A method, system and storage medium for searching short-life cycle products

Abstract

The present application provides a search method, system and storage medium for short-life cycle commodities. By pre-processing the search terms according to certain strategy configurations and then retrieving them, the problem of search recall quantity in business scenarios with small quantity of commodities and short cycle can be solved, while ensuring the accuracy and relevance of the recall results, and improving the overall search conversion rate and user experience. The expansion recall strategy is determined by the number of search rough recall results, and the expansion recall is performed according to the query part of speech, and the expansion recall results are processed by the promotion and demotion strategy. For business scenarios with no search results and few result recalls, inefficient traffic pages are optimized to improve traffic conversion capabilities.

Description

Technical Field

[0001] The present application belongs to the field of computer technology, and more specifically, relates to a search method, system and storage medium for short-life cycle products. Background Art

[0002] In recent years, with the intersection and combination of the two super tracks of social and e-commerce, private domain e-commerce has gradually risen. In order to meet the diverse needs of C-end users, the update and iteration speed of platform products has accelerated and their life cycle has shortened. In this regard, the accuracy of the search system and the optimization of user experience are particularly important.

[0003] The current product search method usually integrates search terms and related terms to perform retrieval and matching to generate recall results. However, for short-life cycle products in private domain e-commerce scenarios, there are problems such as few recall results and results that do not meet expectations, which seriously affects the revenue of the e-commerce platform and the user experience. Summary of the invention

[0004] In view of this, the present application provides a search method, system and storage medium for short-life cycle products, which effectively solve the problems of inaccurate search recall results and shortage of recall quantities caused by the short life cycle of products in private domain e-commerce scenarios.

[0005] The specific technical solutions of this application are as follows:

[0006] The first aspect of the present application provides a method for searching for short-life cycle products, comprising the following steps:

[0007] Get the search terms of the search request and perform query preprocessing;

[0008] Search in ES based on the query preprocessing results to obtain the initial recall results;

[0009] Match the initial recall result with the query to obtain the search rough recall result;

[0010] Determine whether to implement the expanded recall strategy based on the number of search rough ranking recall results. If not, directly output the search rough ranking recall results. If the expanded recall strategy is implemented, expand the recall based on the query part of speech to generate specific recall results.

[0011] The specific recall results are processed with promotion and demotion strategies to obtain the final recall results.

[0012] Preferably, the retrieving in ES according to the query preprocessing result to obtain the initial recall result includes:

[0013] Perform error correction on the search term of the current request according to the error correction dictionary;

[0014] Perform word segmentation search according to the logic of word segmentation to obtain the result of word segmentation recall;

[0015] The recall results of the classification and word segmentation are given the first priority for the initial recall result display.

[0016] Preferably, the file segmentation logic is specifically as follows:

[0017] The following steps are performed in descending order of priority:

[0018] The query is not segmented, and the query and the product title are fuzzy matched without word segmentation;

[0019] Perform word segmentation on the query and take the union of the matching results;

[0020] Perform word segmentation on the query and take the intersection of the matching results.

[0021] Preferably, searching in ES according to the preprocessing result to obtain the initial recall result further includes:

[0022] Each word segment is associated with a synonym dictionary to generate synonyms and near synonyms related to the query, and then retrieved to obtain the synonym recall results;

[0023] Extract query core words for retrieval to obtain core word recall results;

[0024] The recall results are displayed initially according to the priority from high to low, where the priority of the recall results is as follows from high to low: recall results of graded and word segmentation, recall results of synonyms and antonyms, and recall results of core words.

[0025] Preferably, the information matching of the initial recall result and the query to obtain the search rough recall result is specifically:

[0026] The query intent is identified based on the category prediction algorithm and matched with the initial recall result to generate a matching score.

[0027] Sort by matching score, and use the sorting result as the basis for product search display order.

[0028] Preferably, it also includes:

[0029] The query and the initial recall result are weighted by features to calculate the relevance score, where the parameters of feature weighting are set according to categories, entity words, attributes, and styles;

[0030] The matching score and the relevance score are combined to perform sorting, and the sorting results are used as the basis for the order of product search display.

[0031] Preferably, the step of performing expanded recall according to the query part of speech to generate a specific recall result is specifically as follows:

[0032] Predict leaf category data in the corresponding dimension based on the query part of speech, and perform expanded recall based on the leaf data.

[0033] Preferably, the strategy of performing promotion and demotion on a specific recall result is specifically as follows:

[0034] Specific recall results are weighted according to business activities, and the weighted calculation results are used as the basis for the order of product search display.

[0035] A second aspect of the present application provides a method system for searching for short life cycle products, comprising a memory and a processor, wherein the memory comprises a search program for short life cycle products, and when the search program for short life cycle products is executed by the processor, the following steps are implemented:

[0036] Get the search terms of the search request and perform query preprocessing;

[0037] Search in ES based on the query preprocessing results to obtain the initial recall results;

[0038] Match the initial recall result with the query to obtain the search rough recall result;

[0039] Determine whether to implement the expanded recall strategy based on the number of search rough ranking recall results. If not, directly output the search rough ranking recall results. If the expanded recall strategy is implemented, expand the recall based on the query part of speech to generate specific recall results.

[0040] The specific recall results are processed with promotion and demotion strategies to obtain the final recall results.

[0041] Preferably, the retrieving in ES according to the query preprocessing result to obtain the initial recall result includes:

[0042] Perform error correction on the search term of the current request according to the error correction dictionary;

[0043] Perform word segmentation search according to the logic of word segmentation to obtain the result of word segmentation recall;

[0044] The recall results of the classification and word segmentation are given the first priority for the initial recall result display.

[0045] Preferably, the file segmentation logic is specifically as follows:

[0046] The following steps are performed in descending order of priority:

[0047] The query is not segmented, and the query and the product title are fuzzy matched without word segmentation;

[0048] Perform word segmentation on the query and take the union of the matching results;

[0049] Perform word segmentation on the query and take the intersection of the matching results.

[0050] Preferably, searching in ES according to the preprocessing result to obtain the initial recall result further includes:

[0051] Each word segment is associated with a synonym dictionary to generate synonyms and near synonyms related to the query, and then retrieved to obtain the synonym recall results;

[0052] Extract query core words for retrieval to obtain core word recall results;

[0053] The recall results are displayed initially according to the priority from high to low, where the priority of the recall results is as follows from high to low: recall results of graded and word segmentation, recall results of synonyms and antonyms, and recall results of core words.

[0054] Preferably, the information matching of the initial recall result and the query to obtain the search rough recall result is specifically:

[0055] The query intent is identified based on the category prediction algorithm and matched with the initial recall result to generate a matching score.

[0056] Sort by matching score, and use the sorting result as the basis for product search display order.

[0057] Preferably, it also includes:

[0058] The query and the initial recall result are weighted by features to calculate the relevance score, where the parameters of feature weighting are set according to categories, entity words, attributes, and styles;

[0059] The matching score and the relevance score are combined to perform sorting, and the sorting results are used as the basis for the order of product search display.

[0060] Preferably, the step of performing expanded recall according to the query part of speech to generate a specific recall result is specifically as follows:

[0061] Predict leaf category data in the corresponding dimension based on the query part of speech, and perform expanded recall based on the leaf data.

[0062] Preferably, the strategy of performing promotion and demotion on a specific recall result is specifically as follows:

[0063] Specific recall results are weighted according to business activities, and the weighted calculation results are used as the basis for the order of product search display.

[0064] A third aspect of the present application provides a computer-readable storage medium, which includes a search program for short-life cycle products. When the search program for short-life cycle products is executed by a processor, the steps of the search method for short-life cycle products are implemented.

[0065] The benefits brought by the technical solution provided by this application are:

[0066] 1. The search method, system and storage medium for short-life cycle products provided in this application can solve the problem of search recall quantity in business scenarios such as small quantity of products and short cycle by performing query preprocessing according to certain strategy configurations and then performing retrieval, while ensuring the accuracy and relevance of the recall results, and improving the overall search conversion rate and user experience.

[0067] 2. The search method, system and storage medium for short-life cycle products provided in this application determine the expansion recall strategy by searching the number of rough recall results, expand the recall according to the query part of speech, and process the expansion recall results with a promotion and demotion strategy. For business scenarios with no search results or few result recalls, it optimizes inefficient traffic pages and improves traffic conversion capabilities. BRIEF DESCRIPTION OF THE DRAWINGS

[0068] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without paying creative labor.

[0069] Figure 1 This is a flow chart of a method for searching for short-life cycle products according to an embodiment of the present application;

[0070] Figure 2 This is a block diagram of a search system for short-life cycle products according to an embodiment of the present application. DETAILED DESCRIPTION

[0071] In order to make the purpose, features, and advantages of this application more obvious and easy to understand, the technical solutions in the embodiments of this application are clearly and completely described. Obviously, the embodiments described below are only part of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.

[0072] Please refer to Figure 1 , Figure 1 This is a flow chart of a method for searching for short-life cycle products according to an embodiment of the present application.

[0073] A first aspect of an embodiment of the present application provides a method for searching for products with a short life cycle, comprising the following steps:

[0074] S102: Obtain search terms of the search request and perform query preprocessing;

[0075] S104: Search in ES according to the query preprocessing result to obtain the initial recall result;

[0076] S106: Match the initial recall result with the query to obtain a rough search recall result;

[0077] S108: judging whether to implement an expanded recall strategy according to the number of search rough sorting recall results, and directly outputting the search rough sorting recall results if the expanded recall strategy is not implemented; and if the expanded recall strategy is implemented, expanding the recall according to the query part of speech to generate a specific recall result;

[0078] S110: Perform promotion and demotion strategy processing on the specific recall result to obtain the final recall result.

[0079] It should be noted that the results of query preprocessing of the search terms are used as search terms to obtain the initial recall results, and the initial recall results are roughly ranked according to the information relevance of the query. If the number of roughly ranked results cannot meet the recall needs of the current product, an expanded recall strategy is implemented, that is, the search terms are regenerated by expanding the information of the query word type to generate specific recall results. Since the specific recall result is a recall mode configured for a specific demand scenario, it is necessary to calculate the weight of the recall result under dimensions such as business model to improve the pertinence and accuracy of the expanded recall result.

[0080] The search method for short-life cycle products provided in the embodiment of the present application performs query preprocessing on the search terms according to a certain strategy configuration before performing retrieval, and then determines the expansion and recall strategy based on the number of search rough recall results, and performs expansion and recall according to the query word category, and performs promotion and demotion strategy processing on the expansion and recall results. This can solve the problem of search and recall quantity in business scenarios such as small product quantity and short cycle, ensure the accuracy and relevance of recall results, improve the overall search conversion rate and user experience, optimize inefficient traffic pages, and enhance traffic conversion capabilities.

[0081] According to the embodiment of the present application, the searching in ES according to the query preprocessing result to obtain the initial recall result includes:

[0082] Perform error correction on the search term of the current request according to the error correction dictionary;

[0083] Perform word segmentation search according to the logic of word segmentation to obtain the result of word segmentation recall;

[0084] The recall results of the classification and word segmentation are given the first priority for the initial recall result display.

[0085] It should be noted that the purpose of error correction is to correct typos and delete useless fields with low recognition. The purpose of the file segmentation logic is to classify search terms according to certain rules and attributes, and search for matching search results under the corresponding classification. The file segmentation method as the first priority result display has higher search efficiency.

[0086] According to the embodiment of the present application, the file segmentation logic is specifically as follows:

[0087] The following steps are performed in descending order of priority:

[0088] The query is not segmented, and the query and the product title are fuzzy matched without word segmentation;

[0089] Perform word segmentation on the query and take the union of the matching results;

[0090] Perform word segmentation on the query and take the intersection of the matching results.

[0091] It should be noted that the logic of the file segmentation is divided into three layers. The first layer is all. The second layer is to match each term after the query segmentation with the title segmentation to ensure that every word in the query is matched. The third layer is to match the subset after the query segmentation with the title segmentation. Among them, the number of intersections needs to set a threshold. Generally, the number of hits is greater than the number of sub-words after the original word segmentation.

[0092] According to an embodiment of the present application, searching in ES according to the preprocessing result to obtain the initial recall result further includes:

[0093] Each word segment is associated with a synonym dictionary to generate synonyms and near synonyms related to the query, and then retrieved to obtain the synonym recall results;

[0094] Extract query core words for retrieval to obtain core word recall results;

[0095] The recall results are displayed initially according to the priority from high to low, where the priority of the recall results is as follows from high to low: recall results of graded and word segmentation, recall results of synonyms and antonyms, and recall results of core words.

[0096] It should be noted that the synonym dictionary can be divided into relatedness according to the purpose, application scenario and function of the product, and the data can come from the big data information base of products in various fields, and the relevance of synonyms and near synonyms can be adjusted as needed. The query core word refers to extracting the core identifiable word segment in the query, and expanding the recall around the core word segment.

[0097] According to the embodiment of the present application, the initial recall result is matched with the query to obtain the search rough recall result as follows:

[0098] The query intent is identified based on the category prediction algorithm and matched with the initial recall result to generate a matching score.

[0099] Sort by matching score, and use the sorting result as the basis for product search display order.

[0100] It should be noted that the specific prediction range in intent recognition can be determined according to the product category system, and the category prediction algorithm can be determined by the correspondence between product category, product name and application scenario. The intent features in the query are compared with the intent features in the initial recall results to obtain the relevance score for sorting and display, so that users can quickly filter information with similar phrases but different category applications.

[0101] According to an embodiment of the present application, it also includes:

[0102] The query and the initial recall result are weighted by features to calculate the relevance score, where the parameters of feature weighting are set according to categories, entity words, attributes, and styles;

[0103] The matching score and the relevance score are combined to perform sorting, and the sorting results are used as the basis for the order of product search display.

[0104] It should be noted that the feature weight calculation formula can be: y=a*w1+b*w2+...n*wn, where a~n are the feature values ​​of each attribute, and w1~n are the weighted proportions corresponding to the feature values ​​of each attribute.

[0105] According to the embodiment of the present application, the generation of a specific recall result by expanding the recall according to the query part of speech is specifically as follows:

[0106] Predict leaf category data in the corresponding dimension based on the query part of speech, and perform expanded recall based on the leaf data.

[0107] It should be noted that the query part of speech includes information such as category, brand, specification, and field. If it is a category word, the category dimension expansion and recall is performed based on the leaf category results in the query category prediction to ensure the relevance of the expansion and recall results. If it is a brand word, the brand dimension expansion and recall is performed based on similar brands of the brand, where brand similarity is calculated based on the category distribution and price distribution of the products covered by the brand.

[0108] According to the embodiment of the present application, the strategy for performing promotion and demotion on a specific recall result is specifically as follows:

[0109] Specific recall results are weighted according to business activities, and the weighted calculation results are used as the basis for the order of product search display.

[0110] It should be noted that the promotion and demotion strategy is mainly for the business dimension, and all interventions must be carried out according to the priority of the file after the recall results are classified. Business activity weighting means that the products participating in the activities are given priority display, and business activities may include traffic and administrative information. For example, the traffic of key categories and brands of goods is tilted and given priority display; the goods supplied by goods with poor service quality are filtered and sink to the bottom.

[0111] In another embodiment of the present application, the strategy of determining whether to perform an expanded recall is specifically as follows based on the number of search rough sort recall results:

[0112] Determine whether the number of search rough recall results exceeds the expanded recall product quantity threshold. If it exceeds the threshold, the expanded recall strategy will not be implemented. If it is less than the threshold, the expanded recall strategy will be implemented. The expanded recall product quantity threshold is determined according to the length of the product life cycle.

[0113] It should be noted that the threshold for expanding the number of recalled products is a preset value of the system, which can be determined based on the life cycle of the product category or similar products calculated through big data. For products with a short life cycle, the problem of shortage of recall quantity can be further improved, which is especially suitable for modern private domain e-commerce scenarios. At the same time, the number of recalls of products with a long life cycle will not be expanded, so that the recall results maintain a high hit rate and relevance.

[0114] In another embodiment of the present application, it also includes:

[0115] Determine whether to perform the file division operation based on the number of search rough recall results;

[0116] If the number of search rough recall results is greater than the set threshold, the search rough recall results are sorted according to the relevance score and the files are divided according to the sorting results;

[0117] Configure business intervention and functional settings for search items based on gear levels.

[0118] It should be noted that the threshold is set according to the life cycle of the product in different business scenarios. The recalled product set is divided into three levels based on the weighted ranking score reference quantile. Subsequent business interventions and function settings will be configured according to the priority of the divided levels, such as price, sales volume ranking and other functions.

[0119] Please refer to Figure 2 , Figure 2 This is a block diagram of a search system for short-life cycle products according to an embodiment of the present application.

[0120] A second aspect of the embodiment of the present application provides a short life cycle commodity search method system 2, comprising a memory 21 and a processor 22, wherein the memory 21 comprises a short life cycle commodity search program, and when the short life cycle commodity search program is executed by the processor 22, the following steps are implemented:

[0121] Get the search terms of the search request and perform query preprocessing;

[0122] Search in ES based on the query preprocessing results to obtain the initial recall results;

[0123] Match the initial recall result with the query to obtain the search rough recall result;

[0124] Determine whether to implement the expanded recall strategy based on the number of search rough ranking recall results. If not, directly output the search rough ranking recall results. If the expanded recall strategy is implemented, expand the recall based on the query part of speech to generate specific recall results.

[0125] The specific recall results are processed with promotion and demotion strategies to obtain the final recall results.

[0126] The search system for short-life cycle products provided in this application can solve the problem of search recall quantity in business scenarios such as small product quantity and short cycle, while ensuring the accuracy and relevance of the recall results, improving the overall search conversion rate and user experience; expanding the recall according to the query part of speech, and processing the expanded recall results with a promotion and demotion strategy, optimizing inefficient traffic pages and improving traffic conversion capabilities for business scenarios with no search results or few result recalls.

[0127] According to the embodiment of the present application, the searching in ES according to the query preprocessing result to obtain the initial recall result includes:

[0128] Perform error correction on the search term of the current request according to the error correction dictionary;

[0129] Perform word segmentation search according to the logic of word segmentation to obtain the result of word segmentation recall;

[0130] The recall results of the classification and word segmentation are given the first priority for the initial recall result display.

[0131] It should be noted that the purpose of error correction is to correct typos and delete useless fields with low recognition. The purpose of the file segmentation logic is to classify search terms according to certain rules and attributes, and search for matching search results under the corresponding classification. The file segmentation method as the first priority result display has higher search efficiency.

[0132] According to the embodiment of the present application, the file segmentation logic is specifically as follows:

[0133] The following steps are performed in descending order of priority:

[0134] The query is not segmented, and the query and the product title are fuzzy matched without word segmentation;

[0135] Perform word segmentation on the query and take the union of the matching results;

[0136] Perform word segmentation on the query and take the intersection of the matching results.

[0137] It should be noted that the logic of the file segmentation is divided into three layers. The first layer is all. The second layer is to match each term after the query segmentation with the title segmentation to ensure that every word in the query is matched. The third layer is to match the subset after the query segmentation with the title segmentation. Among them, the number of intersections needs to set a threshold. Generally, the number of hits is greater than the number of sub-words after the original word segmentation.

[0138] According to an embodiment of the present application, searching in ES according to the preprocessing result to obtain the initial recall result further includes:

[0139] Each word segment is associated with a synonym dictionary to generate synonyms and near synonyms related to the query, and then retrieved to obtain the synonym recall results;

[0140] Extract query core words for retrieval to obtain core word recall results;

[0141] The recall results are displayed initially according to the priority from high to low, where the priority of the recall results is as follows from high to low: recall results of graded and word segmentation, recall results of synonyms and antonyms, and recall results of core words.

[0142] It should be noted that the synonym dictionary can be divided into relatedness according to the purpose, application scenario and function of the product, and the data can come from the big data information base of products in various fields, and the relevance of synonyms and near synonyms can be adjusted as needed. The query core word refers to extracting the core identifiable word segment in the query, and expanding the recall around the core word segment.

[0143] According to the embodiment of the present application, the initial recall result is matched with the query to obtain the search rough recall result as follows:

[0144] The query intent is identified based on the category prediction algorithm and matched with the initial recall result to generate a matching score.

[0145] Sort by matching score, and use the sorting result as the basis for product search display order.

[0146] It should be noted that the specific prediction range in intent recognition can be determined according to the product category system, and the category prediction algorithm can be determined by the correspondence between product category, product name and application scenario. The intent features in the query are compared with the intent features in the initial recall results to obtain the relevance score for sorting and display, so that users can quickly filter information with similar phrases but different category applications.

[0147] According to an embodiment of the present application, it also includes:

[0148] The query and the initial recall result are weighted by features to calculate the relevance score, where the parameters of feature weighting are set according to categories, entity words, attributes, and styles;

[0149] The matching score and the relevance score are combined to perform sorting, and the sorting results are used as the basis for the order of product search display.

[0150] It should be noted that the feature weight calculation formula can be: y=a*w1+b*w2+...n*wn, where a~n are the feature values ​​of each attribute, and w1~n are the weighted proportions corresponding to the feature values ​​of each attribute.

[0151] According to the embodiment of the present application, the generation of a specific recall result by expanding the recall according to the query part of speech is specifically as follows:

[0152] Predict leaf category data in the corresponding dimension based on the query part of speech, and perform expanded recall based on the leaf data.

[0153] It should be noted that the query part of speech includes information such as category, brand, specification, and field. If it is a category word, the category dimension expansion and recall is performed based on the leaf category results in the query category prediction to ensure the relevance of the expansion and recall results. If it is a brand word, the brand dimension expansion and recall is performed based on similar brands of the brand, where brand similarity is calculated based on the category distribution and price distribution of the products covered by the brand.

[0154] According to the embodiment of the present application, the strategy for performing promotion and demotion on a specific recall result is specifically as follows:

[0155] Specific recall results are weighted according to business activities, and the weighted calculation results are used as the basis for the order of product search display.

[0156] It should be noted that the promotion and demotion strategy is mainly for the business dimension, and all interventions must be carried out according to the priority of the file after the recall results are classified. Business activity weighting means that the products participating in the activities are given priority display, and business activities may include traffic and administrative information. For example, the traffic of key categories and brands of goods is tilted and given priority display; the goods supplied by goods with poor service quality are filtered and sink to the bottom.

[0157] In another embodiment of the present application, the strategy of determining whether to perform an expanded recall is specifically as follows based on the number of search rough sort recall results:

[0158] Determine whether the number of search rough recall results exceeds the expanded recall product quantity threshold. If it exceeds the threshold, the expanded recall strategy will not be implemented. If it is less than the threshold, the expanded recall strategy will be implemented. The expanded recall product quantity threshold is determined according to the length of the product life cycle.

[0159] It should be noted that the threshold for expanding the number of recalled products is a preset value of the system, which can be determined based on the life cycle of the product category or similar products calculated through big data. For products with a short life cycle, the problem of shortage of recall quantity can be further improved, which is especially suitable for modern private domain e-commerce scenarios. At the same time, the number of recalls of products with a long life cycle will not be expanded, so that the recall results maintain a high hit rate and relevance.

[0160] In another embodiment of the present application, it also includes:

[0161] Determine whether to perform the file division operation based on the number of search rough recall results;

[0162] If the number of search rough recall results is greater than the set threshold, the search rough recall results are sorted according to the relevance score and the files are divided according to the sorting results;

[0163] Configure business intervention and functional settings for search items based on gear levels.

[0164] It should be noted that the threshold is set according to the life cycle of the product in different business scenarios. The recalled product set is divided into three levels based on the weighted ranking score reference quantile. Subsequent business interventions and function settings will be configured according to the priority of the divided levels, such as price, sales volume ranking and other functions.

[0165] A third aspect of an embodiment of the present application provides a computer-readable storage medium, which includes a search program for short-life cycle products. When the search program for short-life cycle products is executed by a processor, the steps of the search method for short-life cycle products are implemented.

[0166] For details on the steps in the above short-life cycle product search method, see Figure 1 The description of the method steps will not be repeated here.

[0167] In the several embodiments provided in the present application, it should be understood that the disclosed devices and methods can be implemented in other ways. The device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as: multiple units or components can be combined, or can be integrated into another system, or some features can be ignored or not executed. In addition, the coupling, direct coupling, or communication connection between the components shown or discussed can be through some interfaces, and the indirect coupling or communication connection of the devices or units can be electrical, mechanical or other forms.

[0168] The units described above as separate components may or may not be physically separated, and the components displayed as units may or may not be physical units; they may be located in one place or distributed on multiple network units; some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0169] In addition, all functional units in the embodiments of the present invention may be integrated into one processing unit, or each unit may be separately used as a unit, or two or more units may be integrated into one unit; the above-mentioned integrated units may be implemented in the form of hardware or in the form of hardware plus software functional units.

[0170] Those skilled in the art can understand that: all or part of the steps of implementing the above method embodiments can be completed by hardware related to program instructions, and the aforementioned program can be stored in a computer-readable storage medium. When the program is executed, it executes the steps of the above method embodiments; and the aforementioned storage medium includes: mobile storage devices, read-only memories (ROM, Read-Only Memory), random access memories (RAM, Random Access Memory), disks or optical disks, and other media that can store program codes.

[0171] Alternatively, if the above-mentioned integrated unit of the present invention is implemented in the form of a software function module and sold or used as an independent product, it can also be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the embodiment of the present invention can be essentially or partly reflected in the form of a software product that contributes to the prior art. The computer software product is stored in a storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the methods described in each embodiment of the present invention. The aforementioned storage medium includes: various media that can store program codes, such as mobile storage devices, ROM, RAM, magnetic disks or optical disks.

[0172] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for searching for products with a short life cycle, characterized in that: The steps include: Get the search terms of the search request and perform query preprocessing; Search in ES based on the query preprocessing results to obtain the initial recall results; Match the initial recall result with the query to obtain the search rough recall result; Determine whether to implement the expanded recall strategy based on the number of search rough ranking recall results. If not, directly output the search rough ranking recall results. If the expanded recall strategy is implemented, expand the recall based on the query part of speech to generate specific recall results. Apply promotion and demotion strategies to specific recall results to obtain the final recall results; The initial recall results obtained by searching in ES based on the query preprocessing results include: Perform error correction on the search term of the current request according to the error correction dictionary; Perform word segmentation search according to the logic of word segmentation to obtain the result of word segmentation recall; The recall results of the file and word segmentation are given the first priority for initial recall result display; The logic of file segmentation is as follows: The following steps are performed in descending order of priority: The query is not segmented, and the query and the product title are fuzzy matched without word segmentation; Perform word segmentation on the query and take the union of the matching results; Perform word segmentation on the query and take the intersection of the matching results; The step of searching in ES according to the preprocessing result to obtain the initial recall result also includes: Each word segment is associated with a synonym dictionary to generate synonyms and near synonyms related to the query, and then retrieved to obtain the synonym recall results; Extract query core words for retrieval to obtain core word recall results; The recall results are displayed in the order of priority from high to low. The recall results are prioritized from high to low as follows: recall results of graded and segmented words, recall results of synonyms and antonyms, and recall results of core words; The specific processing of the strategy for raising and lowering the rights of a specific recall result is as follows: Specific recall results are weighted according to business activities, and the weighted calculation results are used as the basis for the order of product search display; Also includes: Determine whether to perform the file division operation based on the number of search rough recall results; If the number of search rough recall results is greater than the set threshold, the search rough recall results are evaluated based on the relevance. Sort by points, and divide files according to the sorting results; Configure business intervention and functional settings for search items based on gear levels.

2. The method for searching for short-life cycle products according to claim 1, characterized in that: The initial recall result is matched with the query to obtain the search rough recall result: The query intent is identified based on the category prediction algorithm and matched with the initial recall result to generate a matching score. Sort by matching score, and use the sorting result as the basis for product search display order.

3. The method for searching for short-life cycle products according to claim 1, characterized in that: Also includes: The query and the initial recall result are weighted by features to calculate the relevance score, where the parameters of feature weighting are set according to categories, entity words, attributes, and styles; The matching score and the relevance score are combined to perform sorting, and the sorting results are used as the basis for the order of product search display.

4. The method for searching for short life cycle products according to claim 1, characterized in that: The specific recall result generated by expanding the recall according to the query part of speech is specifically as follows: Predict leaf category data in the corresponding dimension based on the query part of speech, and perform expanded recall based on the leaf data.

5. A search method system for short life cycle products, characterized in that: The method comprises a memory and a processor, wherein the memory comprises a search program for short life cycle commodities, and when the search program for short life cycle commodities is executed by the processor, the steps of the search method for short life cycle commodities as claimed in any one of claims 1 to 4 are implemented.

6. A computer-readable storage medium, characterized in that: The computer-readable storage medium includes a search program for short-life cycle commodities. When the search program for short-life cycle commodities is executed by a processor, the steps of the search method for short-life cycle commodities as claimed in any one of claims 1 to 4 are implemented.

Images