Information processing method, device, electronic device and storage medium based on large model

Abstract

The present invention provides a large model-based information processing method, device, electronic device and storage medium, which relates to the field of artificial intelligence technology, especially to the technical fields of intelligent search, deep learning, large models, generative models, etc., and can be applied to generative search, document intelligent editing, intelligent assistants, Internet e-commerce, intelligent video browsing and other application scenarios. The method includes: responding to an object's information acquisition request for a target page, determining page information related to the target page; determining structured feedback information corresponding to the object's information acquisition request based on the page information, wherein the structured feedback information is obtained by using the large model to fuse information content with the same text structure level in the page information, and the structured feedback information has a text structure level.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and in particular to technical fields such as intelligent search, deep learning, large models, and generative models. It can be applied to application scenarios such as generative search, intelligent document editing, intelligent assistants, Internet e-commerce, and intelligent video browsing. Background Art

[0002] With the popularity of smart terminal devices such as smartphones and tablet computers, users can easily browse text, video, image and other page information through the display screen of smart terminal devices, which improves the efficiency of obtaining information. Users can also place orders online by executing the products displayed on the page, thus completing the shopping process conveniently. Summary of the invention

[0003] The present invention provides an information processing method, device, electronic equipment, storage medium and program product based on a large model.

[0004] According to one aspect of the present invention, there is provided an information processing method based on a big model, comprising: determining page information related to a target page in response to an information acquisition request of an object; determining structured feedback information corresponding to the information acquisition request of the object based on the page information, wherein the structured feedback information is obtained by fusing information content having the same text structure level in page information using the big model, and the structured feedback information has a text structure level.

[0005] According to another aspect of the present invention, there is provided an information processing device based on a large model, comprising: a page information determination module, for determining page information related to a target page in response to an information acquisition request of an object for the target page; a structured feedback information determination module, for determining structured feedback information corresponding to the information acquisition request of the object based on the page information, wherein the structured feedback information is obtained by fusing information content with the same text structure level in page information using the large model, and the structured feedback information has a text structure level.

[0006] According to another aspect of the present invention, an electronic device is provided, comprising: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method provided according to an embodiment of the present invention.

[0007] According to another aspect of the present invention, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the method provided by an embodiment of the present invention.

[0008] According to another aspect of the present invention, a computer program product is provided, including a computer program, and when the computer program is executed by a processor, the method provided by the embodiment of the present invention is implemented.

[0009] According to an embodiment of the present invention, by obtaining page information related to the target page according to the information acquisition request of the object for the target page, and fusing the information content with the same text structure level in the page information according to the large model, the obtained structured feedback information can more accurately represent the semantic content in the page information, avoiding semantic ambiguity or error. At the same time, based on the text structure level of the structured feedback information, the structured arrangement and integration of the page information can be realized, so that the object can improve the efficiency of information acquisition and information browsing by browsing the structured feedback information, thereby improving the user's interactive experience.

[0010] It should be understood that the contents described in this section are not intended to identify the key or important features of the embodiments of the present invention, nor are they intended to limit the scope of the present invention. Other features of the present invention will become easily understood through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] The accompanying drawings are used to better understand the present invention and do not constitute a limitation of the present invention.

[0012] Figure 1 An exemplary system architecture to which a large model-based information processing method and apparatus can be applied according to an embodiment of the present invention is shown;

[0013] Figure 2 A flowchart of a large model-based information processing method according to an embodiment of the present invention is shown;

[0014] Figure 3 An application scenario diagram of the information processing method based on a large model according to an embodiment of the present invention is shown;

[0015] Figure 4 A schematic diagram showing a principle of determining page information according to an embodiment of the present invention is shown;

[0016] Figure 5 An application scenario diagram of a large model-based information processing method according to another embodiment of the present invention is shown;

[0017] Figure 6 An application scenario diagram of an information processing method based on a large model according to another embodiment of the present invention is shown;

[0018] Figure 7 An application scenario diagram of a large model-based information processing method according to yet another embodiment of the present invention is shown;

[0019] Figure 8A schematic diagram of a large model-based information processing method according to an embodiment of the present invention is shown;

[0020] Fig. 9 A schematic diagram of an operation flow for executing a large model-based information processing method according to an embodiment of the present invention is shown;

[0021] Fig.10 A framework diagram of an information processing system for executing a large model-based information processing method according to an embodiment of the present invention is shown;

[0022] Fig.11 A block diagram showing an information processing device based on a large model according to an embodiment of the present invention; and

[0023] Fig.12 A block diagram of an electronic device suitable for implementing a large model-based information processing method according to an embodiment of the present invention is shown. DETAILED DESCRIPTION

[0024] The following is a description of exemplary embodiments of the present invention in conjunction with the accompanying drawings, including various details of the embodiments of the present invention to facilitate understanding, which should be considered as merely exemplary. Therefore, it should be recognized by those of ordinary skill in the art that various changes and modifications may be made to the embodiments described herein without departing from the scope and spirit of the present invention. Similarly, for clarity and conciseness, the description of well-known functions and structures is omitted in the following description.

[0025] In the technical solution of the present invention, the acquisition, storage and application of user personal information involved all comply with the provisions of relevant laws and regulations, take necessary confidentiality measures, and do not violate public order and good morals.

[0026] The inventors have found that it is difficult for users to obtain the required information accurately and conveniently when browsing pages. It usually requires multiple interactive operations (such as sliding pages) and takes a long time, resulting in a poor user experience.

[0027] The embodiment of the present invention provides an information processing method, device, electronic device, storage medium and program product based on a large model. The information processing method based on a large model includes: in response to an object's information acquisition request for a target page, determining page information related to the target page; according to the page information, determining structured feedback information corresponding to the object's information acquisition request, wherein the structured feedback information is obtained by fusing information content with the same text structure level in the page information using the large model, and the structured feedback information has a text structure level.

[0028] According to an embodiment of the present invention, by obtaining page information related to the target page according to the information acquisition request of the object for the target page, and fusing the information content with the same text structure level in the page information according to the large model, the obtained structured feedback information can more accurately represent the semantic content in the page information, avoiding semantic ambiguity or error. At the same time, based on the text structure level of the structured feedback information, the structured arrangement and integration of the page information can be realized, so that the object can improve the efficiency of information acquisition and information browsing by browsing the structured feedback information, thereby improving the user's interactive experience.

[0029] Figure 1 An exemplary system architecture to which a large model-based information processing method and apparatus can be applied according to an embodiment of the present invention is shown.

[0030] It should be noted that Figure 1 The examples shown are only examples of system architectures to which the embodiments of the present invention can be applied, in order to help those skilled in the art understand the technical content of the present invention, but do not mean that the embodiments of the present invention cannot be used in other devices, systems, environments or scenarios. For example, in another embodiment, an exemplary system architecture to which the information processing method and apparatus based on a large model can be applied may include a terminal device, but the terminal device may implement the information processing method and apparatus based on a large model provided by the embodiments of the present invention without interacting with a server.

[0031] like Figure 1 As shown, the system architecture 100 according to this embodiment may include a first terminal device 101, a second terminal device 102, a third terminal device 103, a network 104, and a server 105. The network 104 is used to provide a medium for a communication link between the first terminal device 101, the second terminal device 102, the third terminal device 103, and the server 105. The network 104 may include various connection types, such as wired and / or wireless communication links, etc.

[0032] The user may use the first terminal device 101, the second terminal device 102, and the third terminal device 103 to interact with the server 105 through the network 104 to receive or send messages, etc. Various communication client applications may be installed on the first terminal device 101, the second terminal device 102, and the third terminal device 103, such as knowledge reading applications, web browser applications, search applications, instant messaging tools, email clients, and / or social platform software, etc. (only as examples).

[0033] The first terminal device 101, the second terminal device 102, and the third terminal device 103 may be various electronic devices having display screens and supporting web browsing, including but not limited to smart phones, tablet computers, laptop computers, desktop computers, and the like.

[0034] The server 105 may be a server that provides various services, such as a background management server (only as an example) that provides support for the content browsed by the user using the first terminal device 101, the second terminal device 102, and the third terminal device 103. The background management server may analyze and process the received data such as user requests, and feed back the processing results (such as web pages, information, or data obtained or generated according to user requests) to the terminal device.

[0035] The server can be a cloud server, also known as a cloud computing server or cloud host, which is a host product in the cloud computing service system to solve the defects of difficult management and weak business scalability in traditional physical hosts and VPS services ("Virtual Private Server", or "VPS" for short). The server can also be a server of a distributed system, or a server combined with blockchain.

[0036] It should be noted that the information processing method based on the large model provided in the embodiment of the present invention can generally be executed by the server 105. Accordingly, the information processing device based on the large model provided in the embodiment of the present invention can generally be set in the server 105. The information processing method based on the large model provided in the embodiment of the present invention can also be executed by a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105. Accordingly, the information processing device based on the large model provided in the embodiment of the present invention can also be set in a server or server cluster that is different from the server 105 and can communicate with the first terminal device 101, the second terminal device 102, the third terminal device 103 and / or the server 105.

[0037] It should be understood that Figure 1 The number of terminal devices, networks and servers in the embodiment is only for illustration. Any number of terminal devices, networks and servers may be provided according to implementation requirements.

[0038] Figure 2 A flow chart of an information processing method based on a large model according to an embodiment of the present invention is shown.

[0039] like Figure 2 As shown, the large model-based information processing method includes operations S210 to S220.

[0040] In operation S210 , in response to an information acquisition request of an object for a target page, page information related to the target page is determined.

[0041] In operation S220, structured feedback information corresponding to the information acquisition request of the object is determined according to the page information.

[0042] According to an embodiment of the present invention, the structured feedback information is obtained by integrating information content having the same text structure level in page information using a large model, and the structured feedback information has a text structure level.

[0043] It should be noted that the text structure level (or simply the structure level) can also be called the text structure hierarchy. The information content of the page information can include data in any format, such as text, images, videos, charts, tables, etc. The information content with the same text structure level can be understood as the information content that expresses similar or identical structural subjects. For example, the information content with the same text structure level can be the text, images, videos, tables, etc. of the same production steps in the dessert production method.

[0044] According to an embodiment of the present invention, different structured feedback information may have a text structure hierarchical relationship, and the text structure hierarchical relationship may be represented in any structural form such as a tree structure or an undirected graph structure.

[0045] According to an embodiment of the present invention, the information acquisition request for the target page may be generated based on the interactive operation of the object on the target page. Alternatively, it may be generated based on other interactive operations of the object, such as generating an information acquisition request for the search information input operation input into the search information input box.

[0046] It should be noted that the interactive operation used to determine the information acquisition request may include any type of operation, for example, it may include gesture interactive operations such as click operations, but is not limited to this. It may also include. The embodiments of the present invention do not limit the specific types of interactive operations, and technical personnel in this field can choose according to actual needs.

[0047] According to an embodiment of the present invention, the target page may include the page that the object is browsing, or the target page may also include the page specified by the information acquisition request. The target page may include any type of page such as a search result page, a video playback page, a detail page, a list page, a map navigation page, etc. The embodiment of the present invention does not limit the specific type of the target page.

[0048] According to an embodiment of the present invention, the page information may include information content in the target page, such as text, images, videos, etc. of the target page. However, the page information may also include information content in associated pages whose similarity to the target page satisfies the page similarity condition. The associated page may be, for example, a jump page linked to the target page, but is not limited thereto, and may also include pages with high similarity to the theme and key information of the target page, such as pages for introducing the same tourist destination, pages associated with the page topic of the target page, and the like.

[0049] According to an embodiment of the present invention, a large model may refer to a deep learning model with large-scale model parameters, and a large model usually contains hundreds of millions, tens of billions, hundreds of billions, trillions, or even more than ten trillion model parameters. The large model may include a large-scale language model (LLM), GPT (Generative Pre-trained Transformer), a large visual model, a multimodal large model, and the like. The large model involved in the embodiment of the present invention may be a general large model, or it may be an expert large model obtained after fine-tuning (Fine Tune) based on demand, and the embodiment of the present invention does not limit this.

[0050] According to an embodiment of the present invention, the structured feedback information may be obtained after the large model processes the page information. For example, the structured feedback information may be obtained by using the large model to fuse the information content with the same text structure level (or text structure level) in the page information, and the structured feedback information may have a text structure level. The structural hierarchical relationship between multiple structured feedback information may be obtained by aggregating and semantically summarizing the information content in the page information based on the semantic understanding ability of the large model. The structured feedback information may be the key information of each structural level output by the large model when the information content of the page information is fully understood. The key information of the same structural level may be the information content with the same structural level in the characterization aggregated page information. The large model may perform semantic processing processes such as deduplication, semantic aggregation, and redundant information deletion on the information content with the same structural level in the page information of the same structural level, so that the structured feedback information output by the large model can clearly display the key information content in the page information based on the structural level and the aggregated key information.

[0051] According to an embodiment of the present invention, structured feedback information may be information that is required to represent an object's interactive operation on a target page. The structured feedback information may be based on one or more structural levels to display information content such as text, images, videos, and tables in the page information. The structural level may include a title level in the text, such as a structural level represented by a first-level title and a second-level title. For another example, the structural level may also include the execution steps of an operation process. Representing the information content in the page information based on the structural level allows the object to quickly obtain the required information content through the structural level of the structured feedback information, thereby reducing the frequency of interactive operations of the object in the information acquisition process, improving the object's information acquisition efficiency, and thereby improving the user experience.

[0052] In one example, the structured feedback information may include a title and a summary text.

[0053] In one example, the structured information may include a title, summary text, and an image.

[0054] According to an embodiment of the present invention, an information acquisition request may be generated based on a user's interactive operation on a target page, and the interactive operation may include at least one gesture operation of a pinch operation, a drag operation, a click operation, and a box selection operation. The pinch operation may include an object's two fingers contacting a target page displayed in an interactive interface in an open manner, and moving the two fingers contacting the interactive interface from an open state to a closed state. The movement of the two fingers from an open state to a closed state may be similar to the two-finger pinching method.

[0055] It should be noted that the pinch operation can be performed based on two or more execution bodies, as long as the pinch operation can be simulated.

[0056] According to an embodiment of the present invention, the drag operation may include an interactive operation action of dragging any page element in the target page. The drag operation may be in any direction.

[0057] According to the embodiment of the present invention, the click operation may include any type of click operation, such as a single click (or single click), multiple clicks (such as double clicks), long press clicks, etc. The embodiment of the present invention does not limit the specific execution method of the click operation or the execution duration.

[0058] According to an embodiment of the present invention, the frame selection operation may include an operation of selecting a page area for a target page.

[0059] According to an embodiment of the present invention, any one of a pinch operation, a drag operation, a click operation, and a box selection operation may be selected to determine an interactive operation for a target page. Alternatively, an operation combination consisting of at least two of the pinch operation, the drag operation, the click operation, and the box selection operation may be selected as an interactive operation for a target page.

[0060] According to an embodiment of the present invention, an information acquisition request may also be generated based on a gesture operation combined with other types of interactive operations. For example, an information acquisition request may be generated based on a gesture operation combined with other types of operations such as a voice input operation, a video input operation, a text input operation, etc. The information acquisition request may carry the operation attributes of the interactive operation. The operation attributes may include the position, type, duration, etc. of the gesture operation, and the operation attributes may also include information such as text, audio, and image input of the interactive operation.

[0061] According to an embodiment of the present invention, the information processing method based on the large model may further include sending structured feedback information to a device related to the object. The object may browse the information content to be obtained more conveniently through a device such as a smart phone.

[0062] According to an embodiment of the present invention, the device can display structured feedback information and can also display target elements related to a target page. The target elements can be obtained by performing page display attribute conversion on the target page.

[0063] According to an embodiment of the present invention, converting the page attributes of the target page may include converting any page display attributes of the target page, such as size, shape, display range, display color, etc. For example, the size of the target page may be reduced, and the obtained target element may be a thumbnail of the target page.

[0064] According to an embodiment of the present invention, converting the page attributes of the target page may also include converting any page display attributes such as size, shape, display range, and display color of page elements in the target page. For example, the cover image of the target page may be selected and the cover image may be reduced or enlarged to obtain the target element.

[0065] According to an embodiment of the present invention, the target element and the structured feedback information are displayed on an interactive interface related to the object. The target element can be used as a prompt mark of the target page to facilitate the object to associate the operation intention related to the target page with the structured feedback information, thereby prompting the object to understand the information related to the operation intention more accurately, clearly and quickly under conditions that affect the object's attention such as being interrupted during browsing the target page, so as to improve the efficiency of information browsing.

[0066] According to an embodiment of the present invention, converting the page display attribute of the target page based on the interactive operation to obtain the target element may include: determining a page operation area from the target page based on the interactive operation; and determining the target element according to the page operation area.

[0067] According to an embodiment of the present invention, the page operation area can be determined based on the positional relationship between the operation position or operation area of ​​the interactive operation and the page area in the target page. For example, the page operation area can be determined to be the comment area of ​​the target page based on the fact that the operation position of the interactive operation in the interactive interface is located in the comment area of ​​the target page.

[0068] According to an embodiment of the present invention, determining the target element according to the page operation area may include determining the target element according to any type of page element such as text, image, etc. in the page operation area. It may include using the page element in the page operation area as the target element. Or it may also include converting any page display attribute such as size, shape, display range, display color, etc. of the page element in the page operation area to obtain the target element. The embodiment of the present invention does not limit the specific method of determining the target element.

[0069] For example, the title area displayed on the target page can be used as the page operation area, and a screenshot of the title area can be taken. The obtained target element can be a screenshot of the title area of ​​the target page.

[0070] For another example, the target element may be determined by converting the page display attributes such as converting the text format and enlarging the text size of the title text in the title area.

[0071] Figure 3 An application scenario diagram of the large model-based information processing method according to an embodiment of the present invention is shown.

[0072] like Figure 3 As shown, the application scenario may include a touch screen display 300 of a terminal device and a cloud server C310. The touch screen display 300 may display a target page 310. The target page 310 may be a recipe page for lotus root and pork ribs soup. The page information of the target page 310 may include a title "Lotus root and pork ribs soup will be more fragrant if made this way", and may also include a text of the steps for making lotus root and a picture showing the steps. The picture showing the steps may include picture 1 and picture 2.

[0073] like Figure 3As shown, the user object can perform a pinch operation O310 on any area of ​​the target page 310. The touch display screen 300 detects the pinch operation O310 and can send the pinch operation instruction to the processor of the terminal device. The processor generates an information acquisition request in response to the pinch operation instruction corresponding to the pinch operation O310, and sends the information acquisition request to the cloud server C310. The cloud server C310 executes the information processing method provided by the embodiment of the present invention in response to the information acquisition request, and uses the large model deployed on the cloud server C310 to process the page information of the target page 310 to generate structured feedback information. The structured feedback information corresponding to the information acquisition request may include first feedback information J311, second feedback information J312, and third feedback information J313. The cloud server C310 can send the first feedback information J311, the second feedback information J312, and the third feedback information J313 to the terminal device so that the terminal device can display it on the touch display screen 300. The first feedback information J311 may be step 1 of making lotus root and pork ribs soup. The first feedback information J311 may include the title of step 1 "Blanching to remove fishy smell", and the step summary of step 1 "Put the pork ribs in cold water, add a little green onion, ginger, and cooking wine, blanch for 5 minutes, remove the foam and wash". The second feedback information J312 may include the title of step 2 "Stir-frying pork ribs", and the step summary of step 2 and the step schematic diagram picture 2. The third feedback information J313 may include the title of step 3 "Stew over low heat", and the step summary of step 3 and the step schematic diagram picture 1. The first feedback information J311, the second feedback information J312, and the third feedback information J313 may be displayed on the touch display screen 300 in the form of titles, step summaries, and display pictures, so that the user can conveniently browse the production points of each step, as well as the display pictures corresponding to the production points of the step, so that the user object can clearly and conveniently obtain the page content in the target page 310 through simple interactive operations, thereby improving the efficiency of information acquisition and information browsing. The terminal device may also determine structured feedback information corresponding to the pinch operation O310 according to the page information of the target page 310 , and generate a target element B310 according to the title in the page information.

[0074] like Figure 3 As shown, the touch display screen 300 can also display a user feedback window J321, which allows the user object to input evaluation content for the structured feedback information generated above, or can also be used to view the evaluation content of other user objects for the structured feedback information of the target page 310, so as to enrich the interaction scene. At the same time, the smart terminal device can also fine-tune the large model by collecting and analyzing the feedback results under the condition of authorization by the user object, thereby improving the prediction accuracy of the structured feedback information.

[0075] It should be noted that Figure 3 The first feedback information J311, the second feedback information J312 and the third feedback information J313 shown can be obtained after processing the page information of the target page 310 based on the large model deployed on the cloud, and cached in the storage area of ​​the cloud server C310. The smart terminal device can send an information acquisition request to the cloud based on the cloud server C310, and then receive the first feedback information J311, the second feedback information J312 and the third feedback information J313 from the cloud storage.

[0076] According to an embodiment of the present invention, determining page information related to a target page may include: performing intent recognition on an object based on an information acquisition request to obtain the object's information acquisition intent; and determining an associated page corresponding to the target page based on the information extension intent.

[0077] According to an embodiment of the present invention, the information acquisition intention may include the information extension intention, which may indicate that the object needs to obtain other information content in addition to the information content recorded on the target page. The object's intention may be identified based on the object's historical interaction behavior (such as historical search information, historical browsing information, etc.) to obtain the information extension intention. Alternatively, the object's information acquisition request may be identified based on the preference attributes configured by the object to obtain the information extension intention. The information acquisition request or data related to the information acquisition request may be processed based on a neural network algorithm, such as a convolutional neural network algorithm, a multi-layer perceptron algorithm, or other neural network algorithm to achieve the intention identification of the object.

[0078] In one example, intent identification can be performed based on the page authority weight of the target page related to the information acquisition request to obtain information extension intent. The page authority weight can characterize the degree of authority of the target page for the subject type and topic type contained in the target page. For example, for the subject type "primary school admission notice", if the target page is information reposted by a social account, it can be determined that the authority weight of the target page for the subject type "primary school admission notice" is lower than a preset authority weight threshold, and in response to the authority weight being lower than the preset authority weight threshold, the information extension intent related to the object can be determined, so as to determine the associated pages related to the target page with an authority weight higher than the preset authority weight threshold.

[0079] In one example, intent recognition can be performed based on the object's object preference attribute. When the similarity between the object preference attribute and the subject of the target page meets a preset similarity threshold, it can be considered that the object has a high degree of interest in the subject of the target page, and the content information of the target page needs to be extended to meet the object's need to obtain extended information content, thereby avoiding the object from obtaining extended information content by clicking or searching other pages related to the subject of the target page, thereby improving the object's information acquisition efficiency.

[0080] According to an embodiment of the present invention, determining the associated pages associated with the target page based on the information extension intention may include retrieving the associated pages with a high similarity to the target page based on the information extension intention. For example, the page theme of the target page may be used for retrieval, and the top k pages in the retrieval ranking results may be used as the associated pages associated with the target page, so that the retrieval function and the large model task execution function may be combined to generate structured feedback information that matches the information acquisition intention of the object, improve the information richness of the structured feedback information, and improve the matching degree between the structured feedback information and the information acquisition needs of the object.

[0081] It should be noted that retrieving related pages with high similarity to the target page based on the information extension intention may include searching from the Internet based on a search engine. However, it is not limited to this, and may also include retrieving related pages associated with the target page in a designated storage area (such as a pre-built page information library).

[0082] According to an embodiment of the present invention, the page information is determined based on the associated pages. For example, the page information that the large model needs to process can be determined from the associated pages. For another example, the page information code corresponding to the associated page can be obtained from a preset storage area based on the key information of the associated page, so as to improve the efficiency of obtaining structured feedback information.

[0083] In one example, the page information may be determined based on the associated page and the target page. For example, the page text in the target page may be concatenated with the page text in the associated page to obtain the page information. Alternatively, any type of information such as images, videos, audio, etc. of the target page and the associated page may be identified and encoded to obtain the page information.

[0084] According to an embodiment of the present invention, performing intent recognition on an object based on an information acquisition request to obtain the information acquisition intent of the object may include: performing similarity analysis based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operations of the object to obtain a type similarity result; and determining the information extension intention when the type similarity result meets a preset information type similarity condition.

[0085] According to an embodiment of the present invention, the information type of the information acquisition request may include the subject type of the object input information carried in the information acquisition request, or may also include the page subject type, page topic type, etc. of the target page related to the information acquisition request. The historical information type may be the page subject type, page topic type, etc. that the object has browsed in the historical period. The type similarity result may represent the similarity between the historical information type and the information type, for example, it may be a similarity value.

[0086] According to an embodiment of the present invention, when the type similarity result meets the preset information type similarity condition, it can indicate that the object has queried or browsed the historical information type in the historical period, and in the case of an information acquisition request generated based on the target page, the object has a need to obtain extended information, thereby obtaining the information extension intention.

[0087] According to an embodiment of the present invention, performing a similarity analysis based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operation of the object, obtaining a type similarity result may include: determining a first similarity between the information type and the first historical information type corresponding to the first historical interaction operation; determining a second similarity between the information type and the second historical information type corresponding to the second historical interaction operation; and in response to the first similarity being lower than a preset similarity threshold and the second similarity being higher than or equal to the preset similarity threshold, determining a type similarity result that meets the information type similarity condition.

[0088] According to an embodiment of the present invention, the first historical interaction operation is a historical interaction operation in a first historical period, the second historical interaction operation is a historical interaction operation in a second historical period, and the timing attribute of the first historical period is earlier than the timing attribute of the second historical period. For example, the first historical interaction operation is an interaction operation performed from December 1, 2023 to January 1, 2024, and the second historical interaction operation may be an interaction operation performed from January 1, 2024 to January 7, 2024.

[0089] According to an embodiment of the present invention, when the first similarity is lower than the preset similarity threshold, it can be indicated that the object did not pay attention to the information type of the target page in an earlier historical period, or in a long historical period. When the second similarity is higher than or equal to the preset similarity threshold, it can be indicated that the object is more concerned about the information type of the target page in a recent historical period, or in a recent period. It can be understood that the object has only recently paid attention to the type of the target page, and there is no information accumulation related to the information type of the target page. Therefore, in response to the first similarity being lower than the preset similarity threshold, and the second similarity being higher than or equal to the preset similarity threshold, a type similarity result that satisfies the information type similarity condition is determined, and it can be determined based on the type similarity result that the object needs to perform information expansion, so that the page information can be enriched by determining the associated pages with a high degree of relevance to the target page, and then the structured feedback information generated by the large model can meet the user's information extension needs, and the object can obtain the information it needs to know through a relatively simple operation while reducing the frequency of object operations, thereby improving the efficiency of information acquisition.

[0090] According to an embodiment of the present invention, based on the information extension intention, determining the associated page corresponding to the target page may include: determining a first associated page based on the information extension intention and the object's historical information acquisition request, the associated page including the first associated page.

[0091] According to an embodiment of the present invention, the historical information acquisition request may be generated by the object in the historical period based on the historical interaction operation, for example, it may be a page browsing request, a search request, etc. of the object in the historical period.

[0092] According to an embodiment of the present invention, in response to the information extension intention, sampling can be performed according to the historical information acquisition request of the object in the preset sampling history period to obtain the sampling history information acquisition request. The history page corresponding to the sampling history information acquisition request is used as the first associated page.

[0093] In one example, N consecutive historical information acquisition requests before the information acquisition request can be used as sampled historical information acquisition requests, and the historical pages corresponding to the N sampled historical information acquisition requests can be used as the first associated page, so as to facilitate the object's intention identification through the historical pages recently browsed by the object, so that the information extension intention can more accurately characterize the type of information that the object has recently paid attention to.

[0094] According to an embodiment of the present invention, determining an associated page corresponding to a target page based on the information extension intention may include: determining a second associated page based on the information extension intention and page key information of the target page.

[0095] According to an embodiment of the present invention, the associated page may include a second associated page.

[0096] According to an embodiment of the present invention, the key page information may include the subject information of the target page, information with a frequency greater than a preset frequency threshold, but is not limited thereto, and may also include information associated with the target page determined based on preset rules, such as text information carried in an information acquisition request corresponding to the target page, etc. The embodiment of the present invention does not limit the specific type of the key page information.

[0097] According to an embodiment of the present invention, in response to an information extension instruction corresponding to an information extension intention, the page key information can be used to retrieve candidate pages with a high similarity in information type with the target page, and the top k candidate pages in the retrieved candidate pages are used as the second associated pages. In this way, information extension for the target page can be achieved through the retrieval function. By using a large model to process the page information of the second associated page with a high degree of relevance to the target page, structured feedback information that can meet the user's information extension needs can be obtained, thereby reducing the amount of information processing of the large model and improving the convenience and timeliness of users in obtaining information.

[0098] According to an embodiment of the present invention, determining the second associated page based on the information extension intention and the page key information of the target page may include: responding to an information extension instruction corresponding to the information extension intention, searching based on the page key information to obtain multiple candidate associated pages; and determining at least one second associated page from the page clustering cluster based on the retrieval sorting attributes corresponding to the clustered candidate associated pages in the page clustering cluster.

[0099] According to an embodiment of the present invention, the candidate associated pages correspond to the search ranking attributes. The search ranking attributes of the candidate associated pages may represent the degree of relevance between the page content and the page key information contained in each of the candidate associated pages. Alternatively, the search ranking attributes may also represent the authority of the topic represented by the page key information.

[0100] According to an embodiment of the present invention, a page cluster is obtained by clustering multiple candidate associated pages. The page contents of each of the multiple candidate associated pages can be processed based on a clustering algorithm, and the multiple candidate associated pages can be divided into N page clusters (N is a positive integer greater than 1). Based on the search ranking attributes corresponding to each of the clustered candidate associated pages in the page cluster, the top k clustered candidate associated pages can be determined from each page cluster as the second associated page.

[0101] According to an embodiment of the present invention, the recall rate of associated pages that match the key information of a target page can be improved by retrieving multiple candidate associated pages, the recalled multiple candidate associated pages can be processed by a clustering algorithm, and the associated pages with higher similarity can be removed by determining the second associated pages from the page clustering clusters, and at the same time, at least one second associated page with a higher retrieval ranking can be determined from the multiple page clustering clusters, so as to improve the degree to which the second associated pages match the information acquisition intention of the object with the help of the retrieval function, and improve the representation accuracy of the structured feedback information.

[0102] Figure 4 A schematic diagram showing a principle for determining page information according to an embodiment of the present invention is shown.

[0103] like Figure 4 As shown, the target page may be a search result page 401, and the search result page 401 may be input into a filtering module 410. The filtering module 410 may include an advertisement filtering component and a negative information filtering component. The advertisement filtering component may filter the advertisement tabs in the search result page 401. The negative information filtering component may filter the page tabs with negative information marks or negative comment information in the search result page 401. The filtered tabs in the search result page 401 may be input into a page information acquisition module 420. The page information acquisition module 420 may include a page text extraction component, a visual recognition component, and an information cleaning component. The page text extraction component may extract the page text of the associated pages associated with the filtered tabs. The visual recognition component may perform visual text recognition on the associated pages associated with the extracted filtered tabs to obtain the page text. The information cleaning component may clean the repeated content, garbled information, and wrong recognition information in the page text. The cleaned text may be input into a user intention recognition module 430. The user intention recognition module 430 may include an object preference attribute recognition component, a search information rewriting component, and an intention recognition component. The object preference attribute recognition component may determine the object preference attribute of the object. The search information rewriting component can rewrite the search information carried in the information acquisition request to obtain the expanded associated search information. The intention recognition component can perform intention recognition based on at least one of the associated search information and the object preference attribute to obtain the information acquisition intention. The information authority evaluation component of the page information screening module 440 can evaluate the authority of the candidate associated pages retrieved by the associated search information to obtain the authority weight. The information similarity evaluation component can deduplicate pages with high similarity in the candidate associated pages. The information gain component can determine the associated pages from the deduplicated candidate associated pages based on the authority weight, and determine the page information from the associated pages. It should be understood that the candidate associated pages can also determine the page information through the filtering module 410 and the page information acquisition module 420.

[0104] According to an embodiment of the present invention, determining page information related to a target page may include: performing intent identification on an object based on operation attributes carried in an information acquisition request to obtain an information acquisition intent; and determining at least one page information from a target page area in the target page based on the operation area intent.

[0105] According to an embodiment of the present invention, the information acquisition intention may also include an operation area intention, the operation area intention corresponds to the target page area in the target page, and the operation attribute is determined based on the interactive operation of determining the information acquisition request. For example, the operation area intention may correspond to the title area of ​​the target page. The information acquisition intention may indicate the object's intention to acquire information from the target page area in the target page.

[0106] According to an embodiment of the present invention, the target page area may include any type of page area such as a video area, a comment area, a recommendation area, etc. The embodiment of the present invention does not limit the specific type of the target page area. By performing intent recognition based on the interactive operation of the object, and determining at least one page information from the target page area of ​​the target page, it is possible to more accurately identify that the object needs to obtain information for the page information in the target page area based on the interactive operation of the object, and generate an information acquisition request based on the operation attribute (such as the page area identifier) ​​of the target page area. The operation area intention is determined by the page area identifier carried by the information acquisition request, and the target page area corresponding to the operation area intention is determined. By using a large model to process the page information obtained from the target page area, the object can browse the current page content of concern more conveniently based on the structured attributes of the obtained structured feedback information.

[0107] It should be noted that the structured feedback information may be summary information obtained by semantically summarizing and generalizing the page information. However, it is not limited to this, and may also include the data processing results of using a large language model to perform data processing tasks on the page information. For example, the structured feedback information may also include the detection results of the large model performing an object detection task on the page image, the analysis results of performing a chart analysis task on the chart data, and so on.

[0108] According to an embodiment of the present invention, the operational attributes of an interactive operation may include the type, location, duration, etc. of the interactive operation, or may also include operational information input by the interactive operation, such as voice information input by a semantic input operation, text information input by a text input operation, etc.

[0109] According to an embodiment of the present invention, the intent of an object is identified based on the operational attributes of an interactive operation, which may include detecting attribute information such as the type, location, and duration of the operation, and comparing the obtained attribute detection results with the attribute conditions corresponding to the target page area. In response to the comparison results satisfying the attribute conditions corresponding to the target page area, the operation area intention corresponding to the target page area can be obtained.

[0110] According to an embodiment of the present invention, the operation attribute may include a positional relationship between an operation position of an interactive operation and a regional element in a target page area. For example, a positional relationship between an interactive operation position and a title element in a title area. For another example, a positional relationship between an interactive operation position and a video frame in a video playback area.

[0111] According to an embodiment of the present invention, the operation attribute may also include the operation area range of the interactive operation, for example, the area range of the sliding operation in the target page, the range of the pinching operation in the target area, and the like.

[0112] According to an embodiment of the present invention, the operation attribute may also include operation attribute information such as text and audio input of the interactive operation. The intention recognition of the object according to the operation attribute of the interactive operation may also include processing the operation attribute information such as text and audio input of the interactive operation based on a neural network algorithm to obtain the operation area intention corresponding to the target page area. The embodiment of the present invention does not limit the specific method of performing the intention recognition.

[0113] It should be noted that the information acquisition request may include one or more operation attributes, and intent identification may be performed based on any one or more operation attributes carried in the information acquisition request to obtain the operation area intention, and the embodiments of the present invention will not be described in detail herein.

[0114] Figure 5 An application scenario diagram of a large model-based information processing method according to another embodiment of the present invention is shown.

[0115] like Figure 5 As shown, the application scenario may include a touch screen display 500 of a terminal device and a cloud server C510, and the touch screen display 500 may include a target page 510. The target page 510 may include a video play frame 511 and a video comment area 512. The video play frame 511 may be playing a video of movie A, and the video comment area 512 may display comment information of multiple users on movie A. For example, the video comment area 512 may include comment information of account 1, account 2, account 3, and account 5.

[0116] like Figure 5As shown, the user object can perform a pinch operation O510 on the video comment area 512 of the target page 510. The touch display screen 500 detects that the operation position of the pinch operation O510 is located in the video comment area 512, and it can be obtained that the object's operation area intention is to process the page information in the video comment area 512. An information acquisition request can be generated according to the area identifier of the video comment area 512. The terminal device can send the information acquisition request to the cloud server C510. The cloud server C510 is deployed with a large model. The large model can obtain multiple comment information from the video comment area 512 as page information corresponding to the information acquisition request based on the operation area intention. Using a large language model to process multiple comment information, the obtained multiple structured feedback information can be the first comment analysis feedback information J511, the second comment analysis feedback information J512 and the third comment analysis feedback information J513. The first comment analysis feedback information J511, the second comment analysis feedback information J512, and the third comment analysis feedback information J513 can be sent to the terminal device, and the terminal device can display the first comment analysis feedback information J511, the second comment analysis feedback information J512, and the third comment analysis feedback information J513 on the touch display screen 500. The target element B510 "A movie screenshot" generated by the terminal device taking a screenshot of the video playback frame 511 can also be displayed on the touch display screen 500. The first comment analysis feedback information J511 can be "Comment Analysis Overview", for example, the overall comment content of multiple accounts on movie A can be obtained by analyzing the respective comment content of multiple accounts. The second comment analysis feedback information J512 may include a title "Positive and Negative Comment Data Analysis" and a comparative bar chart, for example, the number of positive comments and the number of negative comments can be represented based on the comparative bar chart, so that the object can more intuitively obtain the proportional relationship between the positive and negative comments for movie A, and improve the efficiency of the relevant object in obtaining the playback effect of movie A. The third comment analysis feedback information J513 may include a title "Comment User Preference Analysis" and a movie clip heat map. The movie clip heat map can display the playback popularity of movie A at each moment based on the timeline, so as to facilitate the object to quickly obtain the most popular playback clips in movie A, and improve the efficiency of the playback effect analysis of movie A. The first comment analysis feedback information J511, the second comment analysis feedback information J512 and the third comment analysis feedback information J513 can be sent to the terminal device, and the terminal device can display the first comment analysis feedback information J511, the second comment analysis feedback information J512 and the third comment analysis feedback information J513 on the touch display screen 500.

[0117] According to an embodiment of the present invention, by identifying the operation area intention of the object and determining the page information that the large model needs to process based on the target page area corresponding to the operation area intention, the matching degree between the structured feedback information and the information needs of the object can be further improved, and the display probability of redundant structured feedback information can be reduced, thereby improving the information acquisition efficiency of the object and improving the user experience.

[0118] According to an embodiment of the present invention, determining structured feedback information corresponding to an information acquisition request of an object based on page information may include: determining a requirement prompt template corresponding to the feedback information requirement type of the object; and using a large model to process the page information and the requirement prompt template to obtain structured feedback information.

[0119] According to an embodiment of the present invention, the feedback information requirement type may be determined by performing intention recognition on the object based on the information acquisition request, and the feedback information requirement type may characterize the object's requirement intention for the information type of structured feedback information.

[0120] For example, the intent recognition can be performed based on the search text (or query) carried by the information acquisition request to obtain the feedback information requirement type associated with the search text. However, it is not limited to this. The intent recognition can also be performed based on the page information such as the page content, page title, page tag, etc. of the target page related to the information acquisition request to obtain the feedback information requirement type associated with the search text. The embodiments of the present invention do not limit the specific method of determining the feedback information requirement type.

[0121] According to an embodiment of the present invention, the feedback information demand type can characterize the object's demand intention for the information type of structured feedback information. For example, the type of information that the object needs to obtain may include information query type, transaction demand type, local information acquisition type, text creation type, image creation type, etc. Determine the demand prompt template corresponding to the feedback information demand type, and use the big model to process the page information and the demand prompt template. It can be achieved by using the demand prompt template to control the big model to perform data processing tasks corresponding to the feedback information demand type, so that based on the big model's relatively powerful semantic understanding ability and data processing task execution ability, the page information can be accurately processed according to the object's feedback information demand type, and one or more structured feedback information matching the object's information demand type can be obtained, thereby achieving more accurate and convenient information acquisition, and improving information acquisition efficiency and accuracy when the number of interactive operations is small.

[0122] It should be noted that using the big model to process page information and demand prompt templates may include using the big model to process page information and demand prompt templates in the target page. Alternatively, it may also include using the big model to process page information and demand prompt templates in associated pages. Alternatively, it may also include using the big model to process page information in the target page, page information in associated pages, and demand prompt templates. The page information specifically processed by the big model may be determined based on the extended information intent of the object.

[0123] In an example, the feedback information requirement type may be a device operation instruction type, and the requirement prompt template corresponding to the device operation instruction type may be expressed as: "Please analyze the input information enclosed by the following {}, and generate the title of the device operation steps and the specific content of the device operation steps."

[0124] In an example, the feedback information requirement type may be a data analysis type, and the requirement prompt template corresponding to the data analysis type may be expressed as: "Please perform data analysis on the data enclosed by the following {} involved in the page information and generate a bar chart."

[0125] According to an embodiment of the present invention, the feedback information requirement type can be determined based on the following operations: determining the object preference attribute of the object based on the information acquisition request; and determining the feedback information requirement type based on the similarity between the object preference attribute and the page key information of the target page.

[0126] According to an embodiment of the present invention, the object preference attribute may include any attribute that characterizes the object's information acquisition preference, such as an online shopping preference attribute, an information navigation attribute, and a travel preference attribute. The object preference attribute may be determined based on the preference information input by the object. For example, the object may select an attribute tag that meets its information acquisition preference before browsing the target page, so that the object preference information may be determined based on the attribute tag that the object has selected in response to the information acquisition request. However, the object preference attribute may also be determined based on the object's historical interactive operation information (such as historical browsing information, historical text input information, etc.).

[0127] According to an embodiment of the present invention, the page key information of the target page may include the page title, page topic, most frequently appearing text, etc. of the target page, and may also include preset keywords, images and other information. The page key information may be determined based on actual needs.

[0128] According to an embodiment of the present invention, the similarity between the object preference attribute and the page key information of the target page may include the semantic similarity between the object preference attribute and the page key information. The page key information may be associated with the feedback information demand type. In the case where the page key information includes multiple pieces, the page key information having a similarity with the object preference attribute higher than a preset similarity threshold may be determined as the target page key information. According to the feedback information demand type associated with the target page key information, the feedback information demand type of the object is used. By determining the feedback information demand type based on the similarity between the object preference attribute and the page key information of the target page, the object preference attribute may be more accurately associated with the feedback information demand type, so that the object preference of the object may be accurately met according to the demand prompt template corresponding to the feedback information demand type, thereby improving the matching degree between the structural hierarchical attributes, information content, information type, etc. of the structured feedback demand information and the information demand intention of the object, thereby improving the information acquisition efficiency of the object.

[0129] It should be noted that the information acquisition involved in the method provided in the embodiment of the present invention, including but not limited to object preference attributes, interactive operation input information, etc., is obtained under the condition of obtaining the authorization of the object, and the purpose of obtaining the data is to provide the object with structured feedback information that better meets the user's needs. At the same time, necessary data security measures are adopted in the data processing process, including but not limited to encryption, desensitization and other data security measures to avoid data leakage.

[0130] In one embodiment, multiple associated search information matching the search information can be determined based on the search information carried in the information acquisition request. This can expand the semantic scope and semantic accuracy of the search information representation. Based on multiple associated search information as page key information, the object is identified for intent, and at least one feedback information requirement type is obtained. This can obtain feedback information that more accurately represents the object requirements based on the associated search information.

[0131] According to an embodiment of the present invention, the demand prompt template may include a sample (or FewShot) of structured feedback information. The sample may be associated with the feedback information demand type, so as to determine the associated sample based on the feedback information demand type, and then construct a demand prompt template corresponding to the feedback information demand type based on the sample associated with the feedback information demand type. The large model may generate structured feedback information that matches the information demand intention of the object based on the sample included in the demand prompt template, so as to facilitate the object to quickly browse the information content.

[0132] It should be noted that there may be multiple samples associated with the feedback information requirement type, and the sample used to construct the requirement prompt template may be determined based on the scores of each sample corresponding to the feedback requirement type. The score corresponding to the feedback requirement type may be determined based on the scoring operation of the object for the structured feedback information, or may also be determined based on other types of interactive operations of the object for the structured feedback information, so as to form feedback iterative optimization of the user object.

[0133] According to an embodiment of the present invention, using a large model to process page information and a demand prompt template includes: using a target large model determined according to a model calling strategy to process page information and a demand prompt template.

[0134] According to an embodiment of the present invention, the model call strategy is determined according to the feedback delay attribute of the object, and the feedback delay attribute can be determined according to the information volume of the page information. The information volume of the page information can be determined based on the storage space occupied by the page information, or can also be determined based on the format of the information content contained in the page information. For example, in the case where the page information contains information content in the format of MP4, AVI, etc., it can be determined that the page information contains video information, and therefore it can be determined that the information volume of the page information is greater than a preset information volume threshold.

[0135] According to an embodiment of the present invention, the feedback delay attribute can represent the delay time of determining structured feedback information based on page information. When the amount of information in the page information is greater than a preset information amount threshold, the feedback delay attribute can represent a longer delay time. To avoid the object waiting time being too long, a target large model adapted to the amount of information in the page information can be selected based on the call strategy. For example, a target large model with a smaller amount of model parameters can be selected to reduce the delay time of generating structured feedback information, thereby improving the efficiency of users browsing to obtain structured feedback information.

[0136] According to an embodiment of the present invention, a model calling strategy can be determined based on the feedback information requirement type and feedback delay requirement of an object, and the feedback delay requirement can characterize the expected delay time for an object to wait for structured feedback information to be generated. The feedback delay requirement can be obtained by performing intent recognition on an object. For example, intent recognition can be performed based on an object's preference attributes, and the obtained operation requirement intention can include a feedback delay requirement. The model calling strategy can be determined based on the object's feedback information requirement type and feedback delay requirement. The target large model can be called based on the object's feedback information requirement type and feedback delay requirement, thereby enabling the target large model to generate structured feedback information more accurately and quickly under the condition that the object's feedback information requirement type and feedback delay requirement are met, so as to meet the user's actual information acquisition needs.

[0137] According to an embodiment of the present invention, the model calling strategy may indicate the target large model to be called, and may also indicate the calling of a functional small model with a smaller parameter amount to combine with the target large model, so as to complete the generation of structured feedback information. The functional small model may include a deep learning model with a model parameter amount within a preset parameter amount range. For example, it may be a deep learning model with a model parameter amount of one hundred thousand. The functional small model may implement a deep learning model with functions such as target object detection, image text recognition, and audio text recognition. A page information processing method combining a functional small model with a target large model may be selected based on the model calling strategy to generate structured feedback information. Alternatively, it may be selected not to use a functional small model, and only to process page information based on an end-to-end target large model. Alternatively, it may be selected based on the model calling strategy to select a target-specific large model that matches the feedback information requirement type, and then the page information may be processed based on a requirement prompt template corresponding to the feedback information requirement type and a target-specific large model, so as to improve the matching degree between the structured feedback information and the information requirement intention of the object.

[0138] It should be understood that the target-specific large model can be obtained by fine-tuning the general large model based on samples corresponding to the feedback information requirement type.

[0139] According to an embodiment of the present invention, the page information may include video information. For example, the target page may include video information. For another example, the associated page may also include video information.

[0140] According to an embodiment of the present invention, using the target large model determined according to the model calling strategy, processing page information and demand prompt templates may include: using the target large visual model to process video information and video recognition prompt templates to obtain video text corresponding to the video information; and using the target large language model to process video text and feedback information demand prompt templates.

[0141] According to an embodiment of the present invention, the target visual model may be a target visual model for text recognition of video information. The demand prompt template may include a video recognition prompt template, which may be used to control the target visual model to recognize video content displayed in the video information.

[0142] According to an embodiment of the present invention, video content may include text content such as subtitles and titles in the video (such as stickers). However, it is not limited to this. The video content may also include objects such as commodities and buildings in the video. Video text may include text representing video content such as text content and objects in the video. For example, the video text may include the name text of the actor's watch in the video information, the name text of the building in the video background, and so on. The video recognition prompt template can be used to control the target visual large model to perform text recognition on areas where text appears repeatedly in the video information, and can also be used to control the target visual large model to recognize objects such as buildings, characters, commodities, etc. that appear in the video information, so that the target visual large model can output the video text represented in the video information.

[0143] According to an embodiment of the present invention, the feedback information requirement prompt template can be used to control the target large model to process the video text according to the feedback information requirement type. The feedback information requirement prompt template can be a prompt template corresponding to the feedback information requirement type. By processing the page information containing video information by combining the target visual large model with the target large language model, the detection accuracy of the page information and the accuracy of structured semantic aggregation can be improved, and the matching degree between the structured feedback information and the object's requirements can be improved, thereby improving the user experience.

[0144] According to an embodiment of the present invention, the video recognition hint template is determined based on at least one of the first hint marker sequence and the second hint marker sequence.

[0145] According to an embodiment of the present invention, the first prompt mark sequence is used to control the target visual macromodel to identify the video subtitle text based on the display area of ​​the text in the video information. For example, the target visual macromodel can be controlled based on the first prompt mark sequence to identify the video subtitle text in the display area such as the top area, the bottom area, the area where the text appears repeatedly in the video frame in the video information. It should be noted that the video subtitle text may include text that appears in any display area in the video, such as the title, sticker text, bullet text, etc. that appear in the video.

[0146] In one example, the first prompt tag sequence may be: "Identify the barrage texts appearing in the video, and mark different barrage texts with "". The video text may include multiple barrage texts, and the target large language model may analyze the comments on the video information by processing the multiple barrage texts. The obtained structured feedback information may include sentiment analysis information of the comments.

[0147] The second prompt mark sequence is used to control the target visual model to detect the target object on the video information.

[0148] In one example, the second prompt tag sequence may be "Please identify the vehicle in the video and generate text information of the vehicle according to color and brand". The video text may include the color text and brand text of the vehicle. The target large language model may perform vehicle statistics and analysis on the video information by processing the color text and brand text. The obtained structured feedback information may include vehicle color and brand statistics.

[0149] According to an embodiment of the present invention, the page information includes video information, and the structured feedback information includes segment description information associated with the video information segment in the video information. For example, the segment description information may be plot summary information for the video segment, bullet comment summary information, and the like.

[0150] The information processing method based on the large model may further include: in response to a viewing operation on the segment description information, displaying a video information segment corresponding to the segment description information.

[0151] Figure 6 An application scenario diagram of a large model-based information processing method according to yet another embodiment of the present invention is shown.

[0152] like Figure 6 As shown, the application scenario may include a touch screen display 600 of a terminal device and a cloud server C610, and the touch screen display 600 may include a target page 610. The target page 610 may be a video playback page with the theme "The only drawback is that it is only available in City A". The target page 610 may include a travel food video in which a video blogger introduces the current travel situation in City A.

[0153] like Figure 6As shown, the user object can perform an interactive operation O610 for the target page 610, and based on the interactive operation O610, an information acquisition request can be sent to the cloud server C610. The cloud server C610 can determine the page information from the target page 610 according to the information acquisition request, and the page information can include the theme "The only disadvantage is that it is only available in city A", and can also include the video information of the travel food video. The page information can also include an associated page corresponding to the target page 610, such as a navigation map page corresponding to the A Shengjian Shop involved in the video, etc. Using a large model to process the page information related to the target page 610, a plurality of structured feedback information can be obtained. The plurality of structured feedback information can include the first feedback information J621, the second feedback information J622, the third feedback information J623, etc. The first feedback information J621, the second feedback information J622, the third feedback information J623, etc. can each be associated with a video information segment in the travel food video. The first feedback information J621, the second feedback information J622, the third feedback information J623, ... and the like can be sent from the cloud server C610 to the terminal device, and the terminal device displays the first feedback information J621, the second feedback information J622, the third feedback information J623, ... and the like.

[0154] like Figure 6 As shown, the user object can perform the viewing operation O620 by clicking on the position of the third feedback information J623 in the touch display screen 600. The processor of the intelligent terminal device can play the Shengjianpu segment 631 in the travel food video in the jump page 630 of the touch display screen 600 in response to the viewing operation O620, so that the user object can quickly jump to the video segment of interest. At the same time, it can also jump to the associated page 632 corresponding to the third feedback information J623. The associated page 632 can be a navigation page of Shengjianpu in City A. The navigation page may include a navigation map of Shengjianpu. The user can browse the Shengjianpu segment 631 being played by browsing the touch display screen 600, and at the same time come to Shengjianpu for consumption according to the navigation map of Shengjianpu, so as to improve the efficiency of information acquisition and avoid the time waste caused by multiple interactive operations to query information.

[0155] Figure 7 An application scenario diagram of a large model-based information processing method according to yet another embodiment of the present invention is shown.

[0156] like Figure 7As shown, the application scenario may include a touch screen display 700 of a terminal device and a cloud server C710, and the touch screen display 700 may include a target page 710. The target page 710 may be a news page with the theme of "Tourism in City A is booming". The target page 710 may include news text and news images that introduce the current tourism boom in City A, and the news images may include a picture of Garden A and a picture of Garden B corresponding to Garden A and Garden B, respectively.

[0157] like Figure 7 As shown, the user object can perform an interactive operation O710 on the target page 710, and based on the interactive operation O710, an information acquisition request can be sent to the cloud server C710. The cloud server C710 can determine the first page information from the target page 710 according to the information acquisition request, and can also determine the associated page with the target page 710 based on the interactive operation, and determine the second page information from the associated page. By using the large model to process the first page information and the second page information, multiple structured feedback information can be obtained. Multiple structured feedback information can include first feedback information J721, second feedback information J722, third feedback information J723... etc. The first feedback information J721, the second feedback information J722, the third feedback information J723... etc. can each be associated with the associated page that determines the structured feedback information.

[0158] like Figure 7 As shown, the first feedback information J721, the second feedback information J722, the third feedback information J723, etc. can be sent to the terminal device, and the terminal device can display the first feedback information J721, the second feedback information J722, the third feedback information J723, etc. on the touch display screen 700. The first feedback information J721 can be the title "A city travel route recommendation" of multiple structured feedback information. The second feedback information J722 can be a one-day tour scenic spot route recommendation information, for example, it can include the recommended text information of "museum-A garden-B garden-A restaurant-C garden", and the scenic spot navigation map information corresponding to the recommended text information. The scenic spot navigation map information can display the navigation route of "museum-A garden-B garden-A restaurant-C garden" so that users can view it in time. The third feedback information J723 can be the scenic spot introduction summary information of the scenic spot "museum". The third feedback information J723 can include the introduction of the "museum", tickets, reservation links, address links and opening hours and other information related to tourism. In addition, the touch display screen 700 may also include structured feedback information for displaying summary information of scenic spot introductions of other scenic spots, and a target element B720 . The target element B720 may be a page screenshot of the target page 710 .

[0159] like Figure 7As shown, the user object can perform the viewing operation O720 by clicking on the position on the touch display screen 700 corresponding to the "reservation link" structured content in the third feedback information J723. The processor of the smart terminal device can jump to the museum reservation page 730 in response to the position of the viewing operation O720. This can facilitate the user object to quickly plan the travel and avoid the user planning the travel route by searching for travel guides and related pages, thereby improving the user's information acquisition efficiency and interaction efficiency.

[0160] Figure 8 A schematic diagram of a large model-based information processing method according to an embodiment of the present invention is shown.

[0161] like Figure 8 As shown, the information acquisition request 801 can be input into the intention recognition module 810. The intention recognition module 810 can include an operation attribute detection component, a page information classification component and an intention recognition component. The operation attribute detection component can detect the operation attribute of the interactive operation corresponding to the information acquisition request 801, and the page information classification component can classify the page information in the target page and the associated page, and can also extract the page key information in the target page and the associated page. The intention recognition component can perform intention recognition based on at least one of the operation attribute and the page key information to obtain the information acquisition intention. The information acquisition intention may include the feedback information demand type. The prompt information priority module 820 can determine the associated page information matching the feedback information demand type based on the retrieval enhancement module. The prompt template priority component can determine the sample matching the feedback information demand type from the sample library based on the feedback information demand type, and construct the demand prompt template based on the determined sample and the feedback information demand type. The call strategy module 830 can determine the model call strategy based on the feedback information demand type. Based on the model calling strategy, the large model library and the functional small model library can determine the models that need to be assembled. The model assembly component can assemble the target large model and the target functional small model based on the model calling strategy, so as to process the page information and the demand prompt template based on the assembled target large model and the target functional small model to obtain structured feedback information. The structured feedback information can be sent to the terminal device. The user can evaluate based on the displayed structured feedback information, for example, can like or comment. The user interaction module 840 can call the sample splicing component to splice the user's evaluation information. The user interaction driving component can analyze the spliced ​​evaluation information and optimize the sample library in the calling strategy module 830 according to the analysis results. Alternatively, the models in the large model library and the functional model library in the calling strategy module 830 can be fine-tuned to optimize the model. The user interaction driving component can also optimize the configuration parameters of the sample library and the prompt template optimization component in the prompt information selection module 820 according to the analysis results, thereby realizing continuous optimization functions.

[0162] Fig. 9 A schematic diagram of an operational flow for executing a large model-based information processing method according to an embodiment of the present invention is shown.

[0163] like Fig. 9 As shown, the large model-based information processing method can be implemented according to an online path operation 910 and an offline path operation 920. The online path operation 910 may include operations S1.1 to S1.6.

[0164] In operation S1.1, an online summary service may be requested based on a data stream including an information acquisition request.

[0165] Operation S1.2, accessing the structured feedback information in the cache area, directly returning the structured feedback information if the structured feedback information in the cache area meets the requirements of the information acquisition request, and then executing operation S1.5, returning the structured feedback information to the terminal device through the online channel module.

[0166] If the requirements are not met, operation S1.3 is executed to request the large model calling module to call the target large model from the large model library to process the page information related to the information acquisition request and obtain structured feedback information. Operations S1.4 and S1.5 are executed to return the structured feedback information to the terminal device through the online channel module to realize data flow return. In operation S1.6, the data in the cache area can be updated based on the structured feedback information generated online by the large model.

[0167] Offline path operation 920 may include operations S2.1 to S2.5. Operation S2.1 may be triggered in response to a request for storage instruction. Operation S2.1 may control the hotspot mining module to obtain structured feedback information from the online path module. The hotspot mining component may obtain structured feedback information and cache it in the hotspot cache area.

[0168] In operation S2.2, the offline access module can read the structured feedback information from the cache area and determine whether the structured feedback information meets the timeliness condition or the quality condition. If the condition is met, operation S2.3 is executed to request the large model to fuse the structured feedback information using the large model calling module. In operation S2.4, the large model calling module returns the structured feedback information obtained after the large model fusion to the offline access module, and the offline access module executes operation S2.5 to update the structured feedback information in the cache area, so that the user can obtain high-quality structured feedback information with high quality after fusion from the cache area in real time.

[0169] In one embodiment, the information processing method based on the big model can also generate evaluation information through the interactive behaviors of the object on the structured feedback information, such as click rate, dislike rate, sharing, reading time, copying, etc. The retrieval function and the model parameters of the big model are optimized through the evaluation information.

[0170] Fig.10 A framework diagram of an information processing system for executing a large model-based information processing method according to an embodiment of the present invention is shown.

[0171] like Fig.10 As shown, the information processing system 1000 may include an operation response component, a prompt template selection component, a model call strategy component, an online processing component, a cache component, an interaction optimization component, a user interaction component and a pre-storage component.

[0172] The operation response component can be used to detect the operation attributes and object preference attributes of the information acquisition request in response to the information acquisition request. The prompt template selection component can perform intent recognition based on the operation attributes and object preference attributes to obtain information acquisition intentions such as feedback information requirement type and operation area intention. The model call strategy component can call and assemble at least one of the function angle model, the large language model, and the multimodal large model based on the feedback information requirement type. Obtain a target large model for processing page information. The online processing component can be used to obtain page information related to the target page, and call the target large model to process the page information to generate structured feedback information. The cache component can cache based on the storage duration configuration information of the structured feedback information. The pre-storage component can call the structured feedback information corresponding to the information acquisition request of other objects from the cache based on the cache component to improve the information acquisition efficiency of other objects. The user interaction component can collect the interactive operation behavior of the object for the structured feedback information, and send the interactive operation behavior to the interactive optimization component. The interactive optimization component can optimize the model parameters of the large model, the execution configuration parameters of the information processing method such as the model call strategy, etc. through the interactive operation behavior corresponding to each structured feedback information.

[0173] Fig.11 A block diagram of an information processing device based on a large model according to an embodiment of the present invention is shown.

[0174] like Fig.11 As shown, the information processing device 1100 based on the large model includes: a page information determination module 1110 and a structured feedback information determination module 1120.

[0175] The page information determination module 1110 is used to determine page information related to the target page in response to the object's information acquisition request for the target page.

[0176] The structured feedback information determination module 1120 is used to determine the structured feedback information corresponding to the information acquisition request of the object based on the page information, wherein the structured feedback information is obtained by integrating the information content with the same text structure level in the page information using a large model, and the structured feedback information has a text structure level.

[0177] According to an embodiment of the present invention, the structured feedback information determination module includes: a requirement prompt template determination submodule and a structured feedback information acquisition submodule.

[0178] The demand prompt template determination submodule is used to determine the demand prompt template corresponding to the feedback information demand type of the object, wherein the feedback information demand type is determined by identifying the intention of the object based on the information acquisition request, and the feedback information demand type represents the object's demand intention for the information type of structured feedback information.

[0179] The structured feedback information acquisition submodule is used to use the large model to process page information and demand prompt templates to obtain structured feedback information.

[0180] According to an embodiment of the present invention, the feedback information requirement type is determined based on the following operations: determining the object preference attribute of the object based on the information acquisition request; and determining the feedback information requirement type based on the similarity between the object preference attribute and the page key information of the target page.

[0181] According to an embodiment of the present invention, the structured feedback information obtaining submodule includes a first processing unit.

[0182] The first processing unit is used to process page information and demand prompt templates using a target large model determined according to a model calling strategy, wherein the model calling strategy is determined according to a feedback delay attribute of an object, and the feedback delay attribute is determined according to the amount of information in the page information.

[0183] According to an embodiment of the present invention, the page information includes video information.

[0184] According to an embodiment of the present invention, the first processing unit includes: a video text obtaining subunit and a first processing unit.

[0185] The video text acquisition subunit is used to use the target visual large model to process the video information and the video recognition prompt template to obtain the video text corresponding to the video information, wherein the demand prompt template includes a video recognition prompt template, and the video recognition prompt template is used to control the target visual large model to recognize the video content displayed in the video information.

[0186] The first processing unit is used to process the video text and the feedback information requirement prompt template using the target large language model, wherein the feedback information requirement prompt template is used to control the target large model to process the video text according to the feedback information requirement type.

[0187] According to an embodiment of the present invention, a video recognition prompt template is determined based on at least one of the following prompt mark sequences: a first prompt mark sequence, used to control the target visual large model to recognize the video subtitle text based on the display area of ​​the text in the video information; a second prompt mark sequence, used to control the target visual large model to perform target object detection on the video information.

[0188] According to an embodiment of the present invention, the page information determination module includes: a first obtaining submodule and an associated page determination submodule.

[0189] The first acquisition submodule is used to identify the object's intention based on the information acquisition request to obtain the object's information acquisition intention, wherein the information acquisition intention includes the information extension intention, and the information extension intention indicates that the object needs to obtain other information content in addition to the information content recorded in the target page.

[0190] The associated page determination submodule is used to determine the associated page corresponding to the target page based on the information extension intention, wherein the page information is determined based on the associated page.

[0191] According to an embodiment of the present invention, the associated page determination submodule includes a first associated page determination unit.

[0192] The first associated page determining unit is used to determine the first associated page based on the information extension intention and the historical information acquisition request of the object, and the associated pages include the first associated page.

[0193] According to an embodiment of the present invention, the associated page determination submodule includes a second associated page determination unit.

[0194] The second associated page determining unit is used to determine the second associated page based on the information extension intention and the page key information of the target page, and the associated pages include the second associated page.

[0195] According to an embodiment of the present invention, the second associated page determining unit includes: a selected associated page obtaining subunit and a second associated page determining subunit.

[0196] The candidate associated page obtaining subunit is used to respond to the information extension instruction corresponding to the information extension intention, perform a search based on the page key information, and obtain multiple candidate associated pages, and the candidate associated pages correspond to the search sorting attributes.

[0197] The second associated page determination subunit is used to determine at least one second associated page from the page clustering cluster based on the search ranking attribute corresponding to the clustered candidate associated pages in the page clustering cluster, wherein the page clustering cluster is obtained by clustering multiple candidate associated pages.

[0198] According to an embodiment of the present invention, the first obtaining submodule includes: a type similarity result obtaining unit and an information extension intention determining unit.

[0199] The type similarity result obtaining unit is used to perform similarity analysis based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operation of the object to obtain the type similarity result.

[0200] The information extension intention determination unit is used to determine the information extension intention when the type similarity result meets the preset information type similarity condition.

[0201] According to an embodiment of the present invention, the type similarity result obtaining unit includes: a first similarity determining subunit, a second similarity determining subunit and a type similarity result determining subunit.

[0202] The first similarity determination subunit is used to determine a first similarity between the information type and a first historical information type corresponding to a first historical interaction operation, wherein the first historical interaction operation is a historical interaction operation in a first historical period.

[0203] The second similarity determination subunit is used to determine the second similarity between the information type and the second historical information type corresponding to the second historical interaction operation, wherein the second historical interaction operation is a historical interaction operation in a second historical period, and the timing attribute of the first historical period is earlier than the timing attribute of the second historical period.

[0204] The type similarity result determining subunit is used to determine a type similarity result that meets the information type similarity condition in response to the first similarity being lower than a preset similarity threshold and the second similarity being higher than or equal to the preset similarity threshold.

[0205] According to an embodiment of the present invention, the page information determination module includes: an information acquisition intention determination submodule and a page information determination submodule.

[0206] The information acquisition intention determination submodule is used to identify the intention of the object according to the operation attributes carried by the information acquisition request to obtain the information acquisition intention, wherein the information acquisition intention includes the operation area intention, the operation area intention corresponds to the target page area in the target page, and the operation attributes are determined based on the interactive operation of the information acquisition request.

[0207] The page information determination submodule is used to determine at least one page information from a target page area in a target page according to the operation area intention.

[0208] According to an embodiment of the present invention, the operation attribute includes at least one of the following: a positional relationship between an operation position of the interactive operation and an area element in the target page area; and an operation area range of the interactive operation.

[0209] According to an embodiment of the present invention, the page information includes video information, and the structured feedback information includes segment description information associated with a video information segment in the video information.

[0210] According to an embodiment of the present invention, the present invention also provides an electronic device, a readable storage medium and a computer program product.

[0211] According to an embodiment of the present invention, an electronic device includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can execute the method described above.

[0212] According to an embodiment of the present invention, a non-transitory computer-readable storage medium storing computer instructions is provided, wherein the computer instructions are used to enable a computer to execute the method described above.

[0213] According to an embodiment of the present invention, a computer program product includes a computer program, and the computer program implements the above method when executed by a processor.

[0214] Fig.12 A block diagram of an electronic device suitable for implementing an information processing method based on a large model according to an embodiment of the present invention is shown. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workbenches, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices, and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0215] like Fig.12As shown, the device 1200 includes a computing unit 1201, which can perform various appropriate actions and processes according to a computer program stored in a read-only memory (ROM) 1202 or a computer program loaded from a storage unit 1208 into a random access memory (RAM) 1203. In the RAM 1203, various programs and data required for the operation of the device 1200 can also be stored. The computing unit 1201, the ROM 1202, and the RAM 1203 are connected to each other via a bus 1204. An input / output (I / O) interface 1205 is also connected to the bus 1204.

[0216] A number of components in the device 1200 are connected to the I / O interface 1205, including: an input unit 1206, such as a keyboard, a mouse, etc.; an output unit 1207, such as various types of displays, speakers, etc.; a storage unit 1208, such as a disk, an optical disk, etc.; and a communication unit 1209, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 1209 allows the device 1200 to exchange information / data with other devices through a computer network such as the Internet and / or various telecommunication networks.

[0217] The computing unit 1201 may be a variety of general and / or special processing components with processing and computing capabilities. Some examples of the computing unit 1201 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various dedicated artificial intelligence (AI) computing chips, various computing units running machine learning model algorithms, digital signal processors (DSPs), and any appropriate processors, controllers, microcontrollers, etc. The computing unit 1201 performs the various methods and processes described above, such as a large model-based information processing method. For example, in some embodiments, the large model-based information processing method may be implemented as a computer software program, which is tangibly included in a machine-readable medium, such as a storage unit 1208. In some embodiments, part or all of the computer program may be loaded and / or installed on the device 1200 via the ROM 1202 and / or the communication unit 1209. When the computer program is loaded into the RAM 1203 and executed by the computing unit 1201, one or more steps of the large model-based information processing method described above may be performed. Alternatively, in other embodiments, the computing unit 1201 may be configured to execute the large model-based information processing method in any other appropriate manner (eg, by means of firmware).

[0218] Various implementations of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), load programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various implementations can include: being implemented in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0219] The program code for implementing the method of the present invention can be written in any combination of one or more programming languages. These program codes can be provided to a processor or controller of a general-purpose computer, a special-purpose computer or other programmable data processing device, so that the program code, when executed by the processor or controller, enables the functions / operations specified in the flow chart and / or block diagram to be implemented. The program code can be executed entirely on the machine, partially on the machine, partially on the machine as a stand-alone software package and partially on a remote machine, or entirely on a remote machine or server.

[0220] In the context of the present invention, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or equipment, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.

[0221] To provide interaction with a user, the systems and techniques described herein can be implemented on a computer having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the computer. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0222] The systems and techniques described herein may be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or a web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system may be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), and the Internet.

[0223] A computer system may include a client and a server. The client and the server are generally remote from each other and typically interact through a communication network. The relationship of client and server arises through computer programs running on respective computers and having a client-server relationship to each other. The server may be a cloud server, a server in a distributed system, or a server combined with a blockchain.

[0224] It should be understood that the various forms of processes shown above can be used to reorder, add or delete steps. For example, the steps described in the present invention can be executed in parallel, sequentially or in different orders, as long as the desired results of the technical solution disclosed in the present invention can be achieved, and this document does not limit this.

[0225] The above specific implementations do not constitute a limitation on the protection scope of the present invention. It should be understood by those skilled in the art that various modifications, combinations, sub-combinations and substitutions can be made according to design requirements and other factors. Any modification, equivalent substitution and improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A large model-based information processing method, characterized in that: include: In response to a request from an object for obtaining information about a target page, determining page information related to the target page; Determine a demand prompt template corresponding to the feedback information demand type of the object; The page information and the demand prompt template are processed by the big model to obtain structured feedback information. The demand prompt template is used to control the big model to execute a data processing task corresponding to the feedback information demand type. The structured feedback information is obtained by fusing the information content with the same text structure level in the page information by using the big model. The structured feedback information has the text structure level. The feedback information demand type is determined based on the intention recognition of the object by the information acquisition request. The feedback information demand type represents the demand intention of the object for the information type of the structured feedback information. The demand intention of the information type of the structured feedback information represents the type of information that the object needs to obtain.

2. The method according to claim 1, characterized in that: The feedback information requirement type is determined based on the following operations: determining an object preference attribute of the object based on the information acquisition request; and The feedback information requirement type is determined based on the similarity between the object preference attribute and the page key information of the target page.

3. The method according to claim 1, characterized in that Using the large model to process the page information and the demand prompt template includes: The page information and the demand prompt template are processed using a target large model determined according to a model calling strategy, wherein the model calling strategy is determined according to a feedback delay attribute of the object, and the feedback delay attribute is determined according to the information volume of the page information.

4. The method according to claim 3, characterized in that The page information includes video information, Wherein, using the target large model determined according to the model calling strategy, processing the page information and the demand prompt template includes: Processing the video information and the video recognition prompt template using the target visual big model to obtain a video text corresponding to the video information, wherein the demand prompt template includes the video recognition prompt template, and the video recognition prompt template is used to control the target visual big model to recognize the video content displayed in the video information; and The video text and the feedback information requirement prompt template are processed using a target large language model, wherein the feedback information requirement prompt template is used to control the target large model to process the video text according to the feedback information requirement type.

5. The method according to claim 4, characterized in that The video recognition prompt template is determined based on at least one of the following prompt tag sequences: A first prompt mark sequence is used to control the target visual model to identify the video subtitle text based on the display area of ​​the text in the video information; The second prompt mark sequence is used to control the target visual model to perform target object detection on the video information.

6. The method according to claim 1, characterized in that Determining page information related to the target page includes: Performing intent recognition on the object based on the information acquisition request to obtain the information acquisition intent of the object, wherein the information acquisition intent includes an information extension intent, and the information extension intent indicates that the object needs to obtain other information content in addition to the information content recorded on the target page; and Based on the information extension intention, an associated page corresponding to the target page is determined, wherein the page information is determined according to the associated page.

7. The method according to claim 6, characterized in that Based on the information extension intention, determining the associated page corresponding to the target page includes: Based on the information extension intention and the historical information acquisition request of the object, a first associated page is determined, and the associated pages include the first associated page.

8. The method according to claim 6 or 7, characterized in that: Based on the information extension intention, determining the associated page corresponding to the target page includes: Based on the information extension intention and the page key information of the target page, a second associated page is determined, and the associated pages include the second associated page.

9. The method according to claim 8, characterized in that Based on the information extension intention and the page key information of the target page, determining the second associated page includes: In response to the information extension instruction corresponding to the information extension intention, searching based on the page key information to obtain a plurality of candidate associated pages, wherein the candidate associated pages correspond to the search ranking attributes; and At least one second associated page is determined from the page cluster based on a search ranking attribute corresponding to a cluster candidate associated page in the page cluster, wherein the page cluster is obtained by clustering the plurality of candidate associated pages.

10. The method according to claim 6, characterized in that Performing intention recognition on the object based on the information acquisition request to obtain the information acquisition intention of the object includes: Performing similarity analysis based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operation of the object to obtain a type similarity result; and In the case where the type similarity result meets the preset information type similarity condition, the information extension intention is determined, wherein the information type corresponding to the information acquisition request includes at least one of the following: The subject type of the object input information carried in the information acquisition request; The page subject type of the target page related to the information acquisition request; The page topic type of the target page related to the information acquisition request.

11. The method according to claim 10, characterized in that Based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operation of the object, a similarity result of the type is obtained, including: Determining a first similarity between the information type and a first historical information type corresponding to a first historical interaction operation, wherein the first historical interaction operation is a historical interaction operation in a first historical period; determining a second similarity between the information type and a second historical information type corresponding to a second historical interaction operation, wherein the second historical interaction operation is a historical interaction operation in a second historical period, and a timing attribute of the first historical period is earlier than a timing attribute of the second historical period; and In response to the first similarity being lower than a preset similarity threshold, and the second similarity being higher than or equal to the preset similarity threshold, a type similarity result satisfying the information type similarity condition is determined.

12. The method according to claim 1, characterized in that Determining page information related to the target page includes: Performing intent recognition on the object according to the operation attribute carried by the information acquisition request to obtain the information acquisition intent, wherein the information acquisition intent includes an operation area intent, the operation area intent corresponds to a target page area in the target page, and the operation attribute is determined based on determining an interactive operation of the information acquisition request; According to the operation area intention, at least one piece of page information is determined from a target page area in the target page.

13. The method according to claim 12, characterized in that The operation attributes include at least one of the following: The positional relationship between the operation position of the interactive operation and the regional element in the target page area; The operation area range of the interactive operation.

14. The method according to any one of claims 1, 6 or 12, characterized in that: The page information includes video information, and the structured feedback information includes segment description information associated with a video information segment in the video information.

15. An information processing device based on a large model, characterized in that: include: A page information determination module, configured to determine page information related to a target page in response to an object's information acquisition request for the target page; A structured feedback information determination module, configured to determine, based on the page information, structured feedback information corresponding to the information acquisition request of the object, wherein the structured feedback information is obtained by fusing information content having the same text structure level in the page information using a large model, and the structured feedback information has the text structure level; Wherein, the structured feedback information determination module includes: a demand prompt template determination submodule, configured to determine a demand prompt template corresponding to a feedback information demand type of the object, wherein the feedback information demand type is determined by performing intention recognition on the object based on the information acquisition request, the feedback information demand type represents the demand intention of the object for the information type of the structured feedback information, and the demand intention of the information type of the structured feedback information represents the type of information that the object needs to obtain; and The structured feedback information acquisition submodule is used to use the large model to process the page information and the demand prompt template to obtain the structured feedback information. The demand prompt template is used to control the large model to execute the data processing task corresponding to the feedback information demand type.

16. The device according to claim 15, characterized in that The feedback information requirement type is determined based on the following operations: determining an object preference attribute of the object based on the information acquisition request; and The feedback information requirement type is determined based on the similarity between the object preference attribute and the page key information of the target page.

17. The device according to claim 15, characterized in that The structured feedback information obtaining submodule includes: The first processing unit is used to process the page information and the demand prompt template using a target large model determined according to a model calling strategy, wherein the model calling strategy is determined according to a feedback delay attribute of the object, and the feedback delay attribute is determined according to the amount of information in the page information.

18. The device according to claim 17, characterized in that The page information includes video information, Wherein, the first processing unit includes: a video text obtaining subunit, configured to process the video information and the video recognition prompt template using the target visual large model to obtain a video text corresponding to the video information, wherein the demand prompt template includes the video recognition prompt template, and the video recognition prompt template is used to control the target visual large model to recognize the video content displayed in the video information; and The first processing unit is used to process the video text and the feedback information requirement prompt template using the target large language model, wherein the feedback information requirement prompt template is used to control the target large model to process the video text according to the feedback information requirement type.

19. The device according to claim 18, characterized in that The video recognition prompt template is determined based on at least one of the following prompt tag sequences: A first prompt mark sequence is used to control the target visual model to identify the video subtitle text based on the display area of ​​the text in the video information; The second prompt mark sequence is used to control the target visual model to perform target object detection on the video information.

20. The device according to claim 15, characterized in that The page information determination module includes: A first obtaining submodule is configured to perform intent recognition on the object based on the information acquisition request to obtain the information acquisition intent of the object, wherein the information acquisition intent includes an information extension intent, and the information extension intent indicates that the object needs to obtain other information content in addition to the information content recorded on the target page; and The associated page determination submodule is used to determine the associated page corresponding to the target page based on the information extension intention, wherein the page information is determined according to the associated page.

21. The device according to claim 20, characterized in that The associated page determination submodule includes: The first associated page determining unit is configured to determine a first associated page based on the information extension intention and the historical information acquisition request of the object, wherein the associated pages include the first associated page.

22. The device according to claim 20 or 21, characterized in that The associated page determination submodule includes: The second associated page determining unit is used to determine a second associated page based on the information extension intention and the page key information of the target page, and the associated pages include the second associated page.

23. The device according to claim 22, characterized in that The second associated page determining unit includes: a candidate associated page obtaining subunit, for responding to an information extension instruction corresponding to the information extension intention, searching based on the page key information, and obtaining a plurality of candidate associated pages, wherein the candidate associated pages correspond to the search ranking attributes; and The second associated page determination subunit is used to determine at least one second associated page from the page clustering cluster based on the search ranking attribute corresponding to the clustered candidate associated pages in the page clustering cluster, wherein the page clustering cluster is obtained by clustering the multiple candidate associated pages.

24. The device according to claim 20, characterized in that The first obtaining submodule comprises: a type similarity result obtaining unit, configured to perform similarity analysis based on the information type corresponding to the information acquisition request and the historical information type corresponding to the historical interaction operation of the object to obtain a type similarity result; and An information extension intention determining unit is used to determine the information extension intention when the type similarity result meets a preset information type similarity condition, wherein the information type corresponding to the information acquisition request includes at least one of the following: The subject type of the object input information carried in the information acquisition request; The page subject type of the target page related to the information acquisition request; The page topic type of the target page related to the information acquisition request.

25. The device according to claim 24, characterized in that The type similarity result obtaining unit comprises: A first similarity determination subunit is used to determine a first similarity between the information type and a first historical information type corresponding to a first historical interaction operation, wherein the first historical interaction operation is a historical interaction operation in a first historical period; a second similarity determination subunit, configured to determine a second similarity between the information type and a second historical information type corresponding to a second historical interaction operation, wherein the second historical interaction operation is a historical interaction operation in a second historical period, and a timing attribute of the first historical period is earlier than a timing attribute of the second historical period; and The type similarity result determining subunit is configured to determine a type similarity result satisfying the information type similarity condition in response to the first similarity being lower than a preset similarity threshold and the second similarity being higher than or equal to the preset similarity threshold.

26. The device according to claim 15, characterized in that The page information determination module includes: An information acquisition intention determination submodule, configured to perform intention recognition on the object according to the operation attribute carried by the information acquisition request to obtain the information acquisition intention, wherein the information acquisition intention includes an operation area intention, the operation area intention corresponds to the target page area in the target page, and the operation attribute is determined based on determining the interactive operation of the information acquisition request; The page information determination submodule is used to determine at least one piece of page information from a target page area in the target page according to the operation area intention.

27. The device according to claim 26, characterized in that The operation attributes include at least one of the following: The positional relationship between the operation position of the interactive operation and the regional element in the target page area; The operation area range of the interactive operation.

28. The device according to any one of claims 15, 20 or 26, characterized in that The page information includes video information, and the structured feedback information includes segment description information associated with a video information segment in the video information.

29. An electronic device, characterized in that: include: at least one processor; as well as a memory communicatively connected to the at least one processor; wherein, The memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method according to any one of claims 1 to 14.

30. A non-transitory computer-readable storage medium storing computer instructions, characterized in that: The computer instructions are used to cause the computer to execute the method according to any one of claims 1 to 14.

31. A computer program product, characterized in that A computer program is included which, when executed by a processor, implements the method according to any one of claims 1 to 14.

Images