Information search method and apparatus, electronic device, storage medium, and program product

Abstract

Embodiments of the present application disclose an information search method and apparatus, an electronic device, a storage medium, and a program product. The method comprises: in response to a current search operation, which was triggered by a target subject on the basis of target search information, meeting a preset search condition, displaying first search suggestion information in an information search interface; and, in response to a viewing operation triggered on the basis of the first search suggestion information, displaying a search result interface corresponding to the target search information, the search result interface comprising a current search result, the current search result being generating on the basis of the target search information and context corpus information of the target subject, and the context corpus information comprising at least one of the following: search information input by the target subject in each historical search operation, content browsed by the target subject in each historical search result, and content of feedback behavior implemented by the target subject for browsed content.

Description

Information search method, device, electronic device, storage medium and program product

[0001] This application claims priority to the Chinese patent application filed with the China Patent Office on December 26, 2023, with application number 202311816348.3 and application name “An information search method, device, electronic device and storage medium”. Technical Field

[0002] The present application relates to the field of optoelectronic technology, and in particular to an information search method, device, electronic device, storage medium, and program product.

[0003] Background of the Invention

[0004] With the popularization of computers and the development of the Internet, people use the Internet more and more frequently, and computer networks have gradually become an indispensable tool in people's daily lives. Internet search services have been widely used in people's daily lives because they can provide objects with all kinds of information and data, bringing great convenience to people.

[0005] In traditional online search, when someone needs to solve a complex search problem, such as developing a travel itinerary, they must break it down into multiple search terms or phrases. For each search term or phrase, they must conduct a separate online search using a search engine or content platform. After performing multiple searches and browsing content, they then organize and collate the information collected from each search to ultimately create a travel itinerary that meets their needs.

[0006] In summary, for some more complex search problems, the object needs to be searched multiple times across platforms, and the search results need to be summarized by themselves. The entire search process is relatively cumbersome, resulting in low search efficiency.

[0007] Summary of the Invention

[0008] Embodiments of the present application provide an information search method, apparatus, electronic device, storage medium, and program product to improve the efficiency of information search for an object.

[0009] In one aspect, an embodiment of the present application provides an information search method, which is performed by an electronic device and includes:

[0010] In response to the current search operation triggered by the target object based on the target search information meeting the preset search condition, first search suggestion information is displayed in the information search interface, wherein,

[0011] The first search suggestion information includes summary information of the current search result;

[0012] The preset search condition is: the current search operation is related to the historical search operation of the target object; and

[0013] In response to a viewing operation triggered by the first search suggestion information, a search result interface corresponding to the target search information is displayed, the search result interface including the current search result, wherein:

[0014] Generate the current search result based on the target search information and the context corpus information of the target object;

[0015] The contextual corpus information includes at least one of the following: search information input by the target object in each historical search operation, content browsed by the target object in each historical search result, and content of feedback provided by the target object to the browsed content.

[0016] On the other hand, an embodiment of the present application further provides an information search method, which is performed by an electronic device and includes:

[0017] Acquiring contextual corpus information of the target object, the contextual corpus information including at least one of the following: search information input by the target object in each historical search operation, content browsed by the target object in each historical search result, and content of feedback provided by the target object in response to the browsed content;

[0018] After receiving a current search request triggered by target search information, in response to the current search request satisfying a preset search condition, the target search information and the context corpus information are input into a target language model to generate current search results and summary information of the current search results; the preset search condition is that the current search operation is related to a historical search operation of the target object;

[0019] The summary information and the current search results are fed back to the client so that the client displays first search suggestion information containing the summary information in the information search interface, and displays a search result interface containing the current search results in response to a viewing operation triggered by the first search suggestion information.

[0020] On the other hand, an embodiment of the present application further provides an information search device, comprising:

[0021] A first response unit is configured to display first search suggestion information in an information search interface in response to a current search operation triggered by a target object based on target search information satisfying a preset search condition, wherein the first search suggestion information includes summary information of the current search result; the preset search condition is that the current search operation is related to a historical search operation of the target object; and

[0022] A second response unit is used to display a search result interface corresponding to the target search information in response to a viewing operation triggered based on the first search suggestion information, wherein the search result interface includes a current search result, wherein the current search result is generated based on the target search information and the contextual corpus information of the target object; the contextual corpus information includes at least one of the following: the search information input by the target object in each historical search operation, the content browsed by the target object in each historical search result, and the content of the target object's feedback behavior on the browsed content.

[0023] On the other hand, an embodiment of the present application further provides an information search device, comprising:

[0024] an information acquisition unit, configured to acquire contextual corpus information of a target object, the contextual corpus information including at least one of the following: search information input by the target object in each historical search operation, content browsed by the target object in each historical search result, and content of feedback provided by the target object in response to the browsed content;

[0025] a result generating unit configured to, upon receiving a current search request triggered by target search information, input the target search information and the context corpus information into a target language model in response to the current search request satisfying a preset search condition, and generate current search results and summary information of the current search results; the preset search condition being that the current search operation is related to a historical search operation of the target object; and

[0026] A feedback unit is used to feed back the summary information and the current search results to the client, so that the client displays the first search suggestion information containing the summary information in the information search interface, and displays the search result interface containing the current search results in response to a viewing operation triggered by the first search suggestion information.

[0027] On the other hand, an embodiment of the present application further provides an electronic device, comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, the processor performs the steps of any one of the above-mentioned information search methods.

[0028] On the other hand, an embodiment of the present application further provides a computer-readable storage medium, which includes a computer program. When the computer program is run on an electronic device, the computer program is used to enable the electronic device to perform the steps of any one of the above-mentioned information search methods.

[0029] On the other hand, an embodiment of the present application also provides a computer program product, which includes a computer program, and the computer program is stored in a computer-readable storage medium; when the processor of an electronic device reads the computer program from the computer-readable storage medium, the processor executes the computer program, so that the electronic device performs the steps of any one of the above-mentioned information search methods.

[0030] BRIEF DESCRIPTION OF THE DRAWINGS

[0031] FIG1 is an optional schematic diagram of an application scenario in an embodiment of the present application;

[0032] FIG2 is a flowchart of an implementation of an information search method in an embodiment of the present application;

[0033] FIG3 is a schematic diagram of an information search interface in an embodiment of the present application;

[0034] FIG4 is a schematic diagram of a search result interface in an embodiment of the present application;

[0035] FIG5 is a schematic diagram of target content and intelligent recognition information in an embodiment of the present application;

[0036] FIG6 is an animated diagram of a collection target content in an embodiment of the present application;

[0037] FIG7 is a schematic diagram of first search suggestion information in an embodiment of the present application;

[0038] FIG8 is a schematic diagram of an AI result list and map module in an embodiment of the present application;

[0039] FIG9 is a schematic diagram of a search result in an embodiment of the present application;

[0040] FIG10 is a schematic diagram of a first type of additional prompt information in an embodiment of the present application;

[0041] FIG11 is a schematic diagram of a second type of additional prompt information in an embodiment of the present application;

[0042] FIG12 is a schematic diagram of a third type of additional prompt information in an embodiment of the present application;

[0043] FIG13 is a schematic diagram of a fourth type of additional prompt information in an embodiment of the present application;

[0044] FIG14 is a schematic diagram of a first conversational search mode in an embodiment of the present application;

[0045] FIG15 is a schematic diagram of a second conversational search mode in an embodiment of the present application;

[0046] FIG16 is a schematic diagram of a third conversational search mode in an embodiment of the present application;

[0047] FIG17 is a schematic diagram of another information search interface in an embodiment of the present application;

[0048] FIG18 is a schematic diagram of another target content and intelligent recognition information in an embodiment of the present application;

[0049] FIG19 is a schematic diagram of another target content and intelligent recognition information in an embodiment of the present application;

[0050] FIG20 is a schematic diagram of a questioning process in an embodiment of the present application;

[0051] FIG21 is a schematic diagram of an answer interface in an embodiment of the present application;

[0052] FIG22 is a schematic diagram of yet another first search suggestion information in an embodiment of the present application;

[0053] FIG23 is a schematic diagram of yet another first search suggestion information in an embodiment of the present application;

[0054] FIG24 is a schematic diagram of a content summarization process in an embodiment of the present application;

[0055] FIG25 is a schematic diagram of another content summarization process in an embodiment of the present application;

[0056] FIG26 is a flowchart of another information search method according to an embodiment of the present application;

[0057] FIG27 is a flowchart illustrating a network search process for implementing AI multi-round search in an embodiment of the present application;

[0058] FIG28 is a schematic diagram of the structure of an information search device according to an embodiment of the present application;

[0059] FIG29 is a schematic diagram of the structure of another information search device in an embodiment of the present application;

[0060] FIG30 is a schematic diagram of a hardware structure of an electronic device to which an embodiment of the present application is applied;

[0061] FIG31 is a schematic diagram of a hardware structure of another electronic device to which an embodiment of the present application is applied.

[0062] Implementation Method

[0063] To make the purpose, technical solutions, and advantages of the embodiments of this application more clear, the technical solutions of this application will be clearly and completely described below in conjunction with the drawings in the embodiments of this application. Obviously, the described embodiments are part of the embodiments of the technical solutions of this application, but not all of them. Based on the embodiments described in this application document, all other embodiments obtained by ordinary technicians in this field without making creative efforts shall fall within the scope of protection of the technical solutions of this application.

[0064] The following is an introduction to some of the concepts involved in the embodiments of this application.

[0065] Contextual corpus information: It is a new corpus proposed in the embodiment of the present application, which can provide the context used by the object when searching. In this way, the target search information of the current search of the object and the contextual corpus information are integrated to generate the final search results. In the embodiment of the present application, for the target object, the contextual corpus information of the target object includes at least one of the following: the search information input by the target object in each historical search operation, the content browsed by the target object in each historical search result, and the content of the feedback behavior implemented by the target object for the browsed content. Among them, feedback behavior refers to some interactive operation behaviors generated when the target object browses the content, such as active collection, clicking like, like, forwarding, sharing, commenting, etc.

[0066] Search suggestion information: It is a kind of suggestion information (abbreviated as sug) that is dynamically displayed when an object conducts a network search. According to the changes in the search request / search term (query) input by the object, the suggestions generated by the artificial intelligence (AI) also change each time. In the embodiment of the present application, the first search suggestion information refers to the summary information of the current search results displayed when it is determined that the preset search conditions are met during a search. The second search suggestion information refers to the summary information corresponding to the search topic when summarizing.

[0067] Additional prompt information: prompt information related to at least one of the date information and location information in the target search information. In this embodiment of the present application, the additional prompt information includes, but is not limited to, at least one of the following: date-related weather information, date-related travel clothing information, date-related holiday event information, date-related travel restrictions information, location-related play restrictions information, and location-related welfare information.

[0068] Intelligent identification information: refers to information obtained by intelligently identifying the target content selected by the target object, including but not limited to at least one of the following: key information contained in the target content, the type of the target content, and the importance of the target content.

[0069] The embodiments of the present application relate to AI and machine learning technologies, and are designed based on natural language processing (NLP) and machine learning (ML) in artificial intelligence.

[0070] The target language model in the embodiment of the present application is obtained by training using machine learning or deep learning technology, and the model is a large language model (LLM) or a multimodal large language model (MLLM). The trained target language model can generate corresponding answers to questions, enabling network search and intelligent dialogue. Among them, the training method can be online training or offline training, which is not specifically limited here.

[0071] When a user searches for a question through a search engine, the search engine retrieves content from across the entire web, scores and ranks the retrieved content, and displays it to the user in order of priority. However, when a user encounters a more complex search question, traditional web search cannot provide a progressive, in-depth answer based on the context of the user's multiple queries. The user must conduct multiple, cross-platform searches, then independently process the information and content from each search, ultimately summarizing the results that meet their needs. This cumbersome search process results in low search efficiency.

[0072] However, the related art of conversational search using large language models also has certain problems. Taking the aforementioned travel guide scenario as an example, using conversational search using a large language model, when a person asks for a travel itinerary for a particular destination, the large language model can directly provide a final, complete answer based on its own algorithms. However, the places and content mentioned in the answer cannot be tailored to the person's personal preferences, resulting in an unrealistic answer. The person still needs to summarize the information themselves, which makes the entire process rather cumbersome.

[0073] In summary, current web searches or conversational searches are unable to generate more personalized search results based on the subject's own needs and preferences, resulting in the subject having to conduct multiple searches and summarize their own results, resulting in low search efficiency.

[0074] In view of this, the embodiments of the present application propose an information search method, device, electronic device, storage medium and program product. The embodiments of the present application are based on the network search experience. Each time the object searches, the search information input by the object, the content browsed by the object in each search result, and the content of the feedback behavior implemented by the object are all counted to construct the context corpus information of the object. Furthermore, when the object searches based on the target search information, it first analyzes whether the current search operation is related to the previous historical search operation. If so, the context corpus information of the object and the target search information input by the object this time are input into the target language model. The capabilities of the target language model are used to provide the object with the current search results that meet personal preferences after multiple rounds of information processing.

[0075] Furthermore, if the current search operation is determined to be related to a previous historical search operation, the first search suggestion information is dynamically displayed in the information search interface, prompting the subject that the current search is related to the previous search, and providing the subject with a summary of the current search results. After the subject clicks to view the first search suggestion information, the current search results generated after multiple rounds of information processing are displayed. In this way, even for more complex search questions, the subject does not need to conduct multiple cross-platform searches on their own. Instead, the subject's personalized corpus information is combined to automatically generate answers that meet their needs, effectively improving the efficiency of their information search.

[0076] As shown in Figure 1, which is a schematic diagram of an application scenario of an embodiment of the present application, the application scenario diagram includes two terminal devices 110 and a server 120.

[0077] In the embodiments of the present application, terminal device 110 includes, but is not limited to, mobile phones, tablet computers, laptop computers, desktop computers, e-book readers, intelligent voice interaction devices, smart home appliances, and in-vehicle terminals. The terminal device may be installed with a client related to information search, which may be software (such as a browser), a web page, or a mini-program. Server 120 is a backend server corresponding to the software, web page, or mini-program, or a server specifically used for information search, and is not specifically limited in this application.

[0078] It should be noted that the information search method in each embodiment of the present application can be executed by an electronic device, which can be a terminal device 110 or a server 120. That is, the method can be executed by the terminal device 110 or the server 120 alone, or can be executed jointly by the terminal device 110 and the server 120. For example, when the method is executed jointly by the terminal device 110 and the server 120, for example, if an information search-related client, such as a browser, is installed on the terminal device 110, the target object can search in the browser. Each time the target object searches, the server 120 can collect statistics on the search information input this time, the content browsed by the target object in the search results this time, and the content of the target object's feedback behavior (such as like, favorite, forward, share, etc.), and construct the contextual corpus information of the object.

[0079] Furthermore, when the target object is searched based on the target search information, the browser responds to the search operation triggered by the target object based on the target search information, and sends a search request to the server 120 through the terminal device 110. After receiving the search request triggered by the target search information, the server 120 analyzes whether the current search operation is related to the previous historical search operation. If it is determined to be related (that is, the preset search conditions are met), the target search information and context corpus information are input into the target language model to generate the current search results and summary information of the current search results, and feedback is given to the terminal device 110. The terminal device 110 displays the first search suggestion information containing the above summary information in the information search interface through the browser; then, after the browser responds to the viewing operation triggered by the first search suggestion information, it can display the search result interface corresponding to the target search information, and the search result interface contains the above current search results.

[0080] In an optional embodiment, the terminal device 110 and the server 120 can communicate through a communication network. The communication network is a wired network or a wireless network. In an embodiment of the present application, when there are multiple servers, the multiple servers can be composed of a blockchain, and the servers are nodes on the blockchain. As disclosed in the embodiment of the present application, the information search related data involved can be stored on the blockchain, for example, context corpus information, target search information, target content, first question, first answer information, second question, second answer information, third question, third answer information, etc. In addition, the embodiment of the present application can be applied to various scenarios, including but not limited to cloud technology, artificial intelligence, smart transportation, assisted driving and other scenarios.

[0081] It should be emphasized that in the specific implementation of this application, it involves relevant data that the object searches for multiple times when searching for information. When the above embodiments of this application are applied to specific products or technologies, it is necessary to obtain the object's permission or consent, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions.

[0082] Referring to FIG. 2 , which is a flowchart of an implementation of an information search method provided in an embodiment of the present application, the method is executed by an electronic device, for example, the terminal device 110 in FIG. 1 , and more specifically, by a client (such as a browser) in the terminal device 110 . The specific implementation process of the method is as follows S21 to S22:

[0083] S21: In response to the current search operation triggered by the target object based on the target search information meeting the preset search conditions, the first search suggestion information is displayed in the information search interface, wherein the first search suggestion information includes summary information of the current search results, and the preset search conditions are: the current search operation is related to the historical search operations of the target object.

[0084] Specifically, each time a subject conducts a web search, the subject's current search operation is compared with the subject's previous search history to analyze whether there have been any related searches before. Taking the target subject as an example, the current search information entered by the target subject in this search can be recorded as the target search information. Specifically, the target search information can be compared with the search information and search results corresponding to the target subject's previous search operations to analyze whether there are any related results.

[0085] Among them, correlation means that there are certain commonalities between two search operations. The commonalities can specifically mean that the search information contains the same or similar keywords (such as names of people, places, organization names, etc.), the search information has similar semantics, the search information has opposite semantics (can also be understood as a commonality, semantic relevance), etc., which will not be elaborated here.

[0086] In this embodiment of the present application, the search suggestions generated by the AI ​​each time change based on the search information entered by the subject. The AI-generated search suggestions are displayed to the subject as soon as the subject enters the search information, providing a quick preview of the search results. The subject can then determine whether to view the search results generated by the AI ​​in conjunction with the contextual corpus information.

[0087] S22: In response to a viewing operation triggered by the first search suggestion information, a search result interface corresponding to the target search information is displayed, and the search result interface includes the current search result, wherein the current search result is generated based on the target search information and the contextual corpus information of the target object; the contextual corpus information includes at least one of the following: the search information input by the target object in each historical search operation, the content browsed by the target object in each historical search result, and the content of the target object's feedback behavior on the browsed content.

[0088] Search information refers to the search request entered by the target user when conducting an online search, such as the search term entered into the search box. Feedback behavior refers to the interactive actions generated by the target user when browsing content, such as actively saving, clicking "like", "thumbs-up", "repost", "share", and "comment".

[0089] Specifically, during each web search process for a target object, the above-mentioned types of information can be collected, thereby continuously enriching the contextual corpus information corresponding to the target object for use in the subsequent generation of personalized search results.

[0090] In an embodiment of the present application, the generation of the current search results can be achieved based on a target language model (such as a large language model). Specifically, each time the target object performs a network search, the current search operation can be compared with the previous historical search operations. When it is determined that the preset search conditions are met, the target search information input by the target object this time and the currently collected context corpus information are input into the target language model together. The answer generation capability of the target language model is utilized, combined with the context corpus information, to generate a search result (i.e., the current search result) that meets personal preferences after multiple rounds of information processing at one time.

[0091] In an embodiment of the present application, the object can determine whether it is necessary to view the search results generated by AI in combination with the contextual corpus information by browsing the first search suggestion information generated by AI. When the object determines that it needs to view, it can view the current search results generated by AI based on the viewing operation triggered by the first search suggestion information. Specifically, the object can trigger the viewing operation by clicking on the first search suggestion information, long pressing the first search suggestion information, etc., which will not be described in detail here.

[0092] This application builds on the web search experience and incorporates a target language model. This provides contextual information to the target language model, including every search query performed by the subject in the browser, the content viewed in the search results, and the content that the subject actively likes and saves. The target language model can combine this contextual information with the subject's customized context and leverage the capabilities of the target language model to provide search results that meet the subject's preferences after multiple rounds of information processing. This addresses the current issue of single-shot web searches, which are not supported by multiple rounds.

[0093] In addition, this application also solves the problem that the target language model cannot provide answers based on the object personalized corpus. In this application, the active and passive operations of the object on the content in the network search are provided to the target language model as the object personalized corpus, so that the target language model can provide personalized solutions for the object.

[0094] At the same time, this application also solves the problem of mutual integration of network search and target language model: when the object uses network search, it can view the results of AI screening at any time by clicking on the AI ​​search suggestion information.

[0095] The following uses the two scenarios of tourism and word search as examples to illustrate the information search method in the embodiment of the present application with reference to the accompanying drawings:

[0096] (1) Tourism scenes.

[0097] The following example takes the case where the target person first searches for “travel guide to city A” and then searches for “hotels in city A”.

[0098] Refer to Figure 3, which is a schematic diagram of an information search interface in accordance with an embodiment of the present application. Figure 3 illustrates a target user's first search for "travel guide to City A." Assuming the target user has not previously searched for travel or City A-related content, that is, there are no historical searches related to the current search operation, and the preset search conditions are not met. Therefore, the information search interface does not need to display the first search suggestion information.

[0099] After the target object conducts an online search based on the search information "A city travel guide" entered this time, a normal search result interface may be displayed. The search results in the search result interface are the result of the search engine recalling all online content based on "A city travel guide", scoring and sorting the recalled content, and then displaying it in order of sorting priority.

[0100] As shown in Figure 4, it is a schematic diagram of a search result interface in an embodiment of the present application. As shown in Figure 4, the left interface 410 lists several search results that have been recalled and sorted in the above manner, such as search result 1 "Traveled through City A, 9 favorite places", search result 2 "Spent four days in City A, sincere recommendation", and search result 3 "The most beautiful attractions in City A in autumn". The target object can swipe up the interface to view more search results, or click to view a specific search result. If the object selects search result 1, it can jump to the details page 420 of search result 1 shown on the right side of Figure 4.

[0101] In an embodiment of the present application, if the target object selects an article of interest to browse in the search results interface, the content browsed and collected by the target object can be automatically recorded to create a personalized corpus for the object (that is, a form of expression of personalized corpus information in this article).

[0102] It should be noted that in the embodiment of the present application, the target object's favorite content may include not only the content that the target object actively clicks to favorite, but also the content that the target object likes, praises, comments, reposts, shares, etc. In addition, when browsing the search results, the target object can also identify and ask questions about local content in the search results. The embodiment of the present application also supports intelligent collection of this part of the content, enriching the dimension of the target object's favorite content.

[0103] Specifically, when browsing an article, the target object may be interested in some of the local content. For this type of content, AI intelligent recognition can be performed, and AI intelligent collection can also be performed. An optional implementation method is as follows: the client responds to the content selection operation triggered based on the search results interface, and identifies the target content selected this time and the intelligent recognition information corresponding to the target content in the search results interface. Among them, the intelligent recognition information includes but is not limited to at least one of the following: key information contained in the target content, the type of the target content, and the importance of the target content. Among them, the key information contained in the target content may refer to people, places, institutions, special foods, special activities, special events, etc.

[0104] In an embodiment of the present application, for any search operation, the target object can further view the details of the search result interface. For example, the target object views a specific search result in the search result interface and enters the details page of the search result (another search result interface in an embodiment of the present application). Then, the content selection operation is performed on the details page. The client responds to the content selection operation of the target object. In addition to identifying the target content selected this time in the search result interface, it can also further display the intelligent identification information corresponding to the target content.

[0105] Specifically, the intelligent recognition information can be in the form of pictures, text, graphics, videos, icons, labels, animations, etc., which are not specifically limited in this article. The following uses labels as an example for illustration.

[0106] Figure 5 is a schematic diagram of target content and intelligent recognition information in an embodiment of the present application. For example, when a target user browses the article "Traveling through City A, My 9 Favorite Places" listed above, and is interested in a particular attraction and route mentioned therein, they double-click the content area. The interface then selects the paragraph they double-clicked. This is indicated by S501 in the gray area of ​​Figure 5.

[0107] It should be noted here that if the selected range is inaccurate, the target object can manually drag the upper and lower edges of the selection to expand or shrink it to modify the selected range.

[0108] If the target user does not perform any operation, a recognition animation may appear in the interface to prompt the user that "AI is recognizing the selected content" (not shown in Figure 5). After the recognition animation ends, the interface may identify the location and other information identified by the AI, and display them in the form of labels, such as S502 and S503 in the right interface of Figure 5. S502 indicates the recognized location "Palace B", and S503 indicates the recognized local specialty "Dish D".

[0109] In the embodiment of the present application, different styles can be used to mark different intelligent identification information, such as marking different types of locations with different colors, etc., which will not be described in detail here.

[0110] Optionally, in addition to intelligently identifying target content and marking the intelligently identified information, the target content can also be automatically saved. Specifically, a portion of content is selected for the target subject to save. When saving, the AI ​​uses the target language model to analyze and label the saved content, intelligently categorizing the content of interest to the target subject at the time of saving. The categorized tags can be used to search the personalized corpus of the subject during AI conversations. This is described on the server side and will not be repeated here.

[0111] In addition, a corresponding collection animation is displayed in the search results interface. An optional collection animation shows the target content flying into the collection control in the search results interface. During this process, the target content can also gradually shrink, giving the target the impression that the target content is getting smaller and smaller, and finally shrinking enough to fit into the collection control.

[0112] As shown in Figure 6, it is an animated schematic diagram of a collection target content in an embodiment of the present application. The interface shown on the left side of Figure 6 represents a frame in the process of the target content S601 flying into the collection control 602 in the search result interface. In this process, S601 not only includes the target content, but also includes AI-recognized tags, such as B Palace and D Dish. After the target content S601 completely flies into the collection control 602, the style shown in S602 is updated to the style shown in S603.

[0113] In an embodiment of the present application, the target object can click on the favorites control to view the target object's favorites. In the target object's favorites, the intelligent identification information can also be identified in the same or similar manner as described above, such as still identifying it in the form of a label, to facilitate subsequent generation of answers.

[0114] Optionally, the client can also display a corresponding collection interface in response to a viewing operation on the collection content of the target object, and mark different classification tags in the collection interface using different annotation styles; for repeated classification tags, the number of recurrences is marked at a preset position in the collection interface. Specifically, these classification tags can be location tags, time tags, event tags, item tags, etc. For details, please refer to the relevant instructions on the server side and will not be repeated here.

[0115] The above implementation provides a way for users to collect certain content. Furthermore, traditional methods can only collect text or images themselves, but cannot classify or identify the collected content. However, the above implementation allows users to see AI-generated tags for key content when collecting it, making it easier for users to better organize and understand the content.

[0116] It should be noted that Figures 5 and 6 above illustrate the operations performed on search results during the target subject's first search. The target subject can perform these operations on search results during any subsequent search, such as when the target subject subsequently searches for "City A," "Travel," and so on. The specific implementation methods are the same as those in the previous embodiment, and any repetitions will not be repeated here.

[0117] Assume that after searching for "City A travel guide" for the first time, the target user conducts a second search for "City A hotels." This search is related to the previous search for "City A travel guide" and is therefore a search operation related to City A. Therefore, in this embodiment of the present application, when the target user searches for "City A hotels," the preset search conditions are met, and the first search suggestion information can be dynamically displayed on the information search interface during this search process.

[0118] As shown in Figure 7, it is a schematic diagram of displaying the first search suggestion information in an embodiment of the present application. During the search process, the first search suggestion information can be displayed in the information search interface shown in Figure 7, prompting that relevant A city hotels have been screened for the target object. As shown in S701 and S702 in Figure 7, the first search suggestion information shown in S701 indicates the current processing progress, that is, a suitable hotel has not yet been searched, but is in the analysis stage; after the analysis is completed, the first search suggestion information shown in S702 can be displayed, prompting the target object with a summary of the current search results, such as "Combined with your preferred location, we recommend A city hotel for you>" in S702.

[0119] In this embodiment of the present application, the target user can click the first search suggestion information shown in S702 to enter the search results interface and display the AI ​​result list. As shown in Figure 8, it is a schematic diagram of an AI result list and map module in this embodiment of the present application. The AI ​​result list is a list of hotels near attractions, routes, and stores that the target user has previously collected or followed in their browsing experience, which is an AI personalized screening result that meets the needs of the target user.

[0120] It should be noted that, during the information search process, in addition to displaying the first search suggestion information, search suggestion information indicating the progress of search result generation as shown in S701 (also referred to as third search suggestion information) may also be displayed.

[0121] It should also be noted that when searching for a target object multiple times online, the target language model can cross-filter the previous search terms for the target object to filter the current search results for the target object and present them in a suitable format based on the type of content filtered. Alternatively, it can compare multiple search terms and present the compared results.

[0122] For example, if the target search information is related to the geographic location, a map module may be further provided on the search results interface; the map module is used to display at least one of the following: the collection locations related to the geographic location in the collection content of the target object, the recommended locations related to the target search information determined based on preset recommendation rules, and the driving routes between the collection locations and the recommended locations; furthermore, the current search results on the search results interface include information on each recommended location, and at least one web page associated with the recommended location.

[0123] Still taking Figure 8 as an example, S801 is an example of a map module in an embodiment of the present application. In the example shown in Figure 8, the content filtered by AI is related to the geographical location "City A", and the map module shown in S801 is added. The target object's favorite locations and AI-recommended hotel locations are displayed on the map module S801 at the same time, and the route between the two is marked. For example, in S801, the target object's favorite locations include x1 attraction, x2 attraction, x3 attraction, and Palace B, and the AI-recommended locations include P1 Resort Hotel, P2 Sunset Cottage, and P3 Resort Hotel.

[0124] In addition, the map module also marks routes between favorited locations and recommended locations. Specifically, based on the distance between these locations, it can mark a feasible route between locations within a certain distance (closer distance). For example, there is a recommended route between P1 Resort Hotel -> x2 Attraction -> B Palace, a recommended route between P2 Sunset Cottage -> x1 Attraction, and a recommended route between P3 Resort Hotel -> x3 Attraction. These three recommended routes are shown as dotted lines in S801. The target object's favorited locations and AI-recommended locations may overlap, but this is not specifically limited in this article.

[0125] It should be noted that the map module shown in S801 also displays the optional range of hotel prices. For example, the prices of the three hotels recommended in S801 are between 0 and 1,000 yuan. This price range can be obtained based on statistical analysis such as the target object's historical booking records, or it can be a default price range, which is not specifically limited in this article.

[0126] Furthermore, S802 displays a list of hotels selected by the AI. The hotel list includes two elements: basic hotel information (i.e., information about the recommended location) and web pages selected by the AI ​​that mention the hotel (i.e., at least one web page associated with the recommended location). A hotel can be associated with multiple web page results, helping the target audience determine whether they are interested in the current hotel.

[0127] Below, taking each hotel as an example of a web page result, the detailed content of S802 is shown in Figure 9. As shown in Figure 9, it is a schematic diagram of a search result in an embodiment of the present application. The hotel list shown in Figure 9 introduces P1 Resort Hotel, P2 Sunset Cottage and P3 Resort Hotel in sequence, wherein the basic information of the hotel is the name, price, geographical location, some architectural pictures and other information of the hotel. In addition, each hotel also displays at least one web page result. For example, there is one strategy for mentioning P1 Resort Hotel, the content of which comes from platform a, and the content of the web page result is shown in S901 (replaced by ellipsis ...). Similarly, there is one strategy for mentioning P2 Sunset Cottage, the content of which comes from platform b, and the content of the web page result is shown in S902 (replaced by ellipsis ...); there is one strategy for mentioning P3 Resort Hotel, the content of which comes from platform a, and the content of the web page result is shown in S903 (replaced by ellipsis ...).

[0128] In the above implementation, users can more efficiently filter content, eliminating the need to filter across all web results and instead filtering based on existing preferences. Furthermore, AI intelligent filtering can utilize different components, such as the aforementioned map, based on the different forms of user input. This allows for a more intuitive display of content compared to web searches and conversations with target language models.

[0129] Optionally, after the target object triggers a viewing operation based on the first search suggestion information, the search result interface corresponding to the target search information can be displayed. At the same time, considering that the current search results in the search result interface are generated based on the target search information and the contextual corpus information, it is different from the traditional network search method of directly matching and recalling the entire network content based on the target search information. Therefore, the search information in the search information input area in the search result interface can also be updated to more intuitively show the target object the difference between AI intelligent screening and traditional network search.

[0130] The search result interface also includes a search information input area. An optional implementation is: in the search information input area, the display of the target search information is updated to display the target search information and the number of target collection contents; wherein, the target collection content is the content in the collection content of the target object that is related to the current search operation.

[0131] Still taking FIG8 as an example, the search information input area in FIG8 is the search box shown in S803. The search information in the search box should originally be the target search information, such as "A City Hotel" shown in the left interface of FIG8 , but in S803, it is updated to be displayed in two parts, one part is the target search information "A City Hotel", and the other part is "☆4 Favorite Places".

[0132] In the above embodiment, on the AI ​​screening result page, the content in the search box is updated to "search term entered by the target object" and "※ collections", prompting the target object that the current AI screening result is generated by combining the search term entered by the object this time and the locations previously collected by the object, which is more in line with the target object's preferences and more in line with the target object's personalized needs.

[0133] Another optional implementation is to update the target search information displayed in the search information input area to the first search suggestion information. For example, the search information in S803 can be directly updated to "Based on your preferred locations, we recommend hotels in City A for you." In this way, the target subject can be more directly informed that the current AI screening results are generated by combining the search terms entered by the subject and the subject's preferences, which is more in line with the target subject's preferences and more personalized needs.

[0134] It should be noted that the two search information update methods listed above are just simple examples. In addition, any update method that can prompt the object that the current AI screening result is generated by combining the search terms entered by the object this time and the object's preferences is applicable to the embodiments of this application and will not be described one by one here.

[0135] In an embodiment of the present application, intelligent reminders can also be provided for the results of AI screening, that is, in addition to filtering content based on the personalized information provided by the object, additional reminder information can be matched based on the search content. An optional implementation method is as follows: additional prompt information matching the current search results is also displayed on the search results interface; wherein the additional prompt information is: prompt information related to the date information and / or location information in the target search information.

[0136] Among them, during this search process, if date information is involved, additional prompt information related to the date can be further generated; similarly, if location information is involved, additional prompt information related to the location can be further generated; or, when both date information and location information are involved, in addition to the above-mentioned separate generation method, additional prompt information related to both the date and the location can also be generated.

[0137] In the embodiments of the present application, the additional prompt information includes, but is not limited to, at least one of the following: date-related weather information, date-related travel clothing information, date-related holiday event information, date-related travel restrictions information, location-related play restrictions information, and location-related welfare information. Specifically, the above additional prompt information can be generated by matching pre-set prompt templates, and the additional prompt information can be in the form of pictures, text, graphics, videos, icons, labels, animations, etc.

[0138] For example, when the target provides date information, matching prompt templates include, but are not limited to: weather information corresponding to the date, travel clothing information corresponding to the date, major holidays corresponding to the date, and restrictions on appearances corresponding to the date. When the target does not provide date information, matching prompt templates include, but are not limited to, restrictions on tourist attractions and merchant benefits information.

[0139] Figure 10 is a schematic diagram of the first type of additional prompt information in an embodiment of the present application. In the search results interface shown in Figure 10, additional prompt information S1001 and S1002 are displayed before the hotel list in the smart answer. For example, if the target search information provided by the subject includes the date information "z holiday", S1001 lists the weather information corresponding to z holiday, and S1002 lists the travel clothing information corresponding to z holiday.

[0140] FIG11 is a schematic diagram of the second type of additional prompt information in an embodiment of the present application. In the search results interface shown in FIG11 , additional prompt information S1101 is displayed before the hotel list in the smart answer. For example, if the target search information provided by the subject includes the date information "Z Festival," S1101 lists travel restrictions corresponding to Z Festival, specifically "Z Festival coincides with the Y Sports Day. Traffic restrictions will be implemented during this period, so we recommend hotels that are more convenient for walking."

[0141] As shown in Figure 12, it is a schematic diagram of the third additional prompt information in the embodiment of the present application. In the search result interface shown in Figure 12, the additional prompt information S1201 is first displayed before the hotel list of the smart answer. For example, the target search information provided by the object does not contain date information, but contains the location information "City A". During the search process, some attractions, hotels and other places related to City A are further involved. S1201 lists the corresponding travel restriction information of the attractions, and the specific content is "The x4 attractions you plan to visit require women not to wear skirts or off-shoulder tops to enter. Remember to prepare appropriate clothes~".

[0142] As shown in Figure 13, it is a schematic diagram of the fourth type of additional prompt information in the embodiment of the present application. In the search result interface shown in Figure 13, additional prompt information S1301 is first displayed before the hotel list of the smart answer. For example, the target search information provided by the object does not include date information, but includes the location information "City A". During the search process, some attractions, hotels, and other places related to City A are further involved. S1301 lists the corresponding welfare information of the merchant, specifically "The hotel recommended for you has a free shuttle bus to attractions C and X4~".

[0143] In the above embodiment, by displaying additional prompt information, more personalized information is provided to the object, which is more in line with the needs of the object itself and reduces the search path of the object.

[0144] In addition to the above implementations, the information search method in the embodiment of the present application also supports switching the AI ​​screening results to a conversational search mode. If the search result interface is set to also include a first intelligent question and answer entry, the target object can switch modes based on the first intelligent question and answer entry. An optional implementation is as follows: in response to a conversation operation triggered by the first intelligent question and answer entry, the initial network search mode is switched to the conversational search mode; in response to a question operation for the current search result in the conversational search mode, the first question entered for the current search result and the first answer information corresponding to the first question are displayed.

[0145] Specifically, after the target object clicks the first search suggestion information and sees the AI ​​content, he or she can continue the AI ​​dialogue directly with the target language model through the "AI dialogue" entrance, or conduct a network search by clicking the search box.

[0146] As shown in Figure 14, it is a schematic diagram of the first conversational search mode in an embodiment of the present application. As shown in Figure 14, the intelligent question and answer entrance in the search result interface is the "Continue to ask AI" control S1401. The target object clicks S1401 to continue the conversation in the hotel list screened by AI, and the initial network search mode can be switched to the conversational search mode. The target object can continue to ask questions directly based on the current search results. For example, the target object continues to enter questions (which can also be understood as a dialogue instruction), such as S1402 "Help me develop a 3-day and 2-night travel guide for City A with a budget of 4,000 yuan". AI combines the target object's personalized corpus (including collected content, browsed content, etc.), network search content, and target language model to give the answer (i.e., the first answer information) corresponding to the question (i.e., the first question). As shown in S1403 in Figure 14, during the answer generation process, the target object is prompted to "analyze your collection content", such as currently analyzing the "x4 attractions" collected by the target object.

[0147] Specifically, after analyzing all relevant content collected by the target subject, a network search can be performed to search the entire network for content related to travel to City A. As shown in Figure 15, it is a schematic diagram of the second conversational search mode in the embodiment of the present application. S1501 prompts the subject that the current step is "generating a travel guide based on your preferences." Prior to this step, two steps have been completed in sequence, namely "analyzing your collection content" and "searching the entire network for travel to City A." The subject has analyzed 32 collection contents related to the current question and generated an answer by combining the collection contents and the network search content.

[0148] As shown in Figure 16, it is a schematic diagram of the third conversational search mode in the embodiment of the present application. S1601 shown in Figure 16 means that the above three steps are completed, and the specific content of the generated first answer information is displayed after S1601, that is, a 3-day and 2-night travel guide to City A with a budget of 4,000 yuan. As in S1602, the itinerary for each day, as well as the accommodation location, etc. are described in detail. For example, on the first day, visit x4 attractions -> Palace B -> Attraction C, and plan according to time periods such as morning, afternoon, and evening, and list the specific itinerary plan. Similarly, specific itineraries are also given for the second and third days. Please refer to Figure 16 for details, and no further details will be given here.

[0149] The generated travel guide can also use special style symbols (such as a five-pointed star) to indicate that the location is from the subject's collection. Specifically, these symbols can be directly marked at the corresponding location in the travel guide. Or they can be marked together in a unified and fixed location, as shown in S1603 in Figure 16, which uniformly marks the content referenced from the marked collection after the specific itinerary, such as X1 Historic District, B Palace Museum, and P1 Resort Hotel.

[0150] In the above embodiment, after the subject switches to the AI ​​dialogue, each time the AI ​​answers the question, it will combine the results of the network search, the subject's personalized corpus, and the target language model to collaboratively answer the question.

[0151] Optionally, the search result interface may further include at least one answer operation control; the target object may perform related operations on the generated first answer information based on these operation controls. In response to selecting a target answer operation control from the at least one answer operation control, the client performs the operation corresponding to the target answer operation control on the first answer information and displays the result of the operation.

[0152] Still taking FIG. 16 as an example, S1604 is a list of several answer operation controls listed in this application, namely: share link, download PDF, add to favorites. This means that the target object can share, download, add to favorites, and other operations on the generated strategy content.

[0153] In the above implementation, the subject can also share, download, and collect the answers generated by AI. While providing the subject with richer interaction methods, it is also helpful to improve the subject's usage stickiness and provide more convenience for the subject.

[0154] In summary, when the subject needs to solve complex search problems, such as formulating a travel strategy, the subject does not need to conduct multiple, cross-platform searches. This application can provide progressive and in-depth answers through context, and the subject does not need to process the information and content after each search by themselves. Moreover, in the case of formulating a travel guide, the places and content involved are all formulated according to the subject's personal preferences. This application is based on the online search experience and combines the target language model to provide the target language model with contextual corpus information for every search question conducted by the subject in the browser, the content browsed in the search results, and the content that the subject actively likes and collects. The target language model can be combined with the subject's customized corpus and use the capabilities of the target language model to provide the subject with search results that meet personal preferences after multiple rounds of information processing.

[0155] The above uses the tourism scenario as an example to explain the information search method in the embodiment of the present application. The following uses the word query scenario as an example to explain:

[0156] (2) Word query scenario.

[0157] The following example uses the target object to first search for the word "artificial" and then search for the word "naturally":

[0158] Referring to FIG. 17 , which is a schematic diagram of another information search interface in an embodiment of the present application, in information search interface 1710 , a target user searches for the word "artificial" for the first time. Assuming that the target user has not previously searched for other words, that is, there are no historical searches related to the current search operation, and the preset search conditions are not met, therefore, the first search suggestion information does not need to be displayed in information search interface 1710 at this time.

[0159] After the target object conducts a web search based on the search information "artificial" input this time, a normal search result interface may be displayed. The search results in this search result interface are the search results of "artificial" recalled by the search engine based on "artificial", as shown in the right interface 1720 in Figure 17, which introduces the definition of the word and related phrases, examples, etc. in detail, and the target object can browse the search results of the word.

[0160] Similar to the above-mentioned travel scenario, after obtaining the search results this time, the target object can also browse the search results. If the target object is interested in some local content in the search results, for such content, AI intelligent recognition can be performed, or AI intelligent collection can be carried out. An optional implementation method is as follows: The client responds to the content selection operation triggered based on the search result interface, and marks the target content selected this time and the intelligent recognition information corresponding to the target content in the search result interface.

[0161] As shown in FIG. 18, it is a schematic diagram of another target content and intelligent recognition information in an embodiment of the present application. For example, when the target object browses the above search results and is interested in a certain phrase "artificial intelligence 人工智能" mentioned therein, after double-clicking on the content area, the phrase double-clicked by the object will be selected in the interface.

[0162] After the target object has no operation, an identification animation can appear in the interface to prompt the object that "AI is identifying the selected content" (not shown in FIG. 18). After the identification animation ends, information such as the type and importance level of the target content identified by AI can be marked in the interface and displayed in the form of labels. As shown by S1801 in interface 1810, the type of this target content is "phrase" (which also means a short phrase), and the importance level is three stars. In addition, the target content can be automatically collected, and a corresponding collection animation is displayed, such as S1802 shown in the right interface 1820 in FIG. 18 (the content is the same as S1801, not shown), which is a frame of the collection animation in an embodiment of the present application, indicating that the target content flies into the collection control.

[0163] As shown in FIG. 19, it is a schematic diagram of yet another target content and intelligent recognition information in an embodiment of the present application. For example, when the target object browses the above search results and is interested in a certain example sentence "They make up something artificial or untrue.制造人工的或不真实的东西….." mentioned therein, after double-clicking on the content area, the example sentence double-clicked by the object will be selected in the interface.

[0164] After the target object has no operation, an identification animation can appear in the interface to prompt the object that "AI is identifying the selected content" (not shown in FIG. 19). After the identification animation ends, information such as the type and importance level of the target content identified by AI can be marked in interface 1910 and displayed in the form of labels. As shown by S1901, the type of this target content is "example sentence", and the importance level is five stars. In addition, the target content can be automatically collected, and a corresponding collection animation is displayed, such as S1902 shown in the right interface 1920 in FIG. 19 (the content is the same as S1901, not shown), which is a frame of the collection animation in an embodiment of the present application, indicating that the target content flies into the collection control.

[0165] In addition to the aforementioned interaction methods, this application also supports multimodal AI search for any content. While browsing the results page, the user can long-press text or an image to select it and ask a question about the selected content. The target language model will respond based on the selected content and the user's question.

[0166] An optional implementation is as follows: in response to a question operation for the target content, an answer interface is displayed, where the answer interface includes the target content, a second question input for the target content, and second answer information corresponding to the second question.

[0167] In an embodiment of the present application, the target object can also select the target content while browsing the search results and ask questions about the target content through voice input, keyboard input, etc. For example, when the target object is learning a word or browsing any web page content, if the target object has questions about the content in the interface, the target object can directly long press the interface content. After the long press, the voice collection will be aroused, and the target object can directly speak the question. After the target object releases his finger, the selected interface content and the question described by the target object will be asked to the AI ​​together, and the AI ​​will return the result.

[0168] As shown in Figure 20, which is a schematic diagram of a questioning process in an embodiment of the present application, Figure 20 shows that while browsing the search results for the word "artificial", the target subject has questions about some of the content. The target subject can select the content with questions as the target content by double-clicking, long pressing, etc., and then, by double-clicking, long pressing, etc., to invoke voice capture and speak out their questions. When the target subject selects the target content "This artificial beach resort can hold up to 10,000 visitors.", the target subject is prompted "You can directly speak your question~", as shown in S2001 in Figure 20.

[0169] Then, the target object can describe his or her questions in voice, as shown in S2101 in interface 2110. The target object asks "Why use can here?" After the target object releases his or her finger, the selected target content and the second question described by the target object will be asked to the AI ​​together. The AI ​​will feedback the second answer information and display the answer interface 2120 shown on the right side of Figure 21. The answer interface 2120 includes the target content, as shown in S2102, and also includes the second question entered for the target content, as shown in S2103, and the second answer information corresponding to the second question, as shown in S2104.

[0170] In addition, it should be noted that the search information in the search information input area of ​​the answer interface can also be updated to the second question, or updated to the target content + the second question, etc., without specific limitation here. In the above embodiment, the object can be searched multimodally by AI at any location, which can shorten the path of object operation and realize the experience of searching anywhere.

[0171] Optionally, the answer interface may also include a second intelligent question and answer entrance, and the target object can switch modes based on the second intelligent question and answer entrance. An optional implementation method is as follows: in response to a dialogue operation triggered based on the second intelligent question and answer entrance, switch from the initial network search mode to the dialogue search mode; in response to the question operation for the second answer information in the dialogue search mode, display the third question entered for the second answer information and the third answer information corresponding to the third question.

[0172] Still taking Figure 21 as an example, the intelligent question-and-answer entry in the answer interface is the "Continue Asking AI" control shown in S2105. The target object clicks S2105 in the AI-screened hotel list to continue the conversation, and the network search mode can be switched to the conversational search mode. The target object can continue to ask questions directly based on the second answer information mentioned above. Then, the AI ​​combines the target object's personalized corpus (including favorite content, browsed content, etc.), network search content, and the target language model to give the answer (i.e., the third answer information) corresponding to the question (i.e., the third question). The specific implementation method of AI question-and-answer is similar to the relevant parts of Figures 14, 15, and 16, and will not be repeated here.

[0173] In the above embodiment, on the result page returned by AI, the subject can switch to a conversational question and answer session by clicking "Continue to ask AI". After switching to the AI ​​dialogue, each answer by AI will combine the results of the web search, the subject's personalized corpus, and the target language model to collaboratively answer the question.

[0174] Suppose that in a subsequent search, the target user searches for another word, such as "naturally." When the AI ​​analyzes the target user's previous search content as context, it determines that the two words (artificial and naturally) are in the same word learning query and are related, that is, they meet the preset search conditions. During this search, the AI ​​suggestion, i.e., the first search suggestion information, can be dynamically displayed on the information search interface. This suggestion prompts the target user that the two search operations are related, and that "artificial" and "naturally" are antonyms. If the target user is interested, they can click to view more detailed current search results.

[0175] As shown in Figure 22, it is a schematic diagram of another first search suggestion information in an embodiment of the present application. The first search suggestion information "artificial and naturally are antonyms. It..." is directly displayed in the information search interface 2210. The target object can click on the first search suggestion information to further view the corresponding search results, as shown in the information search interface 2220 on the right side of Figure 22, which summarizes some differences between artificial and naturally.

[0176] As shown in Figure 23, it is a schematic diagram of another first search suggestion information in an embodiment of the present application. During the search process, S2301 in the information search interface 2310 displays the current processing progress, that is, the analysis stage; S2302 in the information search interface 2320 displays the first search suggestion information, prompting the target object that artificial and naturally are antonyms, that is, the summary information of the target object's current search result. The target object can click S2302 to enter the search result interface 2330, which displays the AI ​​search results, that is, some differences between artificial and naturally.

[0177] Compared with Figure 22, the information search interface shown in Figure 23 has an additional search suggestion message "Analyzing...", which more vividly reflects that the AI ​​suggestions are dynamically displayed as the analysis process progresses.

[0178] In the above embodiment, when the subject does not find any association between the search terms, the AI ​​screening can proactively provide the subject with related content between different search terms to help the subject better grasp the relevant knowledge.

[0179] In addition, dynamic AI suggestions provide subjects with a quick preview of AI search results combined with context. By browsing AI suggestions, subjects can determine whether they need to view the content generated by AI combined with context. Moreover, AI suggestions appear during the subject's search process, and can prompt the subject with the answer given by AI as soon as the subject enters the input. In addition, the form of AI suggestions does not affect the search term completion suggestions given by the subject when browsing the web search, which can facilitate the subject to switch between web search and AI search.

[0180] Optionally, the information search method in the embodiment of the present application also supports intelligent summarization of previously searched content by topic. When the subject searches for content on multiple different topics, the target language model can be used to summarize the content by topic, giving multiple AI suggestions and their corresponding AI summary content. An optional implementation method is as follows:

[0181] In response to the summary operation triggered by the target object, at least one second search suggestion information is displayed in the information search interface, wherein each second search suggestion information corresponds to a search topic, and the second search suggestion information is the summary information corresponding to the search topic; at least one search topic is obtained by dividing multiple search information related to the target object.

[0182] On this basis, the target object can view the summary content corresponding to any second search suggestion information. Specifically, the client responds to the selection operation of the target search suggestion information in at least one second search suggestion information, and displays the search result summary interface corresponding to the target search suggestion information. The search result summary interface contains the summary content of the search results corresponding to each search information under the corresponding search topic.

[0183] As shown in Figure 24, it is a schematic diagram of a content summarization process in an embodiment of the present application. When the target object enters "summary" in the search information input area of ​​the information search interface 2410 (the search box 2400 in Figure 24), the AI ​​will classify and summarize the content that the target object has queried within the memory range, and display it in the form of multiple suggestions. As shown in S2401 and S2402, the target object has queried a travel guide for city A and multiple words. When the target object enters "summary", the AI ​​will summarize the contents of the two topics respectively, and the two second search suggestion information displayed are S2401 "4000 yuan travel guide for city A" and S2402 "words memorized today". On this basis, the target object can choose to click S2401 to view the specific travel guide; or choose to click S2402 to view the words memorized today.

[0184] After the target object clicks S2401, the travel guide for city A is displayed in the interface 2420, and the search information in the search box S2403 in the interface 2420 is updated to "A city itinerary summary" + "32 related collections".

[0185] Optionally, if the search topic is a travel guide, the corresponding summary content includes an itinerary plan corresponding to at least one travel time period, such as the itinerary plans for the first, second, and third days listed in interface 2420. For example, on the first day, the itinerary includes visiting x4 attractions -> Palace B -> Attraction C, with a detailed itinerary planned for the morning, afternoon, and evening time periods. Similarly, detailed itineraries are provided for the second and third days. For details, please refer to Figure 16 and will not be detailed here.

[0186] In addition, the summary content also marks the collection content referenced by the travel guide, such as the "Content references from marked collections: x1 Historic District, B Palace Museum, P1 Resort Hotel" marked at the bottom of interface 2420, etc.

[0187] As shown in Figure 25, it is a schematic diagram of another content summarization process in an embodiment of the present application. When the target object enters "summary" in the search information input area of ​​the information search interface 2510 (the search box S2400 in Figure 25), the AI ​​will classify and summarize the content that the target object has searched within the memory range and display it in the form of multiple suggestions, as shown in S2401 and S2402 in Figure 25. If the target object chooses to click S2402 to view the words learned today, the AI ​​will summarize all the words that the object has searched today and display the summary content in the interface 2520.

[0188] Optionally, if the search topic is knowledge point learning, the summary content includes each knowledge point divided by importance, as well as a knowledge example that integrates each knowledge point. For example, today's new words, today's phrases, today's examples, etc. listed in interface 2520 in Figure 25 have important programs with one to three stars. In addition, all searched words can be combined into a paragraph of English, so that the target object can master all the words of the day by only memorizing a paragraph of English. As shown in S2404 in Figure 25, the six words searched today are intelligently integrated to generate a knowledge example shown in S2404.

[0189] In the above implementation, intelligent summarization by topic can help subjects categorize and summarize different types of content. Furthermore, for learning subjects, this method can be used to organize the knowledge they have searched for at a specific stage. For subjects working on complex strategy-based problems, if the entire strategy task is interrupted midway, this method can be used to summarize at any time and continue the dialogue with the AI ​​to complete the strategy.

[0190] It should be noted that the above is an introduction to the information search method in the embodiment of the present application from the client side. The following further describes the information search method in the embodiment of the present application from the server side:

[0191] FIG. 26 is a flowchart illustrating another information search method according to an embodiment of the present application. The method is executed by an electronic device, for example, the server 120 in FIG. 1 . The specific implementation process of the method is as follows: S261 to S263:

[0192] S261: Acquire contextual corpus information of the target object, where the contextual corpus information includes at least one of the following: search information input by the target object in each historical search operation, content browsed by the target object in each historical search result, and content of feedback provided by the target object to the browsed content.

[0193] In this embodiment of the present application, data can be continuously collected based on each search process for the target object to update the contextual corpus information of the target object. The contextual corpus information of the target object can be in the form of a database, that is, a personalized corpus. The construction of this personalized corpus is roughly divided into three steps: A. Data collection, B. Data preprocessing, and C. Corpus construction.

[0194] The following is a brief description of the above three processes:

[0195] A. Data collection.

[0196] The contextual corpus information is specifically divided into three categories: the search information of each search operation of the target object, the content browsed by the target object in each search result, and the content of the target object's feedback behavior.

[0197] Among them, the content that the target object browses in each search result and the content for which the target object provides feedback can be used as the target object's collection content.

[0198] The following is a brief description of the process of collecting the target object's favorite content: when collecting the content browsed by the target object in each search result, each content browsed by the target object can be collected, or it can be filtered based on the browsing time. For example, if the target object browses a web page content for more than a specified time, for example, 5 seconds, the web page content will be collected.

[0199] In addition, taking the feedback behavior of collection as an example, when the target object collects an article, the article is collected; when the target object collects a part of the content of an article (such as words, phrases, examples, paragraphs, etc.), the part is collected.

[0200] In the embodiment of the present application, the method of collecting partial content by the target object is not specifically limited, for example, the partial content can be automatically selected after double-clicking. Among them, the technical solution for selecting partial content is implemented based on the text selection engine of the client (the client is a browser as an example in the following).

[0201] For example, when the target double-clicks a word or a paragraph of text, the browser automatically selects the entire word or paragraph based on the target's selection range. Furthermore, the target can adjust the selection margin up or down to change the selected content. The specific implementation process is as follows, including steps Sa1 to Sa7:

[0202] Sa1: When the target object double-clicks a word or a text, the browser calculates the starting and ending positions of the selected text based on the selection range of the target object.

[0203] Sa2: The browser determines the range of the selected text based on the starting position and the ending position.

[0204] Sa3: The browser passes the selected text range to the text selection engine.

[0205] Sa4: The text selection engine traverses all text nodes in the document tree based on the range of the selected text and finds the text node that contains the selected text.

[0206] Sa5: The text selection engine locates the range of selected text in the text node and marks the style of the selected text.

[0207] Sa6: The browser applies the style of the selected text, such as the style of the entire paragraph selection, to the document so that the target object can see the selected text.

[0208] Sa7: If the target object touches the edge of the entire selection frame and moves, the end point of the target object's movement is determined, and the starting and ending positions of the text are recalculated based on the distance the target object moves as the new selection range of the target object.

[0209] It should be noted that the above is only an optional implementation method for selecting local content of a document based on the selection range of the target object listed in the embodiment of this application. Other related methods are also applicable to the embodiment of this application and will not be repeated here.

[0210] B. Data preprocessing.

[0211] In the embodiment of the present application, when constructing the contextual corpus information of the target object by collecting the above data, the collected data may also be preprocessed. The specific implementation method is as follows, including the following steps Sb1 to Sb4:

[0212] Sb1. Remove HTML tags: Use HTML parser or regular expressions to remove HTML tags from web page content, leaving only text content.

[0213] Sb2. Sentence segmentation: Divide the text content into sentences.

[0214] Optionally, a sentence tokenizer can be used to implement this. Common methods are rule-based or machine learning models. This is just a simple example and is not specifically limited in this article.

[0215] Sb3. Word segmentation: divide a sentence into words or phrases.

[0216] Optionally, you can use a word segmentation tool, such as the Natural Language Toolkit (NLTK) and spaCy, to perform word segmentation processing. This is just a simple example and is not specifically limited in this article.

[0217] Sb4. Part-of-speech tagging: tag each word with its part of speech, such as noun, verb, adjective, etc.

[0218] Optionally, a part-of-speech tagger can be used to implement this. Common methods are rule-based or machine learning models. This is just a simple example and is not specifically limited in this article.

[0219] C. Build a corpus.

[0220] After preprocessing the collected data using the above methods, we can construct contextual corpus information corresponding to the target object, also known as a personalized corpus. This is achieved by constructing the preprocessed data into individual text samples and storing them in a corpus. Each text sample can be a location, a word, a phrase, an example sentence, a paragraph, an article, or the content of a webpage.

[0221] Specifically, these text samples can be stored in a suitable text format, such as txt, csv, etc., which is not specifically limited in this article.

[0222] S262: After receiving the current search request triggered by the target search information, in response to the current search request satisfying the preset search conditions, the target search information and context corpus information are input into the target language model to generate the current search results and summary information of the current search results; the preset search conditions are: the current search operation is related to the historical search operations of the target object.

[0223] Refer to Figure 27, which is a flowchart illustrating a network search process implementing AI multi-round search in an embodiment of the present application. Specifically, after a target user enters a search term (query) into a search engine, the search engine returns search results. The target user then browses and saves the search results, storing them as a personalized corpus for the target user. The search term and personalized corpus are then used as context, combined with the target user's new question, and fed into the target language model. The target language model then generates a new answer based on the context.

[0224] In an embodiment of the present application, by combining the target language model, the target object's search terms in the last web search, the interface content browsed after the search, the content corresponding to the target object's active collection and like of the interface feedback behavior, etc. are used as contextual corpus information to create a personalized corpus, and the contextual corpus information generated by each search is provided to the target language model. When the target object asks a question again, the target language model will give a summary of the AI ​​answer in the suggestion interface (i.e., the information search interface) based on the acquired content. When the target object clicks on the AI ​​suggestion information, the target object can see the search answer generated by the target language model based on the target object's personalized corpus information, providing the target object with an experience that turns conventional web searches into multiple rounds of searches.

[0225] Among them, the above summary information is the first search suggestion information, which is a dynamic AI suggestion information. When the target language model determines that the target object's previous search behavior is related, the suggestion information will appear dynamically during the target object's search input process. Other entries in the suggestion interface are still suggested information completions for the current search term, and the target object can still conduct a network search. In other words, the target object can freely choose between network search and target language model search. The specific generation process of this information is as follows Sc1 to Sc5:

[0226] Sc1. Contextual understanding: After the target subject inputs a question (referring to the target search information), the target language model combines the target subject's question with the acquired personalized corpus (also known as context), understands the target subject's intention based on the context, and generates a suitable answer (i.e., the current search result).

[0227] Specifically, this process usually requires encoding the problem into a representation that the model can understand, such as word embedding or vector representation.

[0228] Sc2. Answer generation: The target language model generates an answer based on the context.

[0229] Specifically, it can be a model-based generation algorithm, such as a recurrent neural network (RNN) or a transformer, which generates appropriate suggestion information according to the semantics and grammatical rules of the context.

[0230] Sc3. Continue question suggestion generation: While generating answers, the target language model also generates some possible answer suggestion information. This process is usually based on the output of the generation algorithm and model.

[0231] Specifically, the generation algorithm can select the most appropriate ones from the candidate sentences as suggestion information based on the probability distribution or other rules generated by the model.

[0232] Sc4. Suggestion sorting and filtering: The generated suggestion information needs to be sorted and filtered to provide the most relevant and useful suggestions for the target object.

[0233] Specifically, the ranking can be implemented based on some indicators, such as the confidence of the probability distribution, semantic relevance, etc. The screening can consider some rules or heuristic methods, such as grammatical rules, common question patterns, etc.

[0234] Sc5. The suggested information generated by AI changes each time based on the changes in the search terms input by the target object.

[0235] Taking S702 in FIG. 7 or FIG. 22 as an example, these are examples of several types of first search suggestion information listed in the embodiments of the present application. For details, please refer to the above embodiments, and the repeated parts will not be repeated.

[0236] In the embodiment of this application, the current search results generated by the target language model, combined with the target search information and contextual corpus information, are AI personalized screening results that meet the needs of the object. Specifically, the results are generated in the following way:

[0237] For the target object, the target object can conduct multiple web searches. The target language model can cross-screen the target object's previous search terms to filter the current search results for the target object and display them in a suitable form based on the type of filtered content; or compare multiple search terms with each other and give the compared results.

[0238] The following still takes the above-mentioned travel scenario and word query scenario as examples to respectively describe the specific generation process of the corresponding current search results.

[0239] (1) Tourism scenes.

[0240] In a travel scenario, AI compares the target object's previous search terms and cross-screens them with previously collected places to provide the target object with search results that are more in line with their preferences. An optional implementation is as follows: If the target search information is related to the geographic location, the current search results can be generated in the following way: determine the collected places related to the geographic location in the target object's collection content; input the collected places into the target language model, and use the target language model to learn the characteristic information of the collected places; determine at least one recommended place for the target object based on the characteristic information and the historical behavior of the target object; for each recommended place, collect content related to the recommended place on the network, and extract the information of the recommended place and at least one associated web page from the collected content as the current search result.

[0241] Still using the travel scenarios listed in Figures 3 to 16 as an example, when the target searches for hotels in City A for the second time, the AI ​​​​uses the locations collected by the target in the personalized corpus as a screening basis to find hotels near the target location. The specific process is as follows Sd1 to Sd4:

[0242] Sd1. Input the location information collected by the target object into the target language model, allowing the target language model to learn the characteristics of these locations.

[0243] Sd2. Recommended hotels: Use collaborative filtering or content-based recommendation algorithms to recommend nearby hotels.

[0244] Specifically, the recommendation algorithm based on collaborative filtering refers to recommending items based on the object's historical behavior and the behavior of other objects.

[0245] For example, if subject A and subject B both like a hotel near a certain location, the system will recommend this hotel to subject C.

[0246] Specifically, content-based recommendation algorithms refer to recommending items based on the attributes of the items and the historical behavior of the target object.

[0247] For example, if the target audience likes hotels near a certain location, the system will recommend other similar hotels to the target audience.

[0248] Sd3. Filtering Online Articles: After recommending a hotel, we use web crawler technology to collect relevant articles from the web, such as reviews, ratings, and images. We then use natural language processing technology to analyze and process these articles, extracting useful information as a reminder for the target audience. We also obtain links to the original articles so that the target audience can view them.

[0249] Sd4. Combined with map display: Use the map application programming interface (API) to display the collected locations and recommended hotels on a map.

[0250] Still taking FIG. 8 as an example, S801 is a map module displayed using a map API, including collected locations and recommended hotels. In addition, as shown in FIG. 9 , S901 , S902 , and S903 are articles related to recommended hotels collected from the Internet. For details, please refer to the above embodiment, and repeated parts will not be repeated.

[0251] (2) Word query scenario.

[0252] In the word query scenario, AI compares the target object's previous search terms and lists the similarities and differences between the previous search terms and the current search terms for the target object. An optional implementation method is as follows: If the target search information is related to knowledge point learning, the current search results can be generated in the following way: the knowledge point to be searched corresponding to the target search information is input into the trained knowledge point recognition model to determine whether the associated knowledge points of the knowledge point to be searched have been searched by the target object; the knowledge point to be searched, the associated knowledge points and the interpretation of the knowledge point to be searched are determined as the current search results.

[0253] The following briefly describes the above method by taking the knowledge point to be searched as a word and the knowledge point recognition model as an antonym recognition model as an example. The specific process is as follows Se1 to Se6:

[0254] Se1. Data Collection: First, you need to collect a large number of corpus data containing words and their antonyms. This data can come from existing antonym dictionaries, synonym dictionaries, corpora, or manually annotated datasets.

[0255] Se2. Build a model: Use the collected data to train an antonym recognition model.

[0256] Specifically, machine learning algorithms such as support vector machines (SVM) or deep learning models such as RNN or Transformer models can be used, which are not specifically limited in this article.

[0257] Se3. Feature extraction: For each word, it needs to be converted into a feature representation that can be processed by the machine learning algorithm.

[0258] Specifically, a word vector model (such as Word2Vec or GloVe) can be used to convert words into vector representations, or other feature engineering methods can be used, which are not specifically limited in this article.

[0259] Se4. Training the model: Use the collected data and extracted features to train the antonym recognition model. You can use supervised learning methods, using words and labels indicating whether they are antonyms as training samples.

[0260] Se5. Model evaluation and optimization: Use the test dataset to evaluate the trained antonym recognition model, and optimize and adjust the antonym recognition model based on the evaluation results.

[0261] Se6. Application model: Apply the trained antonym recognition model to actual scenarios.

[0262] Specifically, when a target object queries a word, the antonym recognition model can determine whether its antonym has been input by the target object before. If it has been input before, the antonym recognition model will return the group of words and their related explanations.

[0263] Still taking FIG. 22 as an example, the target object query "artificial" and the previous query "naturally" are antonyms, and the corresponding related explanations are shown in the right interface of FIG. 22 . For details, please refer to the above embodiment, and the repeated parts will not be repeated.

[0264] Optionally, the search results interface also includes additional prompt information that matches the current search results. In an embodiment of the present application, the additional prompt information is: prompt information related to the date information and / or location information in the target search information. The additional prompt information can be determined in the following manner and then fed back to the client, and then displayed to the target object by the client. The specific process is as follows: extract the date information and the location information; based on the date information and the location information, generate prompt information corresponding to each prompt template according to the pre-set prompt templates; sort the generated prompt information according to the preset weight, and use the prompt information whose sorting results are within the specified order range as the additional prompt information; feed back the additional prompt information to the client for display in the search results interface.

[0265] Specifically, the server side may store two types of prompt templates: templates containing date information and templates without date information.

[0266] Among them, templates containing date information include but are not limited to: a weather template, b major holiday template, c travel restriction template, d travel clothing template; templates without date include but are not limited to: e place restriction information template (such as scenic spot travel restriction information template), f merchant welfare information template.

[0267] Each prompt template contains the following content:

[0268] a Weather template includes: local temperature and weather conditions within the date;

[0269] bMajor festival template includes: local festival event information within the date and date range;

[0270] c. The travel restriction template includes: local travel restriction information within a date and date range;

[0271] d. The travel clothing template includes: travel clothing recommendations based on the date and date range, combined with the temperature;

[0272] The e-place restriction information template includes: restricted entry information of relevant places;

[0273] The merchant's welfare information template includes: relevant location welfare information.

[0274] In this embodiment of the present application, when a target user views AI intelligent screening results, the date and location information of the target user's current related search process can be extracted and input into the search engine to sequentially obtain template information of af. The obtained template information is then sorted according to the weight of af. For example, if the specified order range is the top two, the top two templates can be displayed in the final search results interface; if there are fewer than two templates, one template can be displayed; if there are fewer than one template, no template can be displayed.

[0275] Among them, the preset weights corresponding to the prompt templates can be flexibly set according to actual needs. For example, the weights of af can be set to decrease in sequence, or the weights of af can be set to increase in sequence, etc. This article does not make specific restrictions.

[0276] Specifically, the additional prompt information is shown in S1001 and S1002 in FIG. 10 , S1101 in FIG. 11 , S1201 in FIG. 12 , and S1301 in FIG. 13 . For details, please refer to the above embodiments, and repeated parts will not be repeated here.

[0277] In the above embodiment, the display of additional prompt information helps the subject to understand more content at one time, reduces the subject's search path, and further improves the subject's information search efficiency this time.

[0278] S263: Feedback the summary information and current search results to the client, so that the client displays the first search suggestion information containing the summary information in the information search interface, and displays the search result interface containing the current search results in response to the viewing operation triggered by the first search suggestion information.

[0279] As shown in Figures 7, 8, 22, 23, etc., these are schematic diagrams of several types of first search suggestion information and corresponding search result interfaces listed in the embodiments of the present application. For specific implementation methods, please refer to the above embodiments, and the repeated parts will not be repeated.

[0280] In addition, the information search method in the embodiment of the present application also supports switching the AI ​​screening results to a conversational search mode. If the search result interface or the answer interface is set to also include an intelligent question and answer entry, the target object can switch the mode based on the intelligent question and answer entry.

[0281] In the embodiment of the present application, the answer information generated in the AI ​​dialogue is essentially the search results combined with the personalized corpus. After the target object switches to the AI ​​dialogue, each AI answer will combine the results of the network search, the target object's personalized corpus, and the target language model to collaboratively answer the question. The specific process is as follows Se1 to Se3:

[0282] Se1. Establish an index for the personalized corpus of the target object.

[0283] In the embodiment of the present application, after the personalized corpus has been established and its data has been cleaned and preprocessed, it is necessary to establish an index, which is a data structure used for fast retrieval of text data.

[0284] Optionally, you can use open source search engine libraries, such as Elasticsearch and Solr, to index the text data in the personalized corpus.

[0285] Se2. Develop a search plug-in to retrieve text data from the personalized corpus.

[0286] Specifically, the plug-in can use the API of the search engine library to retrieve text data in the personalized corpus through keywords and integrate the search results into the answer.

[0287] Se3. Integration into the target language model: Integrate the search plug-in into the target language model so that the plug-in can be called when answering questions and the contents of the personalized corpus can be displayed.

[0288] In an embodiment of the present application, the personalized corpus is combined with the plug-in technical solution during the AI ​​dialogue process to make the content generation of the target language model more flexible. In addition, the present application also supports the object to choose to turn on or off the plug-in to affect the generated content.

[0289] Optionally, this application also supports multimodal AI search for any content by the object. In any scene of the browsing result page, the object can ask questions about the selected content by long pressing the selected text or picture. The target language model will answer the question based on the selected content and the object's question. An optional implementation method is as follows: after the target object selects content for the current search result, the selected target content is determined according to the selection range of the target object; after detecting the voice of the target object inputting a question for the target content, the voice is converted into text; the text is subjected to semantic analysis and keyword extraction to obtain a text recognition result; the text recognition result and the target content are input into the target language model to obtain the answer information for the question generated by the target language model, and the answer information is fed back to the client so that the client displays an answer interface, which includes the target content, the question and the answer information.

[0290] Taking the client as a browser as an example, the process specifically includes the following steps Sf1 to Sf6:

[0291] Sf1. When the target object long presses a word or a piece of text, the browser calculates the starting and ending positions of the selected text based on the target object's selection range, and then locates and selects the text.

[0292] Sf2. Detect voice input events: Detect voice input events in the browser and respond promptly when the target object starts voice input.

[0293] Sf3. Calling the speech recognition API: When the target object starts voice input, the speech recognition API provided by the mobile phone operating system is called to convert the target object's voice input into text.

[0294] Sf4. Processing recognition results: The speech recognition API returns the recognized text and processes it, such as performing semantic analysis and keyword extraction, to more accurately match search results.

[0295] Sf5: Provide the text obtained in step Sf1 and the question obtained in step Sf4 to the target language model.

[0296] Sf6. Use the target language model to give an answer based on the input content.

[0297] As shown in FIG. 20 and FIG. 21 , please refer to the above embodiments for details, and repeated descriptions will be omitted.

[0298] In the above implementation, multimodal AI search for objects at any location is supported, which can shorten the path of object operation and achieve an experience of searching anywhere.

[0299] It should be noted that in the embodiments of the present application, the target object may also be interested in the local content of the corresponding search results during any search process, and AI intelligent collection can be performed for such content.

[0300] When the target object collects a paragraph of text, the location, name, and other content in that part of the content will be highlighted. For example, Figures 6, 18, and 19 above are examples of several AI smart collections listed in the embodiments of this application. The specific implementation methods can be found in the above embodiments, and the repeated parts will not be repeated here.

[0301] The following is a brief description of the specific implementation of AI smart collection:

[0302] When the target object collects a paragraph, you can label the content in the paragraph based on the following dimensions:

[0303] (1) Location tag, (2) Event tag, (3) Item tag, (4) Time tag.

[0304] The marking methods of the above four dimensions are described below:

[0305] (1) The marking method of location label is as follows:

[0306] In the embodiment of the present application, places can be divided into the following categories: A restaurant, B scenic spot, C store location, D hotel location, E other location. Specifically, it is necessary to label these types of places in the content and mark the number of occurrences of repeated places.

[0307] Specifically, there are many ways to label the above-mentioned locations and mark their occurrence counts. Here are two simple examples:

[0308] Method 1: Training a new location model. The specific implementation of this method includes the following steps Sg1 to Sg6:

[0309] Sg1. Data preparation: Collect travel guide data for various travel destinations and organize them into text format.

[0310] Alternatively, you can use a crawler tool to crawl travel guide data from travel websites or obtain it from other data sources.

[0311] Sg2. Data cleaning: Clean the collected strategy data to remove useless information and noise data.

[0312] Optionally, natural language processing technology can be used to process the strategy text through word segmentation, part-of-speech tagging, named entity recognition, etc. In an embodiment of the present application, the strategy text is pre-processed by word segmentation, part-of-speech tagging, etc., so as to better understand the structure and semantics of the strategy text. Among them, named entity recognition (NER) can be specifically implemented using a named entity recognition model, such as a conditional random field (CRF) based on machine learning or a recurrent neural network (RNN) model based on deep learning, to perform entity recognition on the article. These models will mark out named entities such as places, names, organizations, etc. in the strategy text.

[0313] Sg3. Location Type Classification: Use machine learning or deep learning algorithms, such as Naive Bayes, Support Vector Machines, and Convolutional Neural Networks, to classify the locations in the guide text. Categorize the locations into A: Restaurants, B: Attractions, C: Stores, D: Hotels, and E: Others. Train a classification model for each type.

[0314] Sg4. Model Evaluation and Tuning: Use the test set to evaluate the trained classification model. Evaluation metrics may include accuracy, recall, and F1 score. Based on the evaluation results, fine-tune the classification model, such as adjusting model parameters and increasing the amount of training data.

[0315] Sg5. Location tagging of the target object's collection content: Input the guide text into the tuned classification model, and the classification model will output the label of the location type and the number of occurrences of repeated attractions.

[0316] Sg6. Collection result display and storage: Visually display the output results.

[0317] Specifically, the recognized place noun is compared with the original text, the position where the place noun appears in the original text is determined, and a special style is displayed at the position, as shown in S502 in FIG5 .

[0318] In addition, different types of places can be marked with different colors in the animation display of the target object's collection process; corresponding labels can also be displayed for each type of place in the target object's favorites, and the number of occurrences can be marked next to repeated attractions.

[0319] Method 2: Information comparison. The specific implementation of this method includes the following steps Sh1 to Sh4:

[0320] Sh1. Location word extraction: For the text collected by the target object, natural language processing technology, including word segmentation, part-of-speech tagging, named entity recognition, etc., is used to process the guide text and extract the location words.

[0321] Specifically, you can use open source tools such as Jieba word segmentation, Stanford Named Entity Recognition (StanfordNER), etc.

[0322] Sh2. Identify location categories: For each location word, use network search or comparison with existing data to determine which location category it belongs to: A restaurant, B tourist attraction, C store location, D hotel location, or E other location.

[0323] Specifically, you can use a search engine to search for the place name and determine its type based on the information in the search results. You can also use existing place classification data, such as map software, to perform a comparison.

[0324] Sh3. Duplicate attraction identification: For places of the same type, similarity calculation and other technologies are used to identify duplicate attractions in the guide text.

[0325] Specifically, the location words can be similarly calculated using methods such as cosine similarity and Jaccard similarity to determine whether they are similar. If similar, they are considered to be repeated attractions, and their occurrence times are recorded.

[0326] Sh4. Collection result display and storage: Visually display the output results.

[0327] As above, the recognized place noun is compared with the original text to determine the position where it appears in the original text, and a special style is displayed at the position, as shown in S502 in FIG5 .

[0328] In addition, different types of places can be marked with different colors in the animation display of the target object's collection process; corresponding labels can also be displayed for each type of place in the target object's favorites, and the number of occurrences can be marked next to repeated attractions.

[0329] In summary, the retraining model method in the above method 1 can achieve higher accuracy and customization capabilities. The information comparison method in the above solution 2 can quickly realize the function.

[0330] (2) The method of marking event labels is as follows:

[0331] In the embodiment of the present application, the definition of events includes two categories: events on specific dates (such as historical anniversaries, festival celebrations and other public events), and events on non-specific dates (such as daily float parades, etc.).

[0332] Specifically, the method of event tagging is similar to that of location tagging. A new event model can be trained and used to mark major events that appear in the target object's collection content. Alternatively, information comparison methods can be used to compare network information and identify events in the content.

[0333] (3) The marking method of article labels is as follows:

[0334] In the embodiments of the present application, the definition of items includes but is not limited to specialty goods, specialty foods, specialty toys, etc.

[0335] Specifically, the methods of item tagging are similar to those of location tagging. A new item model can be trained and used to tag items that appear in the target object's collection. Alternatively, information comparison methods can be used to compare network information and identify items in the content.

[0336] (4) The marking method of the time label is as follows:

[0337] In the embodiments of the present application, a large language time information recognition model can be used to identify time information in text, including but not limited to time points, dates, seasons, years, etc. Furthermore, the model can be used to determine whether there are any associated attractions, events, or items at the time points in the text; for those associated attractions, events, or items, corresponding time tags are added.

[0338] It should be noted that the above is based on the example of a paragraph as the content collected by the target object. When the content collected by the target object is a word (or phrase, example sentence, etc.), the word (or phrase, example sentence, etc.) can be labeled in the following way: if the target object selects a word (or phrase), the word (or phrase) can be compared with the existing database to obtain the difficulty score of the word (or phrase); and then the score of the word (or phrase) is displayed in the form of a label in the relevant interface (such as the search result interface, the collection interface, etc.).

[0339] If the target object collects a sentence, the words in the sentence can be scored and the average score can be calculated as the sentence score; then the words, sentences and corresponding difficulty scores are stored together, and the sentence score is displayed in the form of a label in the relevant interface (such as the search result interface, collection interface, etc.).

[0340] It should be noted that the above is just a simple example of the content collected by the target object as paragraphs, words, phrases, sentences, etc. In addition, the content can also be in other forms, and the specific analysis method is similar, so the repetitions will not be repeated.

[0341] Optionally, the information search method in the embodiment of the present application also supports intelligent summarization of previously searched content by topic. When the object searches for content on multiple different topics, the target language model can summarize by topic, give multiple AI suggestions, and their corresponding AI results. An optional implementation method is as follows: after receiving a summary request triggered by the target object, determine the search topics corresponding to the multiple search information related to the target object respectively; summarize the search results of each search topic through the target language model to obtain the summary content corresponding to each search topic; score and sort each summary content, and filter out summary content with a score lower than a preset score threshold; generate second search suggestion information corresponding to each of the remaining summary contents, and feed back the second search suggestion information to the client, the second search suggestion information is the summary information corresponding to the corresponding search topic, so that the client responds to the summary operation and displays at least one second search suggestion information in the information search interface, and responds to the selection operation of the target search suggestion information in the at least one second search suggestion information, and displays the search result summary interface corresponding to the target search suggestion information, and the search result summary interface contains the summary content under the search topic corresponding to the target search suggestion information.

[0342] The specific technical solutions are as follows Sn1~Sn5:

[0343] Sn1. Judge different topics.

[0344] Optionally, you can use the following methods to determine the different subject contents in the search terms entered by the target audience:

[0345] A. Keyword matching: The target language model can extract keywords from the current question (i.e., the target search information) and match them with keywords in the context (i.e., the context corpus information). If the context corpus information contains keywords related to the current question, then this part of the context may be useful.

[0346] B. Semantic Similarity: The target language model can calculate the semantic similarity between the current question and the context. By comparing the semantic representations of the question and the context, it can determine which context has a stronger semantic connection with the current question, thereby determining its usefulness.

[0347] C. Topic Modeling: The target language model can use topic modeling techniques such as Latent Dirichlet Allocation (LDA) or pre-trained language representation models (Bidirectional Encoder Representation from Transformers, BERT) to identify topics in the context. If the topics in the context are highly relevant to the topic of the current question, then this part of the context may be useful.

[0348] Sn2. Tag the content by topic, and use the same tag for the content belonging to the same search topic;

[0349] Sn3. When the target object enters "summary", the target language model summarizes the different subject contents of the target object and summarizes the different subject contents based on the search subject.

[0350] Sn4, scoring and sorting, scoring and sorting the generated summary content, filtering out the content with low scores;

[0351] Here are some common scoring factors:

[0352] A. Keyword matching: Search engines check whether the keywords on a web page match the target object's search query. If the keywords appear frequently in the title, body text, and other tags, the web page may receive a higher score.

[0353] B. Content quality: Search engines evaluate the content quality of web pages, including the accuracy, completeness, and relevance of the information. Web pages with higher content quality usually receive higher scores.

[0354] C. External links: Search engines consider the number and quality of links to a web page from other websites. If other trusted websites link to the web page, the web page may receive a higher score.

[0355] D. Target audience experience: Search engines analyze the target audience's interaction with search results, such as click-through rate, dwell time, and bounce rate. If the target audience is interested in a search result and interacts with it, the webpage may receive a higher score.

[0356] E. Page structure and markup: Search engines analyze the structure and markup of web pages, such as title tags, paragraph tags, and image tags. Good page structure and markup can improve search engines' understanding of the content of the page, thereby affecting the score.

[0357] It should be noted that the above-mentioned several ways of scoring and sorting the generated summary content are just simple examples. In addition, other scoring methods are also applicable to the embodiments of the present application and will not be described in detail here.

[0358] Sn5. Each summary content generates a corresponding AI suggestion (i.e., the second search suggestion information in this article). The content of the AI ​​suggestion is a summary of the corresponding content, and the summary contains the keywords of the topic.

[0359] As shown in FIG. 24 and FIG. 25 , please refer to the above embodiments for details, and repeated descriptions will be omitted.

[0360] In addition, it should be noted that in the embodiments of the present application, the above-mentioned collected content can be used to generate summary content. Specifically, the collected content of the target object can be classified according to preset dimensions, and the classification labels corresponding to each collected content can be marked (see the above-mentioned embodiments for details, such as marking location labels, event labels, item labels, time labels, etc.); then, when the search results of different search topics are summarized by the target language model in units of search topics, and the summary content corresponding to each search topic is obtained, the target collected content related to the target search information in the collected content can be retrieved according to the classification labels; then, the search results of different search topics are summarized based on the target collected content in units of search topics through the target language model to obtain the summary content corresponding to each search topic.

[0361] The following uses the above-mentioned travel scenario and word search scenario as examples to briefly explain them respectively:

[0362] (1) Tourism scenes.

[0363] In an embodiment of the present application, if the target search information is related to a travel guide, the target language model can be used to generate corresponding summary content, such as a travel guide, by calling the location tag. An optional implementation is as follows:

[0364] First, the target object's collection content is analyzed to extract the collection locations related to the travel destination from the target object's collection content; then, based on the extracted collection locations, network information related to the collection locations is collected; combined with the target language model and network information, recommended locations are generated; finally, after the collection locations and recommended locations are formed into a location library, corresponding travel guides are generated based on the locations in the location library.

[0365] Specifically, the calling schemes of the location tags involved in the above embodiment are as follows Si1 to Si6:

[0366] Si1. Favorites content management: In the favorites of the target object, each location is marked with five location labels: A restaurant, B attraction, C shop location, D hotel location, E other, and the number of occurrences of repeated attractions is marked.

[0367] Si2. Extracting the target subject's favorite content: When the target subject requests a guide for a specific attraction, the target language model (such as chatGPT, the full name of which is Chat Generative Pre-trained Transformer) needs to analyze the target subject's favorite content and extract the favorite locations related to the destination.

[0368] Specifically, a machine learning algorithm or a rule engine may be used to analyze and extract the preferences of the target object.

[0369] Si3. Collect data: Based on the extracted relevant locations, collect relevant information from the network. Among them:

[0370] For Category A hotels, information such as the hotel's location, review scores, average spending per person, distinctive signs, and pictures was collected;

[0371] For Category B attractions, collect information such as the location, review ratings, ticket prices, opening hours, and pictures;

[0372] For Class C shops, collect information such as location, review rating, opening hours, average consumption per person, and pictures;

[0373] For Category D hotels, collect information such as location, room type, corresponding price, room status information, and pictures;

[0374] For Category E Other, collect information such as location, review ratings, and pictures.

[0375] Si4, Exploratory Recommendation: Combine the target language model and network information to generate more places based on the places of interest to the target object.

[0376] Optionally, the types generated in this step include but are not limited to the following two:

[0377] Method 1: Based on the location information of the favorite location, collect other travel locations within 1km.

[0378] Method 2: Based on the type of favorite places, use a recommendation algorithm to collect related types of attractions.

[0379] The details are as follows:

[0380] A. Based on the target's preferred category A hotel, collect information about attractions, shops, and hotels within 1 km of the hotel, prioritizing the locations the target has already bookmarked. Then, use the recommendation algorithm to find other hotels similar to category A.

[0381] B. Based on the address information of Category B attractions, we collect information about other nearby locations, including restaurants, shops, hotels, and scenic spots. We prioritize the attractions that the target user has already bookmarked. Then, based on the information about Category B attractions in the target user's preferences, we use recommendation algorithms and other technologies to generate recommendations for attractions that the target user may be interested in but is not yet familiar with. For example, if the target user has bookmarked natural scenery attractions, the recommendation algorithm can recommend related nature reserves or wildlife parks.

[0382] C. Based on the target's preferences for Category C stores, we collect information about attractions, restaurants, and hotels within 1 km of the store, prioritizing those already bookmarked by the target. We then use a recommendation algorithm to find other stores similar to Category C stores.

[0383] D. Based on the target's preferred hotel location, we collect attractions, restaurants, and shops within 1 km of the hotel, prioritizing those already bookmarked by the target. We then use the recommendation algorithm to find other hotels similar to Hotel D.

[0384] E. Based on the target object's preferences in category E, collect scenic spots, restaurants, shops, and hotels within 1 km of the location, and prioritize the locations that the target object has already collected.

[0385] Si5. Establish and sort the location library: The target language model uses the five types of locations collected by the target object, the locations generated by exploratory recommendations, and the guide locations found on the Internet as the location library, and generates a travel guide based on the locations in the location library.

[0386] The priority of the three source categories is: locations collected by the target > locations recommended through exploratory sources > locations recommended through the internet. Within these three source categories, the five location types are prioritized accordingly: the more frequently a location appears, the higher its priority.

[0387] Specifically, for the five location types listed above, locations within each location type are sorted according to their source priority. For example, for the three locations in Category A: Location A1, Location A2, and Location A3, which come from the target's favorites, exploratory recommendations, and online recommendations, respectively, the ranking result is: Location A1 > Location A2 > Location A3. Specifically, for locations from the same source, sorting can be random, by number of occurrences, by location name, and so on, without specific restrictions here.

[0388] Si6. Itinerary planning: Select 5 types of locations from the location library to generate a travel guide.

[0389] Optionally, the travel guide includes an itinerary plan corresponding to at least one travel time period; when generating a corresponding travel guide based on each location in the location library, the travel time set by the target object can be combined to obtain at least one travel time period; then, for each travel time period, based on the source priority corresponding to each location in the location library, a corresponding number of locations of corresponding categories are selected to generate an itinerary plan corresponding to the travel time period.

[0390] For example, a generated travel guide might include one to two B-category attractions, three A-category restaurants, zero to two E-category shops, one D-category hotel, and zero to two E-category other. The appropriate number of locations of corresponding categories is selected based on the priority order in the location database. Based on the specific number of days requested by the target audience and incorporating algorithms such as the target language model's route calculation, a plan for using event tags in the travel guide is generated.

[0391] Still taking S2403 in FIG. 16 and FIG. 24 as an example, these are examples of several travel strategies listed in the embodiments of the present application. For details, please refer to the above embodiments, and the repeated parts will not be repeated.

[0392] In an embodiment of the present application, if the travel guide requirement does not include specific time information, event suggestion information for the target object outside the itinerary plan can be generated based on non-fixed date events related to the travel destination in the collection content; if the travel guide requirement includes specific time information, and the time information coincides with the time of fixed date events related to the travel destination in the collection content, the fixed date events can be added to the corresponding itinerary plan.

[0393] Specifically, the event tag calling schemes involved in the above implementation are as follows Sj1 to Sj4:

[0394] Sj1: When the target object requests a strategy but does not include specific time information, search for events with non-fixed dates in the favorites;

[0395] Sj2. Provide event suggestions for the target audience in the form of tips for non-fixed date events related to the destination, outside of the itinerary;

[0396] Sj3. When the target object's requirements include specific time information, search for fixed-date events in the favorites;

[0397] Sj4. Determine whether the target person's travel date coincides with a fixed date event. If so, schedule the event in the itinerary.

[0398] In an embodiment of the present application, item tags of relevant places in the collection content can also be searched according to the travel strategy requirements proposed by the target object; then, based on the item tags, item recommendations for the target object outside the itinerary are generated; and shops where items can be purchased are determined through network searches, and the target shop locations within a preset distance range of the route formed by the found shops are added to the corresponding itinerary plan.

[0399] Specifically, the usage scenarios of the item tags involved in the above embodiment are as follows Sk1 to Sk3:

[0400] Sk1: The target person asks for a strategy and searches for item tags of relevant locations in the favorites;

[0401] Sk2: Provide suggestions for related items outside the itinerary in the form of tips, such as local specialties that can be purchased;

[0402] Sk3. Search online for shops where you can buy the item, find shops within 1km of your itinerary, and plan the shop locations in your itinerary.

[0403] In an embodiment of the present application, if the travel guide requirements raised by the target object include specific time information, the events, attractions, and items related to the time tags in the collection content are searched; the information of events, attractions, and items that meet the time tags is extracted, and based on the time information, unsuitable attractions, events, and items in the itinerary are screened out for the target object, and relevant suggestions are generated; and additional information is obtained through network search and provided to the target object.

[0404] Specifically, the usage scheme of the time tag involved in the above embodiment is as follows S11 to S14:

[0405] Sl1. When the target object's needs include specific time information, search for events, attractions, and items related to the time tag in the favorites;

[0406] S12. Extract information about events, attractions, and items that match the time tag, and filter out unsuitable attractions, events, and items in the itinerary for the target person based on the time information;

[0407] S13. Provide additional suggestions beyond the itinerary in the form of tips. For example, it is not suitable to go to location H in city X in March.

[0408] S14. You can also use online searches to obtain additional information, such as the weather corresponding to a date, and provide it to the target object in the form of tips.

[0409] (2) Word query scenario.

[0410] The specific summary process is as follows Sm1~Sm5:

[0411] Sm1. Collection content management: Establish a collection content management system in the APP and label each content according to the difficulty tag generated when collecting;

[0412] Sm2, retrieval of favorite content: Retrieve the content that the target object has favorited from the stored data;

[0413] Sm3, AI summarizes the historical content of the target object: Based on the browsing history, AI obtains all the words, phrases and sentences that the target object has searched today, and through the confirmation of the existing database, it assigns a difficulty score to each historical word and sentence;

[0414] Sm4. Label historical content and favorite content based on two dimensions: content type dimension: words, phrases, sentences; content difficulty dimension: one star to five stars;

[0415] Sm5. The target object can filter and summarize knowledge points through two dimensions.

[0416] Based on the same inventive concept, the present application also provides an information search device. As shown in FIG28 , it is a schematic diagram of the structure of the information search device 2800, which may include:

[0417] A first response unit 2801 is configured to display first search suggestion information in the information search interface in response to a current search operation triggered by a target object based on target search information satisfying a preset search condition, wherein the first search suggestion information includes summary information of the current search result; the preset search condition is that the current search operation is related to a historical search operation of the target object; and

[0418] The second response unit 2802 is used to display a search result interface corresponding to the target search information in response to a viewing operation triggered based on the first search suggestion information, wherein the search result interface includes a current search result, wherein the current search result is generated based on the target search information and the contextual corpus information of the target object; the contextual corpus information includes at least one of the following: the search information input by the target object in each historical search operation, the content browsed by the target object in each historical search result, and the content of the target object's feedback behavior on the browsed content.

[0419] Optionally, the search result interface also includes at least one answer operation control; the device also includes: a fourth response unit 2804, which is used to respond to the selection operation of the target answer operation control in the at least one answer operation control, perform the operation corresponding to the target answer operation control for the first answer information, and display the execution result of the operation.

[0420] Optionally, the device also includes: a fifth response unit 2805, which is used to identify the target content selected this time and the intelligent identification information corresponding to the target content in the search result interface in response to the content selection operation triggered based on the search result interface; wherein the intelligent identification information includes at least one of the following: key information contained in the target content, the type of the target content, and the importance of the target content.

[0421] Optionally, the device also includes: a sixth response unit 2806, which is used to display a collection interface in response to a viewing operation of the collection content of the target object, and mark different classification labels in the collection interface using different marking styles; wherein, for repeated classification labels, the number of recurrences is marked at a preset position in the collection interface.

[0422] Optionally, the device also includes: a summary unit 2807, which is used to display at least one second search suggestion information in the information search interface in response to a summary operation triggered by the target object, wherein each second search suggestion information corresponds to a search topic, and the second search suggestion information is summary information corresponding to the search topic; at least one search topic is obtained by dividing multiple search information related to the target object; in response to a selection operation on the target search suggestion information in the at least one second search suggestion information, a search result summary interface corresponding to the target search suggestion information is displayed, and the search result summary interface includes summary content of the search results corresponding to each search information under the search topic corresponding to the target search suggestion information.

[0423] Based on the same inventive concept, the present embodiment also provides another information search device. As shown in FIG29 , it is a schematic diagram of the structure of the information search device 2900, which may include:

[0424] An information acquisition unit 2901 is configured to acquire contextual information of a target object, wherein the contextual information includes at least one of the following: search information input by the target object in each historical search operation, content browsed by the target object in each historical search result, and content of feedback provided by the target object regarding the browsed content;

[0425] A result generating unit 2902 is configured to, upon receiving a current search request triggered by target search information, input the target search information and the context corpus information into a target language model in response to the current search request satisfying a preset search condition, and generate current search results and summary information of the current search results; the preset search condition being that the current search operation is related to a historical search operation for the target object;

[0426] Feedback unit 2903 is used to feed back the summary information and the current search results to the client, so that the client displays the first search suggestion information containing the summary information in the information search interface, and displays the search result interface containing the current search results in response to the viewing operation triggered by the first search suggestion information.

[0427] Optionally, the device also includes: an intelligent processing unit 2904, which is used to determine the selected target content according to the selection range of the target object after the target object selects content for the current search result; after detecting the voice of the target object inputting a question for the target content, convert the voice into text; perform semantic analysis and keyword extraction on the text to obtain a text recognition result; input the text recognition result and the target content into the target language model, obtain the answer information for the question generated by the target language model, and feed it back to the client, so that the client displays an answer interface, and the answer interface includes the target content, the question and the answer information.

[0428] Optionally, the device also includes: a summary unit 2905, which is used to determine the search topics corresponding to the multiple search information related to the target object after receiving the summary request triggered by the target object; summarize the search results of each search topic through the target language model to obtain the summary content corresponding to each search topic; score and sort each summary content, and filter out the summary content with a score lower than a preset score threshold; generate second search suggestion information corresponding to each remaining summary content, and feed back the second search suggestion information to the client, the second search suggestion information is the summary information corresponding to the corresponding search topic, so that the client responds to the summary operation and displays at least one second search suggestion information in the information search interface, and responds to the selection operation of the target search suggestion information in the at least one second search suggestion information to display the search result summary interface corresponding to the target search suggestion information, and the search result summary interface contains the summary content under the search topic corresponding to the target search suggestion information.

[0429] Optionally, the apparatus further includes: a collection unit 2906, configured to classify the target object's collection content according to a preset dimension and label each collection content with a corresponding classification label; the collection content includes at least one of the following: content browsed by the target object in each historical search result, and content for which the target object provides positive feedback on the browsed content;

[0430] The summary unit 2905 is specifically used to: retrieve the target collection content related to the target search information in the collection content according to the classification label; through the target language model, based on the search topic, summarize the search results of different search topics based on the target collection content to obtain the summary content corresponding to each search topic.

[0431] Based on the same inventive concept as the above-mentioned method embodiment, an electronic device is also provided in an embodiment of the present application. In one embodiment, the electronic device may be a server, such as server 120 shown in Figure 1. In this embodiment, the structure of the electronic device may be as shown in Figure 30, including a memory 3001, a communication module 3003, and one or more processors 3002.

[0432] Memory 3001 is used to store computer programs executed by processor 3002. Memory 3001 may mainly include a program storage area and a data storage area. The program storage area may store an operating system and programs required for running instant messaging functions, while the data storage area may store various instant messaging messages and operating instruction sets.

[0433] The processor 3002 may include one or more central processing units (CPUs) or digital processing units, etc. The processor 3002 is configured to implement the above-mentioned information search method when calling the computer program stored in the memory 3001 .

[0434] The communication module 3003 is used to communicate with terminal devices and other servers.

[0435] The specific connection medium between the memory 3001, communication module 3003, and processor 3002 is not limited in the embodiments of the present application. In Figure 30, the memory 3001 and processor 3002 are connected via bus 3004. Bus 3004 is depicted as a bold line in Figure 30. The connection methods between other components are merely illustrative and are not intended to be limiting. Bus 3004 can be divided into an address bus, a data bus, a control bus, etc. For ease of description, Figure 30 depicts only one bold line, but this does not indicate that there is only one bus or only one type of bus.

[0436] The memory 3001 stores a computer storage medium, which stores computer executable instructions for implementing the information search method of the embodiment of the present application. The processor 3002 is used to execute the above information search method, as shown in Figure 26.

[0437] In another embodiment, the electronic device may also be other electronic devices, such as the terminal device 110 shown in FIG1 . In this embodiment, the structure of the electronic device may be as shown in FIG31 , including: a communication component 3110, a memory 3120, a display unit 3130, a camera 3140, a sensor 3150, an audio circuit 3160, a Bluetooth module 3170, a processor 3180, and other components.

[0438] In some possible implementations, various aspects of the information search method provided in the present application can also be implemented in the form of a program product, which includes a computer program. When the program product is run on an electronic device, the computer program is used to enable the electronic device to execute the steps of the information search method according to various exemplary implementations of the present application described above in this specification. For example, the electronic device can execute the steps shown in Figure 2 or Figure 26.

[0439] These computer program commands can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, so that the commands executed on the computer or other programmable device provide steps for implementing the functions specified in one or more processes in the flowchart and / or one or more boxes in the block diagram.

[0440] Although the preferred embodiments of the present application have been described, those skilled in the art may make additional changes and modifications to these embodiments once they have learned the basic creative concept. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments and all changes and modifications that fall within the scope of the present application.

Claims

1. An information search method, executed by an electronic device, the method comprising: In response to a current search operation triggered by a target object based on target search information satisfying a preset search condition, displaying first search suggestion information in an information search interface, wherein, The first search suggestion information includes summary information of current search results; The preset search condition is: the current search operation is related to the historical search operations of the target object; and, In response to a viewing operation triggered based on the first search suggestion information, displaying a search result interface corresponding to the target search information, the search result interface including current search results, wherein, Generating the current search results based on the target search information and the context corpus information of the target object; The context corpus information includes at least one of the following: the search information input in each historical search operation of the target object, the content browsed in each historical search result of the target object, and the content of the feedback behavior implemented by the target object for the browsed content.

2. The method according to claim 1, wherein When the target search information is related to a geographical location, the search result interface further includes a map module; The map module is configured to display at least one of the following: a favorite location related to the geographical location in the favorite content of the target object, a recommended location related to the target search information determined based on a preset recommendation rule, and a driving route between the favorite location and the recommended location; The current search results include information of each recommended location and at least one web page associated with the recommended location.

3. The method according to claim 1 or 2, wherein The search result interface further includes a search information input area, and the step of displaying a search result interface corresponding to the target search information in response to a viewing operation triggered based on the first search suggestion information includes: In the search information input area, updating from displaying the target search information to displaying the target search information and the quantity of target favorite content; wherein the target favorite content is the content in the favorite content of the target object that is related to the current search operation; or In the search information input area, updating from displaying the target search information to the first search suggestion information.

4. The method according to any one of claims 1 to 3, wherein, The search result interface further includes additional prompt information matching the current search results; wherein the additional prompt information is: prompt information related to the date information and / or location information in the target search information.

5. The method according to any one of claims 1 to 4, wherein, The search result interface further includes a first intelligent Q&A entry; the method further includes: In response to a conversation operation triggered based on the first intelligent Q&A entry, switching from an initial web search mode to a conversation search mode; In response to a question operation for the current search results in the conversation search mode, displaying a first question input for the current search results and first answer information corresponding to the first question.

6. The method according to claim 5, wherein, The search result interface further includes at least one answer operation control; then the method further includes: In response to a selection operation on a target answer operation control among the at least one answer operation control, perform an operation corresponding to the target answer operation control on the first answer information, and display the execution result of the operation.

7. The method according to any one of claims 1 to 6, further comprising: In response to a content selection operation triggered based on the search result interface, identify the target content selected this time in the search result interface, and the intelligent recognition information corresponding to the target content; Wherein, the intelligent recognition information includes at least one of the following: key information included in the target content, the type to which the target content belongs, and the importance level of the target content.

8. The method according to claim 7, further comprising: In response to a question operation on the target content, display an answer interface, where the answer interface includes the target content, a second question input for the target content, and second answer information corresponding to the second question.

9. The method according to claim 8, wherein The answer interface further includes a second intelligent question-and-answer entry; the method further comprises: In response to a conversation operation triggered based on the second intelligent question-and-answer entry, switch from an initial network search mode to a conversation search mode; In response to a question operation on the second answer information in the conversation search mode, display a third question input for the second answer information, and third answer information corresponding to the third question.

10. The method according to claim 7, further comprising: Collect the target content, and display a collection animation in the search result interface; wherein, the collection animation indicates that the target content flies into a collection control in the search result interface.

11. The method according to any one of claims 1 to 10, further comprising: In response to a viewing operation on the collection content of the target object, display a collection interface, and in the collection interface, mark different classification labels with different marking styles; Wherein, for repeatedly occurring classification labels, mark the number of repetitions at a preset position in the collection interface.

12. The method according to any one of claims 1 to 11, further comprising: In response to a summary operation triggered by the target object, display at least one second search suggestion information in the information search interface, where Each second search suggestion information corresponds to a search theme, and the second search suggestion information is summary information corresponding to the search theme; At least one search theme is obtained by dividing a plurality of search information related to the target object; In response to a selection operation on a target search suggestion information among the at least one second search suggestion information, display a search result summary interface corresponding to the target search suggestion information, where the search result summary interface includes summary content of search results corresponding to each search information under the search theme corresponding to the target search suggestion information.

13. The method according to claim 12, wherein, If the search theme is a travel guide, the summary content is a travel guide including a travel itinerary corresponding to at least one travel time period, and the collection content referred to by the travel guide is marked. If the search topic is knowledge point learning, the summary content includes each knowledge point divided by importance level, and a knowledge example integrating each knowledge point.

14. An information search method, executed by an electronic device, the method includes: Obtaining context corpus information of a target object, the context corpus information including at least one of the following: search information input in each historical search operation of the target object, content browsed by the target object in each historical search result, and content of the feedback behavior implemented by the target object for the browsed content; After receiving a current search request triggered by target search information, in response to the current search request satisfying a preset search condition, inputting the target search information and the context corpus information into a target language model to generate a current search result and summary information of the current search result; The preset search condition is: the current search operation is related to the historical search operation of the target object; Feeding back the summary information and the current search result to the client, so that the client displays first search suggestion information including the summary information in an information search interface, and in response to a viewing operation triggered by the first search suggestion information, displays a search result interface including the current search result.

15. The method according to claim 14, wherein, If the target search information is related to a geographical location, the inputting the target search information and the context corpus information into the target language model to generate a current search result includes: Determining a collection location related to the geographical location in the collection content of the target object; Inputting the collection location into the target language model to learn feature information of the collection location by using the target language model; Determining at least one recommended location for the target object according to the feature information and the historical behavior of the target object; For each recommended location, collecting content related to the recommended location in the network, and extracting information of the recommended location and at least one associated web page from the collected content as the current search result.

16. The method according to claim 14 or 15, wherein, The search result interface further includes additional prompt information matching the current search result; The additional prompt information is: prompt information related to the date information and / or location information in the target search information; the method further includes: Extracting the date information and the location information; Generating prompt information corresponding to each prompt template based on the date information and the location information according to each preset prompt template; Sorting the generated prompt information according to a preset weight, and using the prompt information within a specified order range of the sorting result as the additional prompt information; Feeding back the additional prompt information to the client for display in the search result interface.

17. The method according to claim 14, wherein If the target search information is related to knowledge point learning and the target language model is a knowledge point recognition model, the inputting the target search information and the context corpus information into the target language model to generate a current search result includes: Input the knowledge points to be searched corresponding to the target search information into the trained knowledge point recognition model to determine that the related knowledge points of the knowledge points to be searched have been searched by the target object; Determine the knowledge points to be searched, the related knowledge points, and the explanations of the knowledge points to be searched as the current search results.

18. The method according to any one of claims 14 to 17, further comprising: After the target object makes a content selection for the current search result, determine the selected target content according to the selection range of the target object; After detecting the voice of the question input by the target object for the target content, convert the voice into text; Perform semantic analysis and keyword extraction on the text to obtain a text recognition result; Input the text recognition result and the target content into the target language model to obtain the answer information generated by the target language model for the question, and feedback it to the client, so that the client displays an answer interface, and the answer interface includes the target content, the question, and the answer information.

19. The method according to any one of claims 14 to 18, further comprising: After receiving the summary request triggered by the target object, respectively determine the search topics corresponding to multiple search information related to the target object; Through the target language model, summarize the search results of each search topic to obtain the summary content corresponding to each search topic; Score and sort the summary contents, and filter out the summary contents with scores lower than the preset score threshold; Generate the second search suggestion information corresponding to each remaining summary content, and feedback the second search suggestion information to the client. The second search suggestion information is the abstract information corresponding to the corresponding search topic, so that the client, in response to the summary operation, displays at least one of the second search suggestion information in the information search interface, and in response to the selection operation for the target search suggestion information in the at least one second search suggestion information, displays the search result summary interface corresponding to the target search suggestion information. The search result summary interface includes the summary content under the search topic corresponding to the target search suggestion information.

20. The method according to claim 19, further comprising: Classify the favorite content of the target object according to a preset dimension, and label the classification label corresponding to each favorite content; The favorite content includes at least one of the following: the content browsed by the target object in each historical search result, and the content for which the target object performs a positive feedback behavior for the browsed content; The step of, through the target language model, summarizing the search results of each search topic to obtain the summary content corresponding to each search topic includes: Retrieve the target favorite content related to the target search information in the favorite content according to the classification label; Through the target language model, taking the search topic as a unit, based on the target favorite content, summarize the search results of different search topics to obtain the summary content corresponding to each search topic.

21. An information search device, comprising: A first response unit, configured to display first search suggestion information in an information search interface in response to a current search operation triggered by a target object based on target search information satisfying a preset search condition, where the first search suggestion information includes summary information of current search results; the preset search condition is that the current search operation is related to the historical search operations of the target object; and, A second response unit, configured to display a search result interface corresponding to the target search information in response to a viewing operation triggered based on the first search suggestion information, where the search result interface includes current search results, and the current search results are generated based on the target search information and the context corpus information of the target object; the context corpus information includes at least one of the following: search information input in each historical search operation of the target object, content browsed in each historical search result of the target object, and content of feedback behaviors implemented by the target object for the browsed content.

22. An information search device, comprising: An information acquisition unit, configured to acquire context corpus information of a target object, where the context corpus information includes at least one of the following: search information input in each historical search operation of the target object, content browsed in each historical search result of the target object, and content of feedback behaviors implemented by the target object for the browsed content; A result generation unit, configured to, after receiving a current search request triggered based on target search information, in response to the current search request satisfying a preset search condition, input the target search information and the context corpus information into a target language model to generate current search results and summary information of the current search results; The preset search condition is that the current search operation is related to the historical search operations of the target object; and, A feedback unit, configured to feedback the summary information and the current search results to a client, so that the client displays first search suggestion information including the summary information in an information search interface, and in response to a viewing operation triggered based on the first search suggestion information, displays a search result interface including the current search results.

23. An electronic device, comprising a processor and a memory, wherein, The memory stores a computer program, and when the computer program is executed by the processor, the processor is caused to execute the steps of any one of claims 1 to 20.

24. A computer-readable storage medium, comprising a computer program, where when the computer program runs on an electronic device, the computer program is used to cause the electronic device to execute the steps of any one of claims 1 to 20.

25. A computer program product, comprising a computer program, where the computer program is stored in a computer-readable storage medium; when a processor of an electronic device reads the computer program from the computer-readable storage medium, the processor executes the computer program, so that the electronic device executes the steps of any one of claims 1 to 20.

Images