Information processing method and device, computer device and storage medium

Abstract

Embodiments of the present application disclose an information processing method and device, computer equipment and a storage medium, wherein the method comprises: displaying a first feedback page of query information, the first feedback page comprising: information content of feedback information corresponding to the query information in a first browsing mode; if a mode switching operation is detected, displaying a second feedback page of the query information; the second feedback page comprising: information content of the feedback information in a second browsing mode; wherein, in the first feedback page and the second feedback page, the information content in one feedback page is all information content of the feedback information, and the information content in the other feedback page is simplified information content obtained based on the all information content. Embodiments of the present application can improve the diversity of browsing modes, thereby improving the convenience of browsing information, and further improving the use viscosity of the application.

Description

Technical Field

[0001] This application relates to the field of Internet technology, specifically to the field of computer technology, and in particular to an information processing method, apparatus, computer equipment, and storage medium. Background Technology

[0002] With the development of internet and computer technologies, an increasing number of applications (APPs) can provide information query services to users. For example, web browsing applications can provide query services for various information such as encyclopedia entries (words or terms collected by dictionary editors, along with their corresponding definitions) and work-related questions. After a user inputs query information (such as words from an encyclopedia entry), the application can output and display corresponding feedback information (such as the definition from the encyclopedia entry) based on that information query service. Currently, how to display the feedback information corresponding to the query information to improve the user stickiness of the application has become a research hotspot. Summary of the Invention

[0003] This application provides an information processing method, apparatus, computer equipment, and storage medium that can enhance the diversity of browsing modes, thereby improving the convenience of browsing information, enabling users to have a good consumption experience when consuming content, and thus increasing the stickiness of the application.

[0004] On one hand, embodiments of this application provide an information processing method, the method comprising:

[0005] The first feedback page displays the query information, and the first feedback page includes: the information content of the feedback information corresponding to the query information in the first browsing mode;

[0006] If a mode switching operation is detected, a second feedback page for the query information is displayed; the second feedback page includes the information content of the feedback information in the second browsing mode;

[0007] In the first feedback page and the second feedback page, one feedback page contains the complete information content of the feedback information, while the other feedback page contains a simplified information content derived from the complete information content.

[0008] On the other hand, embodiments of this application provide an information processing apparatus, the apparatus comprising:

[0009] The first display unit is used to display a first feedback page for query information. The first feedback page includes: information content of the feedback information corresponding to the query information in the first browsing mode.

[0010] The second display unit is used to display a second feedback page of the query information if a mode switching operation is detected; the second feedback page includes the information content of the feedback information in the second browsing mode;

[0011] In the first feedback page and the second feedback page, one feedback page contains the complete information content of the feedback information, while the other feedback page contains a simplified information content derived from the complete information content.

[0012] In another aspect, embodiments of this application provide a computer device, the computer device including an input interface and an output interface, the computer device further including:

[0013] A processor, adapted to implement one or more instructions; and,

[0014] A computer storage medium storing one or more instructions adapted for loading by the processor and executing the following steps:

[0015] The first feedback page displays the query information, and the first feedback page includes: the information content of the feedback information corresponding to the query information in the first browsing mode;

[0016] If a mode switching operation is detected, a second feedback page for the query information is displayed; the second feedback page includes the information content of the feedback information in the second browsing mode;

[0017] In the first feedback page and the second feedback page, one feedback page contains the complete information content of the feedback information, while the other feedback page contains a simplified information content derived from the complete information content.

[0018] In another aspect, embodiments of this application provide a computer storage medium storing one or more instructions, which are adapted to be loaded by a processor and executed as follows:

[0019] The first feedback page displays the query information, and the first feedback page includes: the information content of the feedback information corresponding to the query information in the first browsing mode;

[0020] If a mode switching operation is detected, a second feedback page for the query information is displayed; the second feedback page includes the information content of the feedback information in the second browsing mode;

[0021] In the first feedback page and the second feedback page, one feedback page contains the complete information content of the feedback information, while the other feedback page contains a simplified information content derived from the complete information content.

[0022] In another aspect, embodiments of this application provide a computer program product, which includes a computer program; when the computer program is executed by a processor, it implements the information processing method mentioned above.

[0023] This application embodiment, after obtaining the feedback information corresponding to the query information, can provide multiple browsing modes, such as a first browsing mode and a second browsing mode, thereby enhancing the diversity of browsing modes. Specifically, the first browsing mode allows the user to browse the full or simplified content of the feedback information through the first feedback page; correspondingly, the second browsing mode allows the user to browse the simplified or full content of the feedback information through the second feedback page. Therefore, through the first and second browsing modes, users can not only comprehensively understand the query information through the full content, improving the comprehensiveness of information delivery, but also quickly and effectively understand the query information through the simplified content, improving the effectiveness of information delivery. Furthermore, by allowing users to freely switch between different browsing modes according to their actual needs, the convenience of browsing information can be improved, enabling users to have a good user experience when consuming content, thereby increasing user stickiness. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the description of the embodiments will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 This is a schematic diagram illustrating the interaction between a terminal and a server provided in an embodiment of this application;

[0026] Figure 2 This is a flowchart illustrating an information processing method provided in an embodiment of this application;

[0027] Figure 3a This is a schematic diagram illustrating a display of a second feedback page provided in an embodiment of this application;

[0028] Figure 3b This is a schematic diagram illustrating another way of displaying a second feedback page, provided in an embodiment of this application.

[0029] Figure 3cThis is a schematic diagram illustrating another way of displaying a second feedback page, provided in an embodiment of this application.

[0030] Figure 3d This is a schematic diagram of a content sharing card provided in an embodiment of this application;

[0031] Figure 3e This is a schematic diagram of another content sharing card provided in an embodiment of this application;

[0032] Figure 3f This is a schematic diagram illustrating an update of the browsing identifier provided in an embodiment of this application;

[0033] Figure 3g This is a schematic diagram illustrating an update of historical likes provided in an embodiment of this application;

[0034] Figure 3h This is a schematic diagram illustrating a method for re-displaying a benchmark feedback page, as provided in an embodiment of this application.

[0035] Figure 4 This is a schematic diagram of a process for obtaining target abstract text provided in an embodiment of this application;

[0036] Figure 5a This is a schematic diagram of the structure of a target summary extraction model provided in an embodiment of this application;

[0037] Figure 5b This is a schematic diagram illustrating the working principle of a decoder in a target summary extraction model provided in this application embodiment;

[0038] Figure 6 This is a schematic diagram of the structure of an information processing device provided in an embodiment of this application;

[0039] Figure 7 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application. Detailed Implementation

[0040] The technical solutions in the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings.

[0041] This application proposes an information processing method that, considering the content consumption preferences of users, proposes a method to reorganize and simplify the content structure of feedback information, thereby providing users with multiple browsing modes, such as a long-tail consumption mode and a simplified consumption mode. The long-tail consumption mode refers to a browsing mode that supports displaying all information content. It should be understood that in the long-tail consumption mode, if the terminal screen is small and cannot display all information content at once, a scrolling display method can be used to achieve the display of all information content. The simplified consumption mode refers to a browsing mode that supports displaying simplified information content, which is obtained by simplifying all information content. The simplified information content can summarize the concept or idea of ​​the feedback information with fewer words; for example, the simplified information content may include the target summary text corresponding to all information content. In addition to providing browsing modes such as long-tail consumption mode and simplified consumption mode, this information processing method also allows users to freely switch between different browsing modes according to their actual needs. By using the corresponding browsing mode, users can understand the full or simplified information content of the feedback information corresponding to the query information. This improves the convenience of browsing information and allows users to have a good consumption experience when consuming content, thereby increasing the stickiness of the application and the page dwell time. The so-called page dwell time refers to the user staying on the page without performing any operation (such as swiping up or down).

[0042] In one specific implementation, the information processing method proposed in this application embodiment can be executed by a terminal device (hereinafter referred to as a terminal); the terminal mentioned herein may include, but is not limited to, smartphones, computers (such as tablets, laptops, desktop computers, etc.), smart wearable devices (such as smartwatches, smart glasses), smart voice interaction devices, smart home appliances (such as smart TVs), vehicle terminals, or aircraft, etc. Furthermore, various applications (i.e., APPs) can be installed and run on the terminal, such as information query applications (applications that can provide information query services), social applications, audio and video playback applications, etc.; in this case, the information processing method proposed in this application embodiment can also be executed by an information query application. It should be understood that if a social application or audio and video playback application can also provide information query services, then the method can also be executed by the social application or audio and video playback application. Optionally, the information processing method proposed in this application embodiment can also be executed by an information query mini-program (i.e., a mini-program that provides information query services). A mini-program refers to an application that does not require installation or download and can run within an installed APP.

[0043] In another specific implementation, the information processing method proposed in this application embodiment can also be executed jointly by a terminal and a server; in this case, the terminal and server can communicate through a wired network or a wireless network, without limitation. The server mentioned herein can be an independent physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. For details, see [link to relevant documentation]. Figure 1 As shown: After receiving the query information input by the user, the terminal can send the query information to the server. The server is responsible for determining the corresponding feedback information and simplifying the entire content of the feedback information to obtain simplified information content. Both the simplified information content and the full content of the feedback information are then sent to the terminal, allowing the terminal to output the information content corresponding to the browsing mode selected by the user based on the simplified and full information content. It should be understood that, to improve the timeliness of information feedback, a large amount of feedback information corresponding to query information can be pre-stored on the server, and the simplified information content corresponding to each feedback information can be pre-generated and stored. This allows the server to directly retrieve the corresponding feedback information and the corresponding simplified information content from the storage space and send them to the terminal after receiving the user's query information.

[0044] The following is combined with Figure 2 The flowchart shown further illustrates the information processing method proposed in the embodiments of this application; for ease of explanation, the embodiments of this application mainly use the execution of the information processing method by a terminal as an example for illustration. Please refer to... Figure 2 The information processing method may include the following steps S201-S202:

[0045] S201, Display the first feedback page for the query information. The first feedback page includes the information content of the feedback information corresponding to the query information in the first browsing mode.

[0046] In this context, query information refers to the information input by the target object (the user using the terminal executing the method) to be queried, while the corresponding feedback information is the information used to provide feedback on the query information. For example, the query information could be a word or term from an encyclopedia entry entered by the target object, and the corresponding feedback information could be the definition from that encyclopedia entry; if the target object entered the term "whale eye," the feedback information would be a definition explaining what a whale eye is. Similarly, the query information could be any question entered by the target object, and the corresponding feedback information could be the answer to that question; if the target object entered "What are the principles of machine learning?", the feedback information would be articles about the principles of machine learning.

[0047] In this embodiment, the first browsing mode can be the long-tail consumption mode mentioned above; in this case, the information content of the feedback information corresponding to the query information in the first browsing mode refers to the complete information content of the feedback information. Alternatively, the first browsing mode can be the simplified consumption mode mentioned above; in this case, the information content of the feedback information corresponding to the query information in the first browsing mode refers to the simplified information content of the feedback information.

[0048] S202, if a mode switching operation is detected, a second feedback page for the query information is displayed; wherein, the second feedback page includes: information content of the feedback information in the second browsing mode.

[0049] It should be understood that if the first browsing mode mentioned in step S201 is a long-tail consumption mode, then the second browsing mode mentioned here is a simplified consumption mode. In this case, the information content of the feedback information in the second browsing mode refers to the simplified information content of the feedback information, and mode switching refers to switching from the long-tail consumption mode to the simplified consumption mode. If the first browsing mode mentioned in step S201 is a simplified consumption mode, then the second browsing mode mentioned here is a long-tail consumption mode. In this case, the information content of the feedback information in the second browsing mode refers to the complete information content of the feedback information, and mode switching refers to switching from the simplified consumption mode to the long-tail consumption mode.

[0050] In other words, between the first feedback page and the second feedback page, one feedback page contains the complete feedback information, while the other feedback page contains a simplified version of the complete information. It should be understood that after displaying the second feedback page, the target user can also trigger the terminal to switch modes again by inputting a new mode switching operation to display the first feedback page; that is, this embodiment of the application can support the target user continuously switching between the first browsing mode and the second browsing mode.

[0051] In a specific implementation, any feedback page may include: a first mode component corresponding to the first browsing mode, and a second mode component corresponding to the second browsing mode. These first and second mode components can be located in the bottom navigation bar (bottom bartab) of the feedback page, or in other locations; there is no limitation on this. Furthermore, when any feedback page is displayed, the mode component corresponding to the corresponding browsing mode is selected, while other mode components are unselected. That is, when the first feedback page is displayed, the first mode component is selected, and the second mode component is unselected; when the second feedback page is displayed, the first mode component is unselected, and the second mode component is selected. In this case, the mode switching operation may include selecting the second mode component on the first feedback page. Optionally, the mode switching operation may also be an operation of inputting a specified gesture, which can be set according to actual needs, such as a horizontal swipe to the left; or, the mode switching operation may also be an operation of inputting a voice command to indicate switching from the first browsing mode to the second browsing mode; or, the mode switching operation may also be an operation of pressing or touching physical components on the terminal (such as volume buttons, power buttons), etc.

[0052] Furthermore, the terminal can display the second feedback page of the query information in any of the following ways: Switching from the first feedback page to the second feedback page via page switching; or directly displaying the second feedback page on the first feedback page; in this case, the second feedback page can be an overlay page (i.e., a page located on the first feedback page with a certain degree of transparency) or a page without transparency, without limitation; or, displaying a floating window on the first feedback page and displaying the second feedback page within that window; in this case, the second feedback page can be understood as a subpage displayed independently of the first feedback page.

[0053] Based on the above description, taking the query information as the encyclopedia entry "Whale Eyes," the terminal displays the second feedback page through page switching. The first feedback page corresponds to a first browsing mode of long-tail consumption, while the second feedback page corresponds to a second browsing mode of simplified consumption. An illustration of how the target user triggers the display of the second feedback page by selecting a component in the second mode of the first feedback page can be found here. Figure 3a As shown; a schematic diagram illustrating how the target object triggers the display of a second feedback page on the terminal by inputting a horizontal swipe-to-left gesture can be found here. Figure 3b As shown. It should be understood that, limited by the display size of the terminal screen, Figure 3a and Figure 3bThe first feedback page only displays a portion of the total information. The user can scroll up or down to trigger the display of the remaining information within the first feedback page, thus allowing them to browse the full content. Furthermore, Figure 3a and Figure 3b The "Full Encyclopedia" and "Concise Encyclopedia" components are merely illustrative representations of the styles of the first and second mode components, respectively, and are not intended to limit their functionality; "Full Encyclopedia" corresponds to the first mode component, and "Concise Encyclopedia" corresponds to the second mode component. Additionally, in Figure 3b In the scenario shown, the first mode component and the second mode component may not be included in either of the two pages.

[0054] It should be noted that while the first feedback page includes all information content, the information content on the second feedback page is a simplified version. This simplified content can be obtained by simplifying either the entire feedback information or by simplifying specific target information from the entire feedback information; there is no limitation in either direction. The target information content can be information selected by the target user from the entire feedback information based on their needs. A specific application scenario is as follows: after displaying the first feedback page, if the terminal does not detect a selection operation on the entire feedback information before detecting a mode switching operation, then the simplified information content on the second feedback page is obtained by simplifying the entire feedback information. If a content selection operation is detected before the mode switching operation is detected, the simplified information content on the second feedback page is obtained by simplifying the target information content (e.g., extracting a summary). In other words, before detecting the mode switching operation, the terminal can obtain a content selection operation from all information content and select the target information content based on the content selection operation. Then, after detecting the mode switching operation, the terminal can perform summary extraction on the target information content to obtain the simplified information content and trigger the step of displaying the query information on the second feedback page, as shown below. Figure 3c As shown.

[0055] For ease of description, the feedback page displaying simplified information content in the first and second feedback pages can be referred to as the simplified feedback page; and the feedback page displaying all information content in the first and second feedback pages can be referred to as the baseline feedback page. Furthermore, the baseline feedback page may also include video identifiers for one or more videos related to the queried information. These video identifiers may include, but are not limited to, video cover images, video titles, and video webpage links, etc. Therefore, if the video identifier includes a video cover image, the simplified feedback page may also include the video cover image of one of the videos, as described above. Figures 3a-3c As shown.

[0056] In an optional implementation, the simplified feedback page may further include a sharing component (identified by 31) for sharing simplified information content. In this case, if the target object wants to share the simplified information content with other objects (such as other users or other groups), it can trigger an operation on the sharing component. Correspondingly, during the display of the simplified feedback page, if the terminal detects a trigger operation on the sharing component, it can respond to the trigger operation by displaying a content sharing card 32 corresponding to the simplified information content on the simplified feedback page. The content sharing card 32 may include the simplified information content 321; optionally, the content sharing card 32 may also include a graphic code 322, which, when scanned by any terminal, can output and display the simplified feedback page on any terminal. Further, if the simplified feedback page includes a video cover, the content sharing card may also include a thumbnail of the video cover (identified by 323). Taking a content sharing card 32 that simultaneously includes concise information content 321, a graphic code 322, and a thumbnail as an example, a schematic diagram of displaying the content sharing card 32 can be found here. Figure 3d As shown; it should be understood that in other embodiments, the terminal may also switch from the simplified feedback page to the content sharing interface, thereby displaying the content sharing card in the content sharing interface.

[0057] After displaying the content sharing card, the target object can perform a sharing operation on the content sharing card; correspondingly, if the terminal detects a sharing operation on the content sharing card, it can send the content sharing card to the object indicated by the sharing operation. In one specific implementation, when displaying the content sharing card, the terminal also simultaneously displays one or more object identifiers 33, such as... Figure 3e As shown; in this case, the sharing operation may include selecting at least one object identifier. That is, after the target object selects at least one object identifier, it can trigger the terminal to send the content sharing card to the object indicated by the sharing operation (i.e., the object indicated by each selected object identifier). In another specific implementation, after displaying the content sharing card, the terminal can automatically save the content sharing card to local space; in this case, the sharing operation may include sending the content sharing card in the session interface. That is, the target object can use the terminal to open a session interface with other objects, and send the content sharing card stored in local space to other objects through the session interface.

[0058] In another optional implementation, the simplified feedback page may further include a browsing identifier (identified by 34); this browsing identifier indicates at least one of the following: the historical browsing count 341 of the simplified information content, and the object identifier 342 of all or part of the objects that have historically viewed the simplified information content. Accordingly, if the terminal detects that the simplified information content has been viewed by a new object while displaying the simplified feedback page, it can update the display browsing identifier on the simplified feedback page. The new object can be the target object or an object using another terminal; this is not limited. Furthermore, updating the display browsing identifier may include at least one of the following: updating the historical browsing count, and updating the object identifier; for example, see [link to relevant documentation]. Figure 3f As shown: If the historical pageview count is 65 people, then after the simplified information content is viewed by a new object, the historical pageview count can be updated to 66 people, and the object identifier 343 of the new object can be used to update the displayed object identifier.

[0059] In another optional implementation, the simplified feedback page may further include: the historical number of likes for the simplified information content (identified by 35) and a like component (identified by 36). Accordingly, if the terminal detects that the like component has been triggered while displaying the simplified feedback page, it can update the display of the historical number of likes on the simplified feedback page, such as... Figure 3g As shown. Furthermore, during the display of the simplified feedback page, if it is detected that the simplified information content is simultaneously liked by the target object and at least one other object, the terminal can play a liking animation on the simplified feedback page based on the virtual object corresponding to the target object and the virtual objects corresponding to each of the other objects; and update the historical liking count after the liking animation finishes playing. In other words, when at least two objects simultaneously like the simplified information content, the liking animation can be played based on the virtual objects of at least two objects, which can effectively enhance the fun of liking.

[0060] In another optional implementation, the total information content includes video identifiers for one or more videos related to the query information. In this case, during the display of the baseline feedback page, if any video identifier in the total information content is triggered, a video playback page can be displayed; then, the video corresponding to the triggered video identifier is played on the video playback page, and a sharing component 31 for sharing the simplified information content is displayed on the video playback page. If the sharing component 31 is triggered, a content sharing card 32 corresponding to the simplified information content is displayed on the video playback page. Further, during the display of the content sharing card, if a target interaction operation is detected, the video playback page and the content sharing card are canceled, and the baseline feedback page is redisplayed. The target interaction operation includes a swipe operation along a specified direction on the video playback page; or, the video playback page also includes a page return component 37, and the target interaction operation includes a trigger operation on the page return component 37. Taking the target interaction operation as including a trigger operation on the page return component 37 as an example, the process of the terminal displaying the video playback page and redisplaying the baseline feedback page can be found in [reference needed]. Figure 3h As shown.

[0061] After obtaining the feedback information corresponding to the query information, this embodiment of the application can provide the user with multiple browsing modes, such as a first browsing mode and a second browsing mode, thereby enhancing the diversity of browsing modes. Specifically, the first browsing mode allows the user to browse the full or condensed content of the feedback information through the first feedback page; correspondingly, the second browsing mode allows the user to browse the condensed or full content of the feedback information through the second feedback page. Therefore, through the first and second browsing modes, the user can not only comprehensively understand the query information through the full content, improving the comprehensiveness of information delivery, but also quickly and effectively understand the query information through the condensed content, improving the effectiveness of information delivery. Furthermore, by allowing the user to freely switch between different browsing modes according to actual needs, the convenience of browsing information can be improved, enabling the user to have a good consumption experience when consuming content, thereby increasing the stickiness of the application.

[0062] Based on the above Figure 2The following describes the specific method for obtaining the aforementioned simplified information content, based on the illustrated method embodiments. Specifically, the terminal or server can perform summary extraction processing on all information content of the feedback information to obtain a target summary text, which is then used to construct the simplified information content. In this case, the simplified information content includes the target summary text obtained by summarizing all information content of the feedback information. The characteristic of a summary is that the output text has a much smaller data volume than the input text, but can contain a great deal of effective information. In specific implementations, embodiments of this application can employ an abstract-active summarization strategy to extract the target summary text. See also... Figure 4 As shown, the method for obtaining the target summary text may include the following steps S401-S404:

[0063] S401, construct a vocabulary based on all the information content of the feedback information, and split all the information content of the feedback information into K sentences, where K is a positive integer.

[0064] The vocabulary includes multiple high-frequency words, which are words in the feedback information whose frequency exceeds a preset threshold. Frequency refers to the number of times a word appears in the entire information content. Specifically, the vocabulary can be constructed based on the entire feedback information content as follows: First, the entire feedback information content is segmented into multiple words; then, the frequency of each word in the entire information content is counted. Next, words with frequencies exceeding the preset threshold are identified as high-frequency words, and words with frequencies less than or equal to the preset threshold are identified as low-frequency words (low-frequency words can be marked as UNK). Finally, a vocabulary can be constructed using the identified high-frequency words, with a size of v, where v is a positive integer.

[0065] S402, perform summary extraction processing on each of the K sentences according to the vocabulary to obtain K initial summary texts and the corresponding credibility of each initial summary text.

[0066] In practical implementation, a target summary extraction model can be obtained, which employs an abstractive summarization strategy to extract the summary text. Thus, by leveraging this target summary extraction model, the initial summary text for each sentence and the corresponding credibility of each initial summary text can be obtained. See also... Figure 5aAs shown, a target summarization extraction model can include at least two parts: an encoder and a decoder. The encoder can use single-layer or multi-layer networks such as RNNs (Recurrent Neural Networks), LSTMs (Long Short-Term Memory), or GRUs (Gated Recurrent Units) to encode each word in the input information (e.g., a sentence). Correspondingly, the decoder can be understood as a language model that interprets the input information based on the encoding results output by the encoder, thereby generating a summary text corresponding to the input information. For example, this summarization extraction model can be a seq2seq (sentence-to-sentence) model, or a seq2seq+attention model from Natural Language Processing (NLP).

[0067] For the k-th sentence out of K sentences, it is segmented into M input words, where k∈[1,K] and M is a positive integer. Based on this, using x to represent any input word, the k-th sentence can be represented as a word sequence X. t :X t =|x1, x2, ..., x M |. The initial summary text corresponding to the k-th sentence is determined based on the target words output by the decoder at N time steps. The determination of the target words output by the decoder at time step t (t∈[1,N]) may include the following steps s11-s15:

[0068] s11, at the t-th time step of the decoder, the encoder is called to recursively encode each of the M words to obtain the hidden state of each input word at the t-th time step.

[0069] In the specific implementation, at the t-th time step of the decoder, the first input word out of the M input words can be input into the encoder, which encodes the first input word to obtain the hidden state of the first word at time step t. Then, the second input word out of the M input words can be input into the encoder, which encodes the second input word based on the hidden state corresponding to the first word to obtain the hidden state of the second word at time step t. Then, the third input word out of the M input words can be input into the encoder, which encodes the third input word based on the hidden state corresponding to the second word to obtain the hidden state of the third word at time step t, and so on, until the hidden state of each of the M input words at time step t is obtained.

[0070] s12, based on the target word that the decoder needs to generate at time step t, and the degree of attention given to each input word, determine the attention weight of the hidden state of each input word at time step t.

[0071] In the specific implementation process, for any input word, the attention weights of its hidden states at each historical time step can be obtained. A historical time step refers to a time step preceding the t-th time step. Then, the attention weights of the hidden states of any input word at each historical time step can be summed, and the summation result is used as an attention weight decision factor. After obtaining the attention weight decision factor, the attention weight of the hidden state of any input word at the t-th time step can be determined based on this decision factor and the degree of attention the decoder needs to give to any input word in relation to the target word to be generated at the t-th time step. By summing up the attention weights from previous time steps to determine the attention weights affecting the current time step (i.e., the t-th time step), it is possible to avoid continuing to consider parts that have already obtained high weights, thereby avoiding repetition at the same position and thus avoiding the repeated generation of the same target word, thus mitigating the problem of duplicate generation.

[0072] s13, based on the attention weights of the hidden states of each input word at time step t, integrate the hidden states of the M input words at time step t to obtain the semantic vector of the k-th sentence at time step t. The integration method can be either weighted fusion or weighted concatenation, which is not limited.

[0073] s14, calculate the output probability of each high-frequency word in the output vocabulary at the t-th time step based on the semantic vector, and obtain the probability calculation result.

[0074] In one implementation, an RNNLM (Recurrent Neural Network Language Model) can be used to calculate the output probability of each high-frequency word in the output vocabulary at time step t based on the semantic vector, thus obtaining the probability calculation result. The advantages of RNNLM are as follows: when determining the output probability of each high-frequency word at time step t, it can utilize all the preceding context information (i.e., information about the target words output at each historical time step), instead of only utilizing the information about the target words output at the (t-1)th time step, thereby improving the accuracy of the probability calculation. In this implementation, P... t Let P represent the output probability of any high-frequency word. t This can be specifically represented as P(y) t |{y1, y2, ..., y t-1}, X t The formula is: y = θ; where y = θ. t X is the target word output at time step t. tFor the k-th sentence, θ includes the semantic vector of the k-th sentence at time step t and any high-frequency words. θ can vary depending on the high-frequency words, thus making P... t different.

[0075] In another implementation, it can be based on Figure 5b The decoder, as shown, works by calculating the output probability of each high-frequency word in the output vocabulary at time step t based on the semantic vector, thus obtaining the probability calculation result. Specifically, the target word output by the decoder at time step t-1 (using y) can be obtained. t-1 The first weight W1 corresponding to (represented by h) is the hidden state of the decoder at time step t-1 (using h). t-1 Let W1 represent the second weight W2 corresponding to the semantic vector, and W3 represent the third weight W3 corresponding to the semantic vector. Then, based on the first weight W1, the second weight W2, and the third weight W3, the target word y output by the decoder at time step t-1 is... t-1 The hidden state h of the decoder at time step t-1 t-1 and semantic vector c t A weighted sum is performed to obtain the first weighted sum result; then, the sigmoid function (an activation function) is used to activate the first weighted sum result to obtain the hidden state of the decoder at time step t (using h). t (represented), and then based on the fourth weight W4 and the fifth weight W5, the hidden state h of the decoder at time step t is... t and semantic vector c t A weighted sum is performed to obtain a second weighted sum result; then, the softmax function (an activation function) is used to activate the second weighted sum result, resulting in an activation result P; finally, based on the activation result P, the output probability of each high-frequency word in the output vocabulary at time step t can be calculated to obtain the probability calculation result. For example, the calculation formulas for the hidden state ht and the activation result P of the decoder at time step t can be found in Equations 1.1 and 1.2 below:

[0076] h t =sigmoid(W1y t-1 +W2h t-1 +W3c t )∈R d×1 Formula 1.1

[0077] P = softmax(W4h) t +W5c t Equation 1.2

[0078] Where d is the number of hidden layer neurons. Based on the above description, it can be seen that the decoder needs to use the semantic vector as input at each time step, rather than only introducing the semantic vector at the first time step; furthermore, due to the existence of the attention mechanism, the degree of attention paid to each input word in the input sequence (i.e., the k-th sentence) is different when generating the target word at each time step, therefore the semantic vector given by the encoder at each time step is different. Another noteworthy point is that, in the embodiments of this application, y is not explicitly input when calculating the activation result P. t-1 However, only when calculating the hidden state h t When, the input y t-1 .

[0079] s15, select high-frequency words from the vocabulary based on the probability calculation results, and use them as the target words output by the decoder at the t-th time step.

[0080] In one implementation, generating the summary text can be reduced to solving a conditional probability problem p(word|context), where context represents the input sentence, word represents the target word, and p represents the output probability. Then, under the context condition, the output probability of each high-frequency word in the vocabulary is calculated, and the high-frequency word with the highest output probability is used as the output target word, thereby generating all target words in the summary sequentially. Based on this, the specific implementation of step s15 can be: according to the probability calculation results, select the high-frequency word with the highest output probability from the vocabulary as the target word output by the decoder at time step t. It should be noted that, in this case, the initial summary text corresponding to the k-th sentence is obtained as follows: according to the order of time steps, combine the target words output by the decoder at N time steps to obtain the initial summary text corresponding to the k-th sentence.

[0081] In another implementation, beam search is used to determine the target words, thus obtaining the final initial summary text. The core logic is as follows: When determining the target words for the first time, the top K high-frequency words in the vocabulary are retained as target words in descending order of output probability. Then, when determining the target words for the second time, each of the top K high-frequency words retained in the first instance is combined with each high-frequency word in the vocabulary to obtain Kv bigrams and their corresponding probabilities. The probability of any bigram is equal to the product of the output probabilities of the two high-frequency words that make up that bigram. Thus, the top K bigrams are retained in descending order of probability, and the second high-frequency word in each of the top K bigrams is the target word determined in the second instance. This process is repeated, retaining the top K each time target words are generated. This pruning operation significantly reduces the search space. Based on this, the specific implementation of step s15 can be as follows: According to the output probability from largest to smallest, sort the high-frequency words in the vocabulary based on the probability calculation results, and select the top k high-frequency words from the sorting results as the target words output by the decoder at time step t. It should be noted that, in this case, the initial summary text corresponding to the kth sentence is obtained as follows: determine the N-gram phrases corresponding to each target word among the k target words output at time step N, and select the N-gram phrase with the highest probability as the initial summary text corresponding to the kth sentence.

[0082] Furthermore, to avoid the OOV (Out of Register) problem (i.e., the target word to be generated may not be in the vocabulary), before executing step s15, the relationship between the largest output probability in the probability calculation results and the probability threshold can be compared. If the largest output probability is greater than or equal to the probability threshold, it indicates that the target word to be generated at time step t is in the vocabulary, and step s15 can be triggered. If the largest output probability is less than the probability threshold, it indicates that the target word to be generated at time step t may not be in the vocabulary. In this case, based on the attention weight of the hidden state of each input word at time step t, the input word corresponding to the hidden state with the largest attention weight can be selected from the M input words as the target word output by the decoder at time step t. This can effectively alleviate the OOV problem.

[0083] S403, select a target number of initial summary texts from the K initial summary texts according to the confidence level of each initial summary text, in descending order of confidence level; the target number can be set based on empirical values ​​and is not limited thereto.

[0084] S404, combine the selected target number of initial summary texts to obtain the target summary text.

[0085] Specifically, the target number of initial summary texts can be concatenated according to the order of the sentences corresponding to each selected initial summary text in the feedback information to obtain the target summary text.

[0086] Based on the descriptions of steps S401-S404 above, it should be noted that the target summary extraction model used in the above process can be obtained by training the initial summary extraction model using machine learning / deep learning techniques within Artificial Intelligence (AI) technology. AI technology refers to the theories, methods, techniques, and application systems that utilize digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results. In other words, artificial intelligence is a comprehensive technology within computer science; it primarily aims to understand the essence of intelligence and produce new intelligent machines that can react in a way similar to human intelligence, enabling these machines to possess multiple functions such as perception, reasoning, and decision-making. Machine learning is the core of AI, a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Deep learning, on the other hand, is a technique that utilizes deep neural network systems for machine learning.

[0087] The process of training the initial summary extraction model to obtain the target summary extraction model can be roughly as follows:

[0088] First, S samples, where S is a positive integer, can be obtained for model training. During training, English datasets from CNN (Convolutional Neural Network) / DailyMail (a publicly available database) and a self-maintained Chinese encyclopedia dataset can be used. For example, one or more entries can be obtained from the Chinese encyclopedia database. Each entry is then split into sentences to obtain S sentences. A first summary text is obtained by manually summarizing and abstracting each of the S sentences. Then, S samples are constructed using each sentence and its corresponding first summary text. Each sample includes one sentence and one first summary text. After obtaining the S samples, the initial extraction model can be called to perform summary extraction on the sentences in each sample, obtaining a second summary text for each sample, and the output probability of each target word in each second summary text.

[0089] As mentioned above, the reason for needing to train an initial summary extraction model to obtain a better-performing target summary extraction model is that the target summary extraction model is subsequently called to perform summary extraction on the sentences. Therefore, the input during model training can be the sentence in each sample (which can be understood as a word sequence), and the sentence in the i-th sample can be represented as X. i =|x i 1,x i 2, ..., x i M |, i∈[1,S]; here x i 1. x i 2 and x i M Y refers to the words in the sentence of the i-th sample. Correspondingly, the second summary text output by the model can also be understood as a word sequence. The second summary text output by the model from the sentence of the i-th sample can be represented as Y. i =[y i 1, y i 2, ..., y i N It should be noted that the second summary text Y corresponds to different samples. i The number of words can be different, that is, the second summary text Y corresponding to different samples. i The value of N in the model can be different. Since it does not conform to the definition of a summary if the input and output text lengths of the model are the same, the value of N for each sample can be restricted to be less than the value of M.

[0090] In this embodiment, the objective of model training can be to maximize the output probability of each word in the second summary text; if y represents any word in any second summary text, then the objective of model training can be expressed by the following formula 1.3:

[0091]

[0092] Based on this, after obtaining the output probabilities of each word in each second summary text, the loss value can be minimized using SGD (Stochastic Gradient Descent) to optimize and update the model parameters for model training. Specifically, the loss function can be used to calculate the model loss value based on the output probabilities of each target word in each second summary text. Then, SGD is used to backpropagate the model loss value to obtain the gradient value. The gradient value is then used to optimize the model parameters of the initial summary extraction model so that the initial summary extraction model converges. The converged initial summary extraction model is then used as the target summary extraction model. Here, L represents the loss value. Taking the RNNLM (Recurrent Neural Network Language Model) to obtain the output probabilities of each word in each second summary text as an example, the output probability of any word in any second summary text is P(y). i t |{y i 1, y i 2, ..., y i t-1}, X i t Taking θ as an example, the formula for calculating the model loss value can be shown in Formula 1.4 below:

[0093]

[0094] It should be noted that the specific method by which the initial summary extraction model extracts summaries for sentences in each sample is similar to the aforementioned method of extracting summaries from all information content of the feedback information using the target summary extraction model, and will not be repeated here. Furthermore, when extracting the second summary text of sentences in each sample, the corresponding methods mentioned above can also be used to avoid the OOV (Out of Context) problem and the problem of generating duplicates. As mentioned above, the main principle of the relevant methods used to avoid generating duplicates is: to sum the attention weights from previous time steps to obtain the corresponding attention weight decision factor (using c...). t (This is represented as an example), thereby influencing the decision of attention weights at the current time step through the attention weight decision factor; based on this, it is necessary to add a loss to the attention weight decision factor when calculating the model loss value in the embodiments of this application. Specifically, the corresponding loss value can be calculated using the following formula 1.5:

[0095]

[0096] Here, covloss can be short for coverage loss, which refers to the loss value calculated based on attention weight decision factors; a i tc represents the attention weights involved at the t-th time step corresponding to the i-th sample; i t The attention weight decision factor represents the attention weights involved in influencing the attention weights at the t-th time step corresponding to the i-th sample. Its calculation formula is shown in Formula 1.6 below:

[0097]

[0098] It should be noted that: in formulas 1.5-1.6 above, 't' refers to the time step at the decoder; 'i' represents the token index at the encoder (used to indicate sample information), and if the encoder input has 100 samples, then 'i' ∈ [0, 99]; ∑i represents iterating through each time step at the decoder to compare the two and select the smallest. Taking the first sample and the first time step as an example, formula 1.5 represents starting from a1 1 and c1 1 Select the smallest value from the list as covloss1. If a1 1 and c1 1 Focusing on the same distribution would result in a large covloss1. Therefore, we choose different distributions, selecting the smaller of the two, which leads to a smaller covloss1. The ultimate goal is to ensure that each distribution is distinct, thus avoiding word repetition. In this way, the coverage loss becomes a bounded quantity; based on this, in Equation 1.4 above... This part can be replaced by the following formula 1.7:

[0099]

[0100] Among them, P(w t *) represents the output probability of the target word calculated by the model at time step t; that is, P(w t *) refers to the formula 1.4 above.

[0101] As described above, this application uses machine learning and artificial intelligence to improve model performance, enabling the final target summary extraction model to perform summary extraction more efficiently and improve extraction efficiency. Furthermore, this target summary extraction model can accurately extract useful information, allowing the generation of new words and phrases to form summary text, thereby creating a streamlined consumption model experience. This supports users in freely consuming between long-tail and streamlined consumption models, meeting the needs of different users.

[0102] Based on the description of the relevant embodiments of the above information processing method, this application also proposes an information processing device; specifically, the device may be a computer program (including program code) running on a terminal, and the device may execute... Figure 2 or Figure 4 This shows some of the method steps in the method flow. Please refer to [link / reference]. Figure 6 The device can operate the following units:

[0103] The first display unit 601 is used to display a first feedback page for query information. The first feedback page includes: information content of the feedback information corresponding to the query information in a first browsing mode.

[0104] The second display unit 602 is used to display a second feedback page of the query information if a mode switching operation is detected; the second feedback page includes the information content of the feedback information in the second browsing mode;

[0105] In the first feedback page and the second feedback page, one feedback page contains the complete information content of the feedback information, while the other feedback page contains a simplified information content derived from the complete information content.

[0106] In one implementation, any feedback page includes: a first mode component corresponding to the first browsing mode, and a second mode component corresponding to the second browsing mode;

[0107] Furthermore, when displaying any of the feedback pages, the mode component corresponding to the browsing mode is selected, while other mode components are unselected.

[0108] The mode switching operation includes selecting the second mode component on the first feedback page.

[0109] In another embodiment, the feedback page that displays the simplified information content in the first feedback page and the second feedback page is called the simplified feedback page; the simplified feedback page also includes a sharing component for sharing the simplified information content;

[0110] Accordingly, the first display unit 601 or the second display unit 602 can also be used for:

[0111] During the display of the simplified feedback page, in response to the trigger operation on the sharing component, the content sharing card corresponding to the simplified information content is displayed on the simplified feedback page;

[0112] If a sharing operation is detected for the content sharing card, the content sharing card is sent to the object indicated by the sharing operation.

[0113] In another embodiment, the feedback page that displays the simplified information content in the first feedback page and the second feedback page is called the simplified feedback page; the simplified feedback page also includes a browsing identifier, which is used to indicate at least one of the following: the historical browsing volume of the simplified information content, and the object identifier of all or part of the objects that have historically browsed the simplified information content;

[0114] Accordingly, the first display unit 601 or the second display unit 602 can also be used for:

[0115] If, during the display of the simplified feedback page, it is detected that the simplified information content has been viewed by a new object, the browsing identifier will be updated and displayed on the simplified feedback page.

[0116] In another implementation, the feedback page that displays the simplified information content in the first feedback page and the second feedback page is called the simplified feedback page; the simplified feedback page also includes: the historical number of likes for the simplified information content;

[0117] Accordingly, the first display unit 601 or the second display unit 602 can also be used for:

[0118] During the display of the simplified feedback page, if it is detected that the simplified information content is liked by the target object and at least one other object at the same time, then a like animation is played on the simplified feedback page according to the virtual object corresponding to the target object and the virtual objects corresponding to each other object.

[0119] After the like animation finishes playing, update the historical like count.

[0120] In another implementation, the feedback page that displays all the information content in the first feedback page and the second feedback page is called the baseline feedback page; the all the information content includes: video identifiers of one or more videos related to the query information;

[0121] Accordingly, the first display unit 601 or the second display unit 602 can also be used for:

[0122] During the display of the benchmark feedback page, if any video identifier in the total information content is detected to be triggered, the video playback page is displayed.

[0123] Play the video corresponding to the triggered video identifier on the video playback page, and display a sharing component on the video playback page for sharing the simplified information content;

[0124] If the sharing component is triggered, a content sharing card corresponding to the simplified information content will be displayed on the video playback page.

[0125] In another embodiment, the first display unit 601 or the second display unit 602 may also be used for:

[0126] If a target interaction is detected during the display of the content sharing card, the video playback page and the content sharing card are canceled, and the baseline feedback page is redisplayed.

[0127] The target interactive operation includes: a swiping operation along a specified direction on the video playback page; or, the video playback page also includes a page return component, and the target interactive operation includes: a triggering operation on the page return component.

[0128] In another embodiment, the first feedback page includes all the information content; correspondingly, the first display unit 601 or the second display unit 602 can also be used for:

[0129] Before detecting the mode switching operation, obtain the content selection operation for all information content, and select the target information content in all information content according to the content selection operation;

[0130] After detecting the mode switching operation, the target information content is processed to extract a summary to obtain the simplified information content, and the second feedback page for displaying the query information is triggered.

[0131] In another embodiment, the simplified information content includes: a target summary text obtained by extracting a summary of all information content of the feedback information. The target summary text can be obtained by either the first display unit 601 or the second display unit 602, and specifically as follows:

[0132] A vocabulary is constructed based on all the information content of the feedback information, and the entire information content of the feedback information is divided into K sentences, where K is a positive integer; the vocabulary includes multiple high-frequency words, and the high-frequency words refer to words in the entire information content whose word frequency is greater than a preset threshold.

[0133] Based on the vocabulary, each of the K sentences is extracted into a summary to obtain K initial summary texts and the credibility of each initial summary text.

[0134] In descending order of credibility, a target number of initial summary texts are selected from the K initial summary texts based on the credibility of each initial summary text.

[0135] The initial summary texts of the selected target number are combined to obtain the target summary text.

[0136] In another implementation, for the kth sentence among the K sentences, the kth sentence is segmented into M input words, k∈[1,K], where M is a positive integer; wherein, the initial summary text corresponding to the kth sentence is determined based on the target words output by the decoder at N time steps;

[0137] The target word output by the decoder at time step t is determined as follows:

[0138] At the t-th time step of the decoder, the encoder is invoked to recursively encode each of the M words to obtain the hidden state of each input word at the t-th time step; where t∈[1,N];

[0139] Based on the target word that the decoder needs to generate at the t-th time step, and the degree of attention given to each input word, the attention weight of the hidden state of each input word at the t-th time step is determined.

[0140] Based on the attention weight of the hidden state of each input word at the t-th time step, the hidden states of the M input words at the t-th time step are integrated to obtain the semantic vector of the k-th sentence at the t-th time step.

[0141] The output probability of each high-frequency word in the vocabulary is calculated based on the semantic vector at time step t, and the probability calculation result is obtained. Based on the probability calculation result, high-frequency words are selected from the vocabulary as target words output by the decoder at time step t.

[0142] In another embodiment, when the first display unit 601 or the second display unit 602 determines the attention weight of the hidden state of each input word at the t-th time step based on the attention level of each input word to the target word to be generated by the decoder at the t-th time step, it may specifically be used for:

[0143] For any input word, obtain the attention weight of the hidden state of the input word at each historical time step, where the historical time step refers to the time step before the t-th time step;

[0144] The attention weights of the hidden states of any input word at each historical time step are summed, and the summation result is used as an attention weight decision factor.

[0145] Based on the attention weight decision factor and the degree of attention the decoder needs to generate for any input word at the t-th time step, the attention weight of the hidden state of any input word at the t-th time step is determined.

[0146] In another embodiment, when the first display unit 601 or the second display unit 602 selects high-frequency words from the vocabulary based on the probability calculation result as the target words output by the decoder at the t-th time step, it may specifically be used to:

[0147] Based on the probability calculation results, the high-frequency word with the highest output probability is selected from the vocabulary as the target word output by the decoder at the t-th time step;

[0148] Alternatively, the high-frequency words in the vocabulary can be sorted according to the probability calculation results in descending order of output probability, and the top k high-frequency words in the sorting results can be selected as the target words output by the decoder at the t-th time step.

[0149] In another embodiment, the first display unit 601 or the second display unit 602 may also be used for:

[0150] Compare the relationship between the largest output probability in the probability calculation results and the probability threshold;

[0151] If the maximum output probability is greater than or equal to the probability threshold, then the process of selecting a high-frequency word from the vocabulary based on the probability calculation result is triggered, which is then used as the target word output by the decoder at the t-th time step.

[0152] If the maximum output probability is less than the probability threshold, then based on the attention weight of the hidden state of each input word at the t-th time step, the input word corresponding to the hidden state with the largest attention weight is selected from the M input words as the target word output by the decoder at the t-th time step.

[0153] According to another embodiment of this application, Figure 6The various units in the information processing apparatus shown can be individually or entirely merged into one or more other units, or some of the units can be further divided into multiple functionally smaller units. This achieves the same operation without affecting the technical effects of the embodiments of this application. The above-mentioned units are divided based on logical functions. In practical applications, the function of one unit can also be implemented by multiple units, or the function of multiple units can be implemented by one unit. In other embodiments of this application, the information processing apparatus may also include other units. In practical applications, these functions can also be implemented with the assistance of other units, and can be implemented collaboratively by multiple units.

[0154] According to another embodiment of this application, the following can be achieved by running on a general-purpose computing device, such as a computer, which includes processing elements and storage elements such as a central processing unit (CPU), random access memory (RAM), and read-only memory (ROM), a device capable of performing operations such as... Figure 2 or Figure 4 The computer program (including program code) for each step involved in the corresponding method shown, to construct such... Figure 6 The information processing apparatus shown herein, and the information processing method for implementing the embodiments of this application, are described. The computer program may be recorded on, for example, a computer-readable recording medium, loaded onto the aforementioned computing device via the same medium, and run therein.

[0155] This application embodiment, after obtaining the feedback information corresponding to the query information, can provide multiple browsing modes, such as a first browsing mode and a second browsing mode, thereby enhancing the diversity of browsing modes. Specifically, the first browsing mode allows the user to browse the full or simplified content of the feedback information through the first feedback page; correspondingly, the second browsing mode allows the user to browse the simplified or full content of the feedback information through the second feedback page. Therefore, through the first and second browsing modes, users can not only comprehensively understand the query information through the full content, improving the comprehensiveness of information delivery, but also quickly and effectively understand the query information through the simplified content, improving the effectiveness of information delivery. Furthermore, by allowing users to freely switch between different browsing modes according to their actual needs, the convenience of browsing information can be improved, enabling users to have a good user experience when consuming content, thereby increasing user stickiness.

[0156] Based on the descriptions of the method and device embodiments above, this application also provides a terminal. Please refer to... Figure 7The terminal includes at least a processor 701, an input interface 702, an output interface 703, and a computer storage medium 704. The processor 701, input interface 702, output interface 703, and computer storage medium 704 within the terminal can be connected via a bus or other means. The computer storage medium 704 can be stored in the terminal's memory. The computer storage medium 704 is used to store computer programs, which include program instructions. The processor 701 is used to execute the program instructions stored in the computer storage medium 704. The processor 701 (or CPU (Central Processing Unit)) is the computing and control core of the terminal, suitable for implementing one or more instructions, specifically suitable for loading and executing one or more instructions to achieve corresponding method flows or corresponding functions.

[0157] In one embodiment, the processor 701 described in this application can be used to perform... Figure 2 or Figure 4 The series of processes shown can, for example, perform the following operations: display a first feedback page for the query information, which includes the information content of the feedback information corresponding to the query information in a first browsing mode; if a mode switching operation is detected, display a second feedback page for the query information; the second feedback page includes the information content of the feedback information in a second browsing mode; wherein, in the first feedback page and the second feedback page, there exists a feedback page containing the complete information content of the feedback information, and the other feedback page containing a simplified information content derived from the complete information content, and so on.

[0158] Furthermore, this application embodiment also provides a computer storage medium (memory), which is a memory device in a terminal used to store programs and data. It is understood that the computer storage medium here can include the built-in storage medium in the terminal, or it can include extended storage media supported by the terminal. The computer storage medium provides storage space, which stores the terminal's operating system. Furthermore, this storage space also stores one or more instructions suitable for loading and execution by a processor. These instructions can be one or more computer programs (including program code). It should be noted that the computer storage medium here can be high-speed RAM, or non-volatile memory, such as at least one disk storage device; optionally, it can also be at least one computer storage medium located remotely from the aforementioned processor. In specific implementations, one or more instructions stored in the computer storage medium can be loaded and executed by the processor. Figure 2 or Figure 4The various method steps are shown below.

[0159] This application embodiment, after obtaining the feedback information corresponding to the query information, can provide multiple browsing modes, such as a first browsing mode and a second browsing mode, thereby enhancing the diversity of browsing modes. Specifically, the first browsing mode allows the user to browse the full or simplified content of the feedback information through the first feedback page; correspondingly, the second browsing mode allows the user to browse the simplified or full content of the feedback information through the second feedback page. Therefore, through the first and second browsing modes, users can not only comprehensively understand the query information through the full content, improving the comprehensiveness of information delivery, but also quickly and effectively understand the query information through the simplified content, improving the effectiveness of information delivery. Furthermore, by allowing users to freely switch between different browsing modes according to their actual needs, the convenience of browsing information can be improved, enabling users to have a good user experience when consuming content, thereby increasing user stickiness.

[0160] It should be noted that, according to one aspect of this application, a computer program product or computer program is also provided, which includes computer instructions stored in a computer-readable storage medium. The processor of a terminal reads the computer instructions from the computer-readable storage medium, and executes the computer instructions, causing the terminal to perform the aforementioned... Figure 2 or Figure 4 The methods are provided in various alternative ways in the illustrated method embodiments.

[0161] Furthermore, it should be understood that the above-disclosed embodiments are merely preferred embodiments of this application and should not be construed as limiting the scope of this application. Therefore, any equivalent variations made in accordance with the claims of this application are still within the scope of this application.

Claims

1. An information processing method, characterized in that, include: The first feedback page displays the query information. The first feedback page includes: the information content of the feedback information corresponding to the query information in the first browsing mode. The information content in the first browsing mode is the complete information content of the feedback information. Based on the content selection operation for all information content, the target information content in all information content is selected. The content selection operation is an operation of directly selecting a part of the content in all information content. If a mode switching operation is detected, a second feedback page for the query information is displayed; the second feedback page includes: the information content of the feedback information in the second browsing mode, the information content in the second browsing mode being a simplified information content obtained based on the target information content, the simplified information content being obtained by performing summary extraction processing on the target information content after detecting the mode switching operation; Wherein, after displaying the first feedback page, if no content selection operation for all information content is detected before the mode switching operation is detected, then the simplified information content in the second feedback page is obtained by extracting a summary of all information content of the feedback information. The summary of all information content of the feedback information is processed to obtain target summary text. The target summary text is obtained as follows: a vocabulary is constructed based on all information content of the feedback information, and all information content of the feedback information is split into K sentences, where K is a positive integer; the vocabulary includes multiple high-frequency words, which are words in all information content with a frequency greater than a preset threshold; each of the K sentences is processed to extract a summary according to the vocabulary to obtain K initial summary texts and the corresponding credibility of each initial summary text; according to the credibility from high to low, a target number of initial summary texts are selected from the K initial summary texts; the selected target number of initial summary texts are combined to obtain the target summary text.

2. The method as described in claim 1, characterized in that, Any feedback page includes: a first mode component corresponding to the first browsing mode, and a second mode component corresponding to the second browsing mode; Furthermore, when displaying any of the feedback pages, the mode component corresponding to the browsing mode is selected, while other mode components are unselected. The mode switching operation includes selecting the second mode component on the first feedback page.

3. The method as described in claim 1 or 2, characterized in that, The method further includes: If a sharing operation is detected for a content sharing card, the content sharing card is sent to the object indicated by the sharing operation.

4. The method as described in claim 1 or 2, characterized in that, In the first feedback page and the second feedback page, the feedback page that displays the simplified information content is called the simplified feedback page; the simplified feedback page also includes a browsing identifier, which is used to indicate at least one of the following: the historical browsing volume of the simplified information content, and the object identifier of all or part of the objects that have historically browsed the simplified information content; The method further includes: If, during the display of the simplified feedback page, it is detected that the simplified information content has been viewed by a new object, the browsing identifier will be updated and displayed on the simplified feedback page.

5. The method as described in claim 1 or 2, characterized in that, In the first feedback page and the second feedback page, the feedback page that displays the simplified information content is referred to as the simplified feedback page; The simplified feedback page also includes: the historical number of likes for the simplified information content; The method further includes: During the display of the simplified feedback page, if it is detected that the simplified information content is liked by the target object and at least one other object at the same time, then a like animation is played on the simplified feedback page according to the virtual object corresponding to the target object and the virtual objects corresponding to each other object. After the like animation finishes playing, update the historical like count.

6. The method as described in claim 1 or 2, characterized in that, In the first feedback page and the second feedback page, the feedback page that displays all the information content is referred to as the baseline feedback page; The complete information content includes: video identifiers of one or more videos related to the query information; The method further includes: During the display of the benchmark feedback page, if any video identifier in the total information content is detected to be triggered, the video playback page is displayed. Play the video corresponding to the triggered video identifier on the video playback page, and display a sharing component on the video playback page for sharing the simplified information content; If the sharing component is triggered, a content sharing card corresponding to the simplified information content will be displayed on the video playback page.

7. The method as described in claim 6, characterized in that, The method further includes: If a target interaction is detected during the display of the content sharing card, the video playback page and the content sharing card are canceled, and the baseline feedback page is redisplayed. The target interactive operation includes: a swiping operation along a specified direction on the video playback page; or, the video playback page also includes a page return component, and the target interactive operation includes: a triggering operation on the page return component.

8. The method as described in claim 1, characterized in that, For the k-th sentence among the K sentences, the k-th sentence is segmented into M input words, k∈[1,K], where M is a positive integer; wherein, the initial summary text corresponding to the k-th sentence is determined based on the target words output by the decoder at N time steps; The target word output by the decoder at time step t is determined as follows: At the t-th time step of the decoder, the encoder is invoked to recursively encode each of the M words to obtain the hidden state of each input word at the t-th time step; where t∈[1,N]; Based on the target word that the decoder needs to generate at the t-th time step, and the degree of attention given to each input word, the attention weight of the hidden state of each input word at the t-th time step is determined. Based on the attention weight of the hidden state of each input word at the t-th time step, the hidden states of the M input words at the t-th time step are integrated to obtain the semantic vector of the k-th sentence at the t-th time step. The output probability of each high-frequency word in the vocabulary is calculated based on the semantic vector at time step t, and the probability calculation result is obtained. Based on the probability calculation result, high-frequency words are selected from the vocabulary as target words output by the decoder at time step t.

9. The method as described in claim 8, characterized in that, The step of determining the attention weight of the hidden state of each input word at time step t, based on the attention level of each input word to the target word to be generated by the decoder at time step t, includes: For any input word, obtain the attention weight of the hidden state of the input word at each historical time step, where the historical time step refers to the time step before the t-th time step; The attention weights of the hidden states of any input word at each historical time step are summed, and the summation result is used as an attention weight decision factor. Based on the attention weight decision factor and the degree of attention the decoder needs to generate for any input word at the t-th time step, the attention weight of the hidden state of any input word at the t-th time step is determined.

10. The method as described in claim 8, characterized in that, The step of selecting high-frequency words from the vocabulary based on the probability calculation results as the target words output by the decoder at time step t includes: Based on the probability calculation results, the high-frequency word with the highest output probability is selected from the vocabulary as the target word output by the decoder at the t-th time step; Alternatively, the high-frequency words in the vocabulary can be sorted according to the probability calculation results in descending order of output probability, and the top k high-frequency words in the sorting results can be selected as the target words output by the decoder at the t-th time step.

11. The method according to any one of claims 8-10, characterized in that, The method further includes: Compare the relationship between the largest output probability in the probability calculation results and the probability threshold; If the maximum output probability is greater than or equal to the probability threshold, then the process of selecting a high-frequency word from the vocabulary based on the probability calculation result is triggered, which is then used as the target word output by the decoder at the t-th time step. If the maximum output probability is less than the probability threshold, then based on the attention weight of the hidden state of each input word at the t-th time step, the input word corresponding to the hidden state with the largest attention weight is selected from the M input words as the target word output by the decoder at the t-th time step.

12. An information processing device, characterized in that, include: The first display unit is used to display a first feedback page for query information. The first feedback page includes: the information content of the feedback information corresponding to the query information in a first browsing mode. The information content in the first browsing mode is all the information content of the feedback information. The second display unit is used to select target information content from the total information content based on a content selection operation for the total information content, wherein the content selection operation is an operation of directly selecting a portion of the content from the total information content; if a mode switching operation is detected, a second feedback page for the queried information is displayed; the second feedback page includes: information content of the feedback information in a second browsing mode, wherein the information content in the second browsing mode is a simplified information content obtained based on the target information content, and the simplified information content is obtained by performing summary extraction processing on the target information content after detecting the mode switching operation; Wherein, after displaying the first feedback page, if no content selection operation for all information content is detected before the mode switching operation is detected, then the simplified information content in the second feedback page is obtained by extracting a summary of all information content of the feedback information. The summary of all information content of the feedback information is processed to obtain target summary text. The target summary text is obtained as follows: a vocabulary is constructed based on all information content of the feedback information, and all information content of the feedback information is split into K sentences, where K is a positive integer; the vocabulary includes multiple high-frequency words, which are words in all information content with a frequency greater than a preset threshold; each of the K sentences is processed to extract a summary according to the vocabulary to obtain K initial summary texts and the corresponding credibility of each initial summary text; according to the credibility from high to low, a target number of initial summary texts are selected from the K initial summary texts; the selected target number of initial summary texts are combined to obtain the target summary text.

13. A computer device, comprising an input interface and an output interface, characterized in that, Also includes: A processor, adapted to implement one or more instructions; as well as, A computer storage medium storing one or more instructions, said one or more instructions being adapted to be loaded by the processor and executed as described in any one of claims 1-11.

14. A computer storage medium, characterized in that, The computer storage medium stores one or more instructions, which are adapted to be loaded by a processor and executed as described in any one of claims 1-11.

15. A computer program product, comprising a computer program, characterized in that, When the computer program is executed by a processor, it implements the information processing method as described in any one of claims 1-11.

Images