Information searching method and apparatus, device, medium and program product

Abstract

Embodiments of the present disclosure relate to an information searching method and apparatus, a device, a medium, and a program product. The method comprises: in response to determining that request content in a search request from a client is associated with a predetermined category, acquiring a result identifier for an answer result corresponding to the search request; and content of the answer result for the request content being generated by a machine learning model. The method further comprises, in response to acquiring partial content of the answer result based on the result identifier, sending a first response comprising the partial content to the client, used to display the partial content of the answer result. The method further comprises: in response to receiving from the client an acquisition request for acquiring remaining content of the answer result, acquiring remaining content generated by the machine learning model; and in response to acquiring the remaining content, sending a second response comprising the remaining content to the client, used to display the answer result. By means of the present method, a data recall rate is improved, and time consumption for searching information is reduced.

Description

Methods, apparatus, devices, media, and programs for searching information Technical Field

[0001] The embodiments of this disclosure generally relate to the field of information search technology, and specifically to methods, apparatuses, devices, media, and program products for searching information. Background Technology

[0002] Currently, when searching for information, the returned content is typically obtained by searching a database. For example, if a user wants to query information, they usually use a computing device to send a query request to a server running a search engine. The search engine can then search the database to return the corresponding information to the client. Therefore, users can obtain various types of information they need through various search engines.

[0003] Furthermore, with the development of search information technology, search algorithms are constantly improving. Therefore, search engines can more accurately understand user intent, filtering the most relevant results from massive amounts of data to provide users with highly accurate search answers, greatly satisfying users' requirements for accuracy and timeliness in information acquisition. Consequently, when users search for relevant content, they can find the accurate information they need in a shorter time.

[0004] Summary of the Invention

[0005] Embodiments of this disclosure provide a method, apparatus, device, medium, and program product for searching information.

[0006] According to a first aspect of this disclosure, a method for searching information is provided. The method includes, in response to determining that requested content in a search request from a client is associated with a predetermined category, obtaining a result identifier for an answer result to the search request. The method also includes generating answer result content for the requested content using a machine learning model. The method further includes sending a first response, including the partial content, to the client in response to obtaining the partial content of the answer result based on the result identifier for displaying the partial content of the answer result. The method also includes, in response to receiving a request from the client for obtaining the remaining content of the answer result, obtaining the remaining content generated by the machine learning model. The method further includes, in response to obtaining the remaining content, sending a second response including the remaining content to the client for displaying the answer result including the partial content and the remaining content.

[0007] According to a second aspect of this disclosure, a method for searching information is provided. The method includes sending a search request to a server, the search request including search content. The method further includes displaying the content of the search result in response to receiving a first response from the server for the search content, the search result being generated by a machine learning model based on determining that the search content is associated with a predetermined category. The method further includes sending a retrieval request to the server for retrieving the remaining content of the search result in response to determining that the content of the search result is a partial content of the search result. The method further includes displaying the search result including the partial content and the remaining content in response to receiving a second response from the server including the remaining content of the search result.

[0008] In a third aspect of this disclosure, an apparatus for searching information is provided. The apparatus includes a result identifier acquisition module configured to acquire a result identifier for an answer result to a search request in response to determining that requested content in a search request from a client is associated with a predetermined category; a content generation module configured to generate content for the answer result from the requested content using a machine learning model; a partial content display module configured to send a first response including the partial content to the client for displaying the partial content of the answer result in response to acquiring the partial content of the answer result based on the result identifier; a remaining content acquisition module configured to acquire the remaining content generated by the machine learning model in response to receiving an acquisition request from the client for acquiring the remaining content of the answer result; and a remaining content sending module configured to send a second response including the remaining content to the client for displaying the answer result including the partial content and the remaining content in response to acquiring the remaining content.

[0009] In a fourth aspect of this disclosure, an apparatus for searching information is provided. The apparatus includes a search request sending module configured to send a search request to a server, the search request including search content; a first content display module configured to display the content of the answer result in response to receiving a first response from the server for an answer result related to the search content, the answer result being generated by a machine learning model based on determining that the search content is associated with a predetermined category; an acquisition request sending module configured to send an acquisition request to the server for acquiring the remaining content of the answer result in response to determining that the content of the answer result is a partial content of the answer result; and a second content display module configured to display the answer result including the partial content and the remaining content in response to receiving a second response from the server including the remaining content of the answer result.

[0010] In a fifth aspect of this disclosure, an electronic device is provided, including at least one processor; and a storage device for storing at least one program, which, when executed by at least one processor, causes at least one processor to implement the method according to a first or second aspect of this disclosure.

[0011] In a fourth aspect of this disclosure, a computer-readable storage medium is provided having a computer program stored thereon that, when executed by a processor, implements the method according to a first or second aspect of this disclosure.

[0012] In a fifth aspect of this disclosure, a computer program product is provided. This computer program product includes a computer program that, when executed by a processor, implements the method according to a first or second aspect of this disclosure.

[0013] It should be understood that the content described in this section is not intended to limit the key or essential features of the embodiments of this disclosure, nor is it intended to restrict the scope of this disclosure. Other features of this disclosure will become readily apparent from the following description. Attached Figure Description

[0014] The above and other objects, features and advantages of this disclosure will become more apparent from the more detailed description of exemplary embodiments thereof taken in conjunction with the accompanying drawings, in which the same reference numerals generally represent the same parts.

[0015] Figure 1 illustrates a schematic diagram of an example environment in which some embodiments of the present disclosure may be implemented;

[0016] Figure 2 illustrates a schematic diagram of an example method for searching information according to some embodiments of the present disclosure;

[0017] Figure 3 illustrates a schematic diagram of another example method for searching information according to some embodiments of the present disclosure;

[0018] Figure 4 illustrates a schematic diagram of an example of a process for searching information according to some embodiments of the present disclosure;

[0019] Figure 5 illustrates a schematic diagram of an example of a process for obtaining answer results according to some embodiments of the present disclosure;

[0020] Figure 6 illustrates a schematic diagram of an example of a process for obtaining remaining content according to some embodiments of the present disclosure;

[0021] Figure 7 illustrates a schematic diagram of an example of a streaming protocol for searching information according to some embodiments of the present disclosure;

[0022] Figure 8 illustrates a schematic diagram of another example of a streaming protocol for searching information according to some embodiments of the present disclosure;

[0023] Figure 9 illustrates a schematic diagram of an example of the response results according to some embodiments of the present disclosure;

[0024] Figure 10 illustrates a schematic diagram of an example of a page without network anomalies for searching information according to some embodiments of the present disclosure;

[0025] Figure 11 illustrates a schematic diagram of an example of a weak network anomaly page for searching information according to some embodiments of the present disclosure;

[0026] Figure 12 illustrates a schematic diagram of an example of a fake network anomaly page for searching information according to some embodiments of the present disclosure;

[0027] Figure 13 illustrates a schematic diagram of an example of a server crash exception page for searching information according to some embodiments of the present disclosure;

[0028] Figure 14 illustrates a schematic block diagram of an apparatus for searching information according to some embodiments of the present disclosure;

[0029] Figure 15 illustrates a schematic block diagram of another apparatus for searching information according to some embodiments of the present disclosure;

[0030] Figure 16 illustrates a schematic block diagram of an example device suitable for implementing various embodiments of the present disclosure.

[0031] In the various figures, the same or corresponding reference numerals indicate the same or corresponding parts. Detailed Implementation

[0032] It is understood that the data involved in this technical solution (including but not limited to the data itself, the acquisition or use of the data) shall comply with the requirements of relevant laws, regulations and related provisions.

[0033] It is understood that before using the technical solutions disclosed in the various embodiments of this disclosure, users should be informed of the types, scope of use, and usage scenarios of the personal information involved in this disclosure in an appropriate manner in accordance with relevant laws and regulations, and user authorization should be obtained.

[0034] For example, upon receiving a user's proactive request, a prompt message is sent to the user to explicitly inform them that the requested operation will require the acquisition and use of the user's personal information. This allows the user to independently choose whether to provide personal information to the software or hardware, such as the electronic device, application, server, or storage medium performing the operations of this disclosed technical solution, based on the prompt message.

[0035] As an optional but non-limiting implementation, in response to a user's active request, sending a prompt message to the user can be done via a pop-up window, where the prompt message can be presented in text format. Furthermore, the pop-up window can also include a selection control allowing the user to choose "agree" or "disagree" to provide personal information to the electronic device.

[0036] It is understood that the above notification and user authorization process are merely illustrative and do not constitute a limitation on the implementation of this disclosure. Other methods that comply with relevant laws and regulations may also be applied to the implementation of this disclosure.

[0037] Embodiments of this disclosure will now be described in more detail with reference to the accompanying drawings. While some embodiments of this disclosure are shown in the drawings, it should be understood that this disclosure can be implemented in various forms and should not be construed as limited to the embodiments set forth herein. Rather, these embodiments are provided to provide a more thorough and complete understanding of this disclosure. It should be understood that the accompanying drawings and embodiments of this disclosure are for illustrative purposes only and are not intended to limit the scope of protection of this disclosure.

[0038] In the description of embodiments of this disclosure, the term "comprising" and similar terms should be understood as open-ended inclusion, i.e., "including but not limited to". The term "based on" should be understood as "at least partially based on". The term "one embodiment" or "the embodiment" should be understood as "at least one embodiment". The terms "first", "second", etc., may refer to different or the same objects. Other explicit and implicit definitions may also be included below.

[0039] There are still many problems to be solved in the process of searching for information. When users search for information on a client, such as a smartphone's search interface, they first enter the text of what they want to search for into the browser's search box, and then click on the search box. Subsequently, the search engine displays the search results on the search interface. Currently, search results related to the user's query are usually presented on the browser page. Additionally, related videos, images, etc., are also displayed as results on the browser page.

[0040] In the traditional approach described above, search results are typically retrieved directly from the database and returned to the search engine. For common user queries, the search engine can quickly return relevant content from the database. However, when a user query is a long-tail query with low pageview volume, the search engine may not be able to find suitable information to return because the database contains limited relevant content for this type of query. This leads to a decrease in search recall, and sometimes no search results are obtained, resulting in a poor user experience. Furthermore, in search scenarios, if a large language model is requested online to process the query, it can result in long data retrieval times and a poor user experience.

[0041] To address at least the aforementioned and other potential problems, embodiments of this disclosure propose a method for searching information. In this method, a server first determines whether the content of a search request from a client is associated with a predetermined category, such as a question-and-answer category. If the content is associated with the predetermined category, the server obtains a result identifier for the answer result in response to the search request from a machine learning model. Next, after obtaining the result identifier, the server further obtains a portion of the answer result generated by the machine learning model. The server then sends this portion of the content to the client for display. Subsequently, upon receiving a request from the client for retrieving the remaining content of the answer result, the server retrieves the remaining content generated by the machine learning model. Finally, the server sends the retrieved remaining content to the client to display the answer result including the remaining content and the portion of the content. This method, after determining that the content of a search request is associated with a predetermined category, returns a portion of the answer result generated by the machine learning model to the user. In subsequent operations, the remaining content is sent to the client to supplement the answer result. This solves the problem that the answer result cannot be provided to the user in a timely manner because the machine learning model cannot generate the answer result quickly, improves the recall rate for the search request, reduces the time spent on the search request, and improves the user experience.

[0042] The embodiments of this disclosure will now be described in further detail with reference to the accompanying drawings. Figure 1 illustrates an example environment in which the devices and / or methods of the embodiments of this disclosure may be implemented. In environment 100, server 102 first determines whether the request content 108 of a search request 106 from client 104 is associated with a predetermined category 110. If it is determined that the request content 108 of search request 106 is associated with the predetermined category 110, server 102 may use machine learning model 116 to process the request content 108 to obtain an answer result 112 and obtain a result identifier 114 for the answer result. For example, the answer result 112 may be displayed as a question-and-answer card for the search request. Then, after obtaining the result identifier 114, server 102 further uses the result identifier to obtain the content for the answer result 112 generated by machine learning model 116. If a portion 120 of the answer result 112 is obtained, server 102 sends the portion 120 to client 104 via a first response 124 to be displayed as the answer result 112 for search request 106. Subsequently, after receiving a retrieval request 118 from client 104 for retrieving the remaining content 122 for the answer result 112, server 102 retrieves the remaining content 122 generated by machine learning model 116. Finally, after retrieving the remaining content 122, server 102 sends the remaining content 122 to client 104 via a second response 126 to display the answer result 112, which includes the remaining content 122 and partial content 120, on client 104.

[0043] Examples of server 102 include, but are not limited to, personal computers, server computers, multiprocessor systems, minicomputers, mainframe computers, and distributed computing environments that include any of the above systems or devices.

[0044] Examples of client 104 include, but are not limited to, mobile phones (e.g., smartphones), laptop computers, portable digital assistants (PDAs), e-book readers, portable game consoles, portable media players, game consoles, set-top boxes (STBs), smart TVs (TVs), personal computers, laptop computers, in-vehicle computers (e.g., navigation units), etc.

[0045] The process of obtaining the answer result will be described in detail below. Server 102 first determines whether the request content 108 of the search request 106 from client 104 is associated with a predetermined category 110. For example, the predetermined category may include a question-and-answer category and / or an information / knowledge acquisition category, etc. In some embodiments, server 102 uses a category determination model to determine whether the request content is associated with a predetermined category. This category determination model may be trained using sample data. In some embodiments, the predetermined category is determined by determining whether a predetermined keyword exists among a set of keywords in the request content. In one example, if the keyword "how" exists in the request content, it can be determined that the request content is associated with the question-and-answer category. In another example, if the keyword "automobile development process" exists in the request content, it can be determined that the request content is associated with the information / knowledge acquisition category. The above examples are merely for describing this disclosure and are not intended to specifically limit this disclosure.

[0046] If the requested content is not related to a predetermined category, a conventional search method can be used. For example, a search engine can be used to retrieve relevant information from a database. If the requested content is related to a predetermined category, a machine learning model can be used to process the request content and generate an answer result. For example, this machine learning model could be a large language model. In addition to using the machine learning model to obtain the answer result, a normal database retrieval operation can also be performed through a search engine to retrieve relevant search information from the database.

[0047] Then, after determining that the requested content is associated with a predetermined category, the requested content can be sent to the search engine to obtain a result identifier 114 for the answer result. This result identifier 114 can be used to retrieve the content for the answer result 112 from the repository storing the content generated by the machine learning model 116. If the machine learning model 116 processes quickly, it can obtain the full content of the answer result. At this point, the full content and search results can be directly returned to the client for display as the answer result. However, the machine learning model usually takes a long time to generate the answer result, and when returning the answer result together with the search results obtained by the search engine, the machine learning model 116 may only have generated a portion 120 of the answer result. In this case, the server 102 will form a first response 124 with the obtained portion 120 and other search results and return it to the client 104 for display.

[0048] In some embodiments, after receiving a first response, the client displays the content of the response and / or the search results. Additionally, this can be displayed in a combination of one or more of the following formats: text, images, and video.

[0049] If the response does not contain all the information, client 104 will further send a request 118 to server 102 to retrieve the remaining content of the response. The server will then retrieve the remaining content 122 generated by machine learning model 116 and generate a second response 126 to provide the remaining content to the client for display. Additionally, the process of retrieving this remaining content is implemented via a streaming protocol, such as sending the request 118 and the server 102 providing a streaming response to client 104.

[0050] Finally, after obtaining the remaining content 122, the client displays the remaining content 122 and part of the content 120 as the answer result 112 on the client's search page as a search result.

[0051] This method, after determining that the content of a search request is associated with a predetermined category, returns a portion of the answer result generated by the machine learning model to the user. In subsequent operations, the remaining content is sent to the client to supplement the answer result. This solves the problem that the answer result cannot be provided to the user in a timely manner because the machine learning model cannot generate the answer result quickly, improves the recall rate for the search request, reduces the time spent on the search request, and improves the user experience.

[0052] The foregoing description, with reference to FIG1, illustrates an example environment in which devices and / or methods of some embodiments of the present disclosure may be implemented. The following description, with reference to FIG2, illustrates an example method for searching information according to some embodiments of the present disclosure. The method in FIG2 may be executed by the server 102 in FIG1 or any suitable device.

[0053] As shown in Figure 2, in example method 200, at box 202, after receiving a search request from the client, the server 102 first determines whether the requested content in the search request from the client is related to a predetermined category.

[0054] Upon receiving a search request, server 104 may optionally perform a check. For example, server 104 may retrieve whether the content of the search request meets predetermined requirements. For instance, whether the keywords in the request content are restricted or prohibited keywords. In one example, server 104 stores a keyword database containing restricted keywords or keywords not allowed during the search process. This check is then performed by comparing the keywords in the request content with those in the keyword database. If the request content contains keywords that do not meet the predetermined requirements, server 104 generates a corresponding message for the search request and sends it to the client for display to the user, such as indicating that the search cannot be performed or that the search content does not meet the requirements.

[0055] Server 104 can directly determine whether the content of a search request is related to a predetermined category, or it can determine whether the content of a search request is related to a predetermined category after the search request has passed inspection. For example, server 104 can use a category determination model to predict whether a search request is related to a predetermined category. If the prediction result is greater than a predetermined threshold, it is determined that the content of the book search request is related to the predetermined category.

[0056] In some embodiments, the answer results provided by the server to the client in response to a search request can be implemented in various forms. In one example, the generated answer result is text information. In another example, the generated answer result may include text information and image information. In yet another example, the generated answer result may include video information. The above examples are merely for describing this disclosure and are not intended to specifically limit this disclosure.

[0057] If the content of a search request from a client is associated with a predefined category, then at box 204, server 102 obtains the result identifier for the answer to the search request. After determining that the request content is associated with a predefined category, the request content can be sent to a search engine. In addition to searching its database using the request content, the search engine can further provide the request content to a machine learning model, such as a large language model, to obtain the answer to the request content. At this point, the search engine can obtain the result identifier for that answer. For example, if the answer is a question-and-answer card, the document identifier of the question-and-answer card is obtained.

[0058] Then, at box 206, the machine learning model generates the content of the response to the request. As described above, after determining that the request content is associated with a predetermined category, the server sends the request content to the machine learning model for processing. After obtaining the result identifier for the response, this result identifier can be used to retrieve the content corresponding to the result identifier from the repository used to store the content generated by the machine learning model, such as the full or partial content of the response.

[0059] At box 208, server 102 determines whether partial content of the answer result has been obtained based on the result identifier. If the corresponding content is obtained through the result identifier, it determines whether the obtained content is complete or partial. If the complete content of the answer result is obtained, it indicates that the neural network model has completed the task. In this case, the server can directly return it to the client along with the search results. If it is determined that partial content of the answer result has been obtained, it indicates that the neural network model has not yet completed the task. In this case, at box 210, server 102 still sends a first response including partial content of the answer result to the client for displaying that partial content.

[0060] When retrieving content corresponding to the result identifier from the repository, server 102 can also obtain an identifier indicating whether the content is complete. The server can then send this identifier along with the retrieved content to client 104.

[0061] After receiving the content of the answer result and an indicator indicating whether the content is complete from the server, the client will display the complete content of the answer result along with other search results if the indicator determines that the content is complete. If the indicator determines that the content received by the client is incomplete, i.e., only part of the content is obtained, the client can send a request to the server to retrieve the remaining content.

[0062] In some embodiments, after obtaining the content, it is displayed directly on the client, and may be displayed in text, image, and / or video format. In some embodiments, if the client experiences network problems, an error message indicating that no answer result can be displayed will be shown on the server, and the client will automatically resend the search request to obtain the answer result.

[0063] At box 212, server 102 may further determine whether it has received a request from the client to retrieve the remaining content for the answer result. If a request to retrieve the remaining content for the answer result is received from the client, at box 214, server 102 retrieves the remaining content generated by the machine learning model.

[0064] Upon receiving a request to retrieve the remaining content, server 104 parses the request to obtain a result identifier for the response. Then, based on this result identifier, it retrieves the remaining content for the result identifier from the repository of content generated by the machine learning model.

[0065] In some embodiments, before parsing the retrieval request, the server may first check the parameters in the retrieval request to ensure that the parameters meet predetermined requirements. For example, the server may determine whether the retrieval request includes a result identifier for the answer result. If the retrieval request does not contain this result identifier, the server cannot obtain the information. Additionally, it may also detect whether the retrieval request includes information such as a user identifier. Then, after the retrieval request passes the check, it can be parsed. For example, to ensure data security, the result identifier in the retrieval request is usually encrypted, and the result identifier can be decrypted when parsing the retrieval request.

[0066] Once the result identifier in the request is obtained, the server will further generate a polling task for the answer based on the result identifier. This polling task is used to continuously query the repository used to store the content generated by the machine learning model, thereby continuously obtaining the remaining content for the result identifier.

[0067] Additionally, to better present more useful information to users on the client side, after retrieving the remaining content and completing the polling task for the answer results, the server can further obtain a set of recommended words for the requested content, for example, by obtaining recommended words corresponding to the requested content through a recommendation word model or recommendation word library. Then, the server sends the determined set of recommended words to the client to be displayed along with the answer results.

[0068] Finally, at box 216, server 102 determines whether the remaining content has been retrieved. If the remaining content has not been retrieved, it can be retrieved again. If the remaining content has been retrieved, at box 218, the server sends a second response, including the remaining content, to the client to display the answer result, which includes both the partial and the remaining content.

[0069] In some embodiments, the server may determine a streaming protocol that can be used to transmit data between the client and server for data streaming. Since generating the remaining content takes a relatively long time, streaming can greatly alleviate the problems of insufficient computing resources and poor user experience caused by the time spent generating content.

[0070] If the remaining content can be retrieved, the server will further determine whether to use the streaming protocol to send the remaining content in the response result to the client for display. If the remaining content cannot be retrieved, the server can also send a streaming response to the client according to the streaming protocol to provide the client with fault information and the cause of the fault.

[0071] Since part of the answer results has already been displayed, and the remaining content is transmitted via a streaming protocol and displayed on the client in a streaming format, the results can be presented in a streaming manner because the partial content is retrieved faster than the remaining content. This improves the user experience when performing a search request.

[0072] After successfully retrieving the remaining content, pre-defined operations can be set for the remaining content as needed. In some embodiments, the remaining content can be added in a streaming manner after a portion of the answer result. In other embodiments, some content in the partial content can be replaced with the remaining content, for example, if some content in the partial content is incorrect or does not match a predetermined category, that portion of the content can be replaced with the remaining content. In still other embodiments, some or all of the partial content can be deleted.

[0073] In some embodiments, when the remaining content is not successfully retrieved, the server first determines the fault information of the failure to retrieve the remaining content, for example, by determining the fault code of the fault information to determine the specific cause of the fault, and then sends the fault information to the client through a response message.

[0074] In some embodiments, the server can add likes, dislikes, etc., to the answer results, and these likes and dislikes can be stored locally on the client in real time. Additionally, the likes and dislikes can be updated by refreshing the answer results.

[0075] In some embodiments, keywords in the answer results can also be clicked; for example, long-pressing a keyword can initiate a new search for it.

[0076] In some embodiments, when the content in the answer result cannot be displayed or cannot be displayed due to network issues, an error page message will be displayed in the answer result, such as "Network error, unable to display this page." Additionally, the error page message will vary depending on the specific error caused by different circumstances.

[0077] In some embodiments, the border of the answer result can also be adaptively adjusted, for example, by determining the total height of the border of the answer result based on the total height of the text in the answer result.

[0078] In some embodiments, animation effects can also be added to the answer results. For example, when obtaining the relevant part or remaining content of the result identifier, relevant animation effects can be added to the answer results so that users do not feel bored while waiting.

[0079] This method, after determining that the content of a search request is associated with a predetermined category, returns a portion of the answer result generated by the machine learning model to the user. In subsequent operations, the remaining content is sent to the client to supplement the answer result. This solves the problem that the answer result cannot be provided to the user in a timely manner because the machine learning model cannot generate the answer result quickly, improves the recall rate for the search request, reduces the time spent on the search request, and improves the user experience.

[0080] The above description, with reference to FIG2, illustrates an example method for searching information on a server side according to some disclosed embodiments. The following description, with reference to FIG3, illustrates an example method for searching information on a client side according to some disclosed embodiments. Method 300 can be executed by the client 104 of FIG1 or any suitable computing device.

[0081] In Example 300, at box 302, the client sends a search request to the server after receiving user input. The search request includes the search terms. When searching for information on the client, the user can enter the search terms in the search interface. Then, the client 104 can generate a search request based on the user's input and send the search request to the server 102.

[0082] At box 304, client 104 determines whether it has received a first response from the server regarding the search results. If a first response is received from the server, client 104 displays the content of the response at box 306. This response is obtained after server 102 has confirmed that the requested content is associated with a predetermined category, and then sends the requested content to a machine learning model for processing.

[0083] In some embodiments, the first response received from the server may also include a set of search results. The client retrieves the set of search results and the answer results from the first response. The answer results and the set of search results are then presented to the client's user.

[0084] At box 308, the client further determines whether the content of the response is a partial response. In some embodiments, the first response may include an integrity flag set by the server to indicate whether the content of the response is complete. The client can obtain this integrity flag from the first response. The client then determines whether the content of the response is a partial response based on the integrity flag. For example, the integrity flag may be a first predetermined value indicating completeness, or a second predetermined value indicating incompleteness.

[0085] If it is determined that the content of the response is only a part of the response, at box 310, client 104 sends a request to server 102 to retrieve the remaining content of the response. At this time, the client can also retrieve the result identifier of the response from the first response and then add this result identifier to the retrieval request. Additionally, the client can encrypt the result identifier to generate an encrypted result identifier and then include the encrypted result identifier in the retrieval request. Additionally, client 104 can also include the user identifier and the request content in the retrieval request.

[0086] At box 312, client 104 determines whether it has received a second response from server including the remaining content of the answer result. If it determines that it has received a second response from server 102 including the remaining content of the answer result, at box 314, client 104 displays the answer result including the partial content and the remaining content.

[0087] In some embodiments, the second response may further include a set of operations for the remaining content. For example, add, replace, and / or delete operations. Therefore, when displaying an answer result including partial and remaining content, the client can also obtain the set of operation identifiers, then use the set of operations corresponding to the set of operation identifiers to process the remaining and partial content accordingly, and then display the processed content.

[0088] The above description, in conjunction with Figures 2 and 3, illustrates examples of processes for searching information according to some embodiments of the present disclosure. The following description, in conjunction with Figure 4, illustrates examples of processes for searching information according to some embodiments of the present disclosure. The example in Figure 4 can be executed by the server 102 shown in Figure 1 or any suitable device.

[0089] In Example 400, search module 406 can receive a search request from a client. At box 408, search module 406 obtains a result identifier by communicating with search engine 422. During this process, the search module can determine whether the requested content in the search request is associated with a predetermined category. The requested content is then sent to search engine 422 for searching. If the requested content is not associated with a predetermined category, search engine 422 returns the content retrieved from its database to the client. If the requested content is associated with a predetermined category, search engine 422 can further utilize machine learning model engine 426 to process the requested content, for example, using a large language model engine. Thus, search engine 422 can return a result identifier of the answer generated for the requested content to search module 406. Then, at box 410, search module 406 pre-fetches the content of the question-and-answer results. During this process, search module 406 retrieves the corresponding content from repository 424, which stores the content generated by machine learning model engine 426, using the result identifier. Search module 406 then returns the retrieved answer content to the client for display. In addition to the content of the answer result, the search module 406 may also send a patch operation context related to the answer result to the client. This patch operation context includes the result identifier and an identifier indicating whether the returned content is complete.

[0090] If the client determines, through the patch operation context, that it has obtained the complete content of the answer result, the complete content will be displayed on the client at box 402. In this case, in addition to displaying the complete answer result, other information retrieved from the database by the search engine can also be displayed simultaneously. For example, the answer result can be displayed first, and the search results can be displayed later.

[0091] If the client determines, based on the patch context, that it has not obtained the complete content of the response result, it will initiate a patch operation to retrieve the remaining content of the response result. The client then sends a retrieval request corresponding to the patch operation, which includes return parameters. These return parameters may include an encrypted result identifier.

[0092] When the server's streaming module 412 receives the retrieval request, it parses the parameters at box 414. For example, the server decrypts the encrypted result identifier to determine the result identifier used to retrieve the remaining content. Then, at box 416, it polls to retrieve the answer result content. At this point, the server further uses the retrieved result identifier to generate a polling task for the answer result, which continuously retrieves the remaining content for the result identifier from the repository of content generated by the machine learning model.

[0093] After retrieving the remaining content and completing the polling task for the answer results, a set of recommended words can be retrieved at box 418. This set of recommended words can be sent to the client for display in the answer results. In one example, the retrieval request includes the user's search request content, and the recommended words are generated by the recommended word platform 428 based on the request content.

[0094] Then, at box 420, the polling result is returned. The server will return the remaining content. During this process, the server returns the remaining content to the client in streaming form via a streaming protocol. After receiving the remaining content, the client displays a 404 error in its response.

[0095] Therefore, the above process can avoid the filtering of some answer results due to excessively long writing times for machine learning models, such as large language models.

[0096] The above description, with reference to FIG4, is a schematic diagram of an example flowchart for searching information according to some embodiments of the disclosure. The following description, with reference to FIG5, is a schematic diagram of an example flowchart for obtaining answer results for searching information according to some embodiments of the present disclosure. Example 500 is an example of the search module in Example 400.

[0097] In Example 500, the search module 502 can receive a search request from the client 104. At box 504, the search module performs a pre-check on the request content. For example, it checks for sensitive words or restricted words. If sensitive words or restricted words are present, the search request will fail the pre-check. In this case, an error message response will be returned to the client, such as "This search request cannot be executed" or "Server error."

[0098] After passing the preliminary checks, at box 506, search module 502 obtains the result identifier through search engine 508. During this process, search module 502 can further determine whether the requested content is related to a predetermined category. For example, search module 502 can detect whether the requested content is related to a question-and-answer category or a knowledge acquisition category. In other words, search module 502 is used to determine the likelihood that the requested content can be answered by a machine learning model engine. If the search content is not related to a predetermined category, the search content is sent to the search engine for search retrieval. The search results are then returned to the client for presentation to the user.

[0099] If the search content is determined to be related to a predetermined category, the search module 502 sends the search content to the search engine 508, which then sends it to the machine learning model engine for processing. Since the machine learning model engine needs time to process the search content and provide an answer, the search engine 508 obtains the result identifier corresponding to the answer generated by the machine learning model engine. Additionally, the search engine 508 can further obtain search results obtained from its database. Therefore, the search engine can return the result identifier and the corresponding search results to the search module 502. After obtaining the result identifier, the search module 502 begins generating the answer result at box 510. During this process, the search module 502 parses the result identifier at box 512. For example, it retrieves the result identifier from the information returned by the search engine, including the result identifier. Then, it sends the result identifier to the repository corresponding to the machine learning model engine 514 to obtain the content of the answer result. Then, at box 516, the search module 502 parses the answer result and generates streaming information. For example, this streaming information can be the text of the answer content. Next, the search module further generates a response 518 based on the obtained content. Then, at box 520, return a response to client 104. In the response, you can add an indicator indicating whether the answer is a complete response.

[0100] The above description, with reference to FIG. 5, illustrates an example of a flow for obtaining answer results according to some embodiments of the present disclosure. The following description, with reference to FIG. 6, illustrates an example of a flow for obtaining remaining content according to some embodiments of the present disclosure. Example 600 is directed to the streaming module in Example 400.

[0101] The streaming module 602 can receive a retrieval request from a client for retrieving remaining content. This streaming module 602 includes an Application Programming Interface (API) layer. In this layer, after receiving the retrieval request for remaining content, a pre-check is performed on the retrieval request at box 606. This pre-check checks the parameters in the retrieval request accordingly. For example, it verifies whether the parameters in the retrieval request conform to a format, performs security checks on the parameters, and checks whether it contains parameters necessary to parse the retrieval request. For example, it checks whether the retrieval request contains information such as a result identifier and / or a user identifier. If the retrieval request fails the pre-check, the streaming module returns a response indicating that information cannot be retrieved.

[0102] If the retrieval request passes the pre-check, the request is parsed at box 608. For example, the result identifier in the retrieval request is encrypted to prevent information leakage. The encrypted result identifier is decrypted during the parsing of the retrieval request to obtain the result identifier.

[0103] After obtaining the result identifier, since the machine learning model engine takes a relatively long time to generate the answer, a pre-wait can be performed at box 610 to wait for the machine learning engine 622 to generate the result content. This operation can be asynchronous. Then, at box 612, the streaming module creates a loop task. This loop task then interacts with the machine learning model to poll for the answer at box 620. At this time, the machine learning model engine can poll for the remaining content of the answer result from the repository 624 used to store the content in the machine learning model engine 622. At box 614, the streaming module waits and obtains the polling result. During this process, for example, a channel for data interaction can be sent to the machine learning model to realize data interaction.

[0104] Additionally, the server will return a set of recommended words 616, which is obtained from the recommended word platform 618. Finally, the remaining content and the set of recommended words are streamed back to the client for display via a streaming protocol.

[0105] In some embodiments, the pre-waiting, returning a set of recommended words, and polling for answers are performed asynchronously, while the other processes are performed synchronously. By executing these three processes asynchronously, the return and display of partial content are not affected while waiting for the remaining content to be returned, thus improving the user experience while waiting for the remaining content.

[0106] The above diagram, with reference to FIG. 6, illustrates an example of a process for retrieving remaining content according to some embodiments of the present disclosure. The following diagram, with reference to FIG. 7, illustrates an example of a streaming protocol for searching information according to some embodiments of the present disclosure.

[0107] In Example 700, the answer result can be a question-and-answer card. As shown in box 702, it includes the StreamEventResponse function, the Event function, the ContentPatch function, the ContentOperation function, and the FeaturedAnswerStruct function. The patch represents the remaining content.

[0108] The StreamEventResponse function defines three status codes: 10001, which indicates that the request engine failed; 10002, which indicates that the parsing data failed; and 10003, which indicates that the timeout occurred.

[0109] The Event function defines two states for obtaining the remaining content: EVENT_PATCH=1, which means the remaining content was successfully sent, and EVENT_OP_FAIL=2, which means the remaining content was sent but failed. The specific reason for failure is defined according to the status code.

[0110] In the `ContentPatch` function, `struct_type = 1` represents the content to be modified. Additionally, the `ContentOperation` function defines three operation types: add, replace, and remove, which are `add`, `replace`, and `remove` respectively.

[0111] In the `StreamEventResponse` function, `desc=1` represents the card display style, where 0 = precise Q&A card and 1 = style enhancement card. `status=3` represents the streaming interface operation code, where `status=1` indicates that all Q&A card data is sent this time, and `status=2` indicates that operations are needed for the remaining content. `stream_param=4` represents the fields that the remaining content interface needs to return. In the optional bool `is_stream=5`, a value of 1 indicates that streaming is needed, and a value of 0 indicates that streaming is not needed.

[0112] The above illustration, with reference to FIG. 7, depicts an example of a streaming protocol for searching information according to some embodiments of the present disclosure. The following illustration, with reference to FIG. 8, depicts another example of a streaming protocol for searching information according to some embodiments of the present disclosure.

[0113] In Example 800, as shown in Box 802, this code example is divided into two parts. The first part of the code represents sending the remaining data via patching, that is, sending the remaining content to the client. The second part of the code represents sending the final data via patching, using different operations for different fields, such as performing a replacement operation on the status and setting it to 1.

[0114] The following is a schematic diagram illustrating an example of the response results according to an embodiment of the present disclosure, with reference to FIG9.

[0115] In Example 900, the search box 902 contains the request content "How to sing?". It is determined whether this request content is associated with a predefined category, for example, using a category determination model.

[0116] After determining that the requested content is associated with a predetermined category, a result identifier for the answer is obtained. Then, the obtained result identifier is used to search for the portion of content that matches the result identifier from the repository of content generated by the machine learning model.

[0117] After retrieving a portion of the content, the portion is returned as text to the client's search results box 904 and displayed as text in a text box 906 for that portion of the content. Additionally, the height of this text box is determined based on the height of the text in the portion of the content, such as the specific number of lines of text.

[0118] Furthermore, after displaying a portion of the answer result in text box 906, the remaining content for the requested content will continue to be retrieved from the repository. For example, "Loading more content..." can be added below the text corresponding to the portion of the content in text box 906, so that the successfully retrieved remaining content can be further returned to the client and displayed below the portion of the text box.

[0119] In some embodiments, after successfully retrieving the remaining content and sending it to the client, a "Loading more content..." message can be added below the portion of the content. Alternatively, after the client retrieves the portion of the content, the "Loading more content..." message can be added directly below the portion of the content.

[0120] In some embodiments, after the response results display part of the content and the remaining content, image data 908 and image data 910 related to the requested content are also obtained. This image data is data obtained by the search engine from a database. Additionally, this image data includes a Uniform Resource Locator (URL) and can receive click actions on the image to jump to the page or resource corresponding to the URL.

[0121] In some embodiments, the answer results may also include video data 912, video data 914, and video data 916. This video data is also found and returned to the client by the search engine from its database. In some embodiments, a click action on the video can be received to directly play the corresponding video.

[0122] In some embodiments, when the content of the answer result is being generated, the client switches from the "Home" page to the "Video" page, at which point the display of the answer result content stops. When switching back from the "Video" page to the "Home" page, the answer result is displayed again from the point where it was stopped.

[0123] In this disclosure, modules can be added to the page presented on the client side. In some embodiments, interactive modules can be added to the page. In one example, the interactive module of the page can be displayed when the streaming request for the streaming module ends. The interactive module of the outflow card is only displayed when the first experimental group of the new component is hit, and the display timing is when the outflow card is performing a height-increasing animation. In addition to adding interactive modules, the like status synchronization logic between the inflow and outflow can also be added: when the search page is opened, the like status stored locally is obtained according to the result identifier of the search results issued by the comprehensive search request, and the latest like status is displayed; when liking or disliking, the current status is stored locally, and the data caching logic is consistent with the interactive component of the outflow card. In addition, when returning to the outflow card from the page, the local storage status is obtained, and the like status of the interactive component is refreshed (an API for refreshing the user interface can be added to the component). In another example, like, dislike, and feedback modules are added. The page is displayed when the text content of the inflow page is displayed (after receiving the block request replace=1). The likes and dislikes are read locally based on the result identifier. They are synchronized with the external stream. When returning to the external stream from the details page, the likes or dislikes information of the interactive component needs to be called again.

[0124] In some embodiments, a recommended search module can be added to the page. This module is displayed at the end of the streaming request. Recommended search terms are obtained from the streaming loading module. The server can add a module to return all recommended search terms. Clicking a recommended search term initiates a new search, and a long press brings up a feedback panel, with interactions aligned with the externally streamed recommended search cards. Recommended search data parsing can be performed, and a new module (chunk event = 4) can be added to receive recommended search terms. A search can be initiated by clicking, aligning with the parameter passing logic of the externally streamed recommended search terms, and a long press brings up the feedback panel.

[0125] In some embodiments, an error page can be provided to display error information. A retry page should be displayed when the streaming module cannot retrieve remaining data and when there is no network connection; the prompt text on both pages should conform to search scenario specifications. A unified status component for the search page is mounted in the root layout of the page. For example, the "No network connection" message is displayed for a no-network message. The "Failed retrieval request for the streaming module" message is displayed: "There was a problem." In some embodiments, error pages, no network, weak network, and fake network conditions can also be set, as shown in Figure 10-13.

[0126] The following is a schematic diagram of an example of a page without network anomalies for searching information according to some embodiments of the present disclosure, with reference to FIG10.

[0127] In Example 1000, when a search request is made in the search box 1002, if there is no network connection, the client interface will display the text message "No network connection, click to retry" and a smaller message "Connect to network and retry" will be displayed below, and the retry control 1004 will be displayed on the client interface.

[0128] The above illustration, with reference to FIG10, depicts an example of a page without network anomalies for searching information according to some embodiments of the present disclosure. The following illustration, with reference to FIG11, depicts an example of a page with weak network anomalies for searching information according to some embodiments of the present disclosure.

[0129] In Example 1100, when a search request is made in the search box 1102, if the network is weak, the client interface will display the text message "Your network is unstable, click to retry.", and at the same time, a smaller text message "Network is unstable, please click to retry" will be displayed below, and the retry control 1104 will be displayed on the client interface.

[0130] The above illustration, with reference to FIG11, depicts an example of a weak network anomaly page for searching information according to some embodiments of the present disclosure. The following illustration, with reference to FIG12, depicts an example of a fake network anomaly page for searching information according to some embodiments of the present disclosure.

[0131] In Example 1200, when a search request is executed in the search box 1202, if the network is a fake network, the client interface will display the text message "No network connection, click to retry.", and at the same time, a small text message "Unable to connect to the network, switch to another network and retry" will be displayed below, and the retry control 1204 will be displayed on the client interface.

[0132] The above illustration, with reference to FIG12, depicts an example of a fake network anomaly page for searching information according to some embodiments of the present disclosure. The following illustration, with reference to FIG13, depicts an example of a server crash anomaly page for searching information according to some embodiments of the present disclosure.

[0133] In Example 1300, when a search request is executed in the search box 1302, if the server crashes or malfunctions, the client interface will display the text message "An error has occurred." and below it, the smaller text message "Please try again later." will be displayed, and the retry control 1304 will be displayed on the client interface.

[0134] In some embodiments, for outbound cards, a sub-chain module can be added below the answer results. In this case, the height-expanding animation logic can be modified: originally, the answer page only contained plain text, and the height-expanding animation is implemented within the entire card's container layer. Additionally, the newly added sub-chain module needs to include the sub-chain height, and this information can be considered in the height calculation. To facilitate future iterations of adding modules, the height-expanding animation can be moved to the text view layer, with the card layer's height adapting to the height of its internal elements. Furthermore, clicking the text can jump to a different page, simultaneously refreshing the page text. At this point, a specific module can be used to forward the chunk data received from the comprehensive search page to the Activity component for text refresh.

[0135] In some embodiments, a new loading animation page can also be added. Additionally, a predefined text skeleton component can be used, with settings controlling the display of two lines of text skeleton and allowing the skeleton height to be set. Furthermore, animations can be inserted at the end of the text or dynamic loading effects can be added.

[0136] In some embodiments, label components can be added to the loading page and retry page. Within the page, the layout position where the label component is mounted can be modified; for example, it can be moved to the root layout of the card. This involves some modifications to data transmission, allowing the current answer text to be retrieved at the card level for constructing the label component's data, such as the model of a single feedback panel.

[0137] In some embodiments, a reference subchain module can be used. Reference markers in the text are provided by the server. 角标序号的markdown格式给客户端进行解析渲染。引用角标对应的跳链数据会在回答结果中新增网络文档列表进行接收，流式加载请求的补丁块或剩余内容中也会按照角标顺序逐个下发，客户端实时解析数据后更新回答结果中的网络文档列表。在一个示例中，在外流卡片做撑高动画时根据此时获取到的跳链列表展示引用子链模块，但是这可能会出现引用模块展示链接个数大于外流卡片的角标个数，因为外流卡片的文本截断可能会隐藏掉最后的角标。在另一种方案中，实现新组件实验组，改造文本组件对外提供的能力，获取外流卡片实际展示的文本中包含几个引用角标，根据实际展示的个数从已有的网络文档列表中取出前几个链接进行展示。另外在用户界面交互中，提供的流式布局组件可支持限制最大展示行数，并在最大行数的末尾添加一个视图，可以实现一个通用布局在流式布局组件末尾添加一个展开按钮。点击展开按钮可展开所有子视图，并将末尾按钮的状态更新为可收起。

[0138] 在一些实施例中，可以复用聊天页面的子链逻辑，但是会将其与模型模块解绑。在该实施例中，可以先取得需要展示的所有子链的数据，之后根据子链的文字长度算出当前子链的宽度，如果已经超过了约定的宽度，则选择将其放到下一行，在放的时候动态的调整每个子链的帧。在数据层，在外流页需要实时的展示新增的按钮，在设计的过程中提供附加按钮的操作，提供实时更新的能力。外流卡片根据块实时下发的链接信息，实时的展示链接信息。在内流详情页面中，如果用户在内流页的话，那么收到链接的信息不会实时展示，是先缓存，等收到文字部分结束的标志位(replace＝＝1)之后再进行统一的展示。

[0139] 下面结合图14描述根据本公开的实施例的用于搜索信息的装置1400的示意性框图。

[0140] 如图14所示，装置1400包括结果标识获取模块1402，被配置为响应于确定来自客户端的搜索请求中的请求内容关联于预定类别，获取针对搜索请求的回答结果的结果标识；部分内容获取模块1404，被配置为由机器学习模型针对请求内容生成回答结果的内容；部分内容显示模块1406，被配置为响应于基于所述结果标识获取到回答结果的部分内容，将包括部分内容的第一响应发送给客户端以用于显示回答结果的部分内容；剩余内容获取模块1408，被配置为响应于从客户端接收到用于获取回答结果的剩余内容的获取请求，获取由机器学习模型生成的剩余内容；以及剩余内容发送模块1410，被配置为响应于获取到剩余内容，将包括剩余内容的第二响应发送到客户端以用于显示包括部分内容和剩余内容的回答结果。

[0141] 在一些实施例中，结果标识获取模块1402包括：请求内容发送模块，被配置为响应于来自客户端的搜索请求的请求内容关联于预定类别，将请求内容发送给机器学习模型；以及结果标识获取模块，被配置为响应于请求内容被机器学习模型处理，获取用于标识由机器学习模型生成的回答结果的结果标识。

[0142] 在一些实施例中，装置1400还包括：结果标识发送模块，被配置为将结果标识发送到用于存储由机器学习模型生成的内容的存储库；以及部分内容获取模块，被配置为从存储库获取与结果标识相对应的回答结果的内容。

[0143] 在一些实施例中，剩余内容获取模块1408包括：获取请求接收模块，被配置为接收来自客户端的用于获取回答结果的剩余内容的获取请求；获取请求解析模块，被配置为通过对获取请求进行解析来获取针对回答结果的结果标识；以及剩余内容获取模块，被配置为基于结果标识，获取由机器学习模型生成的剩余内容。

[0144] 在一些实施例中，获取请求解析模块包括：参数检查模块，被配置为对获取请求中的参数进行检查；以及参数解析模块，被配置为响应于确定参数满足预定要求，对获取请求进行解析。

[0145] 在一些实施例中，获取请求包括经加密的结果标识，并且参数解析模块包括：结果标识解密模块，被配置为通过对经加密的结果标识进行解密以获取结果标识。

[0146] 在一些实施例中，剩余内容获取模块还包括：轮询任务生成模块，被配置为基于结果标识，生成针对回答结果的轮询任务；以及剩余内容获取模块，被配置为基于轮询任务，从用于存储由机器学习模型生成的内容的存储库获取剩余内容。

[0147] 在一些实施例中，其中获取请求还包括搜索请求的请求内容，并且剩余内容获取模块还包括：一组推荐词确定模块，被配置为响应于轮询任务被完成，确定针对请求内容的一组推荐词；以及一组推荐词发送模块，被配置为将一组推荐词发送到客户端以与回答结果同时显示。

[0148] 在一些实施例中，剩余内容发送模块1410包括：流式协议确定模块，被配置为确定用于回答结果的流式协议；以及剩余内容提供模块，被配置为基于流式协议，将包括剩余内容的第二响应提供给客户端。

[0149] 在一些实施例中，剩余内容提供模块包括：剩余内容已被成功获取确定模块，被配置为确定剩余内容已被成功获取；一组预定操作确定模块，被配置为确定针对剩余内容要执行的一组预定操作；以及第二响应发送模块，被配置为基于流式协议，将第二响应发送到客户端，第二响应包括剩余内容被成功获取的标识、针对一组预定操作的一组操作标识和剩余内容。

[0150] 在一些实施例中，一组预定操作包括以下中的至少一项：增加操作、替换操作和删除操作。

[0151] 在一些实施例中，剩余内容提供模块还包括：剩余内容未被成功获取确定模块，被配置为确定剩余内容未被成功获取；故障信息确定模块，被配置为确定针对剩余内容未被成功获取的故障信息；以及第三响应发送模块，被配置为基于流式协议，将第三响应发送到客户端，第三响应包括剩余内容未被成功获取的标识以及故障信息的标识。

[0152] 在一些实施例中，装置1400还包括：搜索请求接收模块，被配置为接收来自客户端的搜索请求；检查模块，被配置为通过对搜索请求的请求内容进行检查，来确定请求内容中是否存在预定关键词；以及预定类别确定模块，被配置为响应于确定请求内容中不存在预定关键词，确定搜索请求是否关联于预定类别。

[0153] 在一些实施例中，装置1400还包括：提示信息发送模块，被配置为响应于确定请求内容中存在预定关键词，将针对搜索请求的提示信息发送给客户端。

[0154] 在一些实施实例中，回答结果包括以下中的至少一项：文本数据、图像数据或者视频数据。

[0155] 在一些实施例中，装置1400还包括：搜索结果确定模块，被配置为基于请求内容，确定针对请求内容的一组搜索结果；以及第一响应生成模块，被配置为基于搜索结果和部分内容，生成第一响应。

[0156] 下面结合图15描述根据本公开的实施例的用于搜索信息的装置1500的示意性框图。如图15所示，装置1500包括搜索请求发送模块1502，被配置为向服务器发送搜索请求，搜索请求包括搜索内容；第一内容显示模块1504，被配置为响应于从服务器接收到针对搜索内容的回答结果的第一响应，显示回答结果的内容，回答结果是基于确定搜索内容关联于预定类别而由机器学习模型生成的；获取请求发送模块1506，被配置为响应于确定回答结果的内容是回答结果的部分内容，向服务器发送用于获取回答结果的剩余内容的获取请求；以及第二内容显示模块1508，被配置为响应于从服务器接收到包括回答结果的剩余内容的第二响应，显示包括部分内容和剩余内容的回答结果。

[0157] 在一些实施例中，其中第一响应包括一组搜索结果，并且第一内容显示模块1504包括：结果获取模块，被配置为从第一响应获取一组搜索结果和回答结果；以及第三内容显示模块，被配置为显示回答结果和一组搜索结果。

[0158] 在一些实施例中，装置1500还包括：完整性标识获取模块，被配置为获取第一响应中用于指示回答结果的内容是否完整的完整性标识；以及部分内容确定模块，被配置为基于完整性标识，确定回答结果的内容是回答结果的部分内容。

[0159] 在一些实施例中，装置1500还包括：结果标识获取模块，被配置为从第一响应中获取回答结果的结果标识；以及请求生成模块，被配置为基于结果标识，生成获取请求。

[0160] 在一些实施例中，请求生成模块包括：加密模块，被配置为通过对结果标识进行加密来生成经加密的结果标识；以及获取请求生成模块，被配置为基于经加密的结果标识，生成获取请求。

[0161] 在一些实施例中，第二内容显示模块1508包括：操作确定模块，被配置为基于第二响应中的一组操作标识，确定与一组操作标识相对应的一组操作；以及显示模块，被配置为对剩余数据执行一组操作以用于显示。

[0162] 图16示出了可以用来实施本公开的实施例的示例设备1600的示意性框图。图1中的服务器102和客户端104可以利用设备1600来实现。如图所示，设备1600包括中央处理单元(CPU)1601，其可以根据存储在只读存储器(ROM)1602中的计算机程序指令或者从存储单元1608加载到随机访问存储器(RAM)1603中的计算机程序指令，来执行各种适当的动作和处理。在RAM 1603中，还可存储设备1600操作所需的各种程序和数据。CPU 1601、ROM 1602以及RAM 1603通过总线1604彼此相连。输入 / 输出(I / O)接口1605也连接至总线1604。

[0163] 设备1600中的多个部件连接至I / O接口1605，包括：输入单元1606，例如键盘、鼠标等；输出单元1607，例如各种类型的显示器、扬声器等；存储页面1608，例如磁盘、光盘等；以及通信单元1609，例如网卡、调制解调器、无线通信收发机等。通信单元1609允许设备1600通过诸如因特网的计算机网络和 / 或各种电信网络与其他设备交换信息 / 数据。

[0164] 上文所描述的各个过程和处理，例如方法200、300和示例400、500、600，可由处理单元1601执行。例如，在一些实施例中，方法200、300和示例400、500、600可被实现为计算机软件程序，其被有形地包含于机器可读介质，例如存储单元1608。在一些实施例中，计算机程序的部分或者全部可以经由ROM1602和 / 或通信单元1609而被载入和 / 或安装到设备1600上。当计算机程序被加载到RAM 1603并由CPU 1601执行时，可以执行上文描述的示例方法200、300和示例400、500、600的一个或多个动作。

[0165] 本公开可以是方法、装置、系统和 / 或计算机程序产品。计算机程序产品可以包括计算机可读存储介质，其上载有用于执行本公开的各个方面的计算机可读程序指令。

[0166] 计算机可读存储介质可以是可以保持和存储由指令执行设备使用的指令的有形设备。计算机可读存储介质例如可以是——但不限于——电存储设备、磁存储设备、光存储设备、电磁存储设备、半导体存储设备或者上述的任意合适的组合。计算机可读存储介质的更具体的例子(非穷举的列表)包括：便携式计算机盘、硬盘、随机存取存储器(RAM)、只读存储器(ROM)、可擦式可编程只读存储器(EPROM或闪存)、静态随机存取存储器(SRAM)、便携式压缩盘只读存储器(CD-ROM)、数字多功能盘(DVD)、记忆棒、软盘、机械编码设备、例如其上存储有指令的打孔卡或凹槽内凸起结构、以及上述的任意合适的组合。这里所使用的计算机可读存储介质不被解释为瞬时信号本身，诸如无线电波或者其他自由传播的电磁波、通过波导或其他传输媒介传播的电磁波(例如，通过光纤电缆的光脉冲)、或者通过电线传输的电信号。

[0167] 这里所描述的计算机可读程序指令可以从计算机可读存储介质下载到各个计算 / 处理设备，或者通过网络、例如因特网、局域网、广域网和 / 或无线网下载到外部计算机或外部存储设备。网络可以包括铜传输电缆、光纤传输、无线传输、路由器、防火墙、交换机、网关计算机和 / 或边缘服务器。每个计算 / 处理设备中的网络适配卡或者网络接口从网络接收计算机可读程序指令，并转发该计算机可读程序指令，以供存储在各个计算 / 处理设备中的计算机可读存储介质中。

[0168] 用于执行本公开操作的计算机程序指令可以是汇编指令、指令集架构(ISA)指令、机器指令、机器相关指令、微代码、固件指令、状态设置数据、或者以一种或多种编程语言的任意组合编写的源代码或目标代码，所述编程语言包括面向对象的编程语言-诸如Smalltalk、C++等，以及常规的过程式编程语言-诸如"C”语言或类似的编程语言。计算机可读程序指令可以完全地在用户计算机上执行、部分地在用户计算机上执行、作为一个独立的软件包执行、部分在用户计算机上部分在远程计算机上执行、或者完全在远程计算机或服务器上执行。在涉及远程计算机的情形中，远程计算机可以通过任意种类的网络-包括局域网(LAN)或广域网(WAN)-连接到用户计算机，或者，可以连接到外部计算机(例如利用因特网服务提供商来通过因特网连接)。在一些实施例中，通过利用计算机可读程序指令的状态信息来个性化定制电子电路，例如可编程逻辑电路、现场可编程门阵列(FPGA)或可编程逻辑阵列(PLA)，该电子电路可以执行计算机可读程序指令，从而实现本公开的各个方面。

[0169] 这里参照根据本公开实施例的方法、装置(系统)和计算机程序产品的流程图和 / 或框图描述了本公开的各个方面。应当理解，流程图和 / 或框图的每个方框以及流程图和 / 或框图中各方框的组合，都可以由计算机可读程序指令实现。

[0170] 这些计算机可读程序指令可以提供给通用计算机、专用计算机或其它可编程数据处理装置的处理单元，从而生产出一种机器，使得这些指令在通过计算机或其它可编程数据处理装置的处理单元执行时，产生了实现流程图和 / 或框图中的一个或多个方框中规定的功能 / 动作的装置。也可以把这些计算机可读程序指令存储在计算机可读存储介质中，这些指令使得计算机、可编程数据处理装置和 / 或其他设备以特定方式工作，从而，存储有指令的计算机可读介质则包括一个制造品，其包括实现流程图和 / 或框图中的一个或多个方框中规定的功能 / 动作的各个方面的指令。

[0171] 也可以把计算机可读程序指令加载到计算机、其它可编程数据处理装置、或其它设备上，使得在计算机、其它可编程数据处理装置或其它设备上执行一系列操作步骤，以产生计算机实现的过程，从而使得在计算机、其它可编程数据处理装置、或其它设备上执行的指令实现流程图和 / 或框图中的一个或多个方框中规定的功能 / 动作。

[0172] 附图中的流程图和框图显示了根据本公开的多个实施例的系统、方法和计算机程序产品的可能实现的体系架构、功能和操作。在这点上，流程图或框图中的每个方框可以代表一个模块、程序段或指令的一部分，所述模块、程序段或指令的一部分包含一个或多个用于实现规定的逻辑功能的可执行指令。在有些作为替换的实现中，方框中所标注的功能也可以以不同于附图中所标注的顺序发生。例如，两个连续的方框实际上可以基本并行地执行，它们有时也可以按相反的顺序执行，这依所涉及的功能而定。也要注意的是，框图和 / 或流程图中的每个方框、以及框图和 / 或流程图中的方框的组合，可以用执行规定的功能或动作的专用的基于硬件的系统来实现，或者可以用专用硬件与计算机指令的组合来实现。

[0173] 以上已经描述了本公开的各实施例，上述说明是示例性的，并非穷尽性的，并且也不限于所披露的各实施例。在不偏离所说明的各实施例的范围和精神的情况下，对于本技术领域的普通技术人员来说许多修改和变更都是显而易见的。本文中所用术语的选择，旨在最好地解释各实施例的原理、实际应用或对市场中的技术的技术改进，或者使本技术领域的其它普通技术人员能理解本文披露的各实施例。

Claims

1. A method for searching information, comprising: In response to determining that the requested content in a search request from a client is associated with a predetermined category, a result identifier for the answer result to the search request is obtained; The content of the response result is generated by a machine learning model based on the requested content; In response to obtaining a portion of the answer result based on the result identifier, a first response including the portion of the answer result is sent to the client for displaying the portion of the answer result; In response to receiving a request from the client for retrieving the remaining content of the answer result, the remaining content generated by the machine learning model is retrieved; as well as In response to obtaining the remaining content, a second response including the remaining content is sent to the client for displaying the answer result including the partial content and the remaining content.

2. The method according to claim 1, wherein obtaining the result identifier for the answer result to the search request includes: In response to determining that the request content of the search request from the client is associated with the predetermined category, the request content is sent to the machine learning model; as well as In response to the request content being processed by the machine learning model, a result identifier is obtained to identify the answer result generated by the machine learning model.

3. The method according to claim 1, further comprising: The result identifier is sent to a repository used to store the content generated by the machine learning model; as well as Retrieve the content of the answer result corresponding to the result identifier from the repository.

4. The method of claim 1, wherein obtaining the remaining content generated by the machine learning model comprises: Receive the retrieval request from the client for obtaining the remaining content of the answer result; The result identifier for the answer result is obtained by parsing the request. Based on the result identifier, the remaining content generated by the machine learning model is obtained.

5. The method according to claim 4, wherein obtaining the result identifier for the answer result by parsing the retrieval request includes: The parameters in the acquisition request are checked; as well as In response to determining that the parameters meet the predetermined requirements, the acquisition request is parsed.

6. The method of claim 5, wherein the acquisition request includes an encrypted result identifier, and parsing the acquisition request in response to the parameters meeting a predetermined requirement includes: The result identifier is obtained by decrypting the encrypted result identifier.

7. The method of claim 6, wherein obtaining the remaining content generated by the machine learning model based on the result identifier comprises: Based on the result identifier, a polling task is generated for the answer result; as well as Based on the polling task, the remaining content is retrieved from the repository used to store content generated by the machine learning model.

8. The method of claim 7, wherein the acquisition request further includes the request content of the search request, and the method further includes: In response to the completion of the polling task, a set of recommended words is determined for the requested content; as well as The set of recommended words is sent to the client to be displayed simultaneously with the answer results.

9. The method of claim 1, wherein sending a second response including the remaining content to the client for displaying the answer result including the partial content and the remaining content comprises: Determine the streaming protocol used for the response results; as well as Based on the streaming protocol, a second response, including the remaining content, is provided to the client.

10. The method of claim 9, wherein providing a second response including the remaining content to the client based on the streaming protocol comprises: It has been confirmed that the remaining data has been successfully acquired; Determine a set of predetermined operations to be performed on the remaining data; as well as Based on the streaming protocol, the second response is sent to the client. The second response includes an identifier indicating that the remaining data was successfully acquired, a set of operation identifiers for the set of predetermined operations, and the remaining content.

11. The method of claim 10, wherein the set of predetermined operations includes at least one of the following: an add operation, a replace operation, and a delete operation.

12. The method of claim 10, further comprising: It has been determined that the remaining data was not successfully acquired; The fault information indicates that the remaining data was not successfully acquired; as well as Based on the streaming protocol, a third response is sent to the client, the third response including an identifier that the remaining data was not successfully acquired and an identifier of the fault information.

13. The method according to claim 1, further comprising: Receive the search request from the client; By examining the content of the search request, it can be determined whether the predetermined keywords exist in the content of the search request. as well as In response to determining that the predetermined keyword does not exist in the requested content, it is determined whether the search request is associated with the predetermined category.

14. The method of claim 13, further comprising: In response to determining that the predetermined keyword exists in the requested content, a prompt message for the search request is sent to the client.

15. The method of claim 1, wherein the response result includes at least one of the following: text data, image data, or video data.

16. The method according to claim 1, further comprising: Based on the requested content, determine a set of search results for the requested content; as well as The first response is generated based on the search results and the partial content.

17. A method for searching information, comprising: Send a search request to the server, the search request including the search content; In response to receiving a first response from the server for an answer result for the search content, the content of the answer result is displayed, the answer result being generated by a machine learning model based on determining that the search content is associated with a predetermined category; In response to determining that the content of the answer result is a part of the answer result, a request to obtain the remaining content of the answer result is sent to the server; as well as In response to receiving a second response from the server including the remaining content of the answer result, the answer result including the partial content and the remaining content is displayed.

18. The method of claim 17, wherein the first response comprises a set of search results, and the content displaying the answer results includes: Obtain the set of search results and the answer results from the first response; as well as The answer results and the set of search results are displayed.

19. The method of claim 17, further comprising: Obtain the integrity identifier in the first response that indicates whether the content of the answer result is complete; as well as Based on the integrity identifier, it is determined that the content of the answer result is a part of the answer result.

20. The method of claim 17, further comprising: Obtain the result identifier of the answer result from the first response; as well as The acquisition request is generated based on the result identifier.

21. The method of claim 20, wherein generating the acquisition request based on the result identifier comprises: The encrypted result identifier is generated by encrypting the result identifier; as well as The retrieval request is generated based on the encrypted result identifier.

22. The method of claim 17, wherein displaying the answer result including the partial content and the remaining content comprises: Based on a set of operation identifiers in the second response, determine a set of operations corresponding to the set of operation identifiers; as well as Perform the set of operations on the remaining data for display.

23. An apparatus for searching information, comprising: The result identifier acquisition module is configured to acquire the result identifier of the answer result for the search request in response to determining that the request content in the search request from the client is associated with a predetermined category; The content generation module is configured to generate the content of the response result from the request content using a machine learning model; A partial content display module is configured to, in response to obtaining partial content of the answer result based on the result identifier, send a first response including the partial content to the client for displaying the partial content of the answer result; The remaining content acquisition module is configured to acquire the remaining content generated by the machine learning model in response to receiving an acquisition request from the client for acquiring the remaining content of the answer result; as well as The remaining content sending module is configured to, in response to obtaining the remaining content, send a second response including the remaining content to the client for displaying the answer result including the partial content and the remaining content.

24. An apparatus for searching information, comprising: The search request sending module is configured to send a search request to the server, the search request including search content; A first content display module is configured to display the content of the response result in response to receiving a first response result for the search content from the server, the response result being generated by a machine learning model based on determining that the search content is associated with a predetermined category; The request sending module is configured to send a request to the server to obtain the remaining content of the answer result in response to determining that the content of the answer result is part of the answer result. as well as The second content display module is configured to respond to receiving a content display module from the server, including the content displayed by the server. A second response to the remaining content of the answer result displays the answer result including the partial content and the remaining content.

25. An electronic device, comprising: At least one processor; as well as A storage device for storing at least one program, which, when executed by the at least one processor, causes the at least one processor to implement the method according to any one of claims 1-22.

26. A computer-readable storage medium having a computer program stored thereon, the computer program implementing the method according to any one of claims 1-22 when executed by a processor.

27. A computer program product comprising a computer program that, when executed by a processor, implements the method according to any one of claims 1-22.

Images