An item search method, device, equipment and storage medium

Abstract

An embodiment of the present invention discloses an article search method, device, equipment, and storage medium. The method includes: obtaining input target article search information; performing search recall on the target article search information to determine a candidate article set; based on an article category recognition model, performing category recognition on the target article search information to determine the target article category; based on the article attributes corresponding to the target article category, performing attribute extraction on the target article search information to obtain the extracted target article attribute information; and determining the target article search result based on the target article attribute information and the candidate article set. Through the technical solution of the embodiment of the present invention, automatic search of articles can be achieved, the article search efficiency can be improved, and the accuracy of article search can be ensured.

Description

Technical Field

[0001] Embodiments of the present invention relate to computer technology, and in particular, to an item search method, apparatus, device, and storage medium. Background Art

[0002] With the rapid development of computer technology, item search can be performed based on item information input by a user to obtain the item that the user wants to purchase.

[0003] Currently, when searching for some items with strong professionalism, such as industrial products with professional attributes, it is usually professionals who search by manually matching items to ensure the accuracy of the search.

[0004] However, in the process of implementing the present invention, the inventors found that there are at least the following problems in the prior art:

[0005] The existing manual matching method is time-consuming and laborious, greatly reducing the search efficiency. Summary of the Invention

[0006] Embodiments of the present invention provide an item search method, apparatus, device, and storage medium to achieve automatic item search, improve item search efficiency, and ensure the accuracy of item search.

[0007] In a first aspect, an embodiment of the present invention provides an item search method, including:

[0008] Obtaining input target item search information;

[0009] Performing search recall on the target item search information to determine a candidate item set;

[0010] Based on an item category recognition model, performing category recognition on the target item search information to determine a target item category;

[0011] Based on the item attributes corresponding to the target item category, performing attribute extraction on the target item search information to obtain the extracted target item attribute information;

[0012] Based on the target item attribute information and the candidate item set, determining a target item search result.

[0013] In a second aspect, an embodiment of the present invention further provides an item search apparatus, including:

[0014] A target item search information acquisition module, configured to obtain input target item search information;

[0015] A candidate item set determination module, configured to perform search recall on the target item search information to determine a candidate item set;

[0016] A target item category determination module, configured to perform category recognition on the target item search information based on an item category recognition model to determine a target item category;

[0017] An item attribute information extraction module, configured to perform attribute extraction on the target item search information based on the item attributes corresponding to the target item category to obtain the extracted target item attribute information;

[0018] An item search result determination module, configured to determine a target item search result based on the target item attribute information and the candidate item set.

[0019] In a third aspect, an embodiment of the present invention further provides an electronic device, where the electronic device includes:

[0020] One or more processors;

[0021] A memory, configured to store one or more programs;

[0022] When the one or more programs are executed by the one or more processors, the one or more processors implement the item search method provided in any embodiment of the present invention.

[0023] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium, on which a computer program is stored, and when the program is executed by a processor, the item search method provided in any embodiment of the present invention is implemented.

[0024] One of the above embodiments of the invention has the following advantages or beneficial effects:

[0025] By performing search recall on the input target item search information to determine a candidate item set, performing category recognition on the target item search information based on an item category recognition model to determine a target item category, and performing attribute extraction on the target item search information based on the item attributes corresponding to the target item category to obtain the extracted target item attribute information, a more accurate target item search result can be determined from the candidate item set based on the extracted target item attribute information, ensuring the accuracy of item search. And the search process does not require human participation, realizing automatic search of items and improving the efficiency of item search. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for use in the description of the embodiments or the prior art. Obviously, the drawings in the following description are some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0027] Figure 1 is a flowchart of an item search method provided by an embodiment of the present invention;

[0028] Figure 2 is a flowchart of another item search method provided by an embodiment of the present invention;

[0029] Figure 3 is a schematic structural diagram of an item search device provided by an embodiment of the present invention;

[0030] Figure 4 is a schematic structural diagram of an electronic device provided by an embodiment of the present invention. Detailed implementation manners

[0031] The present invention will be further described in detail below with reference to the accompanying drawings and embodiments. It can be understood that the specific embodiments described herein are only used to explain the present invention, rather than limiting the present invention. Additionally, it should be noted that for the sake of description, only parts related to the present invention rather than all structures are shown in the accompanying drawings.

[0032] Figure 1 is a flowchart of an item search method provided by an embodiment of the present invention. This embodiment is applicable to the situation of searching for items, especially in the application scenario of searching for industrial products. This method can be executed by an item search device, which can be implemented in software and / or hardware and integrated into an electronic device. As Figure 1 shown, the method specifically includes the following steps:

[0033] S110. Obtain the input target item search information.

[0034] Specifically, the user can input the corresponding target item search information in the search input box based on the information of the target item to be searched. The target item search information may include at least one attribute information of the target item. The attribute information may include, but is not limited to, the name information, model information, brand information, structure information, and color information of the target item. It should be noted that the target item search information can be input in any information format, and there is no requirement for the user to input the target item search information in a specific information format. For example, the user can directly input the specific model information, such as 1000, but does not clearly specify that the input information is the model attribute information. Therefore, based on "1000" alone, it is impossible to determine the specific attribute corresponding to this information, which may be model information or length information, etc. This embodiment can accurately search for items without standardizing the input target item search information and item information, that is, in the case of fuzzy input by the user, to ensure the accuracy of item search.

[0035] Exemplarily, the target item searched in this embodiment may refer to industrial products with professional attributes. Industrial products are products used for processing and production or enterprise operation after purchase. Industrial products belong to the business of enterprise-to-business (To B), so more precise item sourcing and item search are required. When purchasing industrial products, based on the information of the target item on the purchase and sales order, the corresponding target item search information can be input in the search box of the shopping platform, so as to search for the purchase link information corresponding to the corresponding target item based on the input target item search information, and then purchase the target item based on this purchase link information.

[0036] S120. Perform search recall on the target item search information to determine a candidate item set.

[0037] Among them, the candidate item set may include multiple candidate items. A candidate item may refer to an item associated with the target item search information. The candidate item set may include different search results of the same candidate item or different search results of different types of candidate items. For example, there may be multiple purchase links for the same item.

[0038] Specifically, based on a preset search recall method, such as a word recall method based on an inverted index or a semantic recall method based on vectors, etc., search recall can be performed in a pre-set item set (such as an item pool) to determine each candidate item that matches the target item search information, and obtain a candidate item set, that is, a search recall pool. For example, each item information in the item pool can be synchronously updated to the elasticsearch database, and the item keywords can be used as a word segmentation dictionary to segment the item information to improve the matching efficiency. Each item information in the item pool can be matched with the input target item search information for information similarity to obtain the recall confidence corresponding to each item information, and the top N (i.e., topN) item information with higher recall confidence can be used as the candidate item set.

[0039] S130. Based on the item category recognition model, perform category recognition on the target item search information to determine the target item category.

[0040] Among them, the categories can be various categories obtained by classifying all items in advance based on item information. Based on the scope of item division, the categories can be divided into first-level categories, second-level categories, third-level categories, etc. Among them, the second-level categories are the categories obtained by further subdividing the first-level categories. The third-level categories are the categories obtained by further subdividing the second-level categories. For example, the first-level category is: household appliances; the second-level categories corresponding to this first-level category can include: televisions, air conditioners, washing machines, refrigerators, etc. The third-level categories corresponding to the second-level category of air conditioners can include: wall-mounted air conditioners, cabinet air conditioners, central air conditioners, etc. The item category recognition model can be any classification model used to recognize item categories. The item category recognition model can be used to recognize categories at any level, such as for recognizing first-level categories, second-level categories, or third-level categories, etc. The item category recognition model can be obtained by pre-training the model based on sample data.

[0041] Specifically, the target item search information can be input into the pre-trained item category recognition model for category recognition, and based on the output of the item category recognition model, the target item category corresponding to the target item search information can be determined. For example, the item category recognition model can directly output the target item category, or can output the recognition probability belonging to each item category, and determine the item category with the highest recognition probability as the target item category.

[0042] Exemplarily, S130 can include: inputting the target item search information into the item category recognition model for recognizing the last-level category, and determining the target item category according to the output of the item category recognition model, where the target item category refers to the last-level category to which the target item belongs.

[0043] Among them, the last-level category can be the last level of the classification of items, that is, the category with the finest division granularity. For example, when the categories can be divided into first-level categories, second-level categories, and third-level categories, the last-level category can refer to the third-level category. Specifically, the item category recognition model can be used to recognize the last-level category that matches the target item search information, so that by recognizing the last-level category with the finest granularity, the proprietary attribute information of the target item can be more accurately located, making the subsequently extracted attribute information more accurate, and further improving the accuracy of item search.

[0044] S140. Extract the attributes of the target item search information based on the item attributes corresponding to the target item category to obtain the extracted target item attribute information.

[0045] Among them, the target item category can correspond to one or more item attributes. Different item categories can have different proprietary attributes or the same general attributes, and the attribute values corresponding to the same general attributes are different. For example, two different item categories both have the attribute of item model, but under different item categories, the model values corresponding to the item model are different.

[0046] Specifically, for each item attribute corresponding to the target item category, each attribute value of the item attribute can be matched with the target item search information. If the match is successful, it indicates that the target item search information contains the target attribute value that matches successfully. At this time, the item attribute and the corresponding target attribute value can be used as the target item attribute information, so as to identify the attribute meaning corresponding to the information in the target item search information, extract all the target item attribute information, and then accurately perform item search when the input of the target item search information is fuzzy.

[0047] S150. Determine the target item search result based on the target item attribute information and the candidate item set.

[0048] Specifically, based on the extracted target item attribute information, candidate items that match the target item attribute information can be determined from the candidate item set as the target item, and the target item page link information corresponding to the target item can be used as the target item search result for display, so that a target item acquisition task, such as an order to purchase the target item, can be generated on the target item page corresponding to the target item page link information.

[0049] Exemplarily, S150 can include: matching the candidate item attribute information corresponding to each candidate item in the candidate item set with the target item attribute information; using the candidate items with successful information matching as the target item to determine the target item search result.

[0050] Specifically, the candidate item attribute information corresponding to each candidate item in the candidate item set can be matched with the target item attribute information, and the candidate items with the target item attribute information can be used as the target item. Thus, based on the target item attribute information, the target item can be more accurately screened out from the candidate item set, ensuring the accuracy of item search.

[0051] In the technical solution of this embodiment, a candidate item set is determined by performing search recall on the input target item search information, and the target item category is determined by performing category recognition on the target item search information based on the item category recognition model. Based on the item attributes corresponding to the target item category, attribute extraction is performed on the target item search information to obtain the extracted target item attribute information. Therefore, based on the extracted target item attribute information, a more accurate target item search result can be determined from the candidate item set, ensuring the accuracy of item search. Moreover, no human intervention is required during the search process, realizing the automatic search of items and improving the item search efficiency.

[0052] Based on the above technical solution, S130 may include: performing category recognition on the target item search information based on the first item category recognition model to determine the first recognition probability belonging to each item category; performing category recognition on the target item search information based on the second item category recognition model to determine the second recognition probability belonging to each item category; and determining the target item category based on the first recognition probability and the second recognition probability belonging to each item category.

[0053] Among them, the first item category recognition model and the second item category recognition model are obtained by training the model based on the co-training method, so that semi-supervised learning can be performed when the labeled data is less, further ensuring the accuracy of model recognition. The first item category recognition model and the second item category recognition model may be two classification models with different recognition methods. Exemplarily, the first item category recognition model may be a classification model based on word vector recognition. The second item category recognition model may be a classification model based on word vector recognition. For example, the first item category recognition model may be, but is not limited to, a classification model composed of a Bert pre-trained model and a multi-layer perceptron (MLP). The second item category recognition model may be, but is not limited to, a FastText classification model.

[0054] Specifically, when there is a situation of chaotic category hanging for items, using this part of the data to train the item category recognition model will affect the training effect of the model and reduce the accuracy of model recognition. In response to this, the co-training method can be used to train two item category recognition models, thereby improving the accuracy of category recognition. For example, label data with higher confidence can be screened out using category keywords, and the label data can be sampled with replacement to obtain D 1 and D 2 two sample data sets, and D 1 and D 2 are respectively processed into a word vector training set D′ 1 and a word vector training set D′ 2 , and D′ 1 and D′2 The first item category recognition model and the second item category recognition model are co-trained respectively. After the co-training is completed, the target item search information can be input into the trained first item category recognition model for category recognition, and based on the output of the first item category recognition model, the first recognition probability belonging to each item category is determined. Similarly, the target item search information can be input into the trained second item category recognition model for category recognition, and based on the output of the second item category recognition model, the second recognition probability belonging to each item category is determined. The first recognition probability and the second recognition probability belonging to each item category are added or averaged to determine the target recognition probability belonging to each item category, and the item category with the highest target recognition probability is used as the target item category. Therefore, by jointly using the first item category recognition model and the second item category recognition model for category recognition, the accuracy of category recognition can be further improved, and further the accuracy of item search can be improved.

[0055] Figure 2 FIG. is a flowchart of another item search method provided by an embodiment of the present invention. Based on the above embodiments, the specific process of attribute extraction is described in detail. The explanations of the same or corresponding terms in the above embodiments are not repeated here.

[0056] See Figure 2 , another item search method provided by this embodiment specifically includes the following steps:

[0057] S210. Obtain the input target item search information.

[0058] S220. Perform search recall on the target item search information to determine a candidate item set.

[0059] S230. Based on the item category recognition model, perform category recognition on the target item search information to determine the target item category.

[0060] S240. Determine the target information matching method and the target entity recognition model corresponding to the target item category.

[0061] Among them, different item categories correspond to different information matching methods. The target information matching method corresponding to the target item category can be an information matching method set in advance based on the attribute extraction rules corresponding to the target item category. The attribute extraction rules can be generated based on expert experience. The target information matching method can include the information matching methods corresponding to each item attribute under the target item category. The information matching method corresponding to each item attribute can be, but is not limited to, a method based on regular expression matching or a string matching method.

[0062] Among them, different item categories correspond to different entity recognition models. The target entity recognition model can be used to label the information input into the model with the item attribute information belonging to the target item category. The target entity recognition model can be any network model for entity recognition. For example, the target entity recognition model can be a network model composed of a Bert pre-trained model and a conditional random field (CRF). The target entity recognition model can be pre-trained based on sample data and corresponding data labels. The data labels can be obtained through manual annotation. To increase generalization and reduce the workload of labeled data, the target information matching method can be used to automatically label the extractable attributes, and the remaining attributes are then manually labeled. The loss function L can be determined based on the loss value of the automatically labeled attributes and the loss value cost of the manually labeled attributes For example, λ is a proportionality coefficient with a value in the range (0.5, 1], so that the target entity recognition model can pay more attention to the manually labeled attributes and further improve the model training effect.

[0063] Specifically, the information matching method corresponding to each item category can be preset based on expert experience, so that when searching for items, the target information matching method corresponding to the target item category can be directly obtained. The entity recognition model corresponding to each item category can be pre-trained, so that when searching for items, the target entity recognition model corresponding to the target item category can be directly obtained, further improving the item search efficiency.

[0064] S250. Based on the target information matching method and / or the target entity recognition model, extract the item attributes of the target item search information under the target item category to obtain the extracted target item attribute information.

[0065] Specifically, in this embodiment, the item attributes of the target item search information under the target item category can be extracted only based on the target information matching method to obtain the extracted target item attribute information. For example, based on the regular expression corresponding to each item attribute under the target item category, the target item search information can be matched, and the item attributes with successful matches and the corresponding target attribute information are used as the target item attribute information; or, each optional attribute information corresponding to each item attribute under the target item category is matched with the target item search information, and the item attributes with successful matches and the corresponding target attribute information are used as the target item attribute information.

[0066] In this embodiment, it is also possible to extract the item attributes under the target item category from the target item search information only based on the target entity recognition model, so as to obtain the extracted target item attribute information. For example, the target item search information can be input into the target entity recognition model to identify the item attributes under the target item category, and based on the output of the target entity recognition model, the meaning of the attribute information in the target item search information can be obtained, thereby obtaining the target item attribute information.

[0067] In this embodiment, it is also possible to extract the item attributes under the target item category from the target item search information by combining the target information matching method and the target entity recognition model, so as to obtain the extracted target item attribute information, thereby enabling more comprehensive and accurate attribute information to be extracted, and further improving the accuracy of item search.

[0068] Exemplarily, "extracting the item attributes under the target item category from the target item search information by combining the target information matching method and the target entity recognition model to obtain the extracted target item attribute information" in S250 may include: extracting the first item attributes under the target item category from the target item search information based on the target information matching method to obtain the extracted first item attribute information; extracting the second item attributes under the target item category from the target item search information based on the target entity recognition model to obtain the extracted second item attribute information; and obtaining the extracted target item attribute information based on the first item attribute information and the second item attribute information.

[0069] Among them, the first item attribute may refer to the item attribute that can be matched by using the target information matching method. For example, the first item attribute may be an item attribute with clear matching rules, such as an attribute that can be enumerated or an attribute with a specific structure of the attribute value, etc. The second item attribute may refer to the item attribute that cannot be matched by using the target information matching method. For example, the second item attribute may be an item attribute with unclear matching rules, such as an attribute that cannot be enumerated, etc.

[0070] Specifically, the first item attribute information under the target item category can be extracted from the target item search information by using regular expressions or string matching. The target item search information can be input into the target entity recognition model to identify the second item attributes under the target item category, and based on the output of the target entity recognition model, the second item attribute information under the target item category can be extracted from the target item search information. Therefore, by using the extraction method that combines the target information matching method and the target entity recognition model, all the target item attribute information can be more accurately extracted from the target item search information, and the accuracy of item search is further improved.

[0071] Exemplarily, based on the target information matching method, the first item attribute extraction is performed on the target item search information under the target item category, and the obtained first item attribute information may include: extracting the attribute information corresponding to the first item attribute in the target item search information based on the regular expression corresponding to the first item attribute under the target item category; or, matching each optional attribute information corresponding to the first item attribute under the target item category with the target item search information to determine the target attribute information corresponding to the first item attribute that matches successfully.

[0072] Specifically, it is possible to match the target item search information based on the regular expression corresponding to each first item attribute under the target item category, and use the first item attribute that matches successfully and the corresponding target attribute information as the first item attribute information; or, match each optional attribute information corresponding to each first item attribute under the target item category with the target item search information, and use the first item attribute that matches successfully and the corresponding target attribute information as the first item attribute information, so that the attribute information can be extracted more accurately using the clear matching rules, further improving the accuracy of item search.

[0073] S260. Determine the target item search result based on the target item attribute information and the candidate item set.

[0074] The technical solution of this embodiment extracts the item attributes of the target item search information under the target item category through the target information matching method and / or the target entity recognition model, so that the target item attribute information can be extracted more accurately, further improving the accuracy of item search.

[0075] The following is an embodiment of the item search device provided by the embodiment of the present invention. This device and the item search methods of the above embodiments belong to the same inventive concept. For the details not described in detail in the embodiment of the item search device, reference can be made to the embodiment of the above item search method.

[0076] Figure 3 The following is a schematic structural diagram of an item search device provided by an embodiment of the present invention. This embodiment is applicable to the situation of item search for a pre-trained model, especially applicable to the situation of item search, especially applicable to the application scenario of searching for industrial products. As Figure 3 shown, the device specifically includes: a target item search information acquisition module 310, a candidate item set determination module 320, a target item category determination module 330, an item attribute information extraction module 340, and a target item search result determination module 350.

[0077] Among them, the target item search information acquisition module 310 is used to acquire the input target item search information; the candidate item set determination module 320 is used to perform search recall on the target item search information to determine a candidate item set; the target item category determination module 330 is used to perform category recognition on the target item search information based on an item category recognition model to determine the target item category; the item attribute information extraction module 340 is used to extract attributes from the target item search information based on the item attributes corresponding to the target item category to obtain the extracted target item attribute information; the target item search result determination module 350 is used to determine the target item search result based on the target item attribute information and the candidate item set.

[0078] The technical solution of this embodiment determines a candidate item set by performing search recall on the input target item search information, performs category recognition on the target item search information based on an item category recognition model to determine the target item category, and extracts attributes from the target item search information based on the item attributes corresponding to the target item category to obtain the extracted target item attribute information. Therefore, based on the extracted target item attribute information, a more accurate target item search result can be determined from the candidate item set, ensuring the accuracy of item search. Moreover, no human intervention is required during the search process, realizing the automatic search of items and improving the item search efficiency.

[0079] Optionally, the target item category determination module 330 is specifically configured to:

[0080] Input the target item search information into the item category recognition model for the recognition of the terminal category, and determine the target item category according to the output of the item category recognition model, where the target item category refers to the terminal category to which the target item belongs.

[0081] Optionally, the target item category determination module 330 is further specifically configured to:

[0082] Perform category recognition on the target item search information based on the first item category recognition model to determine the first recognition probability belonging to each item category; perform category recognition on the target item search information based on the second item category recognition model to determine the second recognition probability belonging to each item category; determine the target item category based on the first recognition probability and the second recognition probability belonging to each item category; where the first item category recognition model and the second item category recognition model are obtained through co-training for model training.

[0083] Optionally, the first item category recognition model is a classification model based on character vector recognition; the second item category recognition model is a classification model based on word vector recognition.

[0084] Optionally, the item attribute information extraction module 340 includes:

[0085] An extraction method determination unit for determining a target information matching method and a target entity recognition model corresponding to the target item category;

[0086] An item attribute information extraction unit for performing item attribute extraction for the target item category on the target item search information based on the target information matching method and / or the target entity recognition model, to obtain the extracted target item attribute information.

[0087] Optionally, the item attribute information extraction unit includes:

[0088] A first item attribute extraction subunit for performing first item attribute extraction for the target item category on the target item search information based on the target information matching method, to obtain the extracted first item attribute information;

[0089] A second item attribute extraction subunit for performing second item attribute extraction for the target item category on the target item search information based on the target entity recognition model, to obtain the extracted second item attribute information;

[0090] An item attribute information determination subunit for obtaining the extracted target item attribute information based on the first item attribute information and the second item attribute information.

[0091] Optionally, the first item attribute extraction subunit is specifically configured to:

[0092] Extract the attribute information corresponding to the first item attribute in the target item search information based on the regular expression corresponding to the first item attribute under the target item category; or, match each optional attribute information corresponding to the first item attribute under the target item category with the target item search information to determine the target attribute information corresponding to the first item attribute with a successful match.

[0093] Optionally, the item search result determination module 350 is specifically configured to:

[0094] Match the candidate item attribute information corresponding to each candidate item in the candidate item set with the target item attribute information; and use the candidate item with a successful information match as the target item to determine the target item search result.

[0095] The item search device provided by the embodiments of the present invention can execute the item search method provided by any embodiment of the present invention, and has corresponding functional modules and beneficial effects for executing the item search method.

[0096] It should be noted that in the embodiments of the above-mentioned item search device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be realized; in addition, the specific names of the functional units are only for the convenience of mutual distinction and do not limit the protection scope of the present invention.

[0097] Figure 4 FIG. 4 is a schematic structural diagram of an electronic device according to an embodiment of the present invention. Figure 4 FIG. 5 shows a block diagram of an exemplary electronic device 12 suitable for implementing the embodiments of the present invention. Figure 4 The electronic device 12 shown is only an example and should not impose any limitation on the functions and usage scope of the embodiments of the present invention.

[0098] As Figure 4 shown, the electronic device 12 is presented in the form of a general-purpose computing device. The components of the electronic device 12 may include, but are not limited to: one or more processors or processing units 16, a system memory 28, and a bus 18 connecting different system components (including the system memory 28 and the processing unit 16).

[0099] The bus 18 represents one or more of several types of bus structures, including a memory bus or a memory controller, a peripheral bus, a graphics acceleration port, a processor, or a local bus using any of the multiple bus structures. For example, these architectures include, but are not limited to, Industry Standard Architecture (ISA) bus, Micro Channel Architecture (MAC) bus, Enhanced ISA bus, Video Electronics Standards Association (VESA) local bus, and Peripheral Component Interconnect (PCI) bus.

[0100] The electronic device 12 typically includes a variety of computer system-readable media. These media can be any available media accessible by the electronic device 12, including volatile and non-volatile media, removable and non-removable media.

[0101] The system memory 28 may include computer system-readable media in the form of volatile memory, such as random access memory (RAM) 30 and / or cache memory 32. The electronic device 12 may further include other removable / non-removable, volatile / non-volatile computer system storage media. By way of example only, the storage system 34 can be used to read and write non-removable, non-volatile magnetic media ( Figure 4 not shown, commonly referred to as a "hard disk drive"). Although Figure 4Not shown in the figure, a disk drive for reading and writing to a removable non-volatile disk (e.g., "floppy disk") and an optical disk drive for reading and writing to a removable non-volatile optical disk (e.g., CD-ROM, DVD-ROM or other optical media) can be provided. In these cases, each drive can be connected to the bus 18 through one or more data medium interfaces. The system memory 28 may include at least one program product having a set (e.g., at least one) of program modules configured to perform the functions of the embodiments of the present invention.

[0102] A program / utility 40 having a set (at least one) of program modules 42 can be stored, for example, in the system memory 28. Such program modules 42 include, but are not limited to, an operating system, one or more application programs, other program modules, and program data. Each or some combination of these examples may include the implementation of a network environment. The program modules 42 generally perform the functions and / or methods in the embodiments described in the present invention.

[0103] The electronic device 12 can also communicate with one or more external devices 14 (e.g., a keyboard, a pointing device, a display 24, etc.), and can also communicate with one or more devices that enable a user to interact with the electronic device 12, and / or communicate with any device that enables the electronic device 12 to communicate with one or more other computing devices (e.g., a network card, a modem, etc.). Such communication can be carried out through the input / output (I / O) interface 22. Moreover, the electronic device 12 can also communicate with one or more networks (e.g., a local area network (LAN), a wide area network (WAN), and / or a public network, such as the Internet) through the network adapter 20. As shown in the figure, the network adapter 20 communicates with other modules of the electronic device 12 through the bus 18. It should be understood that although not shown in the figure, other hardware and / or software modules can be used in combination with the electronic device 12, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems, etc.

[0104] The processing unit 16 executes various functional applications and data processing by running programs stored in the system memory 28, such as implementing the steps of an item search method provided by the embodiments of the present invention. The method includes:

[0105] Obtain the input target item search information;

[0106] Perform a search recall on the target item search information to determine a set of candidate items;

[0107] Based on the item category recognition model, perform category recognition on the target item search information to determine the target item category;

[0108] Extract the attributes of the target item search information based on the item attributes corresponding to the target item category, and obtain the extracted target item attribute information;

[0109] Determine the target item search result based on the target item attribute information and the candidate item set.

[0110] Of course, those skilled in the art can understand that the processor can also implement the technical solutions of the item search method provided in any embodiment of the present invention.

[0111] This embodiment provides a computer-readable storage medium, on which a computer program is stored. When the program is executed by a processor, it implements the steps of the item search method provided in any embodiment of the present invention. The method includes:

[0112] Obtain the input target item search information;

[0113] Perform search recall on the target item search information to determine a candidate item set;

[0114] Based on an item category recognition model, perform category recognition on the target item search information to determine the target item category;

[0115] Extract the attributes of the target item search information based on the item attributes corresponding to the target item category, and obtain the extracted target item attribute information;

[0116] Determine the target item search result based on the target item attribute information and the candidate item set.

[0117] The computer storage medium of the embodiments of the present invention can adopt any combination of one or more computer-readable media. The computer-readable medium can be a computer-readable signal medium or a computer-readable storage medium. The computer-readable storage medium can be, for example, but not limited to: an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination of the above. More specific examples (non-exhaustive list) of the computer-readable storage medium include: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In this document, the computer-readable storage medium can be any tangible medium that contains or stores a program, and the program can be used by or in combination with an instruction execution system, apparatus, or device.

[0118] A computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries computer-readable program code. Such a propagated data signal may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the foregoing. The computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device.

[0119] The program code contained on a computer-readable medium can be transmitted using any appropriate medium, including but not limited to: wireless, wire, optical fiber cable, RF, etc., or any suitable combination of the foregoing.

[0120] The computer program code for performing the operations of the present invention can be written in one or more programming languages or combinations thereof. The programming languages include object-oriented programming languages such as Java, Smalltalk, C++, and also include conventional procedural programming languages such as the "C" language or similar programming languages. The program code can be executed entirely on the user's computer, partially on the user's computer, executed as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In the case of a remote computer, the remote computer can be connected to the user's computer through any type of network, including a local area network (LAN) or a wide area network (WAN), or can be connected to an external computer (e.g., by using an Internet service provider to connect through the Internet).

[0121] Those of ordinary skill in the art should understand that the above-mentioned modules or steps of the present invention can be implemented using a general-purpose computing device. They can be concentrated on a single computing device or distributed over a network composed of multiple computing devices. Optionally, they can be implemented using program code executable by a computer device, so that they can be stored in a storage device and executed by a computing device, or they can be separately fabricated into individual integrated circuit modules, or multiple modules or steps among them can be fabricated into a single integrated circuit module for implementation. Thus, the present invention is not limited to any specific combination of hardware and software.

[0122] Note that the above is only the preferred embodiment of the present invention and the technical principles applied. Those skilled in the art will understand that the present invention is not limited to the specific embodiments described herein, and various obvious changes, re-adjustments and substitutions can be made by those skilled in the art without departing from the protection scope of the present invention. Therefore, although the present invention has been described in more detail through the above embodiments, the present invention is not limited to the above embodiments. Without departing from the concept of the present invention, more other equivalent embodiments can be included, and the scope of the present invention is determined by the scope of the appended claims.

Claims

1. An item search method, characterized in that, comprising: Obtaining the input target item search information; Performing search recall on the target item search information to determine a candidate item set; Inputting the target item search information into an item category recognition model for category recognition, and determining the target item category based on the output of the item category recognition model; Determining the target information matching method and the target entity recognition model corresponding to the target item category; Based on the target information matching method and / or the target entity recognition model, extracting the item attributes of the target item under the target item category from the target item search information to obtain the extracted target item attribute information; Determining the target item search result based on the target item attribute information and the candidate item set.

2. The method according to claim 1, characterized in that, The step of inputting the target item search information into an item category recognition model for category recognition, and determining the target item category based on the output of the item category recognition model includes: Inputting the target item search information into an item category recognition model for recognition of the terminal category, and determining the target item category according to the output of the item category recognition model, where the target item category refers to the terminal category to which the target item belongs.

3. The method according to claim 1, characterized in that, The step of inputting the target item search information into an item category recognition model for category recognition, and determining the target item category based on the output of the item category recognition model includes: Based on a first item category recognition model, performing category recognition on the target item search information to determine the first recognition probability of belonging to each item category; Based on a second item category recognition model, performing category recognition on the target item search information to determine the second recognition probability of belonging to each item category; Determining the target item category based on the first recognition probability and the second recognition probability of belonging to each item category; wherein, the first item category recognition model and the second item category recognition model are obtained by model training based on a collaborative training method.

4. The method according to claim 3, characterized in that, The first item category recognition model is a classification model based on character vector recognition; the second item category recognition model is a classification model based on word vector recognition.

5. The method according to claim 1, characterized in that, The step of extracting the item attributes of the target item under the target item category from the target item search information based on the target information matching method and the target entity recognition model to obtain the extracted target item attribute information includes: Based on the target information matching method, performing first item attribute extraction of the target item under the target item category on the target item search information to obtain the extracted first item attribute information; Based on the target entity recognition model, performing second item attribute extraction of the target item under the target item category on the target item search information to obtain the extracted second item attribute information; Based on the first item attribute information and the second item attribute information, obtain the extracted target item attribute information.

6. The method according to claim 5, wherein, the obtaining of the extracted first item attribute information by performing first item attribute extraction under the target item category on the target item search information based on the target information matching method includes: extracting the attribute information corresponding to the first item attribute in the target item search information based on the regular expression corresponding to the first item attribute under the target item category; or, matching each optional attribute information corresponding to the first item attribute under the target item category with the target item search information to determine the target attribute information corresponding to the first item attribute for which the matching is successful.

7. The method according to any one of claims 1-6, wherein, the determining of the target item search result based on the target item attribute information and the candidate item set includes: matching the candidate item attribute information corresponding to each candidate item in the candidate item set with the target item attribute information; taking the candidate items with successful information matching as target items to determine the target item search result.

8. An item search device, wherein, it includes: a target item search information acquisition module, configured to acquire the input target item search information; a candidate item set determination module, configured to perform search recall on the target item search information to determine a candidate item set; a target item category determination module, configured to input the target item search information into an item category recognition model for category recognition, and determine the target item category based on the output of the item category recognition model; an item attribute information extraction module, configured to determine the target information matching method and the target entity recognition model corresponding to the target item category; based on the target information matching method and / or the target entity recognition model, perform item attribute extraction on the target item search information under the target item category to obtain the extracted target item attribute information; an item search result determination module, configured to determine the target item search result based on the target item attribute information and the candidate item set.

9. An electronic device, wherein, the electronic device includes: one or more processors; a memory, configured to store one or more programs; when the one or more programs are executed by the one or more processors, the one or more processors implement the item search method according to any one of claims 1-7.

10. A computer-readable storage medium, on which a computer program is stored, wherein, when the program is executed by a processor, it implements the item search method according to any one of claims 1-7.

Images