A search method, device, apparatus and storage medium

Abstract

Embodiments of the present application provide a search method, device and equipment and storage medium, which can be applied to various scenes such as cloud technology, artificial intelligence, intelligent transportation and vehicle-mounted devices. The method comprises: in response to a search operation triggered in a search interface for an original search condition, displaying a search recommendation interface, the search recommendation interface comprising: a target search intent and at least one recommended search condition corresponding to the target search intent. In response to a scroll operation triggered in the search recommendation interface for the target search intent, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface, that is, in the search process, the target object is guided to perform a scroll operation, so that the search recommendation interface displays more other recommended search conditions associated with the target search intent, so that the target object can obtain the recommended search condition most matched with the target search intent to perform a search, thereby improving the accuracy and efficiency of the search and greatly improving the search experience of the target object.

Description

Technical Field

[0001] This application relates to the field of artificial intelligence technology, and in particular to a search method, apparatus, device, and storage medium. Background Technology

[0002] With the development of internet technology, the amount of information on the internet is constantly increasing. In order to facilitate the target audience to quickly obtain the information they need from the massive amount of information, information search function has become an indispensable part of many applications.

[0003] When conducting a search, the relevant technology first segments the search terms input by the target object to obtain keywords from the search terms, and then performs a search based on the obtained keywords to obtain the search results corresponding to the search terms.

[0004] However, in the above scheme, when the target object cannot accurately describe the content to be searched, the search terms entered by the target object will be biased, resulting in lower accuracy of the output search results. Summary of the Invention

[0005] This application provides a search method, apparatus, device, and storage medium to improve the accuracy of search results.

[0006] On one hand, embodiments of this application provide a search method, the method comprising:

[0007] In response to a search operation triggered by the original search criteria in the search interface, a search recommendation interface is displayed, the search recommendation interface including: the target search intent corresponding to the original search criteria, and at least one recommended search criterion corresponding to the target search intent;

[0008] In response to a scrolling operation triggered in the search recommendation interface in response to the target search intent, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface.

[0009] In response to the search condition selection operation triggered by the various recommended search conditions displayed in the search recommendation interface, the target search results corresponding to the selected target recommended search condition are displayed.

[0010] On one hand, embodiments of this application provide a search device, which includes:

[0011] The recommendation module is used to display a search recommendation interface in response to a search operation triggered based on the original search conditions in the search interface. The search recommendation interface includes: the target search intent corresponding to the original search conditions, and at least one recommended search condition corresponding to the target search intent.

[0012] The recommendation module is also configured to, in response to a scrolling operation triggered in the search recommendation interface in response to the target search intent, display other recommended search conditions corresponding to the target search intent in the search recommendation interface;

[0013] The search module is used to respond to the search condition selection operation triggered by the various recommended search conditions displayed in the search recommendation interface, and to display the target search results corresponding to the selected target recommended search conditions.

[0014] Optionally, the recommendation module is specifically used for:

[0015] In the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the order they correspond to their respective recommendation search conditions.

[0016] Optionally, the search module is specifically used for:

[0017] In response to a search condition selection operation triggered by the at least one recommended search condition in the search recommendation interface, a search results interface is displayed, the search results interface including: target search results corresponding to the target search condition selected from the at least one recommended search condition.

[0018] Optionally, the search module is specifically used for:

[0019] In response to a search condition selection operation triggered by the other recommended search conditions in the search recommendation interface, a search results interface is displayed, which includes: the target search results corresponding to the target search condition selected from the other recommended search conditions.

[0020] Optionally, the search recommendation interface further includes: the original search conditions, and a connecting line between the original search conditions and the target search intent, wherein the connecting line is used to characterize the association between the original search conditions and the target search intent.

[0021] Optionally, the recommendation module is further configured to:

[0022] Before displaying the search recommendation interface, based on a preset knowledge graph and the original search conditions, the corresponding target search intent is determined, and at least one recommended search condition corresponding to the target search intent is obtained.

[0023] Optionally, the recommendation module is specifically used for:

[0024] Based on the original search criteria, query the preset knowledge graph;

[0025] When it is determined that there is a reference search condition in the knowledge graph that matches the original search condition, the reference search intent corresponding to the reference search condition in the knowledge graph is taken as the target search intent, and at least one associated search condition corresponding to the reference search intent is taken as the at least one recommended search condition.

[0026] Optionally, at least one recommended search condition in the search recommendation interface is displayed according to the recommendation order corresponding to the at least one recommended search condition;

[0027] The recommendation module is also used for:

[0028] Before displaying the search recommendation interface, obtain the recommendation order corresponding to the at least one recommended search condition from the knowledge graph; or,

[0029] Based on the first historical search records of the target object associated with the original search conditions, the search preference features of the target object are determined, and based on the search preference features, the at least one recommended search condition is sorted to obtain the recommendation order corresponding to the at least one recommended search condition.

[0030] Optionally, the recommendation module is further configured to:

[0031] Before obtaining the original search criteria, the following steps are performed for each of the multiple reference search criteria: the search intent is predicted for a reference search criterion using a target recommendation model to obtain the corresponding reference search intent;

[0032] A knowledge graph is constructed based on the multiple reference search conditions, the multiple reference search intentions obtained, and at least one associated search condition corresponding to each of the multiple reference search intentions.

[0033] Optionally, the recommendation module is specifically used for:

[0034] For each of the multiple reference search intents, the following steps are performed: based on the second historical search record associated with a reference search intent, sort at least one associated search condition corresponding to the reference search intent to obtain the recommendation order corresponding to the at least one associated search condition;

[0035] The knowledge graph is constructed based on the multiple reference search conditions, the multiple reference search intentions obtained, at least one associated search condition corresponding to each of the multiple reference search intentions, and the corresponding recommendation order.

[0036] Optionally, the recommendation module is further configured to:

[0037] Obtain multiple sample search conditions and corresponding labeled search intents;

[0038] Using the multiple sample search conditions and corresponding labeled search intentions, the recommendation model to be trained is iteratively trained until the iteration stopping condition is met to obtain the target recommendation model. Each iteration process includes the following steps:

[0039] Based on the sample search conditions, determine the corresponding predicted search intent;

[0040] Based on the predicted search intent and the labeled search intent, a target loss value is determined, and the parameters of the recommendation model to be trained are adjusted using the target loss value.

[0041] On one hand, embodiments of this application provide a computer program including a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the steps of the search method described above.

[0042] On one hand, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the search method described above.

[0043] On one hand, embodiments of this application provide a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer device, cause the computer device to perform the search method steps described above.

[0044] In this embodiment, in response to a search operation triggered by the original search criteria in the search interface, the target search intent corresponding to the original search criteria and at least one recommended search condition corresponding to the target search intent are displayed to the target object. Compared with the original search criteria, the recommended search conditions are more closely matched to the target object's actual search intent. Therefore, when searching based on the recommended search conditions, the accuracy of the search results can be effectively improved. Secondly, in response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. That is, during the search process, the target object is guided to scroll to display more other recommended search conditions related to the target search intent, providing the target object with more selectable recommended search conditions. This allows the target object to obtain the target recommended search condition that best matches the target search intent from among many recommended search conditions, thereby improving the accuracy and efficiency of the search and greatly enhancing the target object's search experience. Attached Figure Description

[0045] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0046] Figure 1 A schematic diagram of a system architecture provided in an embodiment of this application;

[0047] Figure 2 A schematic diagram of a search interface provided in an embodiment of this application;

[0048] Figure 3 A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 1 ;

[0049] Figure 4 A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 2 ;

[0050] Figure 5 A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 3 ;

[0051] Figure 6 A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 4 ;

[0052] Figure 7 A schematic diagram of a search results interface provided in an embodiment of this application;

[0053] Figure 8 A flowchart illustrating a search method provided in this application embodiment. Figure 1 ;

[0054] Figure 9 A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 5 ;

[0055] Figure 10a A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 6 ;

[0056] Figure 10b A schematic diagram of a search recommendation interface provided in an embodiment of this application. Figure 7 ;

[0057] Figure 11 A schematic diagram of the structure of a knowledge graph provided in this application embodiment. Figure 1 ;

[0058] Figure 12 A flowchart illustrating a method for obtaining a second historical search record provided in an embodiment of this application;

[0059] Figure 13 A schematic diagram of the structure of a knowledge graph provided in this application embodiment. Figure 2 ;

[0060] Figure 14 A flowchart illustrating a search method provided in this application embodiment. Figure 2 ;

[0061] Figure 15 A schematic diagram of the structure of a search device provided in an embodiment of this application;

[0062] Figure 16 This is a schematic diagram of the structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0063] To make the objectives, technical solutions, and beneficial effects of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0064] For ease of understanding, the terms used in the embodiments of this invention are explained below.

[0065] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0066] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0067] Machine Learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instructional learning. For example, in the embodiments of this application, machine learning techniques are used to predict the reference search intent corresponding to each of multiple reference search conditions, and based on the multiple reference search conditions, the obtained multiple reference search intents, and at least one associated search condition corresponding to each of the multiple reference search intents, a knowledge graph is constructed.

[0068] Data annotation: Data annotation is the process of labeling metadata such as text, video, and images. The labeled data will be used to train machine learning models.

[0069] Semantic vectors: Converting the symbolic representation of text into a vector representation in a semantic space is a common practice for quantifying and comparing semantics. These methods are usually based on Harris's distributed hypothesis, which states that words in similar contexts usually have similar semantics.

[0070] Data augmentation is a technique that artificially expands a training dataset by generating more equivalent data from a limited dataset.

[0071] Long short-term memory (LSTM) networks are a special type of recurrent neural network (RNN) that can learn long-term dependency information.

[0072] Deep Neural Networks (DNNs) are a framework for deep learning, consisting of a neural network with at least one hidden layer. Similar to shallow neural networks, deep neural networks can also model complex nonlinear systems, but the additional layers provide a higher level of abstraction, thus enhancing the model's capabilities.

[0073] Knowledge graph: A semantic network in which nodes represent entities or concepts, and edges represent various semantic relationships between entities / concepts.

[0074] Search intent: The "need" behind a specific search query.

[0075] The design concept of the embodiments of this application will be introduced below.

[0076] Current technologies for searching first segment the search terms input by the target audience to extract keywords, and then perform a search based on these keywords to obtain the corresponding search results. However, in this approach, when the target audience cannot accurately describe the content being searched, the input search terms will be biased, resulting in lower accuracy of the output search results.

[0077] Analysis revealed that when a target audience is unsure how to clearly describe their search query, even if they enter a search term, they may not be seeking content related to that term, but rather have other search intents. For example, when searching for "weight loss," the target audience might not be looking for content related to "weight loss," but rather content related to "looking thinner." In this case, displaying search terms related to "looking thinner" in the search recommendation interface, while simultaneously guiding the target audience to continuously scroll and update the displayed search terms, allows them to find the most relevant terms to their actual search intent. This effectively improves search accuracy and efficiency, and significantly enhances the search experience.

[0078] In view of this, embodiments of this application provide a search method, the method comprising: displaying a search recommendation interface in response to a search operation triggered on a search interface for an original search condition, wherein the search recommendation interface includes: a target search intent corresponding to the original search condition, and at least one recommended search condition corresponding to the target search intent; displaying other recommended search conditions corresponding to the target search intent on the search recommendation interface in response to a scrolling operation triggered on the search recommendation interface; and then displaying the target search result corresponding to the selected target recommended search condition in response to a search condition selection operation triggered on the search recommendation interface.

[0079] In this embodiment, in response to a search operation triggered by the original search criteria in the search interface, the target search intent corresponding to the original search criteria and at least one recommended search condition corresponding to the target search intent are displayed to the target object. Compared with the original search criteria, the recommended search conditions are more closely matched to the target object's actual search intent. Therefore, when searching based on the recommended search conditions, the accuracy of the search results can be effectively improved. Secondly, in response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. That is, during the search process, the target object is guided to scroll to display more other recommended search conditions related to the target search intent, providing the target object with more selectable recommended search conditions. This allows the target object to obtain the target recommended search condition that best matches the target search intent from among many recommended search conditions, thereby improving the accuracy and efficiency of the search and greatly enhancing the target object's search experience.

[0080] refer to Figure 1 This is a system architecture diagram applicable to the embodiments of this application. The system architecture includes at least a terminal device 101 and a search server 102. The number of terminal devices 101 can be one or more, and the number of search servers 102 can also be one or more. This application does not specifically limit the number of terminal devices 101 and search servers 102.

[0081] The terminal device 101 has a pre-installed target application with search functionality. This target application can be a client application, a web application, a mini-program application, etc. The terminal device 101 can be a smartphone, tablet, laptop, desktop computer, smart home appliance, smart voice interaction device, smart in-vehicle device, etc., but is not limited to these.

[0082] Search server 102 is the backend server of the target application. Search server 102 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms. Terminal device 101 and search server 102 can be directly or indirectly connected via wired or wireless communication; this application does not impose any restrictions on this connection.

[0083] The search method in this application embodiment can be executed by the terminal device 101, the search server 102, or by the interaction between the terminal device 101 and the search server 102.

[0084] Taking the search method in the embodiments of this application executed by terminal device 101 as an example, it includes the following steps:

[0085] In response to a search operation triggered by the original search criteria in the search interface, terminal device 101 displays a search recommendation interface. The search recommendation interface includes: the target search intent corresponding to the original search criteria, and at least one recommended search condition corresponding to the target search intent. In response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed. Subsequently, in response to a search condition selection operation triggered by the displayed recommended search conditions in the search recommendation interface, the target search result corresponding to the selected target recommended search condition is displayed.

[0086] In practical applications, the search method in this application embodiment can be applied to the search of any media content, including but not limited to articles, images, and videos.

[0087] The following example uses article search:

[0088] The terminal device launches the pre-installed search application, which displays the search interface, such as... Figure 2 As shown, the search interface includes a search box and an "OK" button. After entering the original search term "weight loss" in the search box, click the "OK" button.

[0089] The terminal device responds to a search operation triggered by the original search term "weight loss" in the search interface by displaying a search recommendation interface, such as... Figure 3 As shown, the search recommendation interface includes the relationship between the original search term "weight loss" and the target search intent "slimming down". Specifically, the original search term "weight loss" is connected to the recommendation area 301 of the target search intent "slimming down" by a line. The recommendation area 301 represents the "search planet" corresponding to the target search intent "slimming down".

[0090] The terminal device also displays recommended search terms corresponding to the target search intent "slimming" in the search recommendation interface, in the recommended order: "slimming outfit tips", "slimming dress", "slimming pants", etc. Figure 4 As shown, the recommended search terms "slimming outfit tips", "slimming dress", and "slimming pants" are connected to the recommended area 301, which is related to the target search intent "slimming".

[0091] In response to the scrolling operation triggered by recommendation area 301 in the search recommendation interface, the terminal device updates the recommended search terms displayed in the search recommendation interface, namely: "slimming dress", "slimming pants", "slimming makeup tips", etc. Figure 5 As shown, the recommended search terms "slimming dress," "slimming pants," and "slimming makeup tips" are connected to the recommended area 301 for the target search intent "slimming." Due to the limited size of the search recommendation interface, the recommended search term "slimming outfit tips" is hidden. It should be noted that the search recommendation interface can also display all four recommended search terms simultaneously: "slimming outfit tips," "slimming dress," "slimming pants," and "slimming makeup tips." This application does not impose specific limitations on this.

[0092] The target audience clicks on the search recommendation page and sees the target search term "slimming makeup tips," such as... Figure 6 As shown, in response to a click action triggered on the search recommendation interface, the terminal device selects the target search term "slimming makeup tips" from the recommended search terms "slimming dress", "slimming pants", and "slimming makeup tips", and then sends a search request carrying the target search term to the search server.

[0093] The search server retrieves multiple articles related to the target search term "slimming makeup tips" from the search database, and then sends the relevant information from these articles to the terminal device. The terminal device displays the search results interface, such as... Figure 7 As shown, the search results interface includes an article information display area 701 and an article information display area 702. The article information display area 701 includes the article title "Slimming Makeup Tutorial" and the corresponding article thumbnail, while the article information display area 702 includes the article title "5-Minute Slimming Makeup Look" and the corresponding article thumbnail.

[0094] It should be noted that the search method in this application embodiment is not limited to the above-mentioned application scenario, but can also be used in product search scenarios, food delivery search scenarios, merchant information search scenarios, audio and video search scenarios, etc. This application does not make specific limitations in this regard.

[0095] It is understood that in the specific implementation of this application, user-related data such as historical search records and search preference characteristics are involved. When the above embodiments of this application are applied to specific products or technologies, user permission or consent is required, and the collection, use and processing of related data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0096] based on Figure 1 The system architecture diagram shown in this application illustrates the flow of a search method, as provided in this embodiment. Figure 8As shown, the process of this method is executed by a computer device, which can be Figure 1 the terminal device 101 and / or the search server 102 shown in the figure, and includes the following steps:

[0097] Step S801, in response to a search operation triggered for an original search condition in a search interface, display a search recommendation interface.

[0098] Specifically, the search condition can be text (such as a search term), an image (such as a search picture), audio and video (such as a search voice), etc. The original search condition is the search condition input by a target object in an application with a search function, and the target object includes but is not limited to an account number, a device number.

[0099] The search recommendation interface includes: the target search intention corresponding to the original search condition, and at least one recommended search condition corresponding to the target search intention. The target search intention refers to the "requirement" behind a specific search query, and can be represented in forms such as text, image, audio and video. The target search intention itself can also be displayed as a recommended search condition in the search recommendation interface. When the original search condition corresponds to multiple target search intentions, at least one recommended search condition corresponding to each target search intention can be obtained respectively.

[0100] When the original search condition is text, the original search condition may include words that are useless or interfering with the search, such as words like "of", "ah", etc. To improve search efficiency and accuracy, after obtaining the original search condition, the original search condition is segmented, and then keywords are obtained from the segmentation result. Then, based on the obtained keywords, the corresponding target search intention is determined, and at least one recommended search condition corresponding to the target search intention is obtained.

[0101] Optionally, the search recommendation interface further includes: the original search condition and a connection line between the original search condition and the target search intention, and the connection line is used to represent the association relationship between the original search condition and the target search intention.

[0102] Specifically, in the search recommendation interface, the area where the original search condition is located is connected to the area where the target search intention is located in a wired manner. Similarly, in the search recommendation interface, the area where the recommended search condition is located is connected to the area where the target search intention is located in a wired manner, which is used to represent the association relationship between the recommended search condition and the target search intention.

[0103] For example, see Figure 9This is a schematic diagram of a search recommendation interface provided in an embodiment of this application. The search recommendation interface includes the original search term "chicken breast", the target search intent "fitness", and recommended search terms: "fitness exercise tutorial", "classic fitness exercises", and "fitness equipment". The original search term "chicken breast" is connected to the recommendation area 1201 of the target search intent "chicken breast" by a line. At the same time, the recommended search terms "fitness exercise tutorial", "classic fitness exercises", and "fitness equipment" are also connected to the recommendation area 1201 of the target search intent "chicken breast" by lines.

[0104] In this embodiment of the application, the original search conditions are connected to the target search intent by means of a line in the search recommendation interface, so that the target object can clearly understand the relationship between the original search conditions and the target search intent, and then select the appropriate recommended search conditions from the recommended search conditions corresponding to the target search intent to search, thereby improving search efficiency.

[0105] Step S802: In response to the scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface.

[0106] Specifically, because the screen size of a terminal device is limited, while there are many recommended search conditions corresponding to the target search intent, the terminal device cannot display all the recommended search conditions simultaneously. When the search recommendation interface cannot display all the recommended search conditions at the same time, a portion of the recommended search conditions can be displayed first. Then, in response to the scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions that were not previously displayed can be displayed step by step. The scrolling operation can be clockwise, counterclockwise, etc., and all the recommended search conditions corresponding to the target search intent can be displayed sequentially according to the recommendation order.

[0107] When displaying other recommended search criteria in the search recommendation interface, you can simultaneously display previously displayed recommended search criteria and reduce the display size of each recommended search criterion in the search recommendation interface (or you can only reduce the size of previously displayed recommended search criteria) to adapt to the screen size of the terminal device.

[0108] When displaying other recommended search terms in the search recommendation interface, you can also hide previously displayed recommended search terms, so that you don't need to adjust the display size of each recommended search term in the search recommendation interface.

[0109] In some embodiments, the number of recommended search criteria updated in each search recommendation interface is determined based on the scroll angle of the scrolling operation. If, after multiple scrolling operations, the search recommendation interface displays all recommended search criteria, the terminal device can respond to a scrollback operation triggered in the search recommendation interface based on the target search intent, and display the previously displayed recommended search criteria. Alternatively, the terminal device can continue to respond to scrolling operations triggered in the search recommendation interface based on the target search intent, and display each recommended search condition from the beginning.

[0110] For example, the recommended area 1201 for the target search intent "chicken breast" is a scrollable circular area. Figure 9 Based on the search recommendation interface shown, the terminal device responds to the clockwise scrolling operation triggered on the recommendation area 1201 in the search recommendation interface, and updates the recommended search terms displayed in the search recommendation interface, that is, displays the recommended search terms: "classic fitness exercises", "fitness equipment", "fitness meal recipes", such as Figure 10a As shown, "fitness meal recipes" is a newly displayed recommended search term, while the recommended search term "fitness exercise tutorials" is hidden.

[0111] For example, the recommended area 1201 for the target search intent "chicken breast" is a scrollable circular area. Figure 9 Based on the search recommendation interface shown, the terminal device responds to the clockwise scrolling operation triggered on the recommendation area 1201 in the search recommendation interface, and updates the recommended search terms displayed in the search recommendation interface, that is, displays the recommended search terms: "fitness exercise tutorials", "classic fitness exercises", "fitness equipment", "fitness meal recipes", etc. Figure 10b As shown, "fitness meal recipes" is a newly displayed recommended search term.

[0112] Step S803: In response to the search condition selection operation triggered by each of the recommended search conditions displayed in the search recommendation interface, display the target search results corresponding to the selected target recommended search condition.

[0113] In practice, search criteria selection can be performed via click, double-click, long-press, etc. Responding to the search criteria selection action triggered by various recommended search criteria displayed on the search recommendation interface, the system selects the target search criterion from these criteria and then retrieves the target search results matching the target search criterion from the search library. Target search results include, but are not limited to, articles, images, audio, and video. The search library can reside on the terminal device, on a search server, or independently of both.

[0114] In this embodiment, in response to a search operation triggered by the original search criteria in the search interface, the target search intent corresponding to the original search criteria and at least one recommended search condition corresponding to the target search intent are displayed to the target object. Compared with the original search criteria, the recommended search conditions are more closely matched to the target object's actual search intent. Therefore, when searching based on the recommended search conditions, the accuracy of the search results can be effectively improved. Secondly, in response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. That is, during the search process, the target object is guided to scroll to display more other recommended search conditions related to the target search intent, providing the target object with more selectable recommended search conditions. This allows the target object to obtain the target recommended search condition that best matches the target search intent from among many recommended search conditions, thereby improving the accuracy and efficiency of the search and greatly enhancing the target object's search experience.

[0115] One possible implementation involves displaying a search results interface in response to a search condition selection operation triggered for at least one recommended search condition in the search recommendation interface. The search results interface includes: target search results corresponding to the target search condition selected from at least one recommended search condition.

[0116] For example, see Figure 10a This is a schematic diagram of a search recommendation interface provided in an embodiment of this application. When the target object clicks on the recommended search term "classic fitness moves" in the search recommendation interface, the terminal device responds to the search condition selection operation triggered in the search recommendation interface and sends a search request carrying the target search term "classic fitness moves" to the search server.

[0117] The search server retrieves multiple articles related to the target search term "classic fitness moves" from the search database, and then sends the relevant information of these articles to the terminal device. The terminal device then displays this information in the search results interface.

[0118] Another possible implementation involves displaying a search results interface in response to a search condition selection operation triggered by other recommended search conditions in the search recommendation interface. The search results interface includes the target search results corresponding to the target search condition selected from the other recommended search conditions.

[0119] For example, see Figure 10aThis is a schematic diagram of a search recommendation interface provided in an embodiment of this application. "Fitness meal recipes" is a newly displayed recommended search term. When the target object clicks on the recommended search term "Fitness meal recipes" in the search recommendation interface, the terminal device responds to the search condition selection operation triggered in the search recommendation interface and sends a search request carrying the target search term "Fitness meal recipes" to the search server.

[0120] The search server retrieves multiple articles related to the target search term "fitness meal recipes" from the search database, and then sends the relevant information from these articles to the terminal device. The terminal device then displays this information in the search results interface.

[0121] In this embodiment, in response to the scrolling operation triggered by the target object in the search recommendation interface for the target search intent, more recommended search conditions related to the target search intent are displayed. Therefore, the target object can select the target search condition that best matches its own needs from the more recommended search conditions, thereby improving the accuracy of the search results.

[0122] Optionally, before displaying the search recommendation interface, based on a preset knowledge graph and the original search conditions, the corresponding target search intent is determined, and at least one recommended search condition corresponding to the target search intent is obtained.

[0123] Specifically, the knowledge graph can reside on a terminal device, a search server, or be independent of both. Before using a knowledge graph to determine the corresponding target search intent based on the original search conditions and obtain at least one recommended search condition corresponding to the target search intent, the embodiments of this application employ at least the following implementation methods to construct the knowledge graph:

[0124] Implementation Method 1: Obtain multiple reference search conditions. For each reference search condition, perform the following steps: Use a target recommendation model to predict the search intent for one reference search condition to obtain the corresponding reference search intent. Based on the multiple reference search conditions, the obtained multiple reference search intents, and at least one associated search condition corresponding to each of the multiple reference search intents, construct a knowledge graph.

[0125] Specifically, the target recommendation model can be an LSTM model, a DNN model, etc. For each reference search condition, the corresponding reference search intent can be at least one of the other reference search conditions, or it can be a search condition other than the other reference search conditions.

[0126] For each reference search intent, configure at least one related search condition. The related search conditions corresponding to each reference search intent are updated periodically or in real time. Alternatively, for each reference search condition, configure at least one related search condition. The related search conditions corresponding to each reference search condition are also updated periodically or in real time.

[0127] By treating reference search conditions and reference search intents as nodes, and the associated search conditions corresponding to each reference search condition and reference search intent as the attribute information of the nodes, and then connecting the nodes corresponding to multiple reference search conditions and multiple reference search intents based on the association relationship predicted by the target recommendation model, a knowledge graph is obtained.

[0128] For example, such as Figure 11 As shown, multiple reference search conditions are set, including: reference search term 1, reference search term 2, reference search term 3, reference search term 4, and reference search term 5. Among them, reference search term 1 corresponds to related search term A1 and related search term B1; reference search term 2 corresponds to related search term A2 and related search term B2; reference search term 3 corresponds to related search term A3, related search term B3, and related search term C3; reference search term 4 corresponds to related search term A4; and reference search term 5 corresponds to related search term A5, related search term B5, and related search term C5.

[0129] Inputting reference search term 1 into the target recommendation model determines the corresponding reference search intent as reference search term 4. Inputting reference search term 2 into the target recommendation model determines the corresponding reference search intent as reference search term 3. Inputting reference search term 3 into the target recommendation model determines the corresponding reference search intent as reference search term 2. Inputting reference search term 4 into the target recommendation model determines the corresponding reference search intent as reference search term 2. Inputting reference search term 5 into the target recommendation model determines the corresponding reference search intent as reference search term 1.

[0130] Using reference search terms 1, 2, 3, 4, and 5 as nodes in the knowledge graph, and their respective associated search terms as attribute information, the nodes are then connected according to the relationships between the reference search conditions output by the target recommendation model to obtain the knowledge graph.

[0131] In this embodiment, the trained target recommendation model predicts the reference search intent corresponding to each reference search condition, eliminating the need for manual labeling of the reference search intent for each condition, thereby improving the efficiency of knowledge graph construction. Secondly, by constructing a knowledge graph, the target search intent corresponding to the original search conditions input by the target object is determined, improving the understanding of the target object's search intent and thus enhancing the accuracy of search results.

[0132] Implementation Method 2: Obtain multiple reference search conditions. For each of the multiple reference search conditions, perform the following steps: Use a target recommendation model to predict the search intent for a reference search condition and obtain the corresponding reference search intent.

[0133] For multiple reference search intents, perform the following steps respectively: Based on the second historical search record associated with a reference search intent, sort at least one associated search condition corresponding to the reference search intent to obtain the recommendation order corresponding to at least one associated search condition.

[0134] Then, based on multiple reference search conditions, multiple reference search intents obtained, at least one associated search condition corresponding to each of the multiple reference search intents, and the corresponding recommendation order, a knowledge graph is constructed.

[0135] Specifically, the second historical search record includes records of the search process and search results viewed in relation to the reference search intent, such as... Figure 12 As shown, the search process record includes: search data such as search criteria, search time, location information, and account information recorded by the search server during the process of searching for the reference search intent on the terminal device. The search result viewing record includes: behavioral data recorded by the search server during the process of viewing the search results for the reference search intent on the terminal device. Behavioral data includes, but is not limited to, clicking on articles, clicking time, location information, article titles, and article summaries.

[0136] Based on the second historical search record, the probability of each related search condition being searched can be obtained. Then, the related search conditions are sorted in descending order of probability to obtain the recommended order of each related search condition.

[0137] The reference search conditions and reference search intents are used as nodes. The associated search conditions and the recommendation order of the associated search conditions are used as the attribute information of the nodes. Then, based on the association relationship between the reference search conditions and reference search intents predicted by the target recommendation model, the nodes corresponding to multiple reference search conditions and multiple reference search intents are connected to obtain a knowledge graph.

[0138] For example, such as Figure 13As shown, multiple reference search conditions are set, including: reference search term 1, reference search term 2, reference search term 3, reference search term 4, and reference search term 5. Reference search term 1 corresponds to related search terms A1 and B1, and the recommended order of the related search terms is: related search term A1, related search term B1; Reference search term 2 corresponds to related search terms A2 and B2, and the recommended order of the related search terms is: related search term B2, related search term A2; Reference search term 3 corresponds to related search terms A3, B3, and C3, and the recommended order of the related search terms is: related search term B3, related search term A3, related search term C3; Reference search term 4 corresponds to related search term A4; Reference search term 5 corresponds to related search terms A5, B5, and C5, and the recommended order of the related search terms is: related search term B5, related search term A5, and related search term C5.

[0139] Inputting reference search term 1 into the target recommendation model determines the corresponding reference search intent as reference search term 4. Inputting reference search term 2 into the target recommendation model determines the corresponding reference search intent as reference search term 3. Inputting reference search term 3 into the target recommendation model determines the corresponding reference search intent as reference search term 2. Inputting reference search term 4 into the target recommendation model determines the corresponding reference search intent as reference search term 2. Inputting reference search term 5 into the target recommendation model determines the corresponding reference search intent as reference search term 1.

[0140] Using reference search terms 1, 2, 3, 4, and 5 as nodes in the knowledge graph, and their respective associated search terms and recommendation order as attribute information, the knowledge graph is constructed. Then, according to the relationships between the reference search conditions output by the target recommendation model, the nodes corresponding to each reference search condition are connected to obtain the knowledge graph.

[0141] In this embodiment, based on the second historical search record associated with the reference search intent, at least one associated search condition corresponding to the reference search intent is sorted to obtain the recommendation order corresponding to at least one associated search condition. The recommendation order is then added to the knowledge graph as the attribute information of the node. Therefore, when recommending associated search conditions to the target object according to the knowledge graph, associated search conditions suitable for the target object can be recommended first, thereby improving the accuracy and efficiency of the search.

[0142] Optionally, before constructing the knowledge graph, the target object model needs to be trained. The training process includes the following steps:

[0143] Obtain multiple sample search conditions and corresponding labeled search intentions. Then, using multiple sample search conditions and corresponding labeled search intentions, iteratively train the recommendation model to be trained until the iteration stopping condition is met to obtain the target recommendation model. Each iteration process includes the following steps:

[0144] Based on the sample search conditions, the corresponding predicted search intent is determined. Then, based on the predicted search intent and the labeled search intent, the target loss value is determined, and the parameters of the recommendation model to be trained are adjusted using the target loss value.

[0145] Specifically, the sample search conditions are pre-labeled to obtain the labeled search intents corresponding to the sample search conditions. If the labeled search intents obtained from the annotation are biased towards the head or concentrated, the sample search conditions can be clustered based on their semantic vectors to reduce the amount of duplicate labeling of similar data. After obtaining a certain number of labeled search intents, the labeled data can be augmented using nearest neighbor or data augmentation methods.

[0146] Model training requires setting several key parameters, the quality of which directly impacts the model's performance. Therefore, in this embodiment, a grid search method is used to obtain the values ​​of these key parameters. Additionally, during model training, regularization coefficients are set to prevent overfitting, thereby improving the model's generalization ability.

[0147] The target loss value is used to characterize the difference between the predicted search intent and the labeled search intent. Model training aims to minimize this difference. The convergence of the recommendation model is determined based on the target loss value. If convergence is not achieved, the model parameters are adjusted based on the target loss value. The adjusted model is then used for the next round of training. When convergence is achieved, training ends, and the trained target recommendation model is output. Alternatively, in this embodiment, training can also end and the trained target recommendation model output after a preset number of iterations; this application does not impose specific limitations on this.

[0148] In this embodiment, the target recommendation model obtained through training predicts the reference search intent corresponding to each reference search condition, without the need for manual labeling of the reference search intent corresponding to each reference search condition, thereby improving the efficiency and accuracy of knowledge graph construction.

[0149] Optionally, after constructing the knowledge graph, embodiments of this application employ at least the following methods to determine the corresponding target search intent based on the original search conditions, and obtain at least one recommended search condition corresponding to the target search intent:

[0150] Based on the original search criteria, a pre-defined knowledge graph is queried. When it is determined that there are reference search criteria in the knowledge graph that match the original search criteria, the reference search intent corresponding to the reference search criteria in the knowledge graph is taken as the target search intent, and at least one related search condition corresponding to the reference search intent is taken as at least one recommended search condition.

[0151] For example, queries based on the original search terms Figure 11 The query results for the knowledge graph shown are as follows: if the original search term matches the reference search term 1 in the knowledge graph, then the reference search intent corresponding to the reference search term 1 in the knowledge graph (i.e., reference search term 4) is taken as the target search intent, and the related search term A4 corresponding to the reference search term 4 is taken as the recommended search term for the target search intent.

[0152] In this embodiment of the application, a knowledge graph is used to map search conditions and corresponding search intentions. Therefore, when the original search conditions input by the target object are received, the corresponding target search intention can be quickly obtained through the knowledge graph, and at least one recommended search condition corresponding to the target search intention can be obtained from the knowledge graph, so that the target object can search based on the recommended search conditions, thereby improving the accuracy and efficiency of the search.

[0153] Optionally, at least one recommended search condition in the search recommendation interface is displayed according to the recommendation order corresponding to the at least one recommended search condition. In the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed according to their respective recommendation order. The method for obtaining the recommendation order corresponding to at least one recommended search condition is the same as the method for obtaining the recommendation order corresponding to other recommended search conditions. The following describes how to obtain the recommendation order corresponding to each of the at least one recommended search condition. This application provides at least the following implementation methods for obtaining the recommendation order corresponding to each of the at least one recommended search condition:

[0154] Implementation Method 1: Obtain the recommendation order corresponding to at least one recommendation search condition from the knowledge graph.

[0155] For example, queries based on the original search terms Figure 13 The query results for the knowledge graph shown are as follows: If the original search term matches reference search term 2 in the knowledge graph, then the reference search intent corresponding to reference search term 2 (i.e., reference search term 3) in the knowledge graph is taken as the target search intent, and the related search terms A3, B3, and C3 corresponding to reference search term 3 are taken as recommended search terms for the target search intent. Simultaneously, the recommendation order for each recommended search term can also be obtained: related search term B3, related search term A3, and related search term C3.

[0156] Implementation Method 2: Based on the first historical search records of the target object associated with the original search conditions, determine the search preference features of the target object, and based on the search preference features, sort at least one recommended search condition to obtain the recommendation order corresponding to at least one recommended search condition.

[0157] Specifically, the first historical search record includes the target object's search process record and search result viewing record. The search process record includes search data such as search terms, search time, location information, and account information recorded by the target object during each search. The search result viewing record includes behavioral data recorded by the target object during each browsing of search results, including but not limited to article clicks, click time, location information, article titles, and article summaries.

[0158] For example, the recommended search terms might include: "slimming outfit tips," "slimming makeup tips," "slimming dresses," and "slimming pants." Based on the target audience's first historical search records, we can determine their search preferences are related to "outfits." Therefore, recommended search terms with a strong relevance to "outfits" will be ranked higher, while those with a weaker relevance will be ranked lower. The resulting recommendation order would be: "slimming outfit tips," "slimming dresses," "slimming pants," and "slimming makeup tips."

[0159] In this embodiment of the application, after determining the target search intent and at least one recommended search condition corresponding to the target search intent, the recommendation order of at least one recommended search condition is obtained from the knowledge graph, or the recommendation order of at least one recommended search condition is determined based on the search preference characteristics of the target object. Subsequently, at least one recommended search condition is displayed in the recommendation order so as to prioritize recommending search conditions that are more suitable for the target object, thereby improving the accuracy and efficiency of the search.

[0160] To better explain the embodiments of this application, a search method provided by the embodiments of this application is described below in conjunction with a specific implementation scenario. The process of this method can be as follows: Figure 1 The terminal device 101 shown interacts with the search server 102 to perform the following steps, such as: Figure 14 As shown:

[0161] In step S1401, the terminal device responds to the search operation triggered in the search interface based on the original search conditions and obtains the original search terms.

[0162] Specifically, the search application displays a search interface, such as Figure 2As shown, the search interface includes a search box and an "OK" button. After entering the original search term "weight loss" in the search box and clicking the "OK" button, the terminal device responds to the search operation triggered in the search application and retrieves the original search term "weight loss".

[0163] In step S1402, the terminal device sends the original search terms to the search server.

[0164] In step S1403, the search server queries the knowledge graph based on the original search terms to obtain the target search intent.

[0165] Specifically, the search server queries the knowledge graph based on the original search term "weight loss" to obtain the target search intent "look thinner".

[0166] In step S1404, the search server obtains all recommended search terms corresponding to the target search intent and the recommendation order of each recommended search term from the knowledge graph.

[0167] Specifically, the search server retrieves recommended search terms from the knowledge graph corresponding to the target search intent "slimming": "slimming outfit tips", "slimming makeup tips", "slimming dress", "slimming pants", and the recommended order of each search term: "slimming outfit tips", "slimming dress", "slimming pants", "slimming makeup tips".

[0168] In step S1405, the search server sends the target search intent, each recommended search term, and the corresponding recommendation order to the terminal device.

[0169] In step S1406, the terminal device displays the relationship between the original search terms and the target search intent in the search recommendation interface, and displays some recommended search terms in the recommended order.

[0170] Specifically, the search recommendation interface is as follows: Figure 4 As shown, this includes the relationship between the original search term "weight loss" and the target search intent "looking slimmer," as well as the recommended search terms displayed in priority: "slimming outfit tips," "slimming dresses," and "slimming pants." The original search term "weight loss" is connected to the recommended area for the target search intent "looking slimmer" via lines. The recommended search terms "slimming outfit tips," "slimming dresses," and "slimming pants" are also connected to the recommended area for the target search intent "looking slimmer" via lines.

[0171] In step S1407, the terminal device updates the displayed recommended search terms in response to the scrolling operation triggered by the target search intent in the search recommendation interface.

[0172] Specifically, in response to a scrolling operation triggered in the recommendation area of ​​the search recommendation interface based on the target search intent, the terminal device updates the recommended search terms displayed on the search recommendation interface, such as: "slimming dress", "slimming pants", "slimming makeup tips", etc. Figure 5 As shown, the recommended search terms "slimming dress", "slimming pants", and "slimming makeup tips" are connected to the recommended area for the target search intent "slimming".

[0173] In step S1408, the terminal device responds to the search condition selection operation triggered in the search recommendation interface by selecting the target search term from the displayed recommended search terms.

[0174] Specifically, in response to a click action triggered on the search recommendation interface, the terminal device selects the target search term "slimming makeup tips" from the recommended search terms "slimming dress", "slimming pants", and "slimming makeup tips".

[0175] In step S1409, the terminal device sends the target search term to the search server.

[0176] In step S1410, the search server retrieves the target search results from the search database based on the target search term.

[0177] In step S1411, the search server sends the target search results to the terminal device.

[0178] Specifically, the search server retrieves multiple articles related to the target search term "slimming makeup tips" from the search database, and then sends the relevant information of the multiple articles to the terminal device.

[0179] In step S1412, the terminal device displays the target search result in the search results interface.

[0180] Specifically, the terminal device displays a search results interface, such as... Figure 7 As shown, the search results interface includes an article information display area 701 and an article information display area 702. The article information display area 701 includes the article title "Slimming Makeup Tutorial" and the corresponding article thumbnail, while the article information display area 702 includes the article title "5-Minute Slimming Makeup Look" and the corresponding article thumbnail.

[0181] In this embodiment, in response to a search operation triggered by the original search criteria in the search interface, the target search intent corresponding to the original search criteria and at least one recommended search condition corresponding to the target search intent are displayed to the target object. Compared with the original search criteria, the recommended search conditions are more closely matched to the target object's actual search intent. Therefore, when searching based on the recommended search conditions, the accuracy of the search results can be effectively improved. Secondly, in response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. That is, during the search process, the target object is guided to scroll to display more other recommended search conditions related to the target search intent, providing the target object with more selectable recommended search conditions. This allows the target object to obtain the target recommended search condition that best matches the target search intent from among many recommended search conditions, thereby improving the accuracy and efficiency of the search and greatly enhancing the target object's search experience.

[0182] Based on the same technical concept, this application provides a schematic diagram of the structure of a search device, such as... Figure 15 As shown, the search device 1500 includes:

[0183] The recommendation module 1501 is used to display a search recommendation interface in response to a search operation triggered based on the original search conditions in the search interface. The search recommendation interface includes: the target search intent corresponding to the original search conditions, and at least one recommended search condition corresponding to the target search intent.

[0184] The recommendation module 1501 is also configured to, in response to a scrolling operation triggered in the search recommendation interface in response to the target search intent, display other recommended search conditions corresponding to the target search intent in the search recommendation interface.

[0185] The search module 1502 is used to respond to the search condition selection operation triggered by each of the recommended search conditions displayed in the search recommendation interface, and to display the target search results corresponding to the selected target recommended search conditions.

[0186] Optionally, the recommendation module 1501 is specifically used for:

[0187] In the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the order they correspond to their respective recommendation search conditions.

[0188] Optionally, the search module 1502 is specifically used for:

[0189] In response to a search condition selection operation triggered by the at least one recommended search condition in the search recommendation interface, a search results interface is displayed, the search results interface including: target search results corresponding to the target search condition selected from the at least one recommended search condition.

[0190] Optionally, the search module 1502 is specifically used for:

[0191] In response to a search condition selection operation triggered by the other recommended search conditions in the search recommendation interface, a search results interface is displayed, which includes: the target search results corresponding to the target search condition selected from the other recommended search conditions.

[0192] Optionally, the search recommendation interface further includes: the original search conditions, and a connecting line between the original search conditions and the target search intent, wherein the connecting line is used to characterize the association between the original search conditions and the target search intent.

[0193] Optionally, the recommendation module 1501 is further configured to:

[0194] Before displaying the search recommendation interface, based on a preset knowledge graph and the original search conditions, the corresponding target search intent is determined, and at least one recommended search condition corresponding to the target search intent is obtained.

[0195] Optionally, the recommendation module 1501 is specifically used for:

[0196] Based on the original search criteria, query the preset knowledge graph;

[0197] When it is determined that there is a reference search condition in the knowledge graph that matches the original search condition, the reference search intent corresponding to the reference search condition in the knowledge graph is taken as the target search intent, and at least one associated search condition corresponding to the reference search intent is taken as the at least one recommended search condition.

[0198] Optionally, at least one recommended search condition in the search recommendation interface is displayed according to the recommendation order corresponding to the at least one recommended search condition;

[0199] The recommendation module 1501 is also used for:

[0200] Before displaying the search recommendation interface, obtain the recommendation order corresponding to the at least one recommended search condition from the knowledge graph; or,

[0201] Based on the first historical search records of the target object associated with the original search conditions, the search preference features of the target object are determined, and based on the search preference features, the at least one recommended search condition is sorted to obtain the recommendation order corresponding to the at least one recommended search condition.

[0202] Optionally, the recommendation module 1501 is further configured to:

[0203] Before obtaining the original search criteria, the following steps are performed for each of the multiple reference search criteria: the search intent is predicted for a reference search criterion using a target recommendation model to obtain the corresponding reference search intent;

[0204] A knowledge graph is constructed based on the multiple reference search conditions, the multiple reference search intentions obtained, and at least one associated search condition corresponding to each of the multiple reference search intentions.

[0205] Optionally, the recommendation module 1501 is specifically used for:

[0206] For each of the multiple reference search intents, the following steps are performed: based on the second historical search record associated with a reference search intent, sort at least one associated search condition corresponding to the reference search intent to obtain the recommendation order corresponding to the at least one associated search condition;

[0207] The knowledge graph is constructed based on the multiple reference search conditions, the multiple reference search intentions obtained, at least one associated search condition corresponding to each of the multiple reference search intentions, and the corresponding recommendation order.

[0208] Optionally, the recommendation module 1501 is further configured to:

[0209] Obtain multiple sample search conditions and corresponding labeled search intents;

[0210] Using the multiple sample search conditions and corresponding labeled search intentions, the recommendation model to be trained is iteratively trained until the iteration stopping condition is met to obtain the target recommendation model. Each iteration process includes the following steps:

[0211] Based on the sample search conditions, determine the corresponding predicted search intent;

[0212] Based on the predicted search intent and the labeled search intent, a target loss value is determined, and the parameters of the recommendation model to be trained are adjusted using the target loss value.

[0213] In this embodiment, in response to a search operation triggered by the original search criteria in the search interface, the target search intent corresponding to the original search criteria and at least one recommended search condition corresponding to the target search intent are displayed to the target object. Compared with the original search criteria, the recommended search conditions are more closely matched to the target object's actual search intent. Therefore, when searching based on the recommended search conditions, the accuracy of the search results can be effectively improved. Secondly, in response to a scrolling operation triggered by the target search intent in the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. That is, during the search process, the target object is guided to scroll to display more other recommended search conditions related to the target search intent, providing the target object with more selectable recommended search conditions. This allows the target object to obtain the target recommended search condition that best matches the target search intent from among many recommended search conditions, thereby improving the accuracy and efficiency of the search and greatly enhancing the target object's search experience.

[0214] Based on the same technical concept, embodiments of this application provide a computer device, which can be... Figure 1 The terminal devices and / or search servers shown, such as Figure 16 As shown, it includes at least one processor 1601 and a memory 1602 connected to at least one processor. In this embodiment, the specific connection medium between the processor 1601 and the memory 1602 is not limited. Figure 16 Taking the connection between the processor 1601 and the memory 1602 via a bus as an example, the bus can be divided into address bus, data bus, control bus, etc.

[0215] In this embodiment of the application, the memory 1602 stores instructions that can be executed by at least one processor 1601. By executing the instructions stored in the memory 1602, at least one processor 1601 can perform the steps of the search method described above.

[0216] The processor 1601 is the control center of the computer device, capable of connecting various parts of the computer device via various interfaces and lines. It performs information retrieval by running or executing instructions stored in the memory 1602 and accessing data stored in the memory 1602. Optionally, the processor 1601 may include one or more processing units. The processor 1601 may integrate an application processor and a modem processor. The application processor primarily handles the operating system, user interface, and applications, while the modem processor primarily handles wireless communication. It is understood that the modem processor may not be integrated into the processor 1601. In some embodiments, the processor 1601 and the memory 1602 may be implemented on the same chip; in other embodiments, they may be implemented on separate chips.

[0217] Processor 1601 can be a general-purpose processor, such as a central processing unit (CPU), digital signal processor, application-specific integrated circuit (ASIC), field-programmable gate array (FPGA), or other programmable logic device, discrete gate or transistor logic device, or discrete hardware component, capable of implementing or executing the methods, steps, and logic block diagrams disclosed in the embodiments of this application. The general-purpose processor can be a microprocessor or any conventional processor. The steps of the methods disclosed in the embodiments of this application can be directly manifested as being executed by a hardware processor, or executed by a combination of hardware and software modules within the processor.

[0218] Memory 1602, as a non-volatile computer-readable storage medium, can be used to store non-volatile software programs, non-volatile computer-executable programs, and modules. Memory 1602 may include at least one type of storage medium, such as flash memory, hard disk, multimedia card, card-type memory, random access memory (RAM), static random access memory (SRAM), programmable read-only memory (PROM), read-only memory (ROM), electrically erasable programmable read-only memory (EEPROM), magnetic storage, magnetic disk, optical disk, etc. Memory 1602 can be any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer device, but is not limited thereto. In the embodiments of this application, memory 1602 can also be a circuit or any other device capable of implementing storage functions for storing program instructions and / or data.

[0219] Based on the same inventive concept, embodiments of this application provide a computer-readable storage medium storing a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the search method described above.

[0220] Based on the same inventive concept, this application provides a computer program product, which includes a computer program stored on a computer-readable storage medium. The computer program includes program instructions, which, when executed by a computer device, cause the computer device to perform the steps of the search method described above.

[0221] Those skilled in the art will understand that embodiments of the present invention can be provided as methods or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0222] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0223] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0224] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0225] Although preferred embodiments of the invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including both the preferred embodiments and all changes and modifications falling within the scope of the invention.

[0226] Obviously, those skilled in the art can make various modifications and variations to this invention without departing from its spirit and scope. Therefore, if these modifications and variations fall within the scope of the claims of this invention and their equivalents, this invention also intends to include these modifications and variations.

Claims

1. A search method, characterized in that, include: In response to a search operation triggered by the original search criteria in the search interface, a search recommendation interface is displayed. The search recommendation interface includes: the target search intent corresponding to the original search criteria, at least one recommended search condition corresponding to the target search intent, a connecting line between the original search criteria and the target search intent, and a connecting line between the recommended search conditions and the target search intent; the target search intent and the at least one recommended search condition are obtained by querying a preset knowledge graph based on the original search criteria. In response to a scrolling operation triggered in the search recommendation interface in response to the target search intent, other recommended search conditions corresponding to the target search intent are displayed in the search recommendation interface. In response to the search condition selection operation triggered by each of the displayed recommended search conditions in the search recommendation interface, the target search results corresponding to the selected target recommended search condition are displayed. The knowledge graph was obtained in the following way: For multiple reference search conditions, the following steps are performed: A target recommendation model is used to predict the search intent of the reference search conditions to obtain the corresponding reference search intent, which includes at least one of the other reference search conditions; the reference search conditions and reference search intents are used as nodes, and the associated search conditions corresponding to each reference search condition and reference search intent are used as the attribute information of the nodes; based on the association between the reference search conditions and reference search intents, the nodes corresponding to the multiple reference search conditions and the nodes corresponding to the obtained multiple reference search intents are connected to obtain a knowledge graph.

2. The method as described in claim 1, characterized in that, The search recommendation interface displays other recommended search conditions corresponding to the target search intent, including: In the search recommendation interface, other recommended search conditions corresponding to the target search intent are displayed in the order they correspond to their respective recommendation search conditions.

3. The method as described in claim 1, characterized in that, The response to the search condition selection operation triggered by the various recommended search conditions displayed in the search recommendation interface, and the display of the target search results corresponding to the selected target recommended search condition, includes: In response to a search condition selection operation triggered by the at least one recommended search condition in the search recommendation interface, a search results interface is displayed, the search results interface including: target search results corresponding to the target search condition selected from the at least one recommended search condition.

4. The method as described in claim 1, characterized in that, The response to the search condition selection operation triggered by the various recommended search conditions displayed in the search recommendation interface, and the display of the target search results corresponding to the selected target recommended search condition, includes: In response to a search condition selection operation triggered by the other recommended search conditions in the search recommendation interface, a search results interface is displayed, which includes: the target search results corresponding to the target search condition selected from the other recommended search conditions.

5. The method as described in claim 1, characterized in that, The connecting line between the original search criteria and the target search intent is used to characterize the association between the original search criteria and the target search intent.

6. The method as described in claim 1, characterized in that, The target search intent and the at least one recommended search condition are obtained in the following manner: Based on the original search criteria, query the preset knowledge graph; When it is determined that there is a reference search condition in the knowledge graph that matches the original search condition, the reference search intent corresponding to the reference search condition in the knowledge graph is taken as the target search intent, and at least one associated search condition corresponding to the reference search intent is taken as the at least one recommended search condition.

7. The method as described in claim 1, characterized in that, At least one recommended search condition in the search recommendation interface is displayed in the order corresponding to the at least one recommended search condition; Before displaying the search recommendation interface, the following are also included: From the knowledge graph, obtain the recommendation order corresponding to the at least one recommendation search condition; or, Based on the first historical search records of the target object associated with the original search conditions, the search preference features of the target object are determined, and based on the search preference features, the at least one recommended search condition is sorted to obtain the recommendation order corresponding to the at least one recommended search condition.

8. The method as described in claim 1, characterized in that, Also includes: For each of the multiple reference search intents, the following steps are performed: based on the second historical search record associated with a reference search intent, at least one associated search condition corresponding to the reference search intent is sorted to obtain the recommendation order corresponding to the at least one associated search condition.

9. The method as described in claim 8, characterized in that, The target recommendation model is trained using the following method: Obtain multiple sample search conditions and corresponding labeled search intents; Using the multiple sample search conditions and corresponding labeled search intentions, the recommendation model to be trained is iteratively trained until the iteration stopping condition is met to obtain the target recommendation model. Each iteration process includes the following steps: Based on the sample search conditions, determine the corresponding predicted search intent; Based on the predicted search intent and the labeled search intent, a target loss value is determined, and the parameters of the recommendation model to be trained are adjusted using the target loss value.

10. A search device, characterized in that, include: The recommendation module is used to respond to a search operation triggered based on the original search conditions in the search interface and display a search recommendation interface. The search recommendation interface includes: the target search intent corresponding to the original search conditions, at least one recommended search condition corresponding to the target search intent, a connection line between the original search conditions and the target search intent, and a connection line between the recommended search conditions and the target search intent. The target search intent and the at least one recommended search condition are obtained by querying a preset knowledge graph based on the original search conditions. The knowledge graph is obtained by: performing the following steps for multiple reference search conditions: using a target recommendation model to predict the search intent of the reference search conditions to obtain corresponding reference search intents, where the reference search intents include at least one of the other reference search conditions; using the reference search conditions and reference search intents as nodes, and the associated search conditions corresponding to each reference search condition and reference search intent as attribute information of the nodes; and connecting the nodes corresponding to the multiple reference search conditions and the nodes corresponding to the obtained multiple reference search intents based on the association relationship between the reference search conditions and reference search intents to obtain the knowledge graph. The recommendation module is also configured to, in response to a scrolling operation triggered in the search recommendation interface in response to the target search intent, display other recommended search conditions corresponding to the target search intent in the search recommendation interface; The search module is used to respond to the search condition selection operation triggered by the various recommended search conditions displayed in the search recommendation interface, and to display the target search results corresponding to the selected target recommended search conditions.

11. A computer device, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the steps of the method according to any one of claims 1 to 9.

12. A computer-readable storage medium, characterized in that, It stores a computer program executable by a computer device, which, when run on the computer device, causes the computer device to perform the steps of the method according to any one of claims 1 to 9.

13. A computer program product, characterized in that, The computer program product includes a computer program stored on a computer-readable storage medium, the computer program including program instructions that, when executed by a computer device, cause the computer device to perform the method steps of any one of claims 1-9.

Images