Commodity search method, device, electronic device and storage medium

The commodity search method addresses the lack of e-commerce search in live broadcasts by determining user intent, acquiring product results, and adjusting relevance scores, facilitating intelligent sorting and result guidance.

CN115114506BActive Publication Date: 2024-12-06BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
View PDF 3 Cites 0 Cited by

Patent Information

Application Number
CN202110286778.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2021-03-17
Publication Date
2024-12-06
Estimated Expiration
2041-03-17

AI Technical Summary

Technical Problem

There is no effective e-commerce search mechanism based on live broadcasts, limiting the ability to find relevant product live broadcasts through user input information.

Method used

A commodity search method that obtains user input information, determines user intention characteristics, acquires product acquisition results from candidate live broadcast merchants, adjusts relevance scores based on product description and merchant labels, and sorts results intelligently.

Benefits of technology

Enables searching for related product live broadcasts and provides intelligent sorting and active guidance of search results, aligning with user needs.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115114506B_ABST
    Figure CN115114506B_ABST
Patent Text Reader

Abstract

The present disclosure relates to a commodity search method, device, electronic device and storage medium, and belongs to the field of Internet technology. The commodity search method includes: obtaining user input information, and obtaining the user's intention characteristics based on the input information; obtaining multiple commodity acquisition results of multiple candidate live broadcast merchants based on the user's intention characteristics; obtaining commodity description information of multiple commodity acquisition results; determining the correlation scores of multiple commodity acquisition results and the input information based on the input information and the commodity description information; obtaining labels of multiple candidate live broadcast merchants, and adjusting the correlation scores of the commodity acquisition results corresponding to the multiple candidate live broadcast merchants based on the labels of the multiple candidate live broadcast merchants; sorting the multiple commodity acquisition results based on the adjusted correlation scores. In this way, the purpose of searching for related commodity live broadcasts through the user's input information can be achieved, and the intelligent sorting and active guidance of search results can be achieved at the same time.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present disclosure relates to the field of Internet technology, and in particular to a commodity search method, device, electronic device and storage medium. Background Art

[0002] Live streaming is a new form of online social interaction. The anchor uses a camera and other tools to broadcast live video on the live streaming platform. As live streaming on the Internet continues to become more popular, many merchants with a keen sense of smell have begun to use live streaming platforms to promote and sell products online, which is the so-called live streaming with goods.

[0003] The source of live streaming sales may be a by-product of the live streaming format, but with the gradual growth of the live streaming shopping format itself, more and more professional live streaming hosts have emerged on the live streaming platforms. In addition to watching live streaming of content that they are interested in, many viewers are also looking for pure live streaming hosts who sell goods. They watch live streaming sales with the direct purpose of purchasing goods in a manner similar to TV shopping. Summary of the invention

[0004] The present disclosure provides a product search method, device, electronic device and storage medium to at least solve the problem of no e-commerce search based on live broadcast in the related technology, and can achieve the purpose of searching for related product live broadcasts through user input information, while realizing intelligent sorting and active guidance of search results. The technical solution of the present disclosure is as follows:

[0005] According to a first aspect of an embodiment of the present disclosure, there is provided a commodity search method, comprising:

[0006] Obtaining user input information, and obtaining the user's intention characteristics according to the input information;

[0007] Acquire multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics;

[0008] Obtaining product description information of the multiple product acquisition results;

[0009] Determining, based on the input information and the product description information, a correlation score between the plurality of product acquisition results and the input information;

[0010] Obtaining labels of the multiple candidate live broadcast merchants, and adjusting the relevance scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants according to the labels of the multiple candidate live broadcast merchants; and

[0011] The multiple commodity acquisition results are sorted according to the adjusted relevance scores.

[0012] In a possible implementation of the embodiment of the present disclosure, the acquiring the user's intention feature according to the input information includes:

[0013] Acquire the user's intention information according to the input information;

[0014] If the user's intention information is a search intention, search attribute information is further extracted from the input information, and the search attribute information is used as the intention feature.

[0015] In a possible implementation manner of the embodiment of the present disclosure, after further extracting the search attribute information from the input information, the method further includes:

[0016] The current node of the commodity category tree is extracted from the search attribute information.

[0017] In a possible implementation of the embodiment of the present disclosure, the obtaining of multiple product acquisition results of multiple candidate live broadcast merchants according to the intention characteristics of the user includes:

[0018] Generate acquisition keywords according to the intention features;

[0019] Acquisition is performed according to the acquisition keyword to generate an acquisition candidate product set, wherein the acquisition candidate product set includes multiple product acquisition results, and each of the product acquisition results corresponds to a candidate live broadcast merchant.

[0020] In a possible implementation of the embodiment of the present disclosure, the acquiring according to the acquired keyword to generate an acquisition candidate product set includes:

[0021] Obtaining a product acquisition result corresponding to the current node of the product category tree, and adding it to the acquisition candidate product set;

[0022] If the number of product acquisition results in the acquisition candidate product set is less than a preset number, the upper node is searched from the current node along the product category tree, and the product acquisition results corresponding to the upper node are added to the acquisition candidate product set.

[0023] In a possible implementation of the embodiment of the present disclosure, the above-mentioned commodity search method further includes:

[0024] If the number of product acquisition results in the acquired candidate product set is still less than the preset number, the acquisition keywords are screened;

[0025] The product acquisition is performed again according to the obtained keywords after screening, and the obtained product acquisition results are added to the obtained candidate product set.

[0026] In a possible implementation of the embodiment of the present disclosure, the sorting of the plurality of commodity acquisition results according to the adjusted relevance scores includes:

[0027] Obtaining the node level of the node where the multiple product acquisition results are located;

[0028] The plurality of commodity acquisition results are sorted according to the node levels and the adjusted relevance scores.

[0029] In a possible implementation of the embodiment of the present disclosure, the tag is a static tag, wherein obtaining the tags of the multiple candidate live broadcast merchants includes:

[0030] Extracting product description information of the candidate live broadcast merchant;

[0031] The label is generated according to the product description information.

[0032] In a possible implementation of the embodiment of the present disclosure, the tag is a static tag, wherein obtaining the tags of the multiple candidate live broadcast merchants includes:

[0033] Obtaining historical live broadcast content of the candidate live broadcast merchant;

[0034] The tag is generated according to the historical live broadcast content.

[0035] In a possible implementation of the embodiment of the present disclosure, the label is a feature label, wherein obtaining the labels of the plurality of candidate live broadcast merchants includes:

[0036] Obtaining user statistical characteristics and / or commodity sales statistical characteristics of the candidate live broadcast merchant;

[0037] The tag is generated according to the user statistical characteristics and / or commodity sales statistical characteristics.

[0038] In a possible implementation of the embodiment of the present disclosure, determining, based on the input information and the product description information, the correlation scores between the multiple product acquisition results and the user's intention features includes:

[0039] Obtaining product description information of the product acquisition result;

[0040] An initial relevance score is determined according to the input information and the product description information of the product acquisition result.

[0041] In a possible implementation of the embodiment of the present disclosure, the commodity search method further includes:

[0042] The initial relevance score is adjusted according to the intent characteristic.

[0043] In a possible implementation of the embodiment of the present disclosure, the sorting of the plurality of commodity acquisition results according to the adjusted relevance scores includes:

[0044] Obtaining the node level of the node where the multiple product acquisition results are located;

[0045] The plurality of commodity acquisition results are sorted according to the node levels and the adjusted relevance scores.

[0046] According to a second aspect of an embodiment of the present disclosure, there is provided a commodity search device, including:

[0047] A first acquisition module is configured to acquire user input information and acquire the user's intention feature according to the input information;

[0048] A second acquisition module is configured to acquire multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics;

[0049] A third acquisition module is configured to acquire product description information of the plurality of product acquisition results;

[0050] a determination module configured to determine, based on the input information and the product description information, a correlation score between the plurality of product acquisition results and the input information;

[0051] a fourth acquisition module, configured to acquire tags of the plurality of candidate live broadcast merchants, and adjust the relevance scores of the commodity acquisition results corresponding to the plurality of candidate live broadcast merchants according to the tags of the plurality of candidate live broadcast merchants; and

[0052] The sorting module is configured to sort the multiple commodity acquisition results according to the adjusted relevance scores.

[0053] In a possible implementation of the embodiment of the present disclosure, the first acquisition module includes:

[0054] an acquisition unit, configured to acquire the user's intention information according to the input information;

[0055] The first extraction unit is configured to further extract search attribute information from the input information if the user's intention information is a search intention, and use the search attribute information as the intention feature.

[0056] In a possible implementation manner of the embodiment of the present disclosure, the first acquisition module further includes:

[0057] The second extraction unit is configured to extract the current node of the commodity category tree from the search attribute information after the first extraction unit further extracts the search attribute information from the input information.

[0058] In a possible implementation of the embodiment of the present disclosure, the second acquisition module includes:

[0059] A generating unit, configured to generate acquisition keywords according to the intention feature;

[0060] The acquisition unit is configured to acquire according to the acquisition keyword to generate a set of candidate products, wherein the set of candidate products includes multiple product acquisition results, and each of the product acquisition results corresponds to a candidate live broadcast merchant.

[0061] In a possible implementation of the embodiment of the present disclosure, the acquisition unit is specifically configured as follows:

[0062] Obtaining a product acquisition result corresponding to the current node of the product category tree, and adding it to the acquisition candidate product set;

[0063] If the number of product acquisition results in the acquisition candidate product set is less than a preset number, the upper node is searched from the current node along the product category tree, and the product acquisition results corresponding to the upper node are added to the acquisition candidate product set.

[0064] In a possible implementation of the embodiment of the present disclosure, the acquisition unit is further configured to:

[0065] If the number of product acquisition results in the acquired candidate product set is still less than the preset number, the acquisition keywords are screened;

[0066] The product acquisition is performed again according to the obtained keywords after screening, and the obtained product acquisition results are added to the obtained candidate product set.

[0067] In a possible implementation manner of the embodiment of the present disclosure, the tag is a dynamic tag, wherein the fourth acquisition module is specifically configured as follows:

[0068] Obtaining the current live broadcast content or current explanation content of the candidate live broadcast merchant;

[0069] The current live broadcast content or the current explanation content is identified to generate the label.

[0070] In a possible implementation manner of the embodiment of the present disclosure, the tag is a static tag, wherein the fourth acquisition module is specifically configured as follows:

[0071] Extracting product description information of the candidate live broadcast merchant;

[0072] The label is generated according to the product description information.

[0073] In a possible implementation manner of the embodiment of the present disclosure, the tag is a static tag, wherein the fourth acquisition module is specifically configured as follows:

[0074] Obtaining historical live broadcast content of the candidate live broadcast merchant;

[0075] The tag is generated according to the historical live broadcast content.

[0076] In a possible implementation manner of the embodiment of the present disclosure, the tag is a feature tag, wherein the fourth acquisition module is specifically configured as follows:

[0077] Obtaining user statistical characteristics and / or commodity sales statistical characteristics of the candidate live broadcast merchant;

[0078] The tag is generated according to the user statistical characteristics and / or commodity sales statistical characteristics.

[0079] In a possible implementation of the embodiment of the present disclosure, the determining module is specifically configured as follows:

[0080] Obtaining product description information of the product acquisition result;

[0081] An initial relevance score is determined according to the input information and the product description information of the product acquisition result.

[0082] In a possible implementation of the embodiment of the present disclosure, the determining module is further configured to:

[0083] The initial relevance score is adjusted according to the intent characteristic.

[0084] In a possible implementation of the embodiment of the present disclosure, the sorting module is specifically configured as follows:

[0085] Obtaining the node level of the node where the multiple product acquisition results are located;

[0086] The plurality of commodity acquisition results are sorted according to the node levels and the adjusted relevance scores.

[0087] According to a third aspect of an embodiment of the present disclosure, there is provided an electronic device, including:

[0088] processor;

[0089] a memory for storing instructions executable by the processor;

[0090] The processor is configured to execute the instructions to implement the above-mentioned commodity search method.

[0091] According to a fourth aspect of an embodiment of the present disclosure, a computer-readable storage medium is provided. When instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the above-mentioned commodity search method.

[0092] According to a fifth aspect of an embodiment of the present disclosure, there is provided a computer program product, comprising a computer program / instruction, wherein the computer program / instruction implements the above-mentioned commodity search method when executed by a processor.

[0093] The technical solution provided by the embodiments of the present disclosure brings at least the following beneficial effects:

[0094] First, the user's input information is obtained, and the user's intention characteristics are obtained based on the input information, and multiple product acquisition results of multiple candidate live broadcast merchants are obtained based on the user's intention characteristics, and then the product description information of the multiple product acquisition results is obtained, and the correlation scores of the multiple product acquisition results and the input information are determined based on the input information and the product description information, and then the labels of multiple candidate live broadcast merchants are obtained, and the correlation scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants are adjusted based on the labels of the multiple candidate live broadcast merchants, and finally the multiple product acquisition results are sorted according to the adjusted correlation scores. Therefore, through this method, the problem of no e-commerce search based on live broadcast in the related technology is solved, and the purpose of searching for related product live broadcasts through user input information can be achieved, while realizing intelligent sorting and active guidance of search results.

[0095] It is to be understood that the foregoing general description and the following detailed description are exemplary and explanatory only and are not restrictive of the present disclosure. BRIEF DESCRIPTION OF THE DRAWINGS

[0096] The drawings herein are incorporated into and constitute a part of the specification, illustrate embodiments consistent with the present disclosure, and together with the description are used to explain the principles of the present disclosure, and do not constitute improper limitations on the present disclosure.

[0097] Figure 1 is a flow chart of a commodity search method according to an exemplary embodiment;

[0098] Figure 2 is a schematic diagram of a live video content understanding process according to an exemplary embodiment;

[0099] Figure 3 is a schematic diagram of a live audio content understanding process according to an exemplary embodiment;

[0100] Figure 4 is a block diagram of a commodity search device according to an exemplary embodiment; and

[0101] Figure 5 The present invention is a block diagram of an electronic device according to an exemplary embodiment. DETAILED DESCRIPTION

[0102] In order to enable ordinary persons in the art to better understand the technical solutions of the present disclosure, the technical solutions in the embodiments of the present disclosure will be clearly and completely described below in conjunction with the accompanying drawings.

[0103] It should be noted that the terms "first", "second", etc. in the specification and claims of the present disclosure and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present disclosure described herein can be implemented in an order other than those illustrated or described herein. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with the present disclosure. Instead, they are merely examples of devices and methods consistent with some aspects of the present disclosure as detailed in the appended claims.

[0104] The following describes the commodity search method, device, electronic device, and storage medium according to embodiments of the present disclosure with reference to the accompanying drawings.

[0105] The commodity search method provided in the embodiments of the present disclosure may be executed by an electronic device, which may be a PC (Personal Computer), a mobile phone, a tablet computer or a handheld computer, etc., without any limitation herein.

[0106] In the embodiment of the present disclosure, the electronic device may be provided with a processing component, a storage component and a drive component. Optionally, the drive component and the processing component may be integrated, the storage component may store an operating system, an application program or other program modules, and the processing component implements the commodity search method provided in the embodiment of the present disclosure by executing the application program stored in the storage component.

[0107] It should be noted that the commodity search method provided in the embodiments of the present disclosure can be applied to a short video APP (Application) with a live broadcast function, and can also be applied to a dedicated live broadcast APP, without any limitation. In addition, the above-mentioned short video APP with a live broadcast function and the live broadcast APP can both be installed in electronic devices (for example, mobile phones, tablet computers, etc.).

[0108] Figure 1 is a flowchart of a commodity search method according to an exemplary embodiment. Figure 1 As shown, the commodity search method can be used in an electronic device, and includes the following steps.

[0109] In step S11, the user's input information is obtained, and the user's intention characteristics are obtained according to the input information.

[0110] It should be noted that the input information described in this embodiment may include text information input by the user through voice recognition and input content input by the user into the input method system through the input method. The input method system can convert the input content into word candidates for the input text according to the user's current input method, and provide them to the user for selection. The user can input text information through a variety of input methods, such as keyboard, touchpad, mouse, etc. At the same time, the user can also choose any input method to input text information, such as pinyin, Wubi, strokes, handwriting, English, small language keyboard, etc., without any limitation here.

[0111] In the disclosed embodiments, the intent feature may include the product search intent. After obtaining the user's input information, the electronic device (e.g., mobile phone) may first determine whether the user's input information includes the product search intent. If the user's input information includes the product search intent, the user's intent feature may be obtained based on the input information. For example, if the user's input information is "Maybelline red lipstick", it may be determined that the user's input information includes the product search intent; if the user's input information is "live broadcast of beautiful women dancing", it may be determined that the user's input information does not include the product search intent.

[0112] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone. In the process of a user using a mobile phone to operate the short video APP with a live broadcast function, the user can operate the short video APP to call out a search bar, and enter search information (i.e., user input information) in the search bar according to their own shopping needs. For example, enter "Maybelline red lipstick" and then click the search button to search. After the user clicks the search button, the mobile phone can first determine whether the user's input information includes a commodity search intention through the short video APP. If the commodity search intention is included, the user's intention feature (e.g., Maybelline, lipstick) can be obtained according to the input information.

[0113] In step S12, multiple product acquisition results of multiple candidate live broadcast merchants are obtained according to the user's intention characteristics. Among them, the multiple product acquisition results of multiple candidate live broadcast merchants may include live broadcast merchants that are performing "live broadcast with goods" in short video apps or live broadcast apps with live broadcast functions, and the product information being broadcast live.

[0114] In one example, the product search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone. After the mobile phone obtains the user's intention characteristics, it can obtain (recall) multiple product acquisition (recall) results of multiple candidate live broadcast merchants according to the user's intention characteristics.

[0115] In step S13, commodity description information of multiple commodity acquisition results is obtained.

[0116] In the embodiment of the present disclosure, the product description information may include the brand of the product, the title of the product, the name of the product, the description of the product, the type of product, the sales information of the product, and the evaluation information of the product.

[0117] In step S14, the correlation scores between the multiple product acquisition results and the input information are determined according to the input information and the product description information.

[0118] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, after the mobile phone obtains multiple commodity acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics, it can analyze them to obtain the commodity description information of the multiple commodity acquisition results, and then use the preset correlation algorithm to determine the correlation score between the multiple commodity acquisition results and the input information according to the input information (for example, Maybelline red lipstick) and the commodity description information (for example, the commodity title, the name of the commodity, the description of the commodity, etc.). Among them, the preset correlation algorithm can be calibrated according to the actual situation, for example, the preset correlation algorithm can be a traditional correlation determination (calculation) algorithm (for example, BM25 algorithm).

[0119] In step S15, labels of multiple candidate live broadcast merchants are obtained, and the relevance scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants are adjusted according to the labels of the multiple candidate live broadcast merchants.

[0120] It should be noted that the tags described in this embodiment may be dynamic tags, static tags or feature tags, among which dynamic tags may include tags for the live broadcast room, understanding tags according to the content of the live broadcast room, establishing tags for the current content of the live broadcast, or converting the voice of the product being explained into text description to generate corresponding tags for the product that the anchor is explaining; static tags may include tags generated for the product title, product description, word segmentation, and keywords, tags generated for product images and the content of product images, tags generated for the category tree information of the product, tags generated through the anchor's historical live broadcast content and keywords that have been searched for the anchor's live broadcast in history (i.e., anchor tags), etc.; feature tags may include tags generated for the statistical characteristics of the current live broadcast room (such as: number of online people, number of likes, total sales orders), and tags generated for the sales statistical characteristics of the current product.

[0121] In step S16, the multiple commodity acquisition results are sorted according to the adjusted relevance scores.

[0122] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, after determining the correlation scores between the multiple commodity acquisition results and the input information, the mobile phone can obtain the labels of multiple candidate live broadcast merchants (for example, dynamic labels, static labels or feature labels), and adjust the correlation scores of the commodity acquisition results corresponding to the multiple candidate live broadcast merchants according to the labels of the multiple candidate live broadcast merchants. For example, the user search term (the user's input information) "coconut shoes", the e-commerce field knowledge will partially weight the results of the adidas brand, or the user search term (the user's input information) "Maybelline red lipstick", then the product results such as the big anchor who often does lipstick live broadcasts in history and the live broadcasts explaining cosmetics can be weighted. Finally, the mobile phone can sort the multiple product acquisition results according to the adjusted relevance scores. For example, the acquisition results with deeper restrictions in the category tree are always ranked ahead of the acquisition results with shallower restrictions in the category tree (for example: user search terms: "Maybelline red lipstick", the third-level category: "lipstick" products are always ranked ahead of the second-level category: "cosmetics" products); products with brand words and attribute words are always ranked ahead of the acquisition results with relaxed restrictions.

[0123] In the disclosed embodiment, the user's input information is first obtained, and the user's intention characteristics are obtained based on the input information. Then, multiple product acquisition results of multiple candidate live broadcast merchants are obtained based on the user's intention characteristics, and product description information of multiple product acquisition results is obtained. According to the input information and the product description information, the correlation scores of the multiple product acquisition results and the input information are determined. Then, the labels of multiple candidate live broadcast merchants are obtained, and the correlation scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants are adjusted according to the labels of the multiple candidate live broadcast merchants. Finally, the multiple product acquisition results are sorted according to the adjusted correlation scores. In this way, the purpose of searching for related product live broadcasts through the user's input information can be achieved, and the intelligent sorting and active guidance of the search results can be achieved at the same time.

[0124] To clearly illustrate the previous embodiment, in a possible implementation of the presently disclosed embodiment, obtaining the user's intention features based on input information may include obtaining the user's intention information based on the input information, and if the user's intention information is a search intention, further extracting search attribute information from the input information, and using the search attribute information as the intention feature.

[0125] It should be noted that the search intent described in this embodiment may be a product search intent, wherein the product search intent may correspond to the node information of the product category tree at all levels; the search attribute information described in this embodiment may include entity words, brand words, attribute words, etc. corresponding to the search terms.

[0126] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, when a user uses a mobile phone to operate a short video APP with a live broadcast function, the short video APP can be operated to call out a search bar, and search information can be entered in the search bar according to the user's shopping needs, and the user's intention information can be obtained according to the input search information. For example, if the user's search information is "Maybelline red lipstick", the intention information can be considered as a commodity search intention. For example, if the user's search information is "dance live broadcast", the intention information can be considered as a viewing intention. If the user's intention information is a search intention (for example, a commodity search intention), the search attribute information can be further extracted from the input information, and the search attribute information can be used as an intention feature. For example, the search information is: "Maybelline red lipstick", the intention can be understood as: there is a commodity search intention, and the search attribute information can include: entity word: lipstick; brand word: Maybelline; attribute word: red, and the intention feature can include: lipstick, Maybelline, red. Therefore, by extracting the intent features of the user's input information, the user's current needs can be captured intuitively and accurately, so that the final search results are more in line with the user's current needs and provide detailed search terms (ie, intent features) for subsequent searches.

[0127] In order to clearly explain the above embodiment, in a possible implementation of the embodiment of the present disclosure, after further extracting the search attribute information from the input information, it may also include extracting the current node of the commodity category tree from the search attribute information.

[0128] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, after the mobile phone extracts the search attribute information from the input information, it can extract the current node of the commodity category tree from the search attribute information. For example, when the search attribute information includes: entity word: lipstick; brand word: Maybelline; attribute word: red, the extracted commodity category tree can be a first-level category, skin care products / cosmetics / cleaning products; a second-level category, cosmetics; a third-level category, lipstick, and the current node can be: cosmetics-lipstick, or lipstick, without any limitation here. In this way, the accuracy of the search can be improved.

[0129] In the embodiments of the present disclosure, the extraction of the above-mentioned search attribute information and the extraction of the current node of the product category tree can be achieved through an extraction model, wherein the extraction model can be trained in advance and pre-stored in the storage space of the electronic device (e.g., a mobile phone) to facilitate application retrieval. The storage space is not limited to entity-based storage space, such as a hard disk. The above-mentioned storage space can also be the storage space (cloud storage space) of a network hard disk connected to the electronic device (e.g., a mobile phone).

[0130] The training and generation of the extraction model can be performed by a related server, which can be a cloud server or a computer host. A communication connection is established between the server and an electronic device that can execute the commodity search method provided by the disclosed embodiment, and the communication connection can be at least one of a wireless network connection and a wired network connection. The server can send the trained extraction model to the electronic device so that the electronic device can call it when needed, thereby greatly reducing the computing pressure of the electronic device.

[0131] In a possible implementation of the embodiments of the present disclosure, multiple product acquisition results of multiple candidate live broadcast merchants are acquired according to the user's intention characteristics, which may include generating an acquisition keyword according to the intention characteristics, and acquiring according to the acquisition keyword to generate an acquisition candidate product set, wherein the acquisition candidate product set includes multiple product acquisition results, and each product acquisition result corresponds to a candidate live broadcast merchant.

[0132] In one example, the product search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone. After obtaining the user's intention feature, the mobile phone can first generate all the restrictions such as entity words, attribute words, and trademark names in the intention feature to obtain keywords, wherein the obtained keywords can correspond to the products in the deepest level category tree category (for example: the user's input information is: "Maybelline red lipstick", and the corresponding category tree information is: 'First-level category: skin care products / cosmetics / cleaning products; second-level category: cosmetics; third-level category: lipstick', then only all products in the third-level category 'lipstick' can be obtained). Then, the mobile phone obtains according to the obtained keywords to generate a set of candidate products. For example, multiple product acquisition results can be lipsticks of various brands and colors, and their corresponding candidate live merchants (i.e., live merchants that live broadcast lipsticks). As a result, users with shopping needs can find anchors who are selling such products by inputting information (keywords), thereby improving the user experience.

[0133] In order to clearly explain the above embodiment, in a possible implementation of the disclosed embodiment, acquiring according to the acquisition keyword to generate the acquisition candidate product set may include acquiring the product acquisition result corresponding to the current node of the product category tree and adding it to the acquisition candidate product set. If the number of product acquisition results in the acquisition candidate product set is less than the preset number, then searching for the upper node along the product category tree from the current node, and adding the product acquisition result corresponding to the upper node to the acquisition candidate product set. The preset number can be calibrated according to the actual situation.

[0134] In one example, the product search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone. After obtaining the return keyword, the mobile phone can obtain the product acquisition result corresponding to the current node of the product category tree (for example, lipstick), and add it to the acquisition candidate product set. When the number of product acquisition results in the acquisition candidate product set is less than a preset number, the parent node can be found from the current node along the product category tree, and the product acquisition result corresponding to the parent node is added to the acquisition candidate product set (for example, the user's input information is: "Maybelline Red Lipstick", and the corresponding category tree information is: first-level category: skin care products / cosmetics / cleaning products; second-level category: cosmetics; third-level category: lipstick, then all products of the second-level category 'cosmetics' are currently supplemented), thereby expanding the number of products in the acquisition candidate product set.

[0135] Furthermore, in another possible implementation of the embodiment of the present disclosure, if the number of product acquisition results in the candidate product set is still less than a preset number, the acquisition keywords are screened, and the acquisition is performed again according to the acquired keywords after the screening, and the re-acquired product acquisition results are added to the candidate product set.

[0136] In one example, the commodity search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, if the number of commodity acquisition results in the candidate commodity set is still less than the preset number after adding the commodity acquisition results corresponding to the upper node to the set of candidate commodities, the acquisition keywords can be screened (for example, removing the restrictions of some attribute words and trademark names), and the acquisition is performed again according to the acquired keywords after the screening, and the re-acquired commodity acquisition results are added to the set of candidate commodities. For example, the restrictions of the trademark names in the acquisition keywords "Maybelline, red, lipstick" are removed, that is, "Maybelline" in the acquisition keywords is removed and "red, lipstick" is retained, thereby further expanding the number of commodities in the set of candidate commodities.

[0137] It should be noted that the above method of expanding the number of commodities in the candidate commodity set can be repeatedly executed until the number of commodity acquisition results in the candidate commodity set is greater than or equal to a preset number.

[0138] In a possible implementation of the embodiment of the present disclosure, the label may be a dynamic label, wherein obtaining labels of multiple candidate live broadcast merchants may include: obtaining the current live broadcast content or current explanation content of the candidate live broadcast merchants, and identifying the current live broadcast content or current explanation content to generate a label.

[0139] In one example, see Figure 2 and Figure 3 For the live broadcast room, you can understand the label according to the content of the current live broadcast room of the selected live broadcast merchant, and establish the current content label of the live broadcast, such as Figure 2 The output (1) of the video understanding label is shown. The live video stream of the current live broadcast room of the live broadcast merchant is obtained, and the live video frame extraction method is input into the convolutional neural network to output the video understanding label through the convolutional neural network. The label can be hierarchical, for example: the first level: "food / digital / supplies / clothing / game painting", the second level: "fresh food / packaged food / keyboard / men's clothing", the third level: "instant noodles / suits / King of Glory".

[0140] For the content of the product that the anchor is explaining in the live broadcast room (i.e., the current explanation content), the current explanation content is identified to generate a label, such as Figure 3 Output (1) shown, keywords / keywords (labels). Among them, the live video stream of the current live broadcast room of the live broadcast merchant is obtained, and input into the recurrent neural network in the form of audio slices to output text through the recurrent neural network, and then perform text segmentation, grammatical analysis, keyword extraction program and named entity recognition to generate keywords / keywords (labels).

[0141] Therefore, it is possible to generate tags based on the current live broadcast content or the current explanation content, so that the generated tags can better describe the current live broadcast room, thereby improving the accuracy of the search.

[0142] In a possible implementation of the embodiment of the present disclosure, the label may be a static label, wherein obtaining labels of multiple candidate live broadcast merchants may include extracting product description information of the candidate live broadcast merchants and generating a label based on the product description information.

[0143] In another possible implementation of the embodiment of the present disclosure, the label may be a static label, wherein obtaining labels of multiple candidate live broadcast merchants may include obtaining historical live broadcast content of the candidate live broadcast merchants and generating labels based on the historical live broadcast content.

[0144] In one example, for a live broadcast room, product description information of candidate live broadcast merchants can be extracted, such as product titles, product description word segments, keywords, product images, etc., and then the word embedding (word vector) learned from deep learning of these product description information can be used to generate relevant tags, for example, Figure 1 Output (2) shown. Alternatively, labels can be generated directly based on the category tree information of the product. Alternatively, labels can be established based on the anchor's historical live broadcast content (i.e., the historical live broadcast content of the candidate live broadcast merchant), keywords that were searched for the anchor's live broadcast in the past, etc. In this way, labels can be enriched and diversified, further improving the accuracy of the search.

[0145] In a possible implementation of the embodiment of the present disclosure, the label may be a feature label, wherein obtaining labels of multiple candidate live broadcast merchants may include obtaining user statistical characteristics and / or product sales statistical characteristics of the candidate live broadcast merchants, and generating labels based on the user statistical characteristics and / or product sales statistical characteristics.

[0146] In one example, for a live broadcast room, the statistical characteristics of the current live broadcast room (such as the number of online users, the number of likes, the total number of sales orders) and / or the sales statistical characteristics of the current product (such as sales volume, sales evaluation, after-sales service) can be obtained, and tags can be generated based on the statistical characteristics of the current live broadcast room and / or the sales statistical characteristics of the current product. As a result, the generated tags are more consistent with the sales and attention characteristics of the product, so that the search results can take into account both the sales and attention characteristics of the product.

[0147] In a possible implementation of the embodiment of the present disclosure, determining the correlation scores between multiple product acquisition results and the user's intention features based on input information and product description information may include obtaining product description information of the product acquisition results, and determining an initial correlation score based on the input information and the product description information of the product acquisition results.

[0148] Furthermore, in another possible implementation of the embodiment of the present disclosure, the initial relevance score may be adjusted according to the intent feature.

[0149] In one example, the product search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone, wherein, after the mobile phone obtains multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics, it can analyze them to obtain product description information of the multiple product acquisition results, and then use a preset correlation algorithm to determine the correlation score of the multiple product acquisition results with the input information (for example, Maybelline red lipstick) and the product description information (for example, the product title, the name of the product, etc.), or use the input information The intent understanding results (trademark name, attribute words, category tree information) and product annotation information (product categories and keywords marked by sellers) are used to fine-tune the relevance results using knowledge in the e-commerce field (for example, when a user searches for the term "coconut shoes", knowledge in the e-commerce field will partially weight the results of the adidas brand). Alternatively, the relevance results can be adjusted using anchor information, understanding of live broadcast content, and product explanation content (for example, when a user searches for the term (input information) "Maybelline red lipstick", the product results of anchors who often do lipstick live broadcasts and live broadcasts explaining cosmetics will be weighted). As a result, the relevance score obtained can fully meet the needs of users and make the final search results more reasonable and scientific.

[0150] In a possible implementation of the embodiment of the present disclosure, sorting multiple product acquisition results according to the adjusted relevance scores may include obtaining the node levels of the nodes where the multiple product acquisition results are located, and sorting the multiple product acquisition results according to the node levels and the adjusted relevance scores.

[0151] It should be noted that due to the hierarchical acquisition mechanism, the acquisition results with deeper restrictions in the category tree are always ranked ahead of the acquisition results with shallower restrictions in the category tree (for example: when the user searches for the term "Maybelline red lipstick", the third-level category "lipstick" products are always ranked ahead of the second-level category "cosmetics" products); products with brand terms and attribute terms are always ranked ahead of the acquisition results with relaxed restrictions. In addition, the relevance weights of products related to live broadcast content and anchor fields will be amplified, so that the product results searched by users can get the "explaining" service, and the search results with high user clicks and orders can also be weighted to achieve a trade-off between multiple goals.

[0152] In one example, the product search method provided by the embodiment of the present disclosure is applied in a short video APP with a live broadcast function, and the short video APP is installed in a mobile phone. After obtaining the adjusted correlation score, the mobile phone can obtain the node level of the node where the multiple product acquisition results are located, and sort the multiple product acquisition results according to the node level and the adjusted correlation score. Due to the particularity of product search based on the live broadcast scenario, multiple optimization goals will be comprehensively considered in the sorting stage. For example, results with high relevance will be placed at the front as much as possible, products with high product scores will be placed at the front as much as possible, results with high user order volume will be placed at the front as much as possible, and results with high live broadcast user duration will be placed at the front as much as possible (taking into account the live broadcast revenue of the live broadcast platform), thereby realizing intelligent sorting and active guidance of search results, and making the display of search results more in line with user needs.

[0153] In summary, according to the commodity search method provided by the embodiment of the present disclosure, the user's input information is first obtained, and the user's intention characteristics are obtained based on the input information, and then multiple commodity acquisition results of multiple candidate live broadcast merchants are obtained based on the user's intention characteristics, and the commodity description information of the multiple commodity acquisition results is obtained, and the correlation scores of the multiple commodity acquisition results and the input information are determined based on the input information and the commodity description information, and then the labels of multiple candidate live broadcast merchants are obtained, and the correlation scores of the commodity acquisition results corresponding to the multiple candidate live broadcast merchants are adjusted according to the labels of the multiple candidate live broadcast merchants, and finally the multiple commodity acquisition results are sorted according to the adjusted correlation scores. In this way, the purpose of searching for related commodity live broadcasts through the user's input information can be achieved, and the intelligent sorting and active guidance of the search results are achieved at the same time.

[0154] Figure 4 is a block diagram of a commodity search device according to an exemplary embodiment. Figure 4 The product search device 40 may include: a first acquisition module 41, a second acquisition module 42, a third acquisition module 43, a determination module 44, a fourth acquisition module 45 and a sorting module 46.

[0155] The first acquisition module 41 is configured to acquire user input information and acquire user intention features according to the input information.

[0156] The second acquisition module 42 is configured to acquire multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics.

[0157] The third acquisition module 43 is configured to acquire commodity description information of multiple commodity acquisition results.

[0158] The determination module 44 is configured to determine the relevance scores between the multiple product acquisition results and the input information according to the input information and the product description information.

[0159] The fourth acquisition module 45 is configured to acquire tags of multiple candidate live broadcast merchants, and adjust the relevance scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants according to the tags of the multiple candidate live broadcast merchants.

[0160] The sorting module 46 is configured to sort the multiple commodity acquisition results according to the adjusted relevance scores.

[0161] In a possible implementation of the embodiment of the present disclosure, the first acquisition module 41 may include:

[0162] an acquisition unit, configured to acquire user intention information according to input information;

[0163] The first extraction unit is configured to further extract search attribute information from the input information if the user's intention information is a search intention, and use the search attribute information as an intention feature.

[0164] In a possible implementation of the embodiment of the present disclosure, the first acquisition module 42 may further include:

[0165] The second extraction unit is configured to extract the current node of the commodity category tree from the search attribute information after the first extraction unit further extracts the search attribute information from the input information.

[0166] In a possible implementation of the embodiment of the present disclosure, the second acquisition module 42 may include:

[0167] A generating unit, configured to generate acquisition keywords according to the intention feature;

[0168] The acquisition unit is configured to acquire according to the acquisition keyword to generate an acquisition candidate product set, wherein the acquisition candidate product set includes multiple product acquisition results, and each product acquisition result corresponds to a candidate live broadcast merchant.

[0169] In a possible implementation manner of the embodiment of the present disclosure, the acquisition unit is specifically configured as follows:

[0170] Get the product acquisition result corresponding to the current node of the product category tree, and add it to the acquisition candidate product set;

[0171] If the number of product acquisition results in the candidate product set is less than the preset number, the parent node is searched from the current node along the product category tree, and the product acquisition results corresponding to the parent node are added to the candidate product set.

[0172] In a possible implementation of the embodiment of the present disclosure, the acquisition unit is further configured to:

[0173] If the number of product acquisition results in the candidate product set is still less than the preset number, the acquisition keywords are screened;

[0174] The product acquisition is performed again according to the obtained keywords after screening, and the obtained product acquisition results are added to the acquisition candidate product set.

[0175] In a possible implementation of the embodiment of the present disclosure, the tag is a dynamic tag, wherein the fourth acquisition module 45 is specifically configured as follows:

[0176] Obtain the current live broadcast content or current explanation content of the candidate live broadcast merchant;

[0177] The current live broadcast content or the current explanation content is identified to generate a label.

[0178] In a possible implementation of the embodiment of the present disclosure, the tag is a static tag, wherein the fourth acquisition module 45 is specifically configured as follows:

[0179] Extract product description information of candidate live broadcast merchants;

[0180] Generate labels based on product description information.

[0181] In a possible implementation of the embodiment of the present disclosure, the tag is a static tag, wherein the fourth acquisition module 45 is specifically configured as follows:

[0182] Obtain the historical live broadcast content of the candidate live broadcast merchants;

[0183] Generate tags based on historical live broadcast content.

[0184] In a possible implementation of the embodiment of the present disclosure, the tag is a feature tag, wherein the fourth acquisition module 45 is specifically configured as follows:

[0185] Obtaining user statistical characteristics and / or product sales statistical characteristics of candidate live broadcast merchants;

[0186] Generate labels based on user statistical features and / or product sales statistical features.

[0187] In a possible implementation of the embodiment of the present disclosure, the determination module 44 is specifically configured as follows:

[0188] Get the product description information of the product acquisition result;

[0189] An initial relevance score is determined based on the input information and the product description information of the product acquisition results.

[0190] In a possible implementation of the embodiment of the present disclosure, the determination module 44 is further configured to:

[0191] The initial relevance score is adjusted based on the intent characteristics.

[0192] In a possible implementation of the embodiment of the present disclosure, the sorting module 45 is specifically configured as follows:

[0193] Get the node level of the node where multiple product acquisition results are located;

[0194] The multiple product acquisition results are sorted according to the node level and the adjusted relevance score.

[0195] In actual use, the commodity search device provided by the embodiment of the present disclosure can be configured in an electronic device to execute the aforementioned commodity search method. Therefore, regarding the device in the above embodiment, the specific manner in which each module performs operations has been described in detail in the embodiment of the method, and will not be elaborated here.

[0196] In summary, the commodity search device provided by the embodiment of the present disclosure first obtains the user's input information through the first acquisition module, and obtains the user's intention characteristics based on the input information, and obtains multiple commodity acquisition results of multiple candidate live broadcast merchants based on the user's intention characteristics through the second acquisition module, and then obtains the commodity description information of the multiple commodity acquisition results through the third acquisition module, and determines the correlation scores of the multiple commodity acquisition results and the input information through the determination module based on the input information and the commodity description information, and then obtains the labels of multiple candidate live broadcast merchants through the fourth acquisition module, and adjusts the correlation scores of the commodity acquisition results corresponding to the multiple candidate live broadcast merchants according to the labels of the multiple candidate live broadcast merchants, and finally sorts the multiple commodity acquisition results according to the adjusted correlation scores through the sorting module. In this way, the purpose of searching for related commodity live broadcasts through the user's input information can be achieved, and the intelligent sorting and active guidance of search results are achieved at the same time.

[0197] In order to implement the above embodiments, the present disclosure also provides an electronic device.

[0198] The electronic device includes a processor and a memory for storing instructions executable by the processor; wherein the processor is configured to execute the instructions to implement the above-mentioned commodity search method.

[0199] In order to implement the above embodiments, the present disclosure also proposes a computer-readable storage medium.

[0200] When the instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the above-mentioned commodity search method.

[0201] In order to implement the above embodiments, the present disclosure further provides a computer program product, including a computer program / instruction, which implements the above commodity search method when executed by a processor.

[0202] Figure 5 The present invention is a block diagram of an electronic device according to an exemplary embodiment. Figure 5 The electronic device shown is only an example and should not bring any limitation to the functions and scope of use of the embodiments of the present disclosure.

[0203] like Figure 5As shown, the electronic device 50 includes a processor 51, which can perform various appropriate actions and processes according to a program stored in a read-only memory (ROM) 52 or a program loaded from a memory 56 to a random access memory (RAM) 53. Various programs and data required for the operation of the electronic device 50 are also stored in the RAM 53. The processor 51, the ROM 52, and the RAM 53 are connected to each other via a bus 55. An input / output (I / O) interface 55 is also connected to the bus 55.

[0204] The following components are connected to the I / O interface 55: a memory 56 including a hard disk, etc.; and a communication part 57 including a network interface card such as a LAN (Local Area Network) card, a modem, etc., and the communication part 57 performs communication processing via a network such as the Internet; a drive 58 is also connected to the I / O interface 55 as needed.

[0205] In particular, according to an embodiment of the present disclosure, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present disclosure includes a computer program carried on a computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from the network through the communication part 57. When the computer program is executed by the processor 51, the above-mentioned functions defined in the method of the present disclosure are executed.

[0206] In an exemplary embodiment, a storage medium including instructions is also provided, such as a memory including instructions, and the instructions can be executed by the processor 51 of the electronic device 50 to complete the above method. Optionally, the storage medium can be a non-transitory computer-readable storage medium, for example, the non-transitory computer-readable storage medium can be a ROM, a random access memory (RAM), a CD-ROM, a magnetic tape, a floppy disk, an optical data storage device, etc.

[0207] In the present disclosure, a computer-readable storage medium may be any tangible medium containing or storing a program that may be used by or in combination with an instruction execution system, apparatus, or device. In the present disclosure, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. Such propagated data signals may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which may send, propagate, or transmit a program for use by or in combination with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical cable, RF, etc., or any suitable combination of the above.

[0208] Those skilled in the art will readily appreciate other embodiments of the present disclosure after considering the specification and practicing the invention disclosed herein. This application is intended to cover any variations, uses, or adaptations of the present disclosure that follow the general principles of the present disclosure and include common knowledge or customary techniques in the art that are not disclosed in the present disclosure. The specification and examples are intended to be exemplary only, and the true scope and spirit of the present disclosure are indicated by the following claims.

[0209] It should be understood that the present disclosure is not limited to the exact structures that have been described above and shown in the drawings, and that various modifications and changes may be made without departing from the scope thereof. The scope of the present disclosure is limited only by the appended claims.

Claims

1. A commodity search method, characterized in that: include: Obtaining user input information, and obtaining the user's intention characteristics according to the input information; Acquire multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics; Obtaining product description information of the multiple product acquisition results; Determining, based on the input information and the product description information, a correlation score between the plurality of product acquisition results and the input information; Obtaining labels of the multiple candidate live broadcast merchants, and adjusting the relevance scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants according to the labels of the multiple candidate live broadcast merchants; as well as sorting the plurality of product acquisition results according to the adjusted relevance scores; The step of determining the correlation scores between the plurality of product acquisition results and the user's intention features according to the input information and the product description information includes: Obtaining product description information of the product acquisition result; Determine an initial relevance score based on the input information and the product description information of the product acquisition result; The sorting of the plurality of product acquisition results according to the adjusted relevance scores includes: Obtaining the node level of the multiple product acquisition results at the current node of the product category tree; The plurality of commodity acquisition results are sorted according to the node levels and the adjusted relevance scores.

2. The commodity search method according to claim 1, characterized in that: The acquiring the user's intention feature according to the input information includes: Acquire the user's intention information according to the input information; If the user's intention information is a search intention, search attribute information is further extracted from the input information, and the search attribute information is used as the intention feature.

3. The commodity search method according to claim 2, characterized in that: After further extracting the search attribute information from the input information, the method further includes: The current node of the commodity category tree is extracted from the search attribute information.

4. The commodity search method according to claim 2 or 3, characterized in that: The obtaining of multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics includes: Generate acquisition keywords according to the intention features; Acquisition is performed according to the acquisition keyword to generate an acquisition candidate product set, wherein the acquisition candidate product set includes multiple product acquisition results, and each of the product acquisition results corresponds to a candidate live broadcast merchant.

5. The commodity search method according to claim 4, characterized in that: The acquiring according to the acquired keywords to generate an acquisition candidate product set includes: Obtaining a product acquisition result corresponding to the current node of the product category tree, and adding it to the acquisition candidate product set; If the number of product acquisition results in the acquisition candidate product set is less than a preset number, the upper node is searched from the current node along the product category tree, and the product acquisition results corresponding to the upper node are added to the acquisition candidate product set.

6. The commodity search method according to claim 5, characterized in that: Also includes: If the number of product acquisition results in the acquired candidate product set is still less than the preset number, the acquisition keywords are screened; The obtained keywords are retrieved again after being screened, and the obtained commodity acquisition results are added to the obtained candidate commodity set.

7. The commodity search method according to claim 1, characterized in that: The tag is a static tag, wherein obtaining the tags of the plurality of candidate live broadcast merchants includes: Extracting product description information of the candidate live broadcast merchant; The label is generated according to the product description information.

8. The commodity search method according to claim 1, characterized in that: The tag is a static tag, wherein obtaining the tags of the plurality of candidate live broadcast merchants includes: Obtaining historical live broadcast content of the candidate live broadcast merchant; The tag is generated according to the historical live broadcast content.

9. The commodity search method according to claim 1, characterized in that: The label is a feature label, wherein obtaining the labels of the plurality of candidate live broadcast merchants includes: Obtaining user statistical characteristics and / or commodity sales statistical characteristics of the candidate live broadcast merchant; The tag is generated according to the user statistical characteristics and / or commodity sales statistical characteristics.

10. The commodity search method according to claim 1, characterized in that: Also includes: The initial relevance score is adjusted according to the intent characteristic.

11. A commodity search device, characterized in that: include: A first acquisition module is configured to acquire user input information and acquire the user's intention feature according to the input information; A second acquisition module is configured to acquire multiple product acquisition results of multiple candidate live broadcast merchants according to the user's intention characteristics; A third acquisition module is configured to acquire product description information of the plurality of product acquisition results; a determination module configured to determine, based on the input information and the product description information, a correlation score between the plurality of product acquisition results and the input information; A fourth acquisition module is configured to acquire tags of the multiple candidate live broadcast merchants, and adjust the relevance scores of the product acquisition results corresponding to the multiple candidate live broadcast merchants according to the tags of the multiple candidate live broadcast merchants; as well as A sorting module, configured to sort the plurality of commodity acquisition results according to the adjusted relevance scores; Wherein, the determination module is specifically configured as follows: Obtaining product description information of the product acquisition result; Determine an initial relevance score based on the input information and the product description information of the product acquisition result; The sorting module is specifically configured as follows: Obtaining the node level of the multiple product acquisition results at the current node of the product category tree; The plurality of commodity acquisition results are sorted according to the node levels and the adjusted relevance scores.

12. The commodity search device according to claim 11, characterized in that: The first acquisition module includes: an acquisition unit, configured to acquire the user's intention information according to the input information; The first extraction unit is configured to further extract search attribute information from the input information if the user's intention information is a search intention, and use the search attribute information as the intention feature.

13. The commodity search device according to claim 12, characterized in that: The first acquisition module further includes: The second extraction unit is configured to extract the current node of the commodity category tree from the search attribute information after the first extraction unit further extracts the search attribute information from the input information.

14. The commodity search device according to claim 12 or 13, characterized in that: The second acquisition module includes: A generating unit, configured to generate acquisition keywords according to the intention feature; The acquisition unit is configured to acquire according to the acquisition keyword to generate a set of candidate products, wherein the set of candidate products includes multiple product acquisition results, and each of the product acquisition results corresponds to a candidate live broadcast merchant.

15. The commodity search device according to claim 14, characterized in that: The acquisition unit is specifically configured as follows: Obtaining a product acquisition result corresponding to the current node of the product category tree, and adding it to the acquisition candidate product set; If the number of product acquisition results in the acquisition candidate product set is less than a preset number, the upper node is searched from the current node along the product category tree, and the product acquisition results corresponding to the upper node are added to the acquisition candidate product set.

16. The commodity search device according to claim 15, characterized in that: The acquisition unit is further configured to: If the number of product acquisition results in the acquired candidate product set is still less than the preset number, the acquisition keywords are screened; The obtained keywords are retrieved again after being screened, and the obtained commodity acquisition results are added to the obtained candidate commodity set.

17. The commodity search device according to claim 11, characterized in that: The tag is a dynamic tag, wherein the fourth acquisition module is specifically configured to: Obtaining the current live broadcast content or current explanation content of the candidate live broadcast merchant; The current live broadcast content or the current explanation content is identified to generate the label.

18. The commodity search device according to claim 11, characterized in that: The tag is a static tag, wherein the fourth acquisition module is specifically configured as follows: Extracting product description information of the candidate live broadcast merchant; The label is generated according to the product description information.

19. The commodity search device according to claim 11, characterized in that: The tag is a static tag, wherein the fourth acquisition module is specifically configured as follows: Obtaining historical live broadcast content of the candidate live broadcast merchant; The tag is generated according to the historical live broadcast content.

20. The commodity search device according to claim 11, characterized in that: The tag is a feature tag, wherein the fourth acquisition module is specifically configured as follows: Obtaining user statistical characteristics and / or commodity sales statistical characteristics of the candidate live broadcast merchant; The tag is generated according to the user statistical characteristics and / or commodity sales statistical characteristics.

21. The commodity search device according to claim 11, characterized in that: The determining module is further configured to: The initial relevance score is adjusted according to the intent characteristic.

22. An electronic device, characterized in that: include: processor; a memory for storing instructions executable by the processor; The processor is configured to execute the instructions to implement the commodity search method according to any one of claims 1 to 10.

23. A computer-readable storage medium, when instructions in the computer-readable storage medium are executed by a processor of an electronic device, the electronic device is enabled to execute the commodity search method according to any one of claims 1 to 10.

24. A computer program product comprising a computer program / instructions, characterized in that When the computer program / instructions are executed by a processor, the commodity search method according to any one of claims 1 to 10 is implemented.

Citation Information

Patent Citations

  • Method for recording access of user to merchandise information, search method and server

    CN102253936A

  • Personalized recommendation method

    CN102663627A

  • Information display method and device, terminal, server and storage medium

    CN111970523A