Input Method, Apparatus and Device

Abstract

The present application discloses an input method, device and equipment. Among them, the method includes: obtaining the input of a user, and obtaining, according to the input, prompt information that matches the input in a commodity description set, and providing the prompt information to the user. By adopting this processing method, input association information of partial commodity title content input by the user is obtained based on the commodity description set; therefore, the input association relevance of the idle commodity title can be effectively improved.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and specifically relates to an input method and apparatus, as well as an electronic device. Background Art

[0002] With the booming development of the second-hand trading market, more and more ordinary people hang their idle goods on idle trading platforms (such as Xianyu) for sale. However, most of these seller users are not professional sellers, and the commodity information they post is often incomplete. Sellers do not describe or describe unclearly what kind of commodity is being posted, resulting in it being difficult for buyers to search for such a commodity. At the same time, the platform also needs to understand the commodity structure, and then it can recommend idle goods to interested buyers to facilitate the circulation of idle goods and conclude a transaction. To solve the above problems, when a user posts a commodity, intelligent input association for commodity posting makes a drop-down association based on the text edited by the user during posting and associates structured attributes.

[0003] Currently, a typical intelligent input association scheme based on a commodity title is to adopt the search keyword (Query) association technology used in the search box of traditional e-commerce platforms (such as Taobao) to assist in making a drop-down association for the text edited by the seller user on the idle trading platform. For example, when entering "iPhone", drop-down association prompts information such as "iPhone phone case".

[0004] However, in the process of implementing the present invention, the inventors found that the above technical solutions all have at least the following problems: 1) The relevance of the input association for the title of idle goods is relatively low; 2) It is difficult to structure the attributes of idle goods. In summary, how to improve the relevance of the input association and reduce the difficulty of structuring commodity attributes has become an urgent problem for those skilled in the art. Summary of the Invention

[0005] This application provides an input method to solve the problems of relatively low relevance of the input association for commodity titles and difficulty in structuring commodity attributes existing in the prior art. This application also provides an input apparatus and an electronic device.

[0006] This application provides an input method, including:

[0007] Obtain the input of the user;

[0008] According to the input, obtain the prompt information that matches the input in the commodity description set;

[0009] Provide the prompt information to the user.

[0010] Optionally, a corresponding relationship between a query word and the prompt information is stored in the commodity description set.

[0011] Optionally, a correspondence relationship between the vector representation of the query term and the prompt information is also stored in the set of product descriptions.

[0012] Optionally, the prompt information includes at least one product attribute information.

[0013] Optionally, the correspondence relationship is determined by the following steps:

[0014] Extract the query term from the product title according to the query term determination rule;

[0015] Through a language model, according to the product title, determine the N-gram phrases with the query term as the prefix;

[0016] Use the words after the query term in the N-gram phrases as product attribute information to form the correspondence relationship.

[0017] Optionally, the method further includes:

[0018] Determine the idle product title determined based on the at least one product attribute information;

[0019] Structurally store the idle product object according to the product attribute information included in the idle product title.

[0020] Optionally, the query term and the product attribute information belong to different product attributes;

[0021] Alternatively, the query term and the product attribute information belong to the same product attribute, the query term includes: the prefix word of the product attribute; the product attribute information includes: the suffix word of the product attribute.

[0022] Optionally, the method further includes:

[0023] If the input suffix word meets the query term determination rule, use the suffix word as the target query term;

[0024] Obtain the prompt information matching the target query term according to the correspondence relationship.

[0025] Optionally, the method further includes:

[0026] Determine the importance score of the product attribute information;

[0027] Determine the arrangement order of the at least one product attribute information according to the importance score.

[0028] Optionally, the importance score is determined by the following steps:

[0029] Determine the number of times the product attribute information is associated with the query term according to the set of product titles;

[0030] Determine the importance score according to the number of associations.

[0031] Optionally, the method further includes:

[0032] Construct an inverted index of the query term according to the corresponding relationship;

[0033] Determine the hint information matching the input according to the inverted index.

[0034] Optionally, obtaining the user input includes:

[0035] Obtain the text information of the user input;

[0036] And / or, obtain the voice information of the user input;

[0037] And / or, obtain the picture information of the user input;

[0038] And / or, obtain the video information of the user input after that.

[0039] Optionally, the method further includes:

[0040] Pass the user input through a pre-trained model to obtain a vector representation of the input;

[0041] Correspondingly, obtaining the hint information matching the input in the commodity description set according to the input includes:

[0042] Obtain the hint information matching the input in the commodity description set according to the vector representation of the input.

[0043] Optionally, obtaining the hint information matching the input in the commodity description set includes:

[0044] Calculate the distance between the vector representation of the input and the vector representation of the query term stored in the commodity description set;

[0045] Select the hint information corresponding to the vector representation of the query term with a distance less than a preset threshold as the hint information matching the input.

[0046] This application also provides an input device, including:

[0047] An input information acquisition unit for obtaining the input of the user;

[0048] A hint information determination unit for obtaining the hint information matching the input in the commodity description set according to the input;

[0049] A hint information providing unit for providing the hint information to the user.

[0050] The present application also provides an electronic device, including:

[0051] a processor and a memory;

[0052] The memory is used to store a program for implementing the input method. After the device is powered on and runs the program of this method through the processor, the following steps are executed: obtaining the input of the user; obtaining, according to the input, prompt information matching the input in the set of product descriptions; and providing the prompt information to the user.

[0053] The present application also provides an idle product object publishing system, including:

[0054] A server, which is used to determine a set of association relationships between idle product title query terms and product attribute information according to a product title library; for a product title input association request sent by a client, according to the set of association relationships, determine candidate product attribute information associated with the sub-text of the product title edited by the user, and send the candidate product attribute information back to the client; for a product publishing request sent by the client, publish an idle product object at least according to the idle product title carried in the product publishing request;

[0055] A client, which is used to send the input association request, display the candidate product attribute information; determine an idle product title according to the candidate product attribute information, where the idle product title includes a sub-text of the product title and product attribute information determined based on the candidate product attribute information; and send the product publishing request.

[0056] The present application also provides an idle product object publishing method, including:

[0057] Determining a set of association relationships between idle product title query terms and product attribute information according to a product title library;

[0058] For a product title input association request sent by a client, according to the set of association relationships, determine candidate product attribute information associated with the sub-text of the product title edited by the user, and send the candidate product attribute information back to the client;

[0059] For a product publishing request sent by a client, publish an idle product object at least according to the idle product title carried in the product publishing request, where the idle product title includes a sub-text of the product title and product attribute information determined based on the candidate product attribute information.

[0060] Optionally, the determining a set of association relationships between product title query terms and product attribute information according to a product title library includes:

[0061] Extracting product title query terms from the product title according to a query term determination rule;

[0062] Using the N-Gram model, according to the product title, determine the N-gram phrases prefixed with the product title query term;

[0063] Take the words after the product title query term in the N-gram phrases as product attribute information, and form the association relationship between the product title query term and the product attribute information.

[0064] Optionally, the determining of the candidate product attribute information associated with the product title sub-text edited by the user according to the association relationship set includes:

[0065] According to the product title sub-text, determine the target query term;

[0066] According to the association relationship set, determine at least one product attribute information associated with the target query term as the candidate product attribute information.

[0067] Optionally, the determining of the target query term according to the product title sub-text includes:

[0068] If the suffix word of the product title sub-text conforms to the query term determination rule, take the suffix word as the target query term.

[0069] Optionally, the determining of at least one product attribute information associated with the target query term as the candidate product attribute information according to the association relationship set includes:

[0070] According to the association relationship set, determine the product title query term that matches the target query term text;

[0071] Take the product attribute information corresponding to the matching product title query term as the candidate product attribute information.

[0072] Optionally, the method further includes:

[0073] Determine the first text vector of the product title query term;

[0074] The determining of the candidate product attribute information associated with the product title sub-text edited by the user according to the association relationship set includes:

[0075] Determine the second text vector of the product title sub-text;

[0076] Determine the vector distance between the first text vector and the second text vector;

[0077] According to the vector distance, determine the candidate product attribute information.

[0078] Optionally, the determining of the candidate product attribute information according to the vector distance includes:

[0079] Use the product title query terms with a vector distance less than the vector distance threshold as candidate product attribute information.

[0080] Optionally, it further includes:

[0081] Determine the importance scores of each product attribute information associated with the product title query terms;

[0082] Determine the sorting order of the candidate product attribute information according to the importance scores.

[0083] Optionally, the determining the importance scores of each product attribute information associated with the product title query terms includes:

[0084] For each product attribute information associated with the product title query terms, determine the number of associations between the product attribute information and the product title query terms according to the product title library;

[0085] Determine the importance scores according to the number of associations.

[0086] Optionally, it further includes:

[0087] Construct an inverted index of the product title query terms according to the association relationship set;

[0088] Determine the candidate product attribute information associated with the product title sub-text edited by the user according to the inverted index.

[0089] Optionally, it further includes:

[0090] Determine the multiple product attribute information included in the idle product title;

[0091] Structurally store the idle product object according to the multiple product attribute information.

[0092] Optionally, the product title query terms include: product category name, category modifier, brand name, material, style, style element, color, function.

[0093] Optionally, the product title query terms and the product attribute information belong to different product attributes.

[0094] Optionally, the product title query terms and the product attribute information belong to the same product attribute, the product title query terms include: prefix words of the product attribute; the product attribute information includes: suffix words of the product attribute.

[0095] This application also provides a method for publishing an idle product object, including:

[0096] Send a product title input association request for the product title sub-text edited by the user to the server;

[0097] Display the candidate product attribute information associated with the sub - text of the product title sent back by the server;

[0098] Determine the idle product title according to the candidate product attribute information, where the idle product title includes the sub - text of the product title and the product attribute information determined based on the candidate product attribute information;

[0099] Send a product release request for the idle product title to the server.

[0100] This application also provides an idle product object search system, including:

[0101] A server, which is used to determine the association relationship set between the idle product title query term and the product attribute information according to the product title library; for the product title input association request sent by the first client, determine the candidate product attribute information associated with the sub - text of the product title edited by the first user according to the association relationship set, and send the candidate product attribute information back to the first client; for the product release request sent by the first client, release the idle product object of the first user at least according to the idle product title carried in the product release request, and structurally store the idle product object into the idle product object library according to the multiple product attribute information included in the idle product title; and, for the product object search request sent by the second client, determine the idle product object that matches the search term specified by the second user according to the idle product object library, and send the matching idle product object to the second client;

[0102] The first client is used to send the input association request, display the candidate product attribute information; determine the idle product title according to the candidate product attribute information, where the idle product title includes the sub - text of the product title and the product attribute information determined based on the candidate product attribute information; send the product release request;

[0103] The second client is used to send the product object search request and display the matching idle product object.

[0104] This application also provides an idle product title input association system, including:

[0105] A server, which is used to determine the association relationship set between the idle product title query term and the product attribute information according to the product title library; for the product title input association request sent by the client, determine the candidate product attribute information associated with the sub - text of the product title edited by the user according to the association relationship set, and send the candidate product attribute information back to the client;

[0106] A client for sending the input association request and displaying candidate product attribute information; determining an idle product title according to the candidate product attribute information, where the idle product title includes a product title sub-text and product attribute information determined based on the candidate product attribute information.

[0107] This application also provides a computer-readable storage medium storing instructions that, when run on a computer, cause the computer to execute the above various methods.

[0108] This application also provides a computer program product including instructions that, when run on a computer, cause the computer to execute the above various methods.

[0109] Compared with the prior art, this application has the following advantages:

[0110] The input method provided by the embodiment of this application obtains a user's input, obtains prompt information matching the input from a product description set according to the input, and provides the prompt information to the user. By adopting this processing method, input association information of partial product title content input by the user is obtained based on the product description set; therefore, the input association relevance of the idle product title can be effectively improved.

[0111] Furthermore, in the input method provided by the embodiment of this application, a correspondence relationship between a query term and at least one product attribute information is stored in the product description set, so that based on the input prompt information of the idle product title, both the structured attributes (such as brand and model) of the idle product object can be obtained, and the association relationship between the attributes (such as the association between iPhone and 64G) can be constructed, enabling the structure to have additional association capabilities and supplementing product attribute information. Therefore, the difficulty of structured processing of product attributes based on the idle product title can be effectively reduced. At the same time, this processing method of completing the input of the idle product title by prompting candidate product attributes enables the idle product to be better published, thus effectively improving the quality of idle product publication. In this way, for seller users, the exposure rate of the idle product can be increased, prompting the product to be sold as soon as possible; for buyer users, they can buy the idle products they are interested in as soon as possible and have more products to choose from, thus improving the user experience; for the idle trading platform, more idle circulation can be promoted to reach transactions, thereby improving the utilization rate of platform resources, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0112] Figure 1 A flowchart of an embodiment of an input method provided by this application;

[0113] Figure 2 A schematic diagram of an application scenario of an embodiment of an input method provided by this application;

[0114] Figure 3 Schematic diagram of device interaction in an embodiment of an input method provided by this application;

[0115] Figure 4 Schematic diagram of retrieving input association prompt information in an embodiment of an input method provided by this application;

[0116] Figure 5 Another schematic diagram of retrieving input association prompt information in an embodiment of an input method provided by this application. Detailed implementation manners

[0117] Many specific details are set forth in the following description in order to provide a thorough understanding of this application. However, this application can be implemented in many other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the connotation of this application. Therefore, this application is not limited by the specific implementations disclosed below.

[0118] In this application, there are provided an input method and device, a system, method and device for releasing idle commodity objects, a system, method and device for searching idle commodity objects, a system, method and device for input association of idle commodity titles, and an electronic device. Each solution will be described in detail in the following embodiments.

[0119] First Embodiment

[0120] Please refer to Figure 1 , which is a flowchart of an embodiment of the input method of this application. In this embodiment, the method may include the following steps:

[0121] Step S101: Obtain the user's input.

[0122] The user's input may include partial content of the commodity title edited by the user in the commodity title input scenario. Specifically, the input may be partial content of the idle commodity title edited by the user through the client of the idle commodity object release system, or may also be partial content of the commodity title edited by the user in any other commodity title input scenario, such as partial content of the new product commodity title edited by the user through the client of the new product commodity object release system.

[0123] The user's input may be text information, or may also be other forms of information other than text, such as voice information, picture information, video information, etc. The user's input may be any one of various forms of information, such as the user's input only includes picture information or text information. The user's input may also include various forms of information, such as the user's input includes picture information, picture information and voice information.

[0124] In specific implementation, if the user input includes information in other forms than text, the user input in other forms can be first converted into text information, and the converted text information can be used as part of the description information of the product title of the user input. For example, for the product picture or product display video input by the user, the product name can be determined through an image-based product recognition algorithm. For example, according to the picture of an iPhone, the recognized product name is "iPhone". For another example, for the voice information input by the user, the product name said by the user can be recognized through a voice recognition algorithm.

[0125] Please refer to Figure 2 , which is a schematic diagram of the scenario where the method provided in this embodiment is applied to the idle product object release system. The system includes: a server and a client. Among them, the server can be a server deployed on a cloud server or a server dedicated to providing idle product object release services, and can be deployed in a data center. The server can be a cluster server or a single server. The client includes, but is not limited to, a mobile communication device, that is, the commonly said mobile phone or smart phone, and also includes clients such as a personal computer, a PAD, and an iPad. The server and the client can be connected through a network. For example, the client can be connected to the network through WIFI or other means.

[0126] In this embodiment, the user opens the product release page of the idle trading platform (such as Xianyu platform) through the client, and wants to release the information of the idle product in his hand to the platform to sell the idle product. The idle trading platform mainly serves ordinary consumers. In order to encourage such non-professional seller users to release more idle products, compared with the traditional new product e-commerce platform (such as Taobao.com), the platform usually greatly simplifies the operation process of releasing idle products to reduce the difficulty of releasing idle product information. Therefore, the product release page provided by the idle trading platform usually only includes two input boxes: the product title and the product picture. For the new product e-commerce platform for professional sellers, in order to provide comprehensive product information (product details) to buyer users, the product release page provided by the platform not only includes input boxes for the product title (essentially the product name) and the product picture, but also includes input boxes for various attribute information of the product (such as brand, size, etc.). At the same time, since the users of the idle trading platform are not professional sellers, when releasing idle product information, the product title (essentially product description information, which can include not only the product name, but also other attributes of the product, such as color, origin, purchase time, etc. In this scenario, the product title can also be called idle product description information) usually does not describe or describes unclearly what product is being released. By using the system, when the user releases an idle product, it is possible to perform real-time drop-down association based on the text of the product title edited by the user during release, and associate the structured attribute information of the product, that is, to implement the function of the system prompting what information to input while the user is inputting the product title, so as to help non-professional sellers input a more comprehensive idle product title with product information.

[0127] For example, user A has an idle iPhone X to sell through an idle trading platform. After the seller user enters iPhone X in the product title input box on the idle product release page, he may not think of entering the phone color, capacity, and other various product attribute information that is helpful to help the idle product buyer user retrieve the idle product. The system can use the method to associate the attribute information input of the product in real time after the user enters iPhone X, such as providing options for color attributes such as space gray, silver, and red, or providing options for capacity attributes such as 64G and 128G. In this way, the seller user can select the product color, capacity, and other attribute information according to the structured attribute prompt information associated with the product automatically provided by the system, helping the seller user to enter a product title with more comprehensive product attributes (product features). At the same time, the idle trading platform can also understand the idle product in a structured manner, so that when other buyer users enter "iPhone X, 64G, and red" in the platform search box, they can quickly search for the idle product and recommend the idle product to interested buyers, avoiding the inability to retrieve the product due to missing attribute information. In this way, for sellers, the exposure rate of idle goods can be increased, prompting the goods to be sold as soon as possible; for buyers, they can buy the idle goods they are interested in as soon as possible, and have more goods to choose from, thus improving the user experience. At the same time, for idle trading platforms, the pressure on database retrieval can be reduced, and more idle circulation transactions can be facilitated, thereby improving the utilization rate of platform resources.

[0128] It can be seen from the above system that in this embodiment, the execution subject of the method can be the server, and step S101 can be implemented in the following manner: the user edits the product title on the client, and can enter part of the product title (also called product title subtext), such as the product name "iPhone X", and sends a product title input association request to the server through the client; the server receives the request sent by the client and obtains the input from the request.

[0129] Step S103: according to the input, obtaining prompt information matching the input in the commodity description set.

[0130] The set of product description information may include product description information of multiple products, such as product description information of multiple second-hand products (used products), or product description information of multiple new products (first-hand products). The product description information may include multiple product attribute information such as product name, color, material, origin, etc. However, these product attribute information may not be distinguished and stored in a structured form. That is to say, there is only product attribute information, but no attribute name information. Multiple product attribute information are connected together to form a product description information, which can be a second-hand product title or a new product title in the second-hand product scenario.

[0131] In one example, the correspondence between query terms and prompt information is stored in the set of product description information. The prompt information includes, but is not limited to, at least one product attribute information. The product attribute information includes, but is not limited to, product brand, model, category, material, appearance description, etc. The multiple product attribute information included in the prompt information can be attribute information with distinguished attribute names or attribute information without distinguished attribute names.

[0132] Please refer to Figure 3 , which is a schematic diagram of device interaction when the method provided by this application is applied to a second-hand product object publishing system. In this embodiment, the server is used to determine the association relationship set between the second-hand product title query term and the product attribute information according to the product title library. This set may include the correspondence between multiple query terms and product attribute information; for the product title input association request sent by the client, according to the association relationship set, determine the candidate product attribute information (i.e., the prompt information) associated with the product title sub-text edited by the user, and send the candidate product attribute information back to the client; correspondingly, the client is used to send the input association request and display the candidate product attribute information; according to the candidate product attribute information, determine the second-hand product title, and the second-hand product title includes the product title sub-text and the product attribute information determined based on the prompt information.

[0133] Specifically, the client can also be used to send a product publishing request; the server can also be used to publish the second-hand product object at least according to the second-hand product title carried in the product publishing request sent by the client.

[0134] The product title library includes multiple product titles. The product title may include the product name of the product object published on the traditional e-commerce platform, which usually includes a variety of key attribute information that buyers and users are concerned about, such as category name, brand, model, size, capacity, etc. For example, the product name of product object A on Taobao is "Apple / Apple#iPhone8Plus China Telecom, China Mobile, China Unicom 8p Full Netcom 4G Mobile Phone Official Authentic iPhone8plus", where the key attribute information includes: brand "Apple / Apple", category name "mobile phone", sub-category name "iPhone", model "8Plus", operator network "China Telecom, China Mobile, China Unicom 8p Full Netcom 4G", quality "official authentic". The product title may also include product titles that have been published on idle trading platforms, including product titles that provide a more comprehensive description of the characteristics of idle products.

[0135] The association relationship set between the idle product title query words and the product attribute information may include a large number of association relationships between idle product title query words and product attribute information. In one association relationship (i.e., the corresponding relationship), the idle product title query words and the product attribute information have a high correlation, and the two usually appear in pairs, and the product attribute information may be the subsequent adjacent content of the idle product title query words. One idle product title query word may be associated with multiple product attribute information.

[0136] Please refer to Figure 4 , which is a schematic diagram of input association prompt information retrieval in the idle commodity object release system using the method provided by the present application. Figure 4 As can be seen from the above table, the product attribute associated with the query word "iPhone" can be the product model, and the product model attribute value can be X, 11, etc. In addition, the product attribute associated with the query word "iPhone" can also be other attributes, such as capacity, color, etc. The product attribute associated with the query word "iPhone X" can be capacity, and the capacity attribute value can be 64G or 128G, etc. In addition, the product attribute associated with the query word "iPhone X" can also be other attributes, such as place of origin, etc. The product attribute associated with the query word "Huawei" can be the product model, and the product model attribute value can be Mate or P30, etc.

[0137] In this embodiment, the correspondence between the query term and at least one product attribute information can be determined by the following steps: 1) According to the query term determination rule, extract the query term from the product title, such as extracting the product title query term from all or part of the product titles in the idle product library or the new product library; 2) Through a language model, determine the N-gram phrases prefixed with the query term according to the product title; 3) Use the words after the query term in the N-gram phrases as the product attribute information to form the correspondence.

[0138] In specific implementation, the server can form a product title library based on a large number of product names stored in the traditional e-commerce platform. The server can also extract the product title query term from the product title according to the query term determination rule; through the N-Gram language model, determine the N-gram phrases prefixed with the product title query term according to the product title; use the words after the product title query term in the N-gram phrases as the product attribute information to form the association relationship between the product title query term and the product attribute information.

[0139] Applying the query term determination rule, the product title query term can be extracted from the product title. For example, the query term determination rule can include: the product category name is the query term, the category modifier is the query term, the product brand is the query term, the material is the query term, the style is the query term, the style element is the query term, the color is the query term, the function and efficacy are the query terms, the size specification is the query term, the quality and condition are the query terms, and so on. For example, the product name of product object A on Taobao is "Apple / Apple #iPhone 8Plus Telecommunications Mobile Unicom 8p 4G Mobile Phone with Full Network Connectivity, Official Genuine iPhone8plus", and the following product title query terms can be extracted from this product title: "Apple" and "Apple" from the brand, "mobile phone" from the category name, "iPhone" from the sub-category name, "full network connectivity" from the operator network element, and so on.

[0140] After the server determines the rule according to the query term and extracts the product title query term from the product title, it can use the N-Gram model to determine the N-gram phrases with the product title query term as the prefix according to the product title. N-Gram is an algorithm based on the statistical language model. The server performs a sliding window operation of size N on the content in each product title in the product title library by bytes, forming a sequence of byte fragments of length N with the product title query term as the prefix word. Each byte fragment is called a gram, and the occurrence frequencies of all grams can be counted and filtered according to a preset threshold to form a list of key grams. In specific implementation, the corresponding relationship can be determined according to the list of key grams. This model is based on the assumption that the occurrence of the Nth word is only related to the previous N-1 words and has nothing to do with any other words, and the probability of the whole sentence is the product of the occurrence probabilities of each word. These probabilities can be obtained by directly counting the number of times N words appear simultaneously in the corpus. Commonly used are the binary Bi-Gram and the ternary Tri-Gram. Therefore, the word after the product title query term in each gram in the list can be used as the product attribute information, thereby forming the association relationship between the product title query term and the product attribute information.

[0141] For example, the product name of product object A on Taobao is "Apple / Apple #iPhone 8Plus Telecommunications Mobile Unicom 8p Full Netcom 4G Mobile Phone Official Genuine iPhone8plus". Using the 3-Gram model, the following triples can be determined: Apple Apple iPhone, Apple iPhone 8, iPhone 8Plus, Full Netcom 4G Mobile Phone. In each triple, the prefix word is the product title query term, and the suffix word is the product attribute information. For example, in the triple "iPhone 8Plus", iPhone is the query term, and 8Plus is the product attribute information. According to other product titles, the product attribute information associated with the query term "iPhone" can also be X, 12, etc. Another example is that in the triple "Full Netcom 4G Mobile Phone", Full Netcom is the query term, and 4G is the product attribute information. According to other product titles, the product attribute information associated with the query term "Full Netcom" can also be 5G, etc.

[0142] In this embodiment, the query term and the product attribute information may belong to different product attributes. For example, in the triple "iPhone 8 Plus", iPhone is the query term and belongs to the category name attribute, while 8 Plus is the product attribute information and belongs to the model attribute. The query term and the product attribute information may also belong to the same product attribute. The query term includes: the prefix of the product attribute; the product attribute information includes: the suffix of the product attribute. For example, in the triple "4G mobile phone with all-netcom access", "all-netcom access" is the query term and is the prefix of the network attribute, and 4G is the product attribute information and is the suffix of the network attribute.

[0143] In specific implementation, the server may determine the association relationship set between the product title query term and the product attribute information offline according to the product title library.

[0144] In practical applications, the online real-time requirement for the response time of the input association result to be returned is in milliseconds (such as 50 ms). Therefore, in this embodiment, the method of building the index offline and querying online is adopted. In specific implementation, the method may further include the following steps: build the inverted index of the query term according to the corresponding relationship; determine the prompt information that matches the input according to the inverted index, such as at least one product attribute information.

[0145] In this embodiment, the server may build the inverted index of the product title query term according to the association relationship set (i.e., the corresponding relationship); determine the prompt information (such as at least one product attribute information) associated with the sub-text of the product title edited by the user (i.e., the input) according to the inverted index. The inverted index is derived from the need to find records according to the product title query term in practical applications (such as the product attribute information corresponding to the query term). Each item in this index table includes a product title query term and the addresses of the records with this query term. Since it is not the record that determines the query term, but the query term that determines the position of the record, the prompt information (such as product attribute information) associated with the target query term is determined. By adopting this processing method, the retrieval speed of the target query term can be improved, thereby improving the input association speed, further improving the release efficiency of idle product objects, and further improving the exposure rate of idle products.

[0146] After determining the association relationship set (i.e., the corresponding relationship) between the product title query term and the product attribute information, the server can provide the client user with an online service for input association based on the product title. In this embodiment, the server receives a product title input association request sent by the user through the client, and determines candidate product attribute information associated with the sub-text of the product title edited by the user according to the association relationship set (i.e., including the corresponding relationship between multiple query terms and at least one product attribute information), and sends the candidate product attribute information back to the client as prompt information for product title input. In this way, prompt information matching the input can be obtained from the product description set according to the input.

[0147] The prompt information (such as the candidate product attribute information) provided to the user by the method provided in this embodiment may include multiple values of one attribute as optional items for the user to select. The candidate product attribute information may also include multiple values of multiple attributes, which mainly depends on the specific situation of the association relationship set generated from the product title library. Figure 2 It can be seen that for the user input of "iPhone X", the server can generate a query statement based on this term and retrieve the product attribute information corresponding to this term, which may include dark gray, red, etc.

[0148] In an example, the server can determine a target query term according to the user input (such as the sub-text of the product title, that is, part of the product title); according to the association relationship set (i.e., the corresponding relationship), determine at least one product attribute information associated with the target query term as candidate product attribute information; this processing method enables the server to perform an online query on the suffix input by the user in the online query part and obtain relevant attribute information (candidate terms) through full-text matching. This is a text-based processing method that can accurately match the user input for association.

[0149] The target query term is the query term determined from the product title text being edited by the user, that is, the query term determined from the input. In this embodiment, step S103 may include the following sub-steps: If the suffix term of the input meets the query term determination rule, then use the suffix term as the target query term; according to the corresponding relationship, obtain the prompt information matching the target query term.

[0150] In this embodiment, if the server determines that the suffix word of the sub - text of the product title conforms to the query word determination rule, the suffix word is used as the target query word. For example, when the user inputs "iPhone given by others", "iPhone" can be used as the target query word. The query word determination rule may be the same as the above - mentioned query word determination rule, which will not be elaborated here. Specifically, when implementing, the server can determine the query word that matches the target query word text according to the association relationship set; and use the product attribute information corresponding to the matching query word as the candidate product attribute information to be prompted to the user, that is, the prompt information.

[0151] In another example, the product description set may also store the corresponding relationship between the vector representation of the query word and the prompt information. The vector representation can reflect the semantic information of the query word.

[0152] Correspondingly, the method may further include the following steps: obtaining the vector representation of the input through a pre - trained model; correspondingly, step S103 can be implemented in the following manner: obtaining the prompt information that matches the input from the product description set according to the vector representation of the input.

[0153] Specifically, obtaining the prompt information that matches the input from the product description set may include the following sub - steps: calculating the distance between the vector representation of the input and the vector representation of the query word stored in the product description set; selecting the prompt information corresponding to the vector representation of the query word with a distance less than a preset threshold as the prompt information that matches the input.

[0154] The execution subject of the method (such as the server) can determine the vector representation of the query word as the first text vector; and can determine the vector representation of the input as the second text vector; determine the vector distance between the first text vector and the second text vector; and determine the prompt information that matches the input according to the vector distance, such as candidate product attribute information.

[0155] In this embodiment, during the offline phase, the server determines the first text vectors of each query term in the association relationship set; during the online query phase, the server determines the second text vector of the input (such as the sub-text of the product title); determines the vector distance between the first text vector and the second text vector; and determines the candidate product attribute information (i.e., the prompt information) according to the vector distance. By adopting this processing method, it is possible to comprehensively understand the product title input by the seller, so that the obtained input association results are more accurate. For example, the input association prompt information given by the method for the user input of "Qixia apples" includes: 5A quality, etc., and the input association prompt information given by the method for the user input of "iPhone apples" includes: X, 8plus, etc. In addition, this processing method also makes the input association results more comprehensive, avoiding the inability to find query terms with similar semantics due to full-text matching based on the text; therefore, the user experience can be effectively improved.

[0156] The vector representation, also known as the text vector, includes converting the symbolic representation of the text into a vector representation in the semantic space, which can be used to quantitatively compare semantics. Such methods are usually based on Harris's distributional hypothesis, that is, words in similar contexts usually have similar semantics.

[0157] Specifically, during implementation, a pre-trained model, such as a word vector calculation model (Word2vec), can be used to determine the text vector of the query term. Since the word vector calculation model belongs to a relatively mature existing technology, such as BERT (Bidirectional Encoder Representation from Transformers, that is, the encoder of the bidirectional Transformer model), etc., it will not be elaborated here.

[0158] The vector distance can be the Euclidean distance between the first text vector and the second text vector, etc.

[0159] As Figure 5 shown, specifically during implementation, the server can determine one or more query terms whose vector distances are less than the vector distance threshold, and provide all the prompt information corresponding to these query terms to the user. The vector distance threshold can be determined according to application requirements. The larger this threshold is, the more associations are returned, but the lower the relevance; the smaller this threshold is, the fewer associations are returned, but the higher the relevance.

[0160] Specifically, during implementation, the indexed content can be offline indexed by obtaining the vector representation through a pre-trained model (such as BERT). During the query, a vector retrieval engine can be used to obtain the N product title query terms that are the closest and ranked in the front, and the candidate terms (product attribute information) associated with these query terms are sorted in descending order of importance scores, and multiple product attribute information ranked at the top is returned to the user.

[0161] In one example, the method may further include the following steps: determining an importance score of the product attribute information; and determining an arrangement order of the at least one product attribute information according to the importance score.

[0162] In specific implementation, the server may determine the importance scores of the respective product attribute information associated with the query term; and determine the arrangement order of the product attribute information according to the importance scores. By adopting this processing manner, the input prompt information with a greater possibility is displayed in a front position; therefore, the user experience can be effectively improved.

[0163] In one example, the method may further include the following steps: determining the number of times of association between the product attribute information and the query term according to the product description set (such as a product title library); and determining the importance score according to the number of times of association.

[0164] In specific implementation, the server may, for the respective product attribute information associated with the product title query term, determine the number of times of association between the product attribute information and the product title query term according to the product title library; and determine the importance score according to the number of times of association. For example, if the number of occurrences of "iPhone X" is greater than that of "iPhone 8 Plus", it may indicate that there are more users selling "iPhone X". At this time, when the user inputs "iPhone", X can be displayed in front of 8 Plus.

[0165] Step S105: Provide the prompt information to the user.

[0166] In this embodiment, the server sends the prompt information to the client, and the client displays the prompt information for the user to input a product title (product description information) for reference. The user inputs more complete and comprehensive product description information according to the prompt information, that is, the completeness of the product title is improved.

[0167] In one example, the method may further include the following steps: 1) determining an idle product title determined based on the at least one product attribute information; and 2) structurally storing an idle product object according to the product attribute information included in the idle product title, such as storing the idle product object in a relational database.

[0168] After the user completes the title of the idle commodity through input on the client according to the prompt information, the user can send the commodity release request to the server through the client. The request includes at least the title of the idle commodity. The title of the idle commodity may include a sub-text of the commodity title (i.e., the commodity description information input by the user before obtaining the pushed prompt information) and commodity attribute information determined based on at least one commodity attribute information included in the prompt information. The server responds to the commodity release request, generates the idle commodity object, and publishes the object on the idle trading platform for seller users to search, view, place orders for purchase, etc.

[0169] In specific implementation, the server can determine multiple commodity attribute information in the commodity title according to the title of the idle commodity (i.e., the description information of the idle commodity) carried in the request, and structurally store the idle commodity object according to the multiple commodity attribute information.

[0170] As can be seen from the above embodiments, the input method provided by the embodiments of the present application obtains the user's input, obtains the prompt information matching the input from the commodity description set according to the input, and provides the prompt information to the user. By adopting this processing method, input association information of partial commodity title content input by the user is obtained based on the commodity description set. Therefore, the input association relevance of the idle commodity title can be effectively improved.

[0171] Furthermore, in the input method provided by the embodiments of the present application, the corresponding relationship between the query word and at least one commodity attribute information is saved in the commodity description set, so that based on the input prompt information of the idle commodity title, both the structural attributes (such as brand and model) of the idle commodity object can be obtained, and the association relationship between the attributes (such as the association between iPhone and 64G) can be constructed, enabling the structure to have additional association capabilities and supplementing the commodity attribute information. Therefore, the difficulty of structurally processing the commodity attributes based on the idle commodity title can be effectively reduced. At the same time, this processing method of completing the input of the idle commodity title by prompting the candidate words of the commodity attributes enables the idle commodity to be better released, thus effectively improving the quality of the idle commodity release. In this way, for seller users, the exposure rate of the idle commodity can be increased, prompting the commodity to be sold as soon as possible; for buyer users, they can buy the idle commodity they are interested in as soon as possible and have more commodities to choose from, thus improving the user experience; for the idle trading platform, more idle circulation can be promoted to reach transactions, thereby improving the utilization rate of platform resources, etc.

[0172] Second Embodiment

[0173] In the above embodiments, an input method is provided. Correspondingly, the present application also provides an input device. This device corresponds to the method embodiments above. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple. For related parts, refer to the partial description of the method embodiments. The device embodiments described below are merely illustrative.

[0174] The present application further provides an input device, including:

[0175] An input information acquisition unit, configured to obtain user input;

[0176] A prompt information determination unit, configured to obtain, according to the input, prompt information matching the input from a set of product descriptions;

[0177] A prompt information providing unit, configured to provide the prompt information to the user.

[0178] The third embodiment

[0179] In the above embodiments, an input method is provided. Correspondingly, the present application also provides an electronic device. This device corresponds to the method embodiments above. Since the device embodiments are basically similar to the method embodiments, the description is relatively simple. For related parts, refer to the partial description of the method embodiments. The device embodiments described below are merely illustrative.

[0180] An electronic device in this embodiment, the device includes: a processor and a memory; the memory is used to store a program for implementing the input method. After the device is powered on and runs the program of this method through the processor, the following steps are executed: obtaining user input; obtaining, according to the input, prompt information matching the input from a set of product descriptions; providing the prompt information to the user.

[0181] The fourth embodiment

[0182] In the above embodiments, an input method is provided. Correspondingly, the present application also provides a system for publishing idle product objects. This system corresponds to the method embodiments above. Since the system embodiments are basically similar to the method embodiments, the description is relatively simple. For related parts, refer to the partial description of the method embodiments. The system embodiments described below are merely illustrative.

[0183] The system for publishing idle product objects provided in this embodiment includes: a server side, a client side.

[0184] The server side can be a server side deployed on a cloud server or a server dedicated to providing services for publishing idle product objects, and can be deployed in a data center. The server can be a cluster server or a single server.

[0185] The client includes, but is not limited to, mobile communication devices, that is, commonly known as mobile phones or smart phones, and also includes personal computers, PADs, iPads and other clients.

[0186] Please refer to Figure 2 , which is a schematic diagram of the scenario of the idle commodity object publishing system of the present application. The server and the client can be connected through a network. For example, the client can be connected to the network through WIFI and other means. In this embodiment, the user opens the commodity publishing page of the idle trading platform through the client and wants to publish the information of the idle commodity in his hand to the platform to sell the idle commodity. The idle trading platform mainly serves ordinary consumers. In order to encourage such non-professional seller users to publish more idle commodities, compared with the traditional e-commerce platform, the platform usually greatly simplifies the operation process of publishing idle commodities to reduce the difficulty of publishing idle commodity information. Therefore, the commodity publishing page provided by the idle trading platform usually only includes two input boxes: the commodity title and the commodity picture. For the traditional e-commerce platform facing professional sellers, in order to provide comprehensive commodity information (commodity details) to buyer users, the commodity publishing page provided by the platform not only includes input boxes for the commodity title (commodity name) and commodity picture, but also includes input boxes for various attribute information of the commodity (such as brand, size, etc.). At the same time, since the users of the idle trading platform are not professional sellers, when publishing idle commodity information, they usually do not describe or describe unclearly what commodity is being published in the commodity title. By using the system provided in the embodiment of the present application, when the user publishes an idle commodity, the system can perform real-time drop-down association according to the text of the commodity title edited by the user during the publication, and associate the structured attribute information of the commodity, that is, realize the function that the system prompts what information to input while the user inputs the commodity title, so as to help non-professional sellers input a more comprehensive idle commodity title with commodity information.

[0187] For example, User A has an idle iPhone X to sell through the idle trading platform. After the seller enters "iPhone X" in the product title input box on the idle product listing page, the seller may not think of entering various product attribute information such as the color and capacity of the phone, which is beneficial for helping potential buyers of the idle product to retrieve the product. However, the system can, in real time, associate the input of the product's attribute information after the user enters "iPhone X", such as providing optional color attributes like space gray, silver, red, etc., or providing optional capacity attributes like 64G, 128G, etc. In this way, the seller can select product attribute information such as color and capacity based on the structured attribute prompt information associated with the product automatically provided by the system, helping the seller to enter a product title with more comprehensive product attribute information (product features). At the same time, the idle trading platform can also understand the product structure of the idle product. In this way, when other buyers enter "iPhone X, 64G, and red" in the platform search box, they can quickly search for the idle product and recommend the idle product to interested buyers, avoiding the inability to retrieve the product due to missing attribute information. In this way, for the seller, the exposure rate of the idle product can be increased, prompting the product to be sold as soon as possible; for the buyer, they can quickly buy the idle product they are interested in and have more products to choose from, thus improving the user experience. At the same time, for the idle trading platform, more idle circulation can be facilitated to conclude transactions, thereby improving the utilization rate of platform resources, etc.

[0188] Please refer to Figure 3 , which is a schematic diagram of the device interaction of the idle product object publishing system of the present application. In this embodiment, the server is used to determine the association relationship set between the idle product title query term and the product attribute information according to the product title library; for the product title input association request sent by the client, according to the association relationship set, determine the candidate product attribute information associated with the product title sub-text edited by the user, and send the candidate product attribute information back to the client; for the product publishing request sent by the client, publish the idle product object at least according to the idle product title carried by the product publishing request; the client is used to send the input association request, display the candidate product attribute information; according to the candidate product attribute information, determine the idle product title, where the idle product title includes the product title sub-text and the product attribute information determined based on the candidate product attribute information; and send the product publishing request.

[0189] The product title library includes multiple product titles. The product title may include the product name of the product object published on the traditional e-commerce platform, which usually includes a variety of key attribute information that buyers and users are concerned about, such as category name, brand, model, size, capacity, etc. For example, the product name of product object A on Taobao is "Apple / Apple#iPhone8Plus China Telecom, China Mobile, China Unicom 8p Full Netcom 4G Mobile Phone Official Authentic iPhone8plus", where the key attribute information includes: brand "Apple / Apple", category name "mobile phone", sub-category name "iPhone", model "8Plus", operator network "China Telecom, China Mobile, China Unicom 8p Full Netcom 4G", quality "official authentic". The product title may also include product titles that have been published on idle trading platforms, including product titles that provide a more comprehensive description of the characteristics of idle products.

[0190] The association relationship set between the idle product title query words and the product attribute information includes a large number of association relationships between the idle product title query words and the product attribute information. In one association relationship, the idle product title query words and the product attribute information have a high correlation, the two usually appear in pairs, and the product attribute information is the subsequent adjacent content of the idle product title query words. One idle product title query word can be associated with multiple product attribute information.

[0191] Please refer to Figure 4 , which is a schematic diagram of input association prompt information retrieval in an embodiment of the idle commodity object publishing system of the present application. Figure 4 As can be seen from the above table, the product attribute associated with the query word "iPhone" can be the product model, and the product model attribute value can be X, 11, etc.; the product attribute associated with the query word "iPhone X" can be capacity, and the capacity attribute value can be 64G or 128G, etc.; the product attribute associated with the query word "Huawei" can be the product model, and the product model attribute value can be Mate or P30, etc.

[0192] In one example, the server forms a product title library based on a large number of product names stored in a traditional e-commerce platform. The server can be specifically used to extract product title query words from product titles according to query word determination rules; determine N-gram phrases prefixed with product title query words according to the product titles through the N-Gram model; and use the words after the product title query words in the N-gram phrases as product attribute information to form an association relationship between the product title query words and the product attribute information.

[0193] By applying the query word determination rule, the product title query words can be extracted from the product title. For example, the query word determination rule may include: the product category name is the query word, the category modifier is the query word, the product brand is the query word, the material is the query word, the style is the query word, the style element is the query word, the color is the query word, the function effect is the query word, the size specification is the query word, the quality color is the query word, etc. For example, the product name of the product object A on Taobao is "Apple / Apple#iPhone 8Plus China Telecom, China Mobile, China Unicom 8p Full Netcom 4G Mobile Phone Official Authentic iPhone8plus", and the following product title query words can be extracted from the product title: "Apple" and "Apple" from the brand, "Mobile Phone" from the category name, "iPhone" from the sub-category name, "Full Netcom" from the operator network element, etc.

[0194] After the server extracts the product title query word from the product title according to the query word determination rule, it can determine the N-gram phrase prefixed with the product title query word according to the product title through the N-Gram model. N-Gram is an algorithm based on a statistical language model. The server performs a sliding window operation of size N on the content of each product title in the product title library according to bytes, forming a byte segment sequence of length N with the product title query word as the prefix word. Each byte segment is called a Gram, and the occurrence frequency of all Grams can be counted, and filtered according to a pre-set threshold to form a key Gram list. The model is based on the assumption that the appearance of the Nth word is only related to the previous N-1 words, but not to any other words, and the probability of the whole sentence is the product of the probability of occurrence of each word. These probabilities can be obtained by directly counting the number of times N words appear at the same time from the corpus. Commonly used are the binary Bi-Gram and the ternary Tri-Gram. Therefore, the words following the query words in the product title in each Gram in the list can be used as product attribute information, thereby forming an association relationship between the product title query words and the product attribute information.

[0195] For example, the product name of product object A on Taobao is "Apple / Apple #iPhone 8Plus, compatible with China Telecom, China Mobile, and China Unicom, 8p, 4G mobile phone of the whole network, official genuine iPhone8plus". Using the 3-Gram model, the following triples can be determined: Apple iPhone, Apple iPhone 8, iPhone 8 Plus, 4G mobile phone of the whole network, iPhone 8 plus. In each triple, the prefix word is the query term of the product title, and the suffix word is the product attribute information. For example, in the triple "iPhone 8 Plus", iPhone is the query term, and 8 Plus is the product attribute information. According to other product titles, the product attribute information associated with the query term "iPhone" can also be X, 12, etc. Another example is that in the triple "4G mobile phone of the whole network", the whole network is the query term, and 4G is the product attribute information. According to other product titles, the product attribute information associated with the query term "the whole network" can also be 5G, etc.

[0196] The query term of the product title and the product attribute information belong to different product attributes. For example, in the triple "iPhone 8 Plus", iPhone is the category name attribute, and 8 Plus is the model attribute. The query term of the product title and the product attribute information belong to the same product attribute. The query term of the product title includes: the prefix word of the product attribute; the product attribute information includes: the suffix word of the product attribute. For example, in the triple "4G mobile phone of the whole network", the whole network is the prefix word of the network attribute, and 4G is the suffix word of the network attribute.

[0197] In specific implementation, the server can determine the association relationship set between the query term of the product title and the product attribute information in an offline manner according to the product title library.

[0198] In practical applications, the online real-time requirement is that the response time for returning the input association result is in milliseconds (such as 50ms). Therefore, in this embodiment, an offline index construction and online query method is adopted. In this embodiment, the server can also be used to construct an inverted index of the query term of the product title according to the association relationship set; according to the inverted index, determine the candidate product attribute information associated with the sub-text of the product title edited by the user. The inverted index is derived from the need to search for records according to the query term of the product title in practical applications. Each item in this index table includes a query term of the product title and the addresses of each record with this query term. Since it is not the record that determines the query term, but the query term that determines the position of the record, the product attribute information associated with the target query term can be determined. By adopting this processing method, the retrieval speed of the target query term can be improved, thereby improving the input association speed, further improving the release efficiency of idle product objects, and further improving the exposure rate of idle products.

[0199] After determining the set of association relationships between the product title query terms and the product attribute information, the server can provide the client user with an online service for input association based on the product title. In this embodiment, the server receives a product title input association request sent by the user through the client, determines candidate product attribute information associated with the sub-text of the product title edited by the user according to the set of association relationships, and sends the candidate product attribute information back to the client.

[0200] The product title input association request may include the content of the product title that the seller user is editing in real time in the product title input box. Since this content may not be complete and the user has not finished entering it, for the convenience of description, this application refers to the entered content of the product title as the sub-text of the product title. The product title input association request includes at least the sub-text of the product title. Figure 2 It can be seen that the sub-text of the product title can be "iPhone X".

[0201] The server determines candidate product attribute information associated with the sub-text of the product title edited by the user according to the set of association relationships, and sends the candidate product attribute information back to the client.

[0202] The candidate product attribute information may include multiple values of one attribute as optional items for the user to select. The candidate product attribute information may also include multiple values of multiple attributes, which mainly depends on the specific situation of the set of association relationships generated from the product title library. Figure 2 It can be seen that for the "iPhone X" entered by the user, the server can generate a query statement based on this word, retrieve the product attribute information corresponding to this word, which may include dark gray, red, etc.

[0203] In an example, the server may be specifically used to determine a target query term according to the sub-text of the product title; determine at least one product attribute information associated with the target query term according to the set of association relationships as candidate product attribute information; this processing method enables the server to perform an online query on the suffix entered by the user in the online query part, and obtain relevant candidate attribute information (candidate words) through full-text matching. This is a text-based processing method that can accurately match the user input for association.

[0204] The target query term is a query term determined from the product title text being edited by the user. Specifically in implementation, the server may be specifically used to use the suffix term as the target query term if the suffix term of the sub-text of the product title conforms to the query term determination rule. For example, when the user enters "iPhone given by others", "iPhone" can be used as the target query term. The query term determination rule may be the same as the above query term determination rule, which will not be elaborated here.

[0205] In specific implementation, the server can be specifically used to determine a product title query term that matches the target query term text according to the set of association relationships; and use the product attribute information corresponding to the matched product title query term as candidate product attribute information.

[0206] In another example, the server can also be used to determine a first text vector of a product title query term; and can be specifically used to determine a second text vector of a sub-text of the product title; determine the vector distance between the first text vector and the second text vector; and determine candidate product attribute information according to the vector distance.

[0207] In this embodiment, during the offline phase, the server determines the first text vector of each product title query term in the set of association relationships; during the online query phase, it determines the second text vector of the sub-text of the product title; determines the vector distance between the first text vector and the second text vector; and determines candidate product attribute information according to the vector distance. By adopting this processing method, it is possible to comprehensively understand the product title input by the seller, and the resulting input association results are more accurate. For example, the input association prompt information given by the system for the user input of "Qixia apples" includes: 5A quality, etc., and the input association prompt information given for the user input of "iPhone apples" includes: X, 8plus, etc. In addition, this processing method also makes the input association results more comprehensive, avoiding the inability to find query terms with similar semantics due to full-text matching based on text; therefore, it can effectively improve the user experience.

[0208] The text vector, which includes converting the symbolic representation of text into a vector representation in the semantic space, can be used to quantitatively compare semantics. Such methods are usually based on Harris's distributional hypothesis, that is, words in similar contexts usually have similar semantics.

[0209] In specific implementation, the text vector of the query term can be determined through a pre-trained word vector calculation model (such as BERT). Since the word vector calculation model (Word2vec) belongs to a relatively mature existing technology, it will not be elaborated here.

[0210] The vector distance can be the Euclidean distance between the first text vector and the second text vector, etc.

[0211] As Figure 5 shown, in specific implementation, the server can be specifically used to use the product title query term with a vector distance less than the vector distance threshold as candidate product attribute information. The vector distance threshold can be determined according to application requirements. The larger this threshold is, the more associations are returned, but the lower the relevance; the smaller this threshold is, the fewer associations are returned, but the higher the relevance.

[0212] In specific implementation, the indexed content can be vectorized by a pre-trained model (such as BERT) for offline indexing. When querying, a vector retrieval engine can be used to obtain the N commodity title query terms that are closest and ranked at the front. The candidate terms (commodity attribute information) associated with these query terms are sorted in descending order of importance scores, and the top M candidate commodity attribute information is retrieved and returned to the user.

[0213] In this embodiment, the server can also be used to determine the importance scores of the respective commodity attribute information associated with the commodity title query terms; according to the importance scores, determine the sorting order of the candidate commodity attribute information. By adopting this processing method, the input hint information with a higher probability is displayed in the front position; therefore, the user experience can be effectively improved.

[0214] In specific implementation, the server can specifically be used to, for each commodity attribute information associated with the commodity title query terms, determine the number of times the commodity attribute information is associated with the commodity title query terms according to the commodity title library; according to the number of times of association, determine the importance score. For example, if the number of occurrences of "iPhone X" is greater than that of "iPhone 8plus", it may indicate that there are more users selling "iPhone X". At this time, when the user inputs "iPhone", X can be displayed in front of 8plus.

[0215] After the user finishes entering the idle commodity title through the client, the commodity release request can be sent to the server through the client. The request includes at least the idle commodity title. The idle commodity title can include the commodity title sub-text and the commodity attribute information determined based on the candidate commodity attribute information. The server responds to the commodity release request, generates the idle commodity object, and publishes the object to the idle trading platform for seller users to search, view, place orders for purchase, etc.

[0216] In an example, the server can also be used to determine the multiple commodity attribute information included in the commodity title; according to the multiple commodity attribute information, structurally store the idle commodity object, such as storing the idle commodity object in a relational database.

[0217] The idle commodity object publishing system provided by the embodiments of the present application determines the association relationship set between the query terms of the idle commodity titles and the commodity attribute information through the server according to the commodity title library; for the commodity title input association request sent by the client, according to the association relationship set, determines the candidate commodity attribute information associated with the sub-text of the commodity title edited by the user, and sends the candidate commodity attribute information back to the client; for the commodity publishing request sent by the client, publishes the idle commodity object at least according to the idle commodity title carried in the commodity publishing request; sends the input association request through the client to display the candidate commodity attribute information; determines the idle commodity title according to the candidate commodity attribute information, and the idle commodity title includes the sub-text of the commodity title and the commodity attribute information determined based on the candidate commodity attribute information; sends the commodity publishing request; this processing method enables the input prompt information based on the idle commodity title to obtain both the structured attributes of the idle commodity object (such as brand, model), and constructs the association relationship between the attributes (such as the association between iPhone and 64G); therefore, it can effectively improve the relevance of the input association of the idle commodity title. In addition, this processing method also enables the structure to have additional association capabilities and can supplement the commodity attribute information, so it can effectively reduce the difficulty of the structured processing of the commodity attributes based on the idle commodity title. At the same time, this processing method of completing the input of the idle commodity title by prompting the candidate words of the commodity attributes enables the idle commodity to be better published, so it can effectively improve the quality of the idle commodity publishing. In this way, for the seller user, it can increase the exposure rate of the idle commodity and prompt the commodity to be sold as soon as possible; for the buyer user, it can buy the idle commodity they are interested in as soon as possible and have more commodities to choose from, so it can improve the user experience; for the idle trading platform, it can more effectively promote the idle circulation to reach a transaction, thereby improving the utilization rate of the platform resources, etc.

[0218] Although the present application is disclosed above in preferred embodiments, it is not intended to limit the present application. Any person skilled in the art can make possible changes and modifications without departing from the spirit and scope of the present application. Therefore, the protection scope of the present application should be defined by the scope defined in the claims of the present application.

[0219] In a typical configuration, a computing device includes one or more processors (CPUs), an input / output interface, a network interface, and memory.

[0220] The memory may include non-permanent memory in the computer-readable medium, random access memory (RAM), and / or non-volatile memory in the form of, for example, read-only memory (ROM) or flash memory (flash RAM). The memory is an example of a computer-readable medium.

[0221] 1. A computer-readable medium includes permanent and non-permanent, removable and non-removable media that can implement information storage by any method or technology. The information can be computer-readable instructions, data structures, program modules, or other data. Examples of computer storage media include, but are not limited to, phase change memory (PRAM), static random access memory (SRAM), dynamic random access memory (DRAM), other types of random access memory (RAM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), flash memory or other memory technologies, compact disc read-only memory (CD-ROM), digital versatile disc (DVD) or other optical storage, magnetic cassettes, magnetic tape magnetic disk storage or other magnetic storage devices, or any other non-transitory medium that can be used to store information accessible by a computing device. As defined herein, a computer-readable medium does not include transitory media such as modulated data signals and carrier waves.

[0222] 2. Those skilled in the art should understand that the embodiments of the present application can be provided as a method, a system, or a computer program product. Therefore, the present application can take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware aspects. Moreover, the present application can take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk memory, CD-ROM, optical memory, etc.) containing computer-usable program code.

Claims

1. An input method, characterized in that, Including: Obtain the input of the user, where the input includes a sub - text of the product title; According to the input, obtain at least one product attribute information that matches the input from the product description set as the prompt information; the corresponding relationship between the query word and the product attribute information is stored in the product description set, and the corresponding relationship is determined according to the product title library; Provide the prompt information to the user to determine the product title according to the sub - text and the prompt information.

2. The method according to claim 1, wherein, The corresponding relationship between the vector representation of the query word and the prompt information is also stored in the product description set.

3. The method according to claim 1, characterized in that, The corresponding relationship is determined by the following steps: Extract the query word from the product title according to the query word determination rule; Through the language model, determine the N - gram phrase with the query word as the prefix according to the product title; Use the word after the query word in the N - gram phrase as the product attribute information to form the corresponding relationship.

4. The method according to claim 1, characterized in that Also including: Determine the idle product title determined based on the at least one product attribute information; Structurally store the idle product object according to the product attribute information included in the idle product title.

5. The method according to claim 1, wherein The query word and the product attribute information belong to different product attributes; Or, the query word and the product attribute information belong to the same product attribute, the query word includes: the prefix word of the product attribute; the product attribute information includes: the suffix word of the product attribute.

6. The method according to claim 1, characterized in that Obtaining the prompt information that matches the input in the product description set according to the input includes: If the suffix word of the input conforms to the query word determination rule, use the suffix word as the target query word; Obtain the prompt information that matches the target query word according to the corresponding relationship.

7. The method according to claim 1, characterized in that Also including: Determine the importance score of the product attribute information; Determine the sorting order of the at least one product attribute information according to the importance score.

8. The method according to claim 7, wherein The importance score is determined by the following steps: Determine the number of times the product attribute information is associated with the query word according to the product description set; Determine the importance score according to the number of times of association.

9. The method according to claim 1, characterized in that Also including: Construct an inverted index of the query word according to the corresponding relationship; Determine the prompt information that matches the input according to the inverted index.

10. The method according to any one of claims 1-9, wherein, Obtaining the user input includes: Obtain the text information input by the user; And / or, obtain the voice information input by the user; And / or, obtain the picture information input by the user; And / or, obtain the video information input by the user.

11. The method according to claim 10, wherein, The method also includes: Obtain the vector representation of the input by passing the input of the user through a pre - trained model; Correspondingly, obtaining the prompt information that matches the input in the product description set according to the input includes: Obtain the prompt information that matches the input in the product description set according to the vector representation of the input.

12. The method according to claim 11, wherein, Obtaining the prompt information that matches the input in the product description set includes: Calculate the distance between the vector representation of the input and the vector representation of the query word stored in the product description set; Select the hint information corresponding to the vector representation of the query word with a distance less than a preset threshold as the hint information matching the input.

13. An input device, characterized in that, Including: An input information acquisition unit for obtaining a user input, where the input includes a sub-text of a product title; A hint information determination unit for obtaining at least one product attribute information matching the input from a product description set as hint information, where the product attribute information does not include an attribute name; a corresponding relationship between a query word and the hint information is stored in the product description set, and the corresponding relationship is determined according to a product title library; A hint information providing unit for providing the hint information to the user to determine a product title based on the sub-text and the hint information.

14. An electronic device, characterized in that, Including: A processor and a memory; The memory is used to store a program for implementing the input method. After the device is powered on and runs the program of the method through the processor, the following steps are executed: obtaining a user input, where the input includes a sub-text of a product title; obtaining at least one product attribute information matching the input from a product description set as hint information, where the product attribute information does not include an attribute name; a corresponding relationship between a query word and the hint information is stored in the product description set, and the corresponding relationship is determined according to a product title library; providing the hint information to the user to determine a product title based on the sub-text and the hint information.

Images