by outdated edge node caches, and reduce unnecessary back-to-source requests through differentiated update strategies (such as updating low-priority products only when they expire), saving bandwidth and server resources.

[0073] By dynamically adapting to product price fluctuations and user purchase cycles, differentiated edge node caching strategies are formulated to achieve high-frequency refreshes of high-priority products (such as hourly updates for promotional products) and long-cycle updates of low-priority products (such as daily updates for daily necessities). This not only accurately matches the dynamic changes of market-sensitive products and users' high-frequency demands, but also reduces system load through resource hierarchical management. At the same time, relying on edge nodes to prioritize loading high-priority data, it shortens response time and avoids user churn caused by price failures (such as timely synchronization of promotional product caches). Ultimately, it optimizes cache efficiency and user experience while ensuring data timeliness.

[0074] In the embodiment of the present application, when generating the multi-dimensional matching features, the method further includes: Step 1: First, core requirements are extracted by analyzing user input (such as search keywords, product reviews, and customer service conversations). For example, from the query "waterproof phone A under 2,000 yuan," explicit requirements such as "waterproof," "A," and "price range" are identified. At the same time, implicit requirements are supplemented by combining historical user behavior with the current scenario (for example, when a user recently focused on "sports scenes," the demand for "durability" is implicit). The system then extracts standardized attribute rules (such as "waterproof" → "IP68 waterproof," "high cost-effectiveness" → "price / performance ratio") from a preset cross-platform attribute mapping table and dynamically adjusts the rule weights based on platform characteristics (for example, e-commerce platform B prioritizes "authenticity," while e-commerce platform A focuses on "low prices").

[0075] Step 2: Based on the extracted user intent and cross-platform rules, multi-dimensional dynamic filtering conditions are generated. For example, the explicit requirement "waterproof" is converted into "IP68 waterproof = 1.0," "Brand A = 1.0," and "Price ≤ 2000 RMB." The implicit requirement "sports scene" is converted into "screen drop resistance ≥ 80%" and "battery life ≥ 4000 mAh." Condition priorities are also adjusted based on platform rules (for example, e-commerce platform B prioritizes "authenticity guarantee = 1.0"). The system then converts these conditions into executable query statements (such as SQL statements or search engine query expressions), supporting combined operators (such as "AND" and "OR") and nested priorities (for example, "(waterproof AND Brand A) OR (high cost-effectiveness AND price < 1500 RMB)").

[0076] Step 3 gradually narrows the candidate products through a tiered screening mechanism. The first tier involves hard filtering based on core attributes, retaining only products that meet the user's explicit needs (e.g., "IP68 waterproof," "Brand A," and "Price ≤ 2,000 RMB"). The second tier involves soft sorting based on implicit attributes, weighting the first tier results by implicit needs (e.g., "screen drop resistance" and "battery life") to prioritize highly matched products. The third tier involves platform validation, supplemented by additional validation based on the user's platform (e.g., e-commerce platform B) (e.g., "Authenticity Guarantee = 1.0"), to filter out products that don't meet platform requirements. If the user's intent explicitly mentions an attribute (e.g., "waterproof"), that attribute receives the highest priority across all tiers. If conflicting user intent is detected (e.g., "low price" and "authenticity guarantee"), dynamic balancing is achieved through weighting (e.g., "low price weight 0.6, authenticity guarantee weight 0.4").

[0077] By combining users' explicit and implicit needs with cross-platform attribute rules, dynamic filtering conditions are generated and a hierarchical screening mechanism is adopted. This not only accurately captures users' deep needs that are not clearly expressed (such as "sports scenes" implies "durability"), but also eliminates semantic differences by unifying multi-platform attribute expressions (such as "waterproof" → "IP68"), ensuring the consistency of cross-platform search results. At the same time, the hierarchical screening mechanism first filters core attributes and then sorts implicit needs, which not only reduces computing resource consumption, but also prioritizes the display of key attributes explicitly mentioned by users (such as "A"), ultimately achieving efficient matching and accurate sorting, significantly improving user experience and search efficiency.

[0078] In an embodiment of the present application, when updating the matching strategy based on the feedback information, the method further includes: Step 1: First, collect user behavior data on search results from multiple platforms, including clicks, purchases, and favorites. For example, if a user clicks on a product but doesn't purchase it, it may indicate dissatisfaction with the price or after-sales service; purchase behavior indicates that the product attributes (such as price and features) meet user needs.

[0079] The system also extracts standardized attributes from product data (e.g., "Waterproof Rating = IP68," "Price = 1999 RMB," "Brand = A") and combines them with a cross-platform attribute mapping table (e.g., "Waterproof" → "IP68," "High Cost-Effectiveness" → "Price / Performance Ratio") for unified representation. Subsequently, the system analyzes the correlation between user behavior and product attributes using statistical methods or machine learning models. For example, click behavior may reveal a user's focus on "price" or "after-sales service," purchase behavior may indicate that "screen drop resistance" and "battery life" are core drivers, and collection behavior may reflect a user's preference for "brand" or "design." Ultimately, the system generates dynamic rules based on the analysis results.

[0080] Step 2: Dynamically adjust semantic mapping and attribute alignment logic based on the generated attribute association enhancement rules. Regarding semantic mapping optimization, the accuracy of attribute expression is improved by updating the synonym replacement mechanism. For example, if a user prefers "IP68" over "waterproof," the mapping priority of "waterproof" to "IP68" is increased; if a user is more sensitive to "cost-effectiveness," the mapping weight of "price / performance ratio" is strengthened.

[0081] At the same time, semantic associations are expanded in combination with user behavior. For example, if a user frequently searches for "sports mobile phone" and then buys "long battery life + waterproof" products, a strong association will be established between "sports scene" and "battery life" and "waterproof". In terms of attribute alignment optimization, the consistency of attributes on different platforms is verified. For example, after a user purchases an "IP68 waterproof mobile phone" from e-commerce platform b, if a similar product on e-commerce platform a is only marked as "waterproof", it is mandatory to add the "IP68" attribute. In addition, attribute weights are adjusted according to user behavior. For example, if the purchase rate increases significantly after the user clicks "price < 2,000 yuan", the weight of the "price" attribute will be increased; if the user has a high collection rate for "brand", the brand matching priority will be strengthened.

[0082] In step 3, the system continuously monitors the impact of the optimized semantic mapping and attribute alignment logic on search results (e.g., click-through rate, conversion rate, and complaint rate). For example, if the optimized "IP68" mapping increases user click-through rate by 10%, the rule is retained. If the "brand" weight adjustment results in user churn, the change is rolled back. Simultaneously, attribute association enhancement rules are regularly updated based on monitoring results. For example, if user interest in "environmentally friendly materials" increases, a new association rule is added between the "environmentally friendly" attribute and the "user collection rate," and the semantic mapping table is adjusted (e.g., "environmentally friendly" → "degradable materials").

[0083] Furthermore, the optimized rules are synchronized across all platforms to ensure consistency in attribute representation and user behavior across platforms. For example, if a user adjusts the "brand" weight on e-commerce platform B, e-commerce platform A will also update the matching logic for the "brand" attribute.

[0084] By associating user operation behaviors such as clicks, purchases, and favorites with the cross-platform attribute characteristics of products, dynamic attribute association rules are generated, and based on this, semantic mapping (e.g., "IP68" takes precedence over "waterproof") and attribute alignment logic are optimized, eliminating the differences in attribute expressions between different platforms, and ensuring that users obtain a consistent search experience on platforms such as a e-commerce platform and b e-commerce platform. At the same time, through a closed-loop feedback mechanism, the optimization effect is continuously monitored (e.g., improving click-through rate, adapting to emerging needs such as "environmentally friendly materials"), and the matching strategy is dynamically adjusted to avoid the lag of static rules, thereby achieving precise optimization based on user behavior, ensuring cross-platform consistency, and quickly responding to market changes, comprehensively improving the accuracy of search matching and user experience.

[0085] In the embodiment of the present application, when parsing natural query language, the method further includes: Step 1, first extract key attributes (such as "waterproof", "high cost performance", "A") and their multilingual expressions (such as the English "waterproof" corresponding to the Chinese "防水") from the user's natural language input (such as search terms, customer service conversations). Subsequently, call the cross-platform attribute semantic mapping table to map the fuzzy natural language attributes (such as "waterproof") to standardized attributes (such as "IP68-level waterproof"), and unify the expressions between different languages (such as mapping "waterproof" to "IP68").

[0086] At the same time, dynamically adjust the attribute weights in combination with the characteristics of each platform (e.g., the a e-commerce platform prefers "low price", and the b e-commerce platform emphasizes "genuine product guarantee") (e.g., the priority of "waterproof" on the b e-commerce platform is higher than that on the a e-commerce platform). Finally, based on the semantic mapping results, construct a multi-language compatible standardized query template to ensure that whether the user inputs the Chinese "waterproof mobile phone below 2000 yuan" or English, the system generates query templates with the same structure, thereby eliminating language and platform differences and improving the consistency of cross-platform search.

[0087] In step 2, after generating a standardized query template, the system further analyzes the attribute description conventions of each platform (e.g., e-commerce platform A emphasizes "low price," while e-commerce platform B emphasizes "authenticity") and builds a platform attribute rule base (e.g., "waterproof" is labeled "strong waterproof performance" on e-commerce platform A, while it is labeled "IP68 waterproof" on e-commerce platform B). Through a semantic expansion mechanism, the system dynamically converts the standardized template into query features that conform to platform conventions. For example, on e-commerce platform A, "IP68 waterproof" is expanded to "strong waterproof performance" and labeled "low price," while on e-commerce platform B, it is expanded to "IP68 certified" and emphasizes "authenticity" while retaining core attributes (e.g., "A" and "price ≤ 2000 yuan") to ensure cross-platform result consistency. Furthermore, the system adjusts the display priority of query features based on platform characteristics (e.g., e-commerce platform B prioritizes "authenticity") or supplements implicit attributes (e.g., e-commerce platform A automatically adds the "promotion" tag), thereby adapting to platform preferences and improving search result relevance and user click-through rate.

[0088] By uniformly mapping natural user queries (e.g., "waterproof" or "waterproof") in multiple languages ​​to standardized attributes (e.g., "IP68 waterproof") and dynamically generating adaptive query features based on the attribute description conventions of each platform, this approach achieves precise cross-platform search matching. Its core value lies in: eliminating language and platform differences (e.g., "waterproof" → "IP68"), ensuring consistent search results across different languages ​​or platforms; dynamically adapting to platform preferences (e.g., e-commerce platform B prioritizes "authentic products"), improving the relevance of search results to user needs; and enhancing the user search experience. Through standardized templates and semantic extension mechanisms, users receive consistent and platform-specific search results across different platforms (e.g., e-commerce platform A emphasizes "low price," e-commerce platform B emphasizes "certification"), significantly improving click-through rates and satisfaction.

[0089] The present application embodiment discloses a cross-platform e-commerce resource dynamic matching search system, referring to Figure 2 ,include: Semantic mapping relationship establishment module 001 performs real-time analysis of product description information from different e-commerce platforms at the data access end, extracts the core attributes of the products, and generates a unified semantic representation. Product description information includes text and images, and core attributes include name, specifications, and price. The semantic representation is based on a context-aware synonym replacement mechanism and attribute association rules to establish semantic mapping relationships between cross-platform attributes. The standardized data is then transmitted to the central database through a distributed consistency synchronization mechanism. Conflict resolution strategy generation module 002, based on the product attribute information in the standardized data and the user's historical behavior characteristics, dynamically generates a conflict resolution strategy when cross-platform data conflicts occur. The conflict resolution strategy is executed by the local pre-processing module deployed on the edge node. The local pre-processing module quickly responds to sudden data and batch processes periodic data. The target edge node selection module 003 obtains the conflict-resolved data content after the conflict is resolved, selects the target edge node based on the data content and the user's geographic location information in the computing architecture composed of the edge node and the central server, and dynamically configures the cache data set of the target edge node; Update module 004 obtains the natural query language input by the user, parses the corresponding user intent information based on the natural query language, generates corresponding multi-dimensional matching features based on the user intent information and the product data in the configured cache, determines the candidate product set through approximate nearest neighbor retrieval based on the multi-dimensional matching features, outputs the corresponding search results based on the candidate product set, obtains user feedback information on the search results, and uses the feedback information to update the intent parsing model and matching strategy.

[0090] An embodiment of the present application further discloses a cross-platform e-commerce resource dynamic matching and search system, comprising a processor, wherein the processor runs a program of any one of the above-described cross-platform e-commerce resource dynamic matching and search methods.

[0091] An embodiment of the present application also discloses a storage medium storing a program of any one of the above-mentioned cross-platform e-commerce resource dynamic matching and search methods.

[0092] Although the embodiments of the present application have been shown and described above, it can be understood that the above embodiments are exemplary and cannot be understood as limitations on the present application. Ordinary technicians in this field can change, modify, replace and modify the above embodiments within the scope of the present application.

Claims

1. A cross-platform e-commerce resource dynamic matching search method, characterized in that: include: At the data access end, product descriptions from different e-commerce platforms are parsed in real time to extract core product attributes and generate a unified semantic representation. Product descriptions include text and images, and core attributes include name, specifications, and price. The semantic representation establishes semantic mapping relationships between cross-platform attributes based on a context-aware synonym replacement mechanism and attribute association rules. The standardized data is then transmitted to the central database through a distributed consistency synchronization mechanism. Based on the product attribute information in the standardized data and the user's historical behavior characteristics, a conflict resolution strategy is dynamically generated when a cross-platform data conflict occurs. The conflict resolution strategy is executed by a local pre-processing module deployed on the edge node, wherein the local pre-processing module quickly responds to sudden data and batch processes periodic data; After the conflict is resolved, the conflict-resolved data content is obtained, and in the computing architecture composed of the edge node and the central server, the target edge node is selected based on the data content and the user's geographic location information, and the cache data set of the target edge node is dynamically configured; Obtain natural query language input by the user, parse corresponding user intent information based on the natural query language, generate corresponding multi-dimensional matching features based on the user intent information combined with product data in the configured cache, determine a set of candidate products through approximate nearest neighbor retrieval based on the multi-dimensional matching features, output corresponding search results based on the candidate product set, obtain user feedback information on the search results, and update the intent parsing model and matching strategy based on the feedback information.

2. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: When generating a unified semantic representation, the method further includes: Correlate and map product core attributes with preference tags in user historical behavior to extract corresponding cross-platform behavior patterns; Based on the cross-platform behavior pattern, corresponding dynamic attribute weights are generated, the priorities of different attributes in the semantic representation are calculated through the dynamic attribute weight generation module, and the priorities of different attributes in the semantic representation are dynamically updated through the attribute priority adjustment module.

3. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: In conflict resolution strategies, methods also include: Obtaining the source platform of the conflicting data and the geographic location information of the target user, and extracting the correlation features between the source platform and the region based on the source platform and the geographic location information; Based on the association features, cross-regional data consistency verification rules are dynamically generated, wherein the cross-regional data consistency verification rules give priority to using the platform data with the highest matching degree with the user's region as a conflict resolution benchmark.

4. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: When dynamically configuring the cache data set of the target edge node, the method further includes: Extracting cross-platform price fluctuation trends and user purchase periodicity characteristics of products, and prioritizing cache updates of products based on the cross-platform price fluctuation trends and user purchase periodicity characteristics; Based on the cache update priority, a short-cycle forced refresh strategy is executed for commodities with frequent price fluctuations, and a long-cycle passive update strategy is executed for commodities with stable prices.

5. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: When generating the multi-dimensional matching features, the method further includes: Retrieving user intent information and cross-platform attribute association rules, and generating corresponding dynamic attribute filtering conditions based on the user intent information and the cross-platform attribute association rules; Based on the dynamic attribute filtering conditions, the range of candidate products is gradually narrowed down through a hierarchical screening mechanism, with priority given to matching the core attributes explicitly mentioned in the user's intent.

6. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: When updating the matching strategy based on the feedback information, the method further includes: Obtaining user operations on search results, where user operations include clicks, purchases, and favorites on search results, obtaining cross-platform attribute features corresponding to products, performing correlation analysis based on the user operations and the cross-platform attribute features, and generating user behavior-driven attribute correlation enhancement rules; Based on the attribute association enhancement rules, the synonym replacement mechanism and the attribute alignment logic are updated through the semantic mapping optimization mechanism.

7. The cross-platform e-commerce resource dynamic matching search method according to claim 1, characterized in that: When parsing natural query language, the method also includes: Acquire semantic mapping relationships between cross-platform attributes, and generate a multi-language compatible standardized query template based on the semantic mapping relationships between the natural query language and the cross-platform attributes; Based on the standardized query template, the attribute description habits of different platforms are automatically adapted through a semantic expansion mechanism to generate cross-platform compatible query features.

8. A cross-platform e-commerce resource dynamic matching search system, characterized by: include: The semantic mapping relationship establishment module parses product descriptions from different e-commerce platforms in real time at the data access end, extracts the core attributes of the products, and generates a unified semantic representation. Product descriptions include text and images, and core attributes include name, specifications, and price. The semantic representation uses a context-aware synonym replacement mechanism and attribute association rules to establish semantic mapping relationships between cross-platform attributes. The standardized data is then transmitted to the central database through a distributed consistency synchronization mechanism. A conflict resolution strategy generation module, which dynamically generates a conflict resolution strategy when a cross-platform data conflict occurs based on the product attribute information in the standardized data and the user's historical behavior characteristics. The conflict resolution strategy is executed by a local pre-processing module deployed on the edge node, wherein the local pre-processing module quickly responds to sudden data and batch processes periodic data; A target edge node selection module, after conflict resolution, obtains the conflict-resolved data content, selects a target edge node based on the data content combined with the user's geographic location information in a computing architecture jointly constructed by edge nodes and central servers, and dynamically configures the cache data set of the target edge node; An update module obtains natural query language input by a user, parses corresponding user intent information based on the natural query language, generates corresponding multi-dimensional matching features based on the user intent information combined with product data in a configured cache, determines a set of candidate products through approximate nearest neighbor retrieval based on the multi-dimensional matching features, outputs corresponding search results based on the set of candidate products, obtains user feedback on the search results, and uses the feedback information to update the intent parsing model and matching strategy.

9. A cross-platform e-commerce resource dynamic matching search system, characterized by: It includes a processor, in which a program of the cross-platform e-commerce resource dynamic matching search method as described in any one of claims 1 to 7 is running.

10. A storage medium, characterized in that: A program storing the cross-platform e-commerce resource dynamic matching search method according to any one of claims 1 to 7.

Images