Recommendation word determination method and device

Abstract

The invention relates to the field of computers, in particular to a recommendation word determination method and device. The method comprises the steps of obtaining a first text; determining a candidate word set according to the first text; the candidate word set comprises at least one text associated with the semantic feature of the first text; searching n preset texts contained in a commodity knowledge base from the candidate word set; wherein the commodity knowledge base comprises a plurality of preset texts, and the plurality of preset texts are respectively used for describing attributes of commodities in the preset commodity base; a first corresponding relation between the first text and the n preset texts is established, and the first corresponding relation is used for indicating that the n preset texts serve as recommendation words corresponding to the first text when the first text is input into the search interface.

Description

Technical Field

[0001] The present disclosure relates to the field of data processing, and in particular to a method and device for determining a recommendation word. Background Art

[0002] Currently, when using shopping apps, users often use the app's search function to find products of interest. In this search function, after receiving a user's input word into the search interface, the app can determine one or more recommended words based on the input word and display one or more recommended words on the interface for the user to select. When the user selects one of the recommended words, the app can use the recommended word as a search term and output the search results. In this way, users can browse the search results to find products of interest.

[0003] However, in the prior art, when using application-provided recommended terms as search terms, the products found often do not meet the user's needs, or even no products are found at all, which seriously affects the user experience. Therefore, how to determine the recommended terms for the input terms so that users can more efficiently and quickly find the products of interest when searching using the recommended terms is a problem that needs to be solved. Summary of the Invention

[0004] In order to solve the above technical problems, the present disclosure provides a method and device for determining a recommendation word.

[0005] In a first aspect, the present disclosure provides a method for determining recommended words, comprising: obtaining a first text; determining a set of candidate words based on the first text; the set of candidate words including at least one text associated with the semantic features of the first text; searching for n preset texts contained in a product knowledge base from the set of candidate words; wherein the product knowledge base includes multiple preset texts, and the multiple preset texts are respectively used to describe the attributes of products in a preset product library; establishing a first correspondence between the first text and the n preset texts, the first correspondence being used to indicate that when the first text is entered into a search interface, the n preset texts are used as recommended words corresponding to the first text.

[0006] In this method, a product knowledge base is constructed, which may include multiple preset texts, each of which describes the attributes of a product in the preset product library. For example, assuming a shopping platform includes 10,000 products for sale, texts describing the attributes of these 10,000 products (e.g., attributes such as brand, product name, product model, and color) can be collected to form the product knowledge base. Furthermore, when determining a recommendation term corresponding to a first text, at least one text associated with the semantic features of the first text (i.e., a set of candidate terms) can be determined based on the first text. Furthermore, n preset texts included in the product knowledge base can be searched from the set of candidate terms. These n preset texts can then be used as the recommendation term corresponding to the first text. This makes it easier for users to find products of interest when searching using these recommended terms because the n preset texts are associated with the semantic features of the first text. Furthermore, because the n preset texts describe the attributes of products in the preset product library, users can at least find the products described by the recommended terms when searching using these recommended terms, avoiding the situation where no products are found. Especially when there are not many types of goods on sale (for example, the types of goods on sale are less than 100,000), the number of preset texts in the product knowledge base is limited. Therefore, the method provided by the embodiment of the present disclosure can more efficiently and quickly determine the recommended words corresponding to the text.

[0007] In some implementations, the method further includes: obtaining historical search data; the historical search data includes multiple historical search records, and each historical search record includes: the search terms used in a search behavior and the behavior result information; the behavior result information is used to indicate whether the search is successful; based on the behavior result information in each historical search record, determining the first historical search record in which the search was unsuccessful from the historical search data; obtaining the first text, including: obtaining the first text as the search term used in the first historical search record.

[0008] In some implementations, the behavior result information is specifically used to indicate: whether there are any related products that the user is interested in for the current search behavior; and the first historical search record in which the search was unsuccessful, including: the first historical search record in which there are no related products.

[0009] In some implementations, the method further includes: determining a second historical search record that meets a first preset condition from historical search data; the first preset condition includes: the search is successful, and the search time interval with the first historical search record is less than a time threshold, and the user corresponding to the first historical search record is the same; determining a set of candidate words based on the first text, including: inputting the second historical search record and the first text into a recommendation word generation model to obtain a set of candidate words output by the recommendation word generation model; wherein the recommendation word generation model is used to generate text associated with the semantic features of the first text based on the second historical search record.

[0010] In some implementations, the method further includes: determining k products that meet a second preset condition from a preset product library based on n preset texts; wherein the second preset condition includes: being associated with at least one preset text among the n preset texts; searching for p preset texts used to describe the attributes of the k products from a set of candidate words; establishing a second correspondence between the first text and the p preset texts, the second correspondence being used to use the p preset texts as recommended words corresponding to the first text when the first text is entered into a search interface.

[0011] In some implementations, establishing a first correspondence between a first text and n preset texts includes: storing the first correspondence between the first text and n preset texts in a recommended vocabulary library; the recommended vocabulary library is used to store the correspondence between input words in the input search interface and recommended words.

[0012] In some implementations, the method further includes: obtaining an exception message; the exception message is used to indicate a search behavior in which no product is found, and the exception message also includes a target recommendation word used in the search behavior; and according to the exception message, deleting the target recommendation word from the recommendation word library.

[0013] In a second aspect, a device for determining a recommended word is provided, comprising: an acquisition unit for acquiring a first text; a processing unit for determining a set of candidate words based on the first text; the set of candidate words includes at least one text associated with the semantic features of the first text; the processing unit is further used to search for n preset texts contained in a product knowledge base from the set of candidate words; wherein the product knowledge base includes multiple preset texts, and the multiple preset texts are respectively used to describe the attributes of the products in the preset product library; the processing unit is further used to establish a first correspondence between the first text and the n preset texts, and the first correspondence is used to indicate that when the first text is entered into a search interface, the n preset texts are used as recommended words corresponding to the first text.

[0014] In some implementations, the acquisition unit is further used to acquire historical search data; the historical search data includes multiple historical search records, and each historical search record includes: the search term used in a search behavior and the behavior result information; the behavior result information is used to indicate whether the search is successful; the processing unit is further used to determine the first historical search record in which the search was unsuccessful from the historical search data based on the behavior result information in each historical search record; the acquisition unit is used to acquire the first text, including: an acquisition unit 401, used to acquire the first text as the search term used in the first historical search record.

[0015] In some implementations, the behavior result information is specifically used to indicate: whether there are any related products that the user is interested in for the current search behavior; and the first historical search record in which the search was unsuccessful, including: the first historical search record in which there are no related products.

[0016] In some implementations, the processing unit is further used to determine a second historical search record that meets a first preset condition from the historical search data; the first preset condition includes: the search is successful, and the search time interval with the first historical search record is less than a time threshold, and the user corresponding to the first historical search record is the same; the processing unit is used to determine a set of candidate words based on the first text, including: a processing unit is used to input the second historical search record and the first text into a recommendation word generation model to obtain a set of candidate words output by the recommendation word generation model; wherein the recommendation word generation model is used to generate text associated with the semantic features of the first text based on the second historical search record.

[0017] In some implementations, the processing unit is further used to determine, from a preset product library, k products that meet a second preset condition based on n preset texts; wherein the second preset condition includes: being associated with at least one preset text among the n preset texts; the processing unit is further used to search for p preset texts for describing the attributes of the k products from a set of candidate words; the processing unit is further used to establish a second correspondence between the first text and the p preset texts, and the second correspondence is used to use the p preset texts as recommended words corresponding to the first text when the first text is entered into the search interface.

[0018] In some implementations, the processing unit is also used to establish a first correspondence between the first text and n preset texts, including: the processing unit is also used to store the first correspondence between the first text and n preset texts in a recommended vocabulary library; the recommended vocabulary library is used to store the correspondence between the input words in the input search interface and the recommended words.

[0019] In some implementations, the acquisition unit is further used to obtain an exception message; the exception message is used to indicate a search behavior in which no product was found, and the exception message also includes the target recommendation word used in the search behavior; the processing unit is further used to delete the target recommendation word from the recommendation word library based on the exception message.

[0020] In a third aspect, an electronic device is provided, comprising: a memory and a processor, wherein the memory is used to store a computer program and the processor is used to enable the electronic device to implement the method as described in the first aspect or any one of the implementation methods of the first aspect when executing the computer program.

[0021] In a fourth aspect, a computer-readable storage medium is provided, on which a computer program is stored. When the computer program is executed by a computing device, the computing device implements the method as described in the first aspect or any one of the implementation methods in the first aspect.

[0022] In a fifth aspect, a computer program product is provided. When the computer program product is run on a computer, the computer implements the method as described in the first aspect or any one of the implementations of the first aspect.

[0023] The technical solution provided by the embodiments of the present disclosure has the following advantages over the prior art:

[0024] In this method, a product knowledge base is constructed, which may include multiple preset texts, each of which describes the attributes of a product in the preset product library. For example, assuming a shopping platform includes 10,000 products for sale, texts describing the attributes of these 10,000 products (e.g., attributes such as brand, product name, product model, and color) can be collected to form the product knowledge base. Furthermore, when determining a recommendation term corresponding to a first text, at least one text associated with the semantic features of the first text (i.e., a set of candidate terms) can be determined based on the first text. Furthermore, n preset texts included in the product knowledge base can be searched from the set of candidate terms. These n preset texts can then be used as the recommendation term corresponding to the first text. This makes it easier for users to find products of interest when searching using these recommended terms because the n preset texts are associated with the semantic features of the first text. Furthermore, because the n preset texts describe the attributes of products in the preset product library, users can at least find the products described by the recommended terms when searching using these recommended terms, avoiding the situation where no products are found. Especially when there are not many types of goods on sale (for example, the types of goods on sale are less than 100,000), the number of preset texts in the product knowledge base is limited. Therefore, the method provided by the embodiment of the present disclosure can more efficiently and quickly determine the recommended words corresponding to the text. BRIEF DESCRIPTION OF THE DRAWINGS

[0025] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present disclosure and, together with the description, serve to explain the principles of the present disclosure.

[0026] In order to more clearly illustrate the embodiments of the present disclosure or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0027] Figure 1 A schematic diagram of a search interface provided in an embodiment of the present disclosure;

[0028] Figure 2 This is a schematic diagram of a device for determining a recommendation word according to an embodiment of the present disclosure;

[0029] Figure 3 This is a flowchart of a method for determining a recommendation word provided by an embodiment of the present disclosure;

[0030] Figure 4 A second flowchart of a method for determining a recommendation word provided in an embodiment of the present disclosure;

[0031] Figure 5 A third flow chart of a method for determining a recommendation word provided in an embodiment of the present disclosure;

[0032] Figure 6 A fourth flowchart of a method for determining a recommendation word provided in an embodiment of the present disclosure;

[0033] Figure 7 A fifth flowchart of a method for determining a recommendation word provided in an embodiment of the present disclosure;

[0034] Figure 8 A sixth flow chart of a method for determining a recommendation word provided in an embodiment of the present disclosure;

[0035] Figure 9 This is a second structural diagram of a device for determining a recommendation word provided by an embodiment of the present disclosure;

[0036] Figure 10 A schematic structural diagram of an electronic device provided in an embodiment of the present disclosure. DETAILED DESCRIPTION

[0037] In order to more clearly understand the above-mentioned objectives, features and advantages of the present disclosure, the scheme of the present disclosure will be further described below. It should be noted that the embodiments of the present disclosure and the features therein can be combined with each other in the absence of conflict.

[0038] In the following description, many specific details are set forth to facilitate a full understanding of the present disclosure, but the present disclosure may also be implemented in other ways different from those described herein; it is obvious that the embodiments in the specification are only part of the embodiments of the present disclosure, rather than all of the embodiments.

[0039] First, the relevant terms in the embodiments of the present disclosure are introduced:

[0040] Input words refer to the text entered into the search interface through user operations in the search function. For example Figure 1 In the search interface 10 shown in (a), the text "copper" can be input into the search bar 101 through user operation. The text "copper" input into the search bar 101 through user operation is the input word.

[0041] Recommended words refer to one or more texts displayed in the search interface for users to select when the input word is entered into the search interface in the search function. When the user selects one of the texts, the user can search based on the text. For example Figure 1 As shown in (b), when the user enters the text "copper" into the search bar 101, multiple texts (i.e., "copper wire," "copper pipe," etc.) are displayed in the display area 103 for the user to select. When the user selects a text, they can search based on the text. The text displayed in the display area 103 is a recommended word.

[0042] Search terms refer to the text used as the basis for search in the search function. In actual application, the search terms can be the above input terms or the above recommended terms. For example Figure 1 In (a), when the user enters the text "copper" into the search bar 101 and clicks the "search" control 102, "copper" can be used as the search basis to obtain the corresponding search results and display the search results in the interface. At this time, the input word "copper" becomes the search word for this search behavior. For example Figure 1 In (b), when the user selects a recommended word "copper wire" in the display area 103, "copper wire" can be used as a search basis to obtain corresponding search results and display the search results in the interface. At this time, the recommended word "copper wire" becomes the search word for this search behavior.

[0043] The present disclosure will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0044] In the embodiments of the present disclosure, it is considered that a product knowledge base can be constructed, wherein the product knowledge base can include multiple preset texts, each of which is used to describe the attributes of the products in the preset product library. For example, if a shopping platform includes 10,000 products for sale, the texts describing the attributes of these 10,000 products (for example, attributes may include brand, product name, product model, color, etc.) can be collected to form a product knowledge base.

[0045] Furthermore, when determining the recommendation words corresponding to a text (hereinafter referred to as the first text), on the one hand, at least one text associated with the semantic features of the first text (hereinafter referred to as the candidate word set) can be determined based on the first text; on the other hand, n preset texts contained in the product knowledge base can be searched from the candidate word set. These n preset texts can then be used as the recommendation words corresponding to the first text. In this way, on the one hand, since the n preset texts are texts associated with the semantic features of the first text, when the user uses these recommendation words to search, it is easier to search for the products of interest; on the other hand, since the n preset texts are used to describe the attributes of the products in the preset product library, when the user uses these recommendation words to search, at least the products described by the recommendation words can be searched, avoiding the situation where no products can be found. Especially when there are not many types of products on sale (for example, the types of products on sale are less than 100,000), the number of preset texts in the product knowledge base is limited. Therefore, the method provided by the embodiment of the present disclosure can more efficiently and quickly determine the recommendation words corresponding to the texts.

[0046] The specific implementation process of the technical solution provided by the embodiment of the present disclosure is described in detail below with reference to examples:

[0047] First, an embodiment of the present disclosure provides a method for determining a recommendation word. The execution subject of the method for determining a recommendation word may be a device for determining a recommendation word. When the device for determining a recommendation word is running, it can be used to execute all or part of the steps in the method for determining a recommendation word provided by the embodiment of the present disclosure. In actual application, the functions of the device for determining a recommendation word may be implemented by a personal computer (including desktop computers, laptop computers, handheld computers, and notebook computers, etc.), or electronic devices such as smart phones and servers; or, the functions of the device for determining a recommendation word may also be implemented by some hardware / software devices in the electronic devices. The embodiment of the present disclosure does not impose any special restrictions on the specific form of the device for determining a recommendation word.

[0048] For example, Figure 2As shown, it is a structural diagram of a recommendation word determination device 20 provided by an embodiment of the present disclosure. Among them, the recommendation word determination device 20 includes: a user interface layer 201, an application logic layer 202, a data layer 203, an AI and intelligent module 204 and an underlying service 205. Among them, the user interface layer 201 is responsible for interacting with the user, specifically for receiving input words and displaying recommended words and search results. The application logic layer 202 is used to process the input words entered by the user, and perform data point reporting, recommendation word matching and search optimization, etc. The data layer 203 is used to store user behavior data, recommendation word library and product knowledge base, etc. The AI ​​and intelligent module 204 is used to generate candidate words through a recommendation word generation model. The underlying service 205 is used to provide data query and analysis through databases such as Microsoft relational database management service (Microsoft SQL Server, MS SQL) or (React JavaScript Library, React), and perform basic services such as uploading, sharding, integration and cleaning.

[0049] like Figure 3 As shown, the method for determining recommended words provided by the embodiment of the present disclosure may include:

[0050] S301: The recommendation word determination device obtains a first text.

[0051] For example, the first text may be an input word entered into the search interface through a user operation.

[0052] S302: The recommendation word determination device determines a candidate word set based on the first text.

[0053] The candidate word set includes at least one text associated with the semantic feature of the first text.

[0054] For example, the recommendation word determination device can input the first text into a large language model (LLM) so that the semantic features of the first text can be analyzed by the large language model and the user's possible search intention can be predicted, thereby obtaining at least one text output by the large language model that is associated with the semantic features of the first text.

[0055] For example, when the first text is "copper", the candidate word set may include texts such as "copper pipe", "copper wire", "copper flange", "copper angle valve", "copper coin", "socket" and "electric wire".

[0056] S303: The recommendation word determination device searches for n preset texts contained in the product knowledge base from the candidate word set.

[0057] The product knowledge base includes a plurality of preset texts, and the plurality of preset texts are respectively used to describe the attributes of the products in the preset product library.

[0058] For example, when the method provided by the embodiments of the present disclosure is applied in a shopping platform, if the shopping platform includes 10,000 kinds of goods on sale, the texts describing the attributes of these 10,000 kinds of goods (such as brand, product name, product model, color, etc.) can be collected to form a product knowledge base.

[0059] By searching the candidate word set for n preset texts contained in the product knowledge base, texts not in the product knowledge base can be filtered out from the candidate word set. For example, taking the first text "copper" as an example, when the method provided by the embodiment of the present disclosure is applied to a shopping platform selling decoration accessories, texts such as "copper coins" that are not related to the product being sold can be filtered out from the candidate word set in S303.

[0060] S304: The recommendation word determination device establishes a first correspondence between the first text and n preset texts, where the first correspondence is used to indicate that when the first text is input into the search interface, the n preset texts are used as recommendation words corresponding to the first text.

[0061] In some implementations, such as Figure 4 As shown, S304 may include:

[0062] S304a: The recommendation word determination device stores a first correspondence between the first text and n preset texts in a recommendation word library.

[0063] The recommended word library is used to store the correspondence between the input word entered into the search interface and the recommended word. For example, the recommended word library includes multiple texts as the input word and the recommended words corresponding to the multiple texts. The recommended words corresponding to the multiple texts in the recommended word library can be determined according to the above-mentioned method of the embodiment of the present disclosure.

[0064] In this way, when the recommendation word determination device receives the input word entered by the user into the search interface, it can use the recommendation word library to quickly determine the recommended word corresponding to the input word, thereby quickly and accurately providing recommended words that meet the user's search intention during the user search process.

[0065] It is understood that in the above implementation, the first correspondence between the first text and the n preset texts is established by storing the first text and the n preset texts in the recommended word library. In actual application, the first correspondence between the first text and the n preset texts can also be established by other methods.

[0066] For example, when the recommendation word determination device is running on an electronic device used by a user, the first text can be the input word currently entered by the user into the search bar. Furthermore, the recommendation word determination device can search the candidate word set for n preset texts contained in the product knowledge base according to the above-mentioned process S301-S303, and then the recommendation word determination device can directly display the n preset texts as recommendation words on the interface. At this time, the process of the recommendation word determination device displaying the n preset texts as recommendation words on the interface in response to the user inputting the first text can also be understood as an implementation method of establishing a first correspondence between the first text and the n preset texts.

[0067] In some implementations, such as Figure 5 As shown, the method may further include:

[0068] S305: The recommendation word determination device obtains historical search data.

[0069] The historical search data includes multiple historical search records, each of which includes: the search terms used in a search behavior and behavior result information. The behavior result information is used to indicate whether the search is successful.

[0070] In some designs, the behavior result information is specifically used to indicate whether there are any related products that the user is interested in for this search behavior. A successful search can be understood as: the user is interested in the products in this search behavior; an unsuccessful search can be understood as: the user is not interested in the products in this search behavior.

[0071] For example, when the user collects, adds products to the shopping cart, or places an order to purchase the product during this search behavior, it can be determined that there are products that the user is interested in during this search behavior, and thus the search is successful; conversely, when the user does not collect, add products to the shopping cart, or place an order to purchase the product during this search behavior, it can be determined that there are no products that the user is interested in during this search behavior, and thus the search is unsuccessful.

[0072] In some other designs, the action result information is specifically used to indicate whether the search results corresponding to the current search action include at least one product. A successful search can be understood as: the search results corresponding to the current search action include at least one product; an unsuccessful search can be understood as: the search results corresponding to the current search action do not include any products.

[0073] For example, when the search term used in this search behavior is a typo or pinyin, after the user clicks the search control, the search results may not contain any products, and it can be determined that the search was unsuccessful; conversely, if the search results include at least one product, the search is determined to be successful.

[0074] S306: The recommendation word determination device determines the first unsuccessful search record from the historical search data based on the behavior result information in each historical search record.

[0075] Specifically, in some designs, when a successful search means that there are products that the user is interested in in this search behavior, and when a failed search means that there are no products that the user is interested in in this search behavior, the first historical search record may be a historical search record in which there are no products that the user is interested in in this search behavior.

[0076] In other designs, when the search is successful and the search results corresponding to the current search behavior include at least one product, and the search is unsuccessful and the search results corresponding to the current search behavior do not include any product, the first historical search record can be a historical search record in which the search results corresponding to the current search behavior do not include any product.

[0077] Furthermore, S301 specifically includes:

[0078] S301a: The recommendation word determination device obtains a first text that is a search word used in a first historical search record (hereinafter referred to as search word X).

[0079] In the above implementation, it is considered that the user's historical search history can be collected, and the search term X used in the first historical search record in which the search was unsuccessful can be determined from the historical search history. Then, based on the contents of S301-S304 described above in the embodiment of the present disclosure, a recommended term corresponding to the search term X can be determined. In this way, when the user enters the search term X again, the recommended term corresponding to the search term X can be displayed to the user for selection, thereby guiding the user to search for the product of interest and reducing the possibility of further search failure.

[0080] Especially when the search term X is a misspelling, pinyin or text in other languages, the above implementation method can determine the recommended words corresponding to the misspelling, pinyin or text in other languages, thereby increasing the probability of successful search by the user.

[0081] In some designs, such as Figure 6 As shown, the method may further include:

[0082] S307: The recommendation word determination device determines a second historical search record that meets the first preset condition from the historical search data.

[0083] The first preset condition includes: the search is successful, the search time interval between the first historical search record and the search record is less than a time threshold, and the user corresponding to the first historical search record is the same.

[0084] Furthermore, the above S302 may include:

[0085] S302a: The recommendation word determination device inputs the second historical search record and the first text into a recommendation word generation model to obtain a candidate word set output by the recommendation word generation model.

[0086] The recommendation word generation model is used to generate text associated with the semantic features of the first text based on the second historical search record. For example, the recommendation word generation model can be a large language model.

[0087] For example, a user first searches for a product of interest using search term A but fails to find it (this search corresponds to the first historical search record), and then searches for a product of interest using search term B (this search corresponds to the second historical search record). The second historical search record and the first text can then be input into the recommendation word generation model, which can then use the second historical search record to analyze the user's search intent, thereby generating text that is associated with the semantic features of the first text and better meets the user's search intent.

[0088] In some implementations, when a recommended word library is used to store the correspondence between an input word in a search interface and a recommended word, such as Figure 7 As shown, the method may further include:

[0089] S308: The recommendation word determination device obtains an abnormal message.

[0090] Among them, the abnormal message is used to indicate the search behavior in which no product is found, and the abnormal message also includes the target recommendation words used in the search behavior.

[0091] For example, in actual applications, historical search data can be periodically acquired. This historical search data includes multiple historical search records generated within a period. These multiple historical search records can then be screened to identify those corresponding to search behaviors that did not result in a product search. Furthermore, an exception message can be generated based on these historical search records.

[0092] The target recommended term used in a search is specifically the recommended term displayed to the user and selected as the search term during the current search. For example, if a user enters the word "copper" and multiple recommended terms such as "copper wire" and "copper tube" are displayed, and the user selects "copper wire" as the search term, "copper wire" will be the recommended term used in the current search.

[0093] S309: The recommendation word determination device deletes the target recommendation word from the recommendation word library according to the abnormal message.

[0094] In this implementation method, it is taken into consideration that: when a user uses a target recommendation word in a search behavior and no product is found in this search, it usually means that the product corresponding to the target recommendation word has been removed from the shelf or there is an error in the target recommendation word. Therefore, the target recommendation word can be deleted from the recommendation word library through the above implementation method, thereby avoiding recommending the target recommendation word to the user again in subsequent searches.

[0095] In some implementations, after searching for n preset texts contained in the product knowledge base from the candidate word set (ie, S303), as shown in FIG. Figure 8 As shown, the method may further include:

[0096] S310 , the recommendation word determination device determines k commodities that meet a second preset condition from a preset commodity library based on n preset texts.

[0097] The second preset condition includes: being associated with at least one preset text among the n preset texts.

[0098] For example, if n preset texts include text a, text b, and text c, you can search the preset product library using text a, text b, and text c, respectively. Assuming the search yields products t1, t2, t3, t4, and t5, then products t1, t2, t3, t4, and t5 are associated with at least one of text a, text b, and text c.

[0099] S311 . The recommendation word determination device searches for p preset texts for describing the attributes of k products from the candidate word set.

[0100] For example, assuming that k items include item t1, item t2, item t3, item t4, and item t5, then the preset text corresponding to item t1 is searched from the candidate word set (referred to as p1 preset texts), the preset text corresponding to item t2 is searched from the candidate word set (referred to as p2 preset texts), the preset text corresponding to item t3 is searched from the candidate word set (referred to as p3 preset texts), the preset text corresponding to item t4 is searched from the candidate word set (referred to as p4 preset texts), and the preset text corresponding to item t5 is searched from the candidate word set (referred to as p5 preset texts). Furthermore, the set of p1 preset texts, p2 preset texts, p3 preset texts, p4 preset texts, and p5 preset texts can be taken as p preset texts.

[0101] S312. The recommendation word determination device establishes a second correspondence between the first text and p preset texts. The second correspondence is used to use the p preset texts as recommendation words corresponding to the first text when the first text is input into the search interface.

[0102] In the above implementation, it is taken into consideration that when determining the recommendation words for the first text, the recommendation words corresponding to the products corresponding to the above n preset texts can be used as the recommendation words for the first text, so that the range of recommendation words provided to users is wider, thereby improving search efficiency.

[0103] Based on the same inventive concept, as an implementation of the above method, the embodiment of the present application also provides a recommendation word determination device, which corresponds to the above method embodiment. For ease of reading, this embodiment will no longer describe the details of the above method embodiment one by one, but it should be clear that the recommendation word determination device in this embodiment can correspond to all or part of the steps performed by the recommendation word determination device in the above method embodiment. Figure 9 As shown, the recommendation word determination device 40 includes:

[0104] An acquiring unit 401 is configured to acquire a first text;

[0105] Processing unit 402 is configured to determine a candidate word set based on the first text; the candidate word set includes at least one text associated with the semantic feature of the first text;

[0106] The processing unit 402 is further configured to search for n preset texts contained in a product knowledge base from the candidate word set; wherein the product knowledge base includes multiple preset texts, each of which is used to describe the attributes of a product in the preset product base;

[0107] The processing unit 402 is further configured to establish a first correspondence between the first text and n preset texts, wherein the first correspondence indicates that when the first text is input into the search interface, the n preset texts are used as recommended words corresponding to the first text.

[0108] In some implementations, the acquisition unit 401 is further configured to acquire historical search data; the historical search data includes multiple historical search records, each of which includes: a search term used in a search behavior and behavior result information; the behavior result information is used to indicate whether the search is successful;

[0109] The processing unit 402 is further configured to determine a first unsuccessful search record from the historical search data based on the behavior result information in each historical search record;

[0110] The acquisition unit 401 is configured to acquire a first text, including: the acquisition unit 401 is configured to acquire a first text as a search term used in a first historical search record.

[0111] In some implementations, the behavior result information is specifically used to indicate: whether there are any related products that the user is interested in for the current search behavior; and the first historical search record in which the search was unsuccessful, including: the first historical search record in which there are no related products.

[0112] In some implementations, the processing unit 402 is further configured to determine, from the historical search data, a second historical search record that satisfies a first preset condition; the first preset condition including: the search is successful, the search time interval between the second historical search record and the first historical search record is less than a time threshold, and the user corresponding to the second historical search record is the same;

[0113] The processing unit 402 is configured to determine a candidate word set based on the first text, including:

[0114] Processing unit 402 is used to input the second historical search record and the first text into the recommendation word generation model to obtain a candidate word set output by the recommendation word generation model; wherein the recommendation word generation model is used to generate text associated with the semantic features of the first text based on the second historical search record.

[0115] In some implementations, the processing unit 402 is further configured to determine, from the preset commodity library, k commodities that meet a second preset condition based on the n preset texts; wherein the second preset condition includes: being associated with at least one preset text in the n preset texts;

[0116] The processing unit 402 is further configured to search the candidate word set for p preset texts for describing the attributes of the k products;

[0117] The processing unit 402 is further configured to establish a second correspondence between the first text and p preset texts, wherein the second correspondence is used to use the p preset texts as recommended words corresponding to the first text when the first text is input into the search interface.

[0118] In some implementations, the processing unit 402 is further configured to establish a first correspondence between the first text and n preset texts, including:

[0119] The processing unit 402 is further configured to store the first correspondence between the first text and the n preset texts in a recommendation word library; the recommendation word library is configured to store the correspondence between an input word in the search interface and a recommendation word.

[0120] In some implementations, the acquisition unit 401 is further configured to acquire an exception message; the exception message is used to indicate a search behavior in which no product was found, and the exception message also includes a target recommendation word used in the search behavior;

[0121] The processing unit 402 is further configured to delete the target recommended word from the recommended word library according to the abnormal message.

[0122] The recommendation word determination device 40 provided in the embodiment of the present application can execute part or all of the steps executed in the method provided in any of the above embodiments. Its implementation principle and technical effect are similar and will not be repeated here.

[0123] Based on the same inventive concept, an embodiment of the present application also provides an electronic device. Figure 10 A schematic diagram of the structure of an electronic device provided in an embodiment of the present application, such as Figure 10 As shown, the electronic device provided in this embodiment includes: a memory 801 and a processor 502, the memory 501 is used to store computer programs, and the processor 502 is used to execute part or all of the steps executed by the method recommendation word determination device provided in the above embodiment when executing the computer program.

[0124] Based on the same inventive concept, an embodiment of the present application also provides a computer-readable storage medium, which stores a computer program. When the computer program is executed by a processor, the computing device implements part or all of the steps performed by the recommendation word determination device in the method provided in the above embodiment.

[0125] Based on the same inventive concept, an embodiment of the present application also provides a computer program product. When the computer program product runs on a computer, it enables the computing device to implement some or all of the steps performed by the recommendation word determination device in the method provided in the above embodiment.

[0126] Those skilled in the art will appreciate that the embodiments of the present application may be provided as methods, systems, or computer program products. Therefore, the present application may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present application may take the form of a computer program product implemented on one or more computer-usable storage media containing computer-usable program code.

[0127] The processor may be a central processing unit (CPU), other general-purpose processors, digital signal processors (DSP), application-specific integrated circuits (ASIC), field-programmable gate arrays (FPGA), other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc.

[0128] The memory may include non-permanent memory in a computer-readable medium, random access memory (RAM) and / or non-volatile memory in the form of read-only memory (ROM) or flash RAM. The memory is an example of a computer-readable medium.

[0129] Computer-readable media includes both permanent and non-permanent, removable and non-removable storage media. Storage media can implement any method or technology for storing information, which 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 disk storage or other magnetic storage devices, or any other non-transmission media that can be used to store information that can be accessed by a computing device. As defined herein, computer-readable media does not include transitory computer-readable media, such as modulated data signals and carrier waves.

[0130] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the above embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the above embodiments, or make equivalent replacements for some or all of the technical features therein, and these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for determining a recommendation word, characterized in that: The method comprises: Get the first text; Determining a candidate word set based on the first text; the candidate word set includes at least one text associated with the semantic feature of the first text; Searching for n preset texts contained in a product knowledge base from the candidate word set; wherein the product knowledge base includes a plurality of preset texts, each of which is used to describe the attributes of a product in the preset product base; A first correspondence between the first text and the n preset texts is established, where the first correspondence is used to indicate that when the first text is input into a search interface, the n preset texts are used as recommended words corresponding to the first text.

2. The method according to claim 1, characterized in that The method further comprises: Obtaining historical search data; the historical search data includes multiple historical search records, each of which includes: a search term used in a search behavior and behavior result information; the behavior result information is used to indicate whether the search is successful; Determining a first unsuccessful search record from the historical search data based on the behavior result information in each of the historical search records; The obtaining of the first text includes obtaining the first text as a search term used in the first historical search record.

3. The method according to claim 2, characterized in that The behavior result information is specifically used to indicate whether there are any related products that the user is interested in during the current search behavior; and the first historical search record in which the search was unsuccessful includes: the first historical search record in which there are no related products.

4. The method according to claim 2, characterized in that The method further comprises: Determining, from the historical search data, a second historical search record that meets a first preset condition; the first preset condition including: a search is successful, a search time interval with the first historical search record is less than a time threshold, and the user corresponding to the first historical search record is the same; The step of determining a candidate word set according to the first text includes: The second historical search record and the first text are input into a recommendation word generation model to obtain the candidate word set output by the recommendation word generation model; wherein the recommendation word generation model is used to generate text associated with the semantic features of the first text based on the second historical search record.

5. The method according to any one of claims 1 to 4, characterized in that The method further comprises: Determining, from the preset commodity library, k commodities that meet a second preset condition based on the n preset texts; wherein the second preset condition includes: being associated with at least one preset text among the n preset texts; Searching for p preset texts for describing the attributes of the k items from the candidate word set; A second correspondence between the first text and the p preset texts is established, and the second correspondence is used to use the p preset texts as recommended words corresponding to the first text when the first text is input into a search interface.

6. The method according to any one of claims 1 to 4, characterized in that The establishing of a first correspondence between the first text and the n preset texts includes: The first correspondence between the first text and the n preset texts is stored in a recommendation word library; the recommendation word library is used to store the correspondence between the input word of the input search interface and the recommended word.

7. The method according to claim 6, characterized in that The method further comprises: Obtaining an exception message; the exception message is used to indicate a search behavior in which no product was found, and the exception message also includes a target recommendation word used in the search behavior; According to the abnormal message, the target recommended word is deleted from the recommended word library.

8. A device for determining a recommended word, characterized in that: include: An acquiring unit, configured to acquire a first text; a processing unit, configured to determine a candidate word set based on the first text; the candidate word set including at least one text associated with a semantic feature of the first text; The processing unit is further configured to search, from the candidate word set, n preset texts contained in a product knowledge base; wherein the product knowledge base includes a plurality of preset texts, each of which is used to describe the attributes of a product in the preset product base; The processing unit is further configured to establish a first correspondence between the first text and the n preset texts, wherein the first correspondence indicates that when the first text is input into a search interface, the n preset texts are used as recommended words corresponding to the first text.

9. An electronic device, characterized in that: include: A memory and a processor, wherein the memory is used to store a computer program, and the processor is used to enable the electronic device to implement the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium, characterized in that The computer-readable storage medium stores a computer program, which, when executed by a computing device, enables the computing device to implement the method according to any one of claims 1 to 7.

Images