Content recommendation methods, devices, equipment and storage media

By receiving voice input and parsing intent on short video platforms, a seamless transition from initial media content to recommended content sequences is achieved, solving the real-time and contextual requirements of passive recommendations and improving the accuracy of recommendations and the continuity of user interaction.

CN123123646APending Publication Date: 2026-11-17BEIJING DAJIA INTERNET INFORMATION TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202611114381.5
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2026-07-24
Publication Date
2026-11-17

AI Technical Summary

Technical Problem

The passive recommendation algorithms of existing short video and content recommendation platforms are unable to meet users' immediate and contextual needs, and the efficiency of text search input is low, interrupting users' immersive browsing experience.

Method used

By receiving voice input during content playback and performing semantic parsing to generate intent tags, it enables seamless switching from initial media content to a recommended content sequence containing at least two recommended content items, and supports voice interaction to proactively express immediate needs.

Benefits of technology

While ensuring an immersive browsing experience, it improves the accuracy of content recommendations, response speed, and continuity of user interaction, simplifies the operation process, and enhances the user experience.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN123123646A_ABST
    Figure CN123123646A_ABST
Patent Text Reader

Abstract

This application discloses a content recommendation method, apparatus, device, and storage medium. It includes: playing initial media content on a content playback page; and, in response to receiving a first voice input, switching the initial media content to play one of the recommended contents in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained through semantic parsing of the first voice input; and the first recommended content sequence includes at least two recommended contents. This solution allows users to proactively express their immediate needs and obtain continuous relevant content recommendations through voice interaction without interrupting the current content playback experience, avoiding interruptions in the interaction process, and improving the accuracy of content recommendations and the immersive browsing experience.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

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

[0002] Currently, short video and content recommendation platforms generally employ passive recommendation algorithms based on users' historical behavior, or offer active acquisition methods such as text search and category browsing. However, passive recommendations struggle to meet users' immediate and contextual needs, while text search is inefficient and costly, disrupting the user's immersive browsing experience. Summary of the Invention

[0003] In view of this, this application provides a content recommendation method, apparatus, device, and storage medium.

[0004] This application provides the following solution: Firstly, a content recommendation method is provided, the method comprising: Play the initial media content on the content playback page; In response to receiving a first voice input, the initial media content is switched to play one of the recommended contents in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained by semantic parsing the first voice input, and the first recommended content sequence contains at least two recommended contents.

[0005] Optionally, the method further includes: In response to a swipe operation on the currently displayed recommended content on the content playback page, the previous or next recommended content in the first recommended content sequence is switched.

[0006] Optionally, the method further includes: When playing a recommended content from the first recommended content sequence, in response to receiving a second voice input, the currently playing recommended content is switched to playing a recommended content from the second recommended content sequence; wherein, the second recommended content sequence is generated based on the intent tag obtained by semantic parsing the second voice input and the context information of the first recommended content sequence, and the second recommended content sequence contains at least two recommended content items.

[0007] Optionally, the step of switching the initial media content to play one of the recommended contents in the first recommended content sequence in response to receiving the first voice input includes: In response to receiving a first voice input, the system sends the first voice input to the server and receives response data returned by the server in response to the first voice input; wherein, when the semantic correlation between the intent tag and the initial media content is lower than a preset threshold, the response data is the first recommended content sequence, and when the semantic correlation between the intent tag and the initial media content is not lower than the preset threshold, the response data is additional information matching the first voice input; When the response data is the first recommended content sequence, the initial media content is switched to play one of the recommended contents in the first recommended content sequence; The method further includes: When the response data is the additional information, the additional information is overlaid on the content playback page.

[0008] Optionally, the intent tag includes at least one of the following: content recommendation intent, product query intent, local service intent, and information query intent.

[0009] Optionally, if the intent tag is the content recommendation intent, then the recommended content is media content, and the display format of the recommended content includes: The recommended content is played on the content playback page.

[0010] Optionally, if the intent tag is the product query intent, then the display format of the recommended content includes: Play a video explaining the target product on the content playback page; In addition, purchase entry information is superimposed on the playback interface of the product explanation video; wherein, the purchase entry information includes at least one of the following: product price information, purchase button, product details page link or add to cart entry.

[0011] Optionally, if the intent tag is the local service intent, then the display format of the recommended content includes: Play service-related media content on the content playback page; In addition, service discount information is overlaid on the playback interface of the media content associated with the service; wherein the service discount information includes at least one of the following: group purchase discount information, store location information, distance information, coupon information, voucher information, or discount activity information.

[0012] Optionally, if the intent tag is the information query intent, then the display format of the recommended content includes: An information card is displayed on the content playback page, the information card containing query results for the first voice input; Furthermore, on the content playback page, media content associated with the query results is arranged sequentially after the information card, and the associated media content is played in response to the user's swipe operation.

[0013] Optionally, the response to receiving the first voice input includes: In response to a triggering operation on the first voice input control on the content playback page, a voice acquisition prompt message is displayed on the content playback page, and the first voice input is acquired; The voice acquisition prompts include at least one of the following: sound wave effects, volume indication, and text transcription display during the voice recognition process.

[0014] Secondly, a content recommendation method is provided, the method comprising: Receive the first voice input sent by the user terminal, which is collected during the initial media content playback process on the content playback page; Semantic parsing is performed on the first voice input data to obtain intent tags. Based on the intent tags, a first recommended content sequence is generated, which contains at least two recommended content items. The first recommended content sequence is sent to the user terminal, so that the user terminal switches the initial media content played on the content playback page to play one of the recommended contents in the first recommended content sequence.

[0015] Optionally, the method further includes: Receive a second voice input sent by the user terminal, the second voice input being collected during the process of the user terminal playing a recommended content from the first recommended content sequence; Semantic parsing is performed on the second voice input to obtain the second intent label; Based on the context information of the second intent tag and the first recommended content sequence, a second recommended content sequence is generated, wherein the second recommended content sequence contains at least two recommended content items; The second recommended content sequence is sent to the user terminal so that the user terminal switches the currently playing recommended content to playing one of the recommended contents in the second recommended content sequence.

[0016] Optionally, generating the first recommended content sequence includes: The first voice input is subjected to speech recognition to obtain the voice text. Perform semantic parsing on the speech text to determine the intent tag; Based on the intent tags, original recommendation data is obtained from the corresponding content sources to form the first recommended content sequence.

[0017] Optionally, the step of semantically parsing the speech text to determine the intent tag includes: Perform semantic parsing on the spoken text; When the speech text is identified to contain a reference word referring to the initial media content, the contextual features of the initial media content at the moment when the first speech input is triggered are obtained; The intent label is determined by performing joint semantic parsing on the spoken text and the contextual features.

[0018] Optionally, obtaining the original recommendation data from the corresponding content source based on the intent tag includes: When the intent tag is a content recommendation intent, target media content that matches the voice text is obtained from the corresponding content source and used as the original recommendation data. When the intent tag is a local service intent, the user's location information is obtained, and based on the local service intent and the user's location information, service-related media content and service offer information that match the voice text are obtained from the corresponding content source as the original recommendation data; When the intent tag is a product query intent, the target product explanation content and purchase entry information that match the voice text are obtained from the corresponding data source as the original recommendation data; When the intent tag is an information query intent, the corresponding query result and the media content associated with the query result are obtained from the corresponding data source, and the information card containing the query result and the media content associated with the query result are used as the original recommendation data.

[0019] Optionally, generating the first recommended content sequence based on the intent tag includes: Determine the semantic correlation between the intent tag and the initial media content currently being played by the client; When the semantic relevance is lower than a preset threshold, the first recommended content sequence is generated; The method further includes: When the semantic relevance is not lower than the preset threshold, additional information matching the first voice input is generated; The additional information is sent to the user terminal so that the user terminal can overlay the additional information on the content playback page.

[0020] Thirdly, a content recommendation device is provided, the device comprising: The page display unit is configured to play initial media content on the content playback page; The monitoring unit is configured to monitor voice input on the content playback page; The page display unit is further configured to: in response to receiving a first voice input, switch the initial media content to play a recommended content in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained by semantic parsing the first voice input, and the first recommended content sequence contains at least two recommended content items.

[0021] Fourthly, a content recommendation device is provided, the device comprising: The receiving unit is configured to receive a first voice input sent by the user terminal, which is collected during the initial media content playback process on the content playback page; The generation unit is configured to perform semantic parsing on the first voice input data to obtain intent tags, and generate a first recommended content sequence based on the intent tags, wherein the first recommended content sequence contains at least two recommended content items; The sending unit sends the first recommended content sequence to the user terminal, so that the user terminal switches the initial media content played on the content playback page to play one of the recommended contents in the first recommended content sequence.

[0022] Fifthly, a computer-readable storage medium is provided having a computer program stored thereon, which, when executed by a processor, implements the steps of the method described in any one of the first or second aspects above.

[0023] Sixthly, an electronic device is provided, comprising: One or more processors; and A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any one of the first or second aspects above.

[0024] A seventh aspect provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any one of the first or second aspects.

[0025] According to the specific embodiments provided in this application, the following technical effects are disclosed: This application embodiment receives voice input during media content playback and, based on the intent parsing of the voice input, seamlessly switches the initial media content to a recommended content sequence within the current playback page. This sequence contains at least two recommended content items, allowing users to actively express their immediate needs through voice interaction and obtain a continuously switchable recommended content sequence without interrupting their current browsing experience. This avoids the interaction interruption caused by exiting the playback page for text search or selecting content from a list, as is common in traditional solutions. It also overcomes the shortcomings of passive recommendations in responding to users' immediate needs, thereby significantly improving the accuracy, response speed, and continuity of interaction between users and recommended content while maintaining an immersive browsing experience.

[0026] Of course, any product implementing this application does not necessarily need to achieve all of the advantages described above at the same time. Attached Figure Description

[0027] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0028] Figure 1 This is a schematic diagram of the system architecture applicable to the embodiments of this application.

[0029] Figure 2 A flowchart illustrating the content recommendation method provided in this application embodiment.

[0030] Figure 3 This is a schematic diagram of a content playback page provided in an embodiment of this application.

[0031] Figure 4 This is a schematic diagram illustrating the page switching display of the content recommendation intent provided in the embodiments of this application.

[0032] Figure 5 This is a schematic diagram illustrating the page switching display of the product search intent provided in the embodiments of this application.

[0033] Figure 6 This is a schematic diagram illustrating the page switching of the local service intent provided in the embodiments of this application.

[0034] Figure 7 This is a schematic diagram showing the page switching display of the information query intent provided in the embodiments of this application.

[0035] Figure 8 A flowchart illustrating the content recommendation method provided in this application embodiment.

[0036] Figure 9 A schematic block diagram of a content recommendation device provided in an embodiment of this application.

[0037] Figure 10 A schematic block diagram of a content recommendation device provided in an embodiment of this application.

[0038] Figure 11 A schematic block diagram of an electronic device provided in an embodiment of this application. Detailed Implementation

[0039] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of this application are within the scope of protection of this application.

[0040] The terminology used in the embodiments of this invention is for the purpose of describing particular embodiments only and is not intended to limit the invention. The singular forms “a,” “the,” and “the” as used in the embodiments of this invention and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise.

[0041] It should be understood that the term "and / or" used in this article is merely a description of the relationship between related objects, indicating that three relationships can exist. For example, A and / or B can represent: A existing alone, A and B existing simultaneously, and B existing alone. Additionally, the character " / " in this article generally indicates that the preceding and following related objects have an "or" relationship.

[0042] Depending on the context, the word "if" as used here can be interpreted as "when," "when," "in response to determination," or "in response to detection." Similarly, depending on the context, the phrase "if determination" or "if detection (of the stated condition or event)" can be interpreted as "when determination," "in response to determination," "when detection (of the stated condition or event)," or "in response to detection (of the stated condition or event)."

[0043] Currently, short video and content recommendation platforms generally employ passive recommendation algorithms based on users' historical behavior, or offer active acquisition methods such as text search and category browsing. However, passive recommendations struggle to meet users' immediate and contextual needs, while text search is inefficient and costly, disrupting the user's immersive browsing experience.

[0044] For example, when a user is watching a home renovation video, if they suddenly have the idea of ​​"wanting to see similar high-end renovation videos" or "wanting to find the paint shown in the video," the user can only exit the current full-screen playback interface, switch to the search function, and search by text input. This not only interrupts the browsing experience, but also the text input is inefficient for expressing vague intentions. The entire operation chain, from exiting, inputting, searching to filtering results, takes much longer than the interval of a single swipe, causing the recommended content to fail to adjust in real time according to the user's current focus.

[0045] In view of this, this application provides a new approach. To facilitate understanding of this application, the system architecture on which this application is based will first be described. Figure 1 This is a schematic diagram of the system architecture applicable to the embodiments of this application, such as... Figure 1 As shown, the system architecture may include: user terminal, terminal device and server terminal.

[0046] In this embodiment, the user terminal is set on the terminal device. The user terminal involved can be a client running on the terminal device, a mini-program, or a web application running through a browser, etc., and is used to play the initial media content on the content playback page, collect user voice input, and initiate content recommendation requests.

[0047] Terminal devices can include, but are not limited to, smart mobile terminals, wearable devices, PCs (Personal Computers), and smart home devices. Smart mobile devices can include devices such as mobile phones, tablets, laptops, PDAs (Personal Digital Assistants), and connected car terminals. Wearable devices can include devices such as smartwatches, smart glasses, smart bracelets, VR (Virtual Reality) devices, AR (Augmented Reality) devices, and mixed reality devices (devices that support both virtual and augmented reality). Smart home devices can include devices such as smart TVs and smart refrigerators with displays.

[0048] The user can interact with the server via the network. The server can be a platform providing voice parsing, intent recognition, and content recommendation services. It can be a single server, a server cluster consisting of multiple servers, or a cloud server. A cloud server, also known as a cloud computing server or cloud host, is a hosting product in the cloud computing service system, designed to address the shortcomings of traditional physical hosts and Virtual Private Servers (VPS) services, such as high management difficulty and weak service scalability.

[0049] As one possible approach, after receiving the first voice input from the user, the server performs semantic parsing on the first voice input to obtain an intent tag, generates a corresponding first recommended content sequence based on the intent tag, and returns it to the user; the user then switches the initial media content on the content playback page to play one of the recommended contents from the first recommended content sequence.

[0050] It should be understood that Figure 1 The number of client terminals, terminal devices, and servers shown is merely illustrative. Depending on implementation needs, there can be any number of client terminals, terminal devices, and servers.

[0051] Figure 2 A flowchart illustrating the content recommendation method provided in this application embodiment, which can be generated by... Figure 1 The user-side execution in the system shown is as follows. Figure 2 As shown, the method may include the following steps: Step 201: Play the initial media content on the content playback page.

[0052] Step 202: In response to receiving the first voice input, switch the initial media content to play one of the recommended contents in the first recommended content sequence; wherein, the first recommended content sequence is generated based on the intent tag obtained by semantic parsing the first voice input, and the first recommended content sequence includes at least two recommended contents.

[0053] As can be seen, this embodiment of the application receives voice input during media content playback and, based on the intent parsing result of the voice input, seamlessly switches the initial media content to a recommended content in the recommended content sequence directly on the current playback page. This sequence contains at least two recommended content items, allowing users to actively express their immediate needs through voice interaction and obtain a recommended content sequence that can be continuously switched without interrupting their current browsing experience. This avoids the interaction interruption caused by exiting the playback page to perform text search or select content from a list in traditional solutions. At the same time, it overcomes the defect that passive recommendations cannot respond to users' immediate needs, thereby significantly improving the accuracy of content recommendations, response speed, and the continuity of interaction between users and recommended content while ensuring an immersive browsing experience.

[0054] The following describes in detail each step of the above process and the effects that can be further produced, with reference to the embodiments.

[0055] First, the above step 201, namely "playing the initial media content on the content playback page", will be described in detail with reference to the embodiments.

[0056] In the embodiments of this application, the initial media content includes, but is not limited to, videos, images in an image gallery, etc.

[0057] Videos can be either long or short. The distinction between long and short videos mainly lies in their length and content. Long videos are typically several tens of minutes or longer; their content is more complete and in-depth, and the quality is relatively high. Examples include movies, TV series, and documentaries. Short videos are usually a few minutes or less; their content is usually concise and refined, focusing on highlighting a specific point or quickly conveying information. Short videos emphasize rapid production and dissemination, stressing timeliness. Examples include funny videos, lifestyle videos, advertising videos, and short dramas.

[0058] In addition, the videos mentioned above can be either landscape or portrait. A landscape video is one where the width of the video frame is greater than its height, which better matches the natural field of vision of the human eye and is more suitable for viewing on widescreen devices (such as TVs and computers). A portrait video, on the other hand, is one where the height of the video frame is greater than its width, making it more suitable for viewing on narrow-screen devices (such as smartphones and tablets) and more popular in social media and short video platforms.

[0059] The images in the collection can be regular images or long images. Regular images are usually single, static images that users can view without any additional action. Long images, on the other hand, are a special type of image composed of continuous images. Users need to scroll or swipe to view the entire content of the long image. Long images typically combine multiple image elements to form a coherent narrative.

[0060] It should be noted that, in the embodiments of this application, "playback" refers to converting the content of a work into images, sounds, or a combination of both that can be directly perceived by humans.

[0061] The following describes in detail step 202, namely, "in response to receiving the first voice input, the initial media content is switched to play one of the recommended contents in the first recommended content sequence; wherein the first recommended content sequence is generated based on the intent tag obtained by semantic parsing the first voice input, and the first recommended content sequence includes at least two recommended contents", with reference to the embodiments.

[0062] In this embodiment, when a user browses initial media content through the content playback page on the user's device, they can actively express their immediate needs via voice. This first voice input, as an actively triggered interactive command, carries the user's current intent information. After receiving the first voice input, the user's device sends it to the server for parsing. The server performs voice parsing on the first voice input to obtain an intent tag, and generates a first recommended content sequence containing at least two recommended items based on the intent tag. This first recommended content sequence is a sequence containing at least two recommended items and is returned to the user's device. The user's device then switches from the currently playing initial media content to playing one of the recommended items in this sequence, thereby achieving a seamless switch from passive browsing to actively obtaining recommended content without interrupting the playback experience.

[0063] Specifically, step 202 above can be implemented as follows: When a user browses initial media content on the content playback page, by performing a trigger operation on the voice input control on the content playback page (e.g., pressing and holding the microphone icon button), the user's terminal responds to the trigger operation by calling the microphone hardware to capture the user's first voice input.

[0064] Optionally, during the voice input collection process, the user terminal can display voice collection prompts on the content playback page to provide feedback on the current voice collection status. These prompts may be at least one of the following: sound wave effects, volume indicators, or text transcription displays during the voice recognition process; this embodiment does not limit the specific type of prompt.

[0065] As an example, see Figure 3 The content playback page 100 includes a voice input control 101. This voice input control 101 is a persistent interactive element on the content playback page 100, always displayed or at least displayed in a specific state of the playback page, providing users with a clear and convenient voice interaction entry point. This voice input control 101 is typically presented in the form of a button, for example... Figure 3 The microphone icon is shown in the middle of the bottom navigation bar. Naturally, this voice input control 101 can also be placed in other locations on the page, such as the side or floating display, as long as it is presented on the content playback page, it falls within the protection scope of this application.

[0066] By setting a permanent voice input control 101 on the content playback page and using it as the sole entry point for the entire recommendation process, users can easily initiate voice interaction without jumping to other pages or switching applications while immersing themselves in content consumption. This reduces the difficulty of operation and ensures the integrity and consistency of the interactive experience.

[0067] After the user receives the first voice input, it sends the first voice input to the server. Upon receiving the first voice input, the server performs semantic parsing on the first voice input to obtain an intent tag, generates a first recommended content sequence based on the intent tag, and sends the first recommended content sequence to the user so that the user can switch the initial media content played on the content playback page to play one of the recommended contents in the first recommended content sequence.

[0068] Intent tags are categorized labels assigned after a structured understanding of user needs, determining the type of content to recommend. For example, when a user says "I want to go to the beach," the semantic analysis might yield an intent tag of "travel" or "content recommendation." Based on this tag, a series of beach travel-related videos can be generated as the first recommended content sequence. In other words, a set of recommended content that can be continuously browsed can be generated based on a single user voice command.

[0069] In an optional embodiment, to enable users to conveniently browse the first recommended content sequence continuously after voice-triggered recommendations, when the user switches the content playback page to play one of the recommended contents in the first recommended content sequence, the user can continuously switch between other recommended contents in the sequence by swiping on the content playback page. Specifically, in response to a swipe operation (e.g., an up or down swipe gesture) on the currently displayed recommended content on the content playback page, the currently displayed recommended content is switched to the previous or next recommended content in the first recommended content sequence. For example, when the user swipes up, the user switches the currently playing recommended content to the next recommended content in the sequence; when the user swipes down, the client switches the currently playing recommended content to the previous recommended content in the sequence.

[0070] This swipe operation is consistent with the operation paradigm of switching media content in a single-column information stream. Users do not need to relearn new interaction methods or return to any list interface to make selections. They can continuously browse all recommended content in the first recommended content sequence in the recommendation results, thus obtaining a coherent and efficient browsing experience.

[0071] In an optional embodiment, the intent tags obtained from server-side semantic parsing include at least one of the following: content recommendation intent, product query intent, local service intent, and information query intent. Different intent tags correspond to different types of recommended content and display formats.

[0072] Among them, content recommendation intent corresponds to the user's need to browse more media content on specific topics, such as "recommend some funny cat videos"; product search intent corresponds to the user's need to learn about or purchase physical or virtual goods, such as "I want to buy a black hoodie"; local service intent corresponds to the user's need to find local restaurants, entertainment, and lifestyle service stores, such as "where are some good hot pot restaurants nearby"; and information search intent corresponds to the user's need to obtain factual information, such as "what's the weather like today" or "what new movies are showing recently".

[0073] When a user receives the first recommended content sequence and switches to play one of the recommended items, if the intent tag is content recommendation intent, then the recommended content itself is media content (such as a short video). In this case, the user directly plays the recommended content on the content playback page, for example, in a single-column video stream format in full screen. Users can use up and down swiping to continuously switch between the previous or next recommended content in the first recommended content sequence on the current playback page, forming an immersive and continuous content consumption experience.

[0074] As an example, such as Figure 4 As shown, Figure 4 The content playback page in (a) is playing a home renovation video, which is the initial media content. At this time, the user triggers the voice input control 101 and inputs the voice message "Recommend some funny cat videos". The server generates the first recommended content sequence related to "funny cat videos" and returns it to the user, such as... Figure 4 As shown in (b), the user switches the home renovation video on the content playback page to play the first funny cat video in the sequence. The user can then swipe up to browse other funny cat videos in the sequence.

[0075] As can be seen from the above examples, in this embodiment of the application, users do not need to exit the current playback page to conduct text search or manually filter from the list. They only need to use voice input once to obtain a continuously arranged sequence of recommended content in the current immersive playback environment, and complete the browsing of each recommended content in the sequence through familiar swiping operations. This greatly simplifies the operation path for users to actively obtain personalized recommended content and enhances the immersion and continuity of browsing.

[0076] If the intent tag is a product search intent, the recommended content will be displayed in the following ways: a video explaining the target product (such as a review or demonstration video created by a creator) will be played on the content playback page, while a purchase entry information will be overlaid on the playback interface of the target product explanation video. This purchase entry information can be at least one of the following: product price information, a purchase button, a link to the product details page, or an add-to-cart entry. While watching the product explanation video, users can directly jump to the product details page or complete the purchase operation by clicking on the overlaid purchase entry information without leaving the current playback page.

[0077] As an example, such as Figure 5 As shown, Figure 5 The content playback page in (a) is playing a home renovation video, which serves as the initial media content. At this point, the user triggers the voice input control 101 and inputs the voice command, "I want to buy a black hoodie." Based on this voice command, the server generates a first recommended content sequence related to "black hoodie" and returns it to the user. This first recommended content sequence contains at least two recommended items. For example, the first recommended item might be a video explaining a black hoodie from a certain brand and its purchase entry information; the second recommended item might be a video explaining a black hoodie from another brand and its purchase entry information; the third recommended item might be a video showcasing how to wear a black hoodie and its purchase entry information, and so on. The user switches the home renovation video on the content playback page to the first black hoodie explanation video 521 in this sequence, and simultaneously overlays purchase entry information 522 on top of this black hoodie explanation video. This purchase entry information 522 can be at least one of the following: product price information, a purchase button, a link to the product details page, or an add-to-cart entry. For example, as shown below... Figure 5 As shown in (b), a semi-transparent purchase button (i.e. purchase information) 522 is overlaid in the area below the black hoodie demonstration video 521. The purchase button 522 is marked with product price information (such as "¥199") and the words "Buy Now".

[0078] If a user is interested in the product while watching the video explaining the black hoodie, they can directly click the purchase button 522 that is displayed overlaid on the video playback interface. In response to this click, the client will either pop up a half-screen product details page on the current content playback page 100 or jump to the product details page, allowing the user to complete the purchase without exiting the current playback experience.

[0079] As can be seen from the above examples, in this embodiment of the application, users do not need to exit the current playback page to conduct text searches or browse the product list. They can obtain a sequence of recommended content containing multiple related product explanation videos without interrupting the viewing experience by simply using voice input. They can also continuously browse each recommended content by swiping. At the same time, they can directly complete the purchase through the overlaid purchase entry during the viewing process, which greatly simplifies the operation link from content seeding to product conversion and improves browsing efficiency and conversion experience.

[0080] If the intent tag is for local services, the recommended content will be displayed in the following ways: playing service-related media content (such as a restaurant review video) on the content playback page, while simultaneously displaying service discount information overlaid on the playback interface of the service-related media content. This service discount information can be at least one of the following: group-buying discount information, store location information, distance information, coupon information, voucher information, or discount promotion information. While watching the restaurant review video, users can directly place a group-buying order or obtain more store details by clicking on the overlaid service discount information.

[0081] As an example, such as Figure 6 As shown, Figure 6 The content playback page shown in (a) is playing a home renovation video, which is the initial media content. A voice input control 101 (e.g., a microphone icon button) is located at the bottom of the content playback page. While watching the home renovation video, the user feels hungry as noon approaches and has an immediate need to go out for lunch, wanting to find a good hot pot restaurant nearby. The user then triggers the voice input control 101 (e.g., pressing and holding the microphone icon button) and inputs the first voice input: "What are some good hot pot restaurants nearby with group deals?"

[0082] In response to the trigger operation, the user terminal calls the microphone hardware to collect the user's first voice input and sends the collected first voice input to the server. After receiving the first voice input, the server performs speech recognition on the first voice input to obtain the voice text "What are some good hot pot restaurants nearby with group deals"; then it performs semantic analysis on the voice text to identify the user's core intent as "local services" and the specific request as "hot pot group deals".

[0083] In response to a local service intent, the server obtains the user's current location information via GPS or cell tower triangulation. Based on this location information and the query criteria of "hot pot group buying," it retrieves service-related media content and service offers matching the voice / text from the local life service platform, forming a first recommended content sequence. This first recommended content sequence is then returned to the user as response data. This first recommended content sequence contains at least two recommended items. For example, the first recommended item might be a video review of hot pot restaurant A and related group buying information; the second recommended item might be a video review of hot pot restaurant B and related group buying information; the third recommended item might be a video review of hot pot restaurant C and related group buying information, and so on.

[0084] The user client switches the home renovation video playing on the content playback page to the first recommended content in the first recommended content sequence (i.e., the hot pot restaurant A's store visit video and related group-buying information).

[0085] like Figure 6 As shown in (b), since the intent tag is a local service intent, the client displays the recommended content in full-screen format after switching. In addition to playing the service-related media content 621 (i.e., a video review of hot pot restaurant A, such as introducing the restaurant's signature dishes, interior environment, customer reviews, etc.) on the content playback page, it also overlays service discount information 622 on the playback interface of the service-related media content 621. This service discount information 622 can be at least one of the following: group-buying discount information (e.g., "Double set meal group-buying price 158 yuan"), store location information (e.g., "800 meters from current location"), distance information (e.g., "approximately 10 minutes walk away"), coupon information, voucher information (e.g., "50 yuan voucher available for a limited time"), or discount activity information (e.g., "30 yuan off for purchases over 200 yuan"). For example, as... Figure 6 As shown in (b), a semi-transparent group-buying information card (i.e., service discount information) 622 is overlaid in the area below the service-related media content 621. The group-buying information card 622 is marked with the name of hot pot restaurant A, details of the group-buying package, comparison of the original price and the group-buying price, distance to the store, and a "Buy Now" button.

[0086] While watching the video review of hot pot restaurant A, users can learn about the restaurant's food quality, interior environment, and other practical information. If they are interested in the restaurant, they can directly click the "Buy Now" button on the group-buying information card 622 displayed on the video playback screen. In response to this click, the client will either pop up a half-screen group-buying order page on the current content playback page or redirect to the group-buying details page, allowing users to complete the group-buying order without exiting the current playback experience.

[0087] After watching the first recommended content (a video review of hot pot restaurant A and related group-buying information), users can swipe up on the content playback page. In response, the user's device switches the currently displayed recommended content from hot pot restaurant A to the next recommended content in the first recommended content sequence (i.e., the second recommended content—a video review of hot pot restaurant B and related group-buying information). This process continues, allowing users to continuously swipe up on the current content playback page 100 to browse various recommended content in the first recommended content sequence (such as video reviews of hot pot restaurants B and C and related group-buying information), enabling a direct and visual comparison of multiple hot pot restaurants and allowing for better consumption decisions.

[0088] As can be seen from the above examples, in this embodiment of the application, users do not need to exit the current playback page to conduct text search or manually search in the map application. They can obtain a sequence of recommended content containing multiple store visit videos and related discount information without interrupting the viewing experience by simply using voice input. They can also continuously browse each recommended content by swiping, and at the same time, they can directly complete group purchase orders by using the service discount information displayed on the screen during the viewing process. This greatly simplifies the operation link from content browsing to local life service consumption and improves the accuracy and conversion efficiency of local service recommendations.

[0089] If the intent tag is information query intent, the recommended content will be displayed in the following ways: First, an information card will be displayed on the content playback page, containing the query results (such as a weather summary) for the initial voice input. Following this information card on the content playback page are media content related to the query results (such as outfit suggestion videos or travel tips videos), which will play in response to the user's swipe gesture. In other words, the user first sees a structured information card, quickly obtains the core answer, and then seamlessly switches to watching media content related to the query results by swiping upwards, achieving a smooth transition from information query to content consumption.

[0090] As an example, such as Figure 7 As shown, Figure 7 The content playback page shown in (a) is playing a home renovation video, which is the initial media content. A voice input control 101 (e.g., a microphone icon button) is located at the bottom of the content playback page. At this moment, the user plans to go shopping the next day and wants to know the weather forecast to decide what to wear. Therefore, the user triggers the voice input control 101 (e.g., presses and holds the microphone icon button) and inputs the first voice input: "What's the weather like in Beijing tomorrow?"

[0091] In response to the triggered operation, the user terminal calls the microphone hardware to collect the user's first voice input and sends the collected first voice input to the server. After receiving the first voice input, the server performs speech recognition on the first voice input to obtain the voice text "What will the weather be like in Beijing tomorrow?". Then, it performs semantic analysis on the voice text to identify the user's core intent as "information query", specifically the query content "the weather in Beijing tomorrow". It then calls the weather service API (Application Programming Interface) to obtain real-time weather data for Beijing tomorrow (such as temperature, weather conditions, wind force, humidity, etc.) as the query result. Simultaneously, it retrieves media content associated with the query result from the media content library (such as a "rainy day travel outfit suggestion video" or a "high temperature prevention guide video" corresponding to "tomorrow's weather"). The information card containing the query result and the media content associated with the query result are combined as raw recommendation data to form a first recommendation content sequence, which is then returned to the user terminal. This first recommendation content sequence contains at least two recommended items; for example, the first recommended item is an information card containing Beijing's weather information for tomorrow, and the second recommended item is an outfit suggestion video associated with that weather information.

[0092] The user client switches the home renovation video playing on the content playback page to play the first recommended content in the first recommended content sequence (i.e., the information card containing tomorrow's weather information for Beijing).

[0093] like Figure 7 As shown in (b), since the intent tag is for information query intent, the user's terminal first displays an information card 710 on the content playback page after the switch. This information card 710 contains the query result for the first voice input—a summary of Beijing's weather information for tomorrow, such as "Beijing, July 20, sunny turning cloudy, 25°C~34°C, southerly wind 3~4, good air quality, suitable for travel." The information card 710 is presented in a structured form, allowing users to easily obtain the core answer. On the content playback page, media content 720 related to the query result can be played below the information card 710, such as "High Temperature Outfit Suggestions Video" or "Summer Sun Protection Guide Video." This related media content 720 switches in response to the user's swipe operation.

[0094] As can be seen from the above examples, in this embodiment of the application, users do not need to exit the current playback page to conduct text search or switch to a dedicated weather application. They can obtain structured information cards in the current immersive playback environment with just one voice input, quickly obtain core query results, and seamlessly connect to media content associated with the query results for extended browsing through swiping operations. This achieves an organic integration of efficient information acquisition and in-depth content consumption, expanding the depth of content consumption while ensuring query efficiency.

[0095] This solution allows the system to organize interface elements in a targeted manner based on the user's intent, whether it is to "watch videos," "buy things," "find services," or "search for information." This delivers core information in the most intuitive and efficient way, avoiding the low browsing efficiency caused by a one-size-fits-all approach to all information in a unified single-column interface. It improves information delivery efficiency and commercial conversion value.

[0096] Furthermore, in one embodiment, when the user is playing a specific piece of recommended content from the first recommended content sequence, the user can trigger the voice input control again to input a second voice input to express further needs. The user then sends the collected second voice input to the server.

[0097] After receiving the second voice input, the server first performs speech recognition to obtain the corresponding speech text. Then, it performs semantic parsing on the speech text to obtain the second intent tag. Unlike the first round of interaction, when generating the second recommended content sequence, the server not only obtains matching raw recommendation data from the content source based on the second intent tag, but also performs joint analysis in conjunction with the contextual information of the first recommended content sequence. This contextual information includes at least one of the following: the intent tag corresponding to the first recommended content sequence, the content characteristics of the recommended content already played in the first recommended content sequence (such as product category, price range, or service type), the user's browsing behavior in the first recommended content sequence (such as the position already scrolled to, the duration of stay on the current recommended content, whether there is a click purchase entry, etc.), and the user's current playback position in the sequence.

[0098] For example, after a user receives a sequence of recommended running shoes with video explanations by voice inputting "Recommend some shoes suitable for long-distance running" in the first round of interaction, and then says "Something cheaper" during browsing, the server recognizes the intent of the second voice input as "price filtering." Combining this with the contextual information that "the current recommended content is running shoes, with a price range of 800 to 1200 yuan," the server generates a second recommended content sequence—a running shoe recommendation sequence with a price range of 400 to 700 yuan—and returns it to the user. The user then switches the currently playing recommended content to one of the recommended items in the second recommended content sequence.

[0099] For example, when a user receives a sequence of running shoe recommendations in the first round of interaction and is browsing a video explaining a particular shoe, they might say, "Are there any red ones left?" The server recognizes the intent of the second voice input as "color filtering," and, combining this with the product identifier corresponding to the currently playing recommendation, queries the product database for the red version of that running shoe. It then generates a second sequence of recommendations containing information about the red running shoe and returns it to the user. The user then switches the currently playing recommendation to one of the recommendations in the second sequence.

[0100] For example, after browsing a home renovation video, a user receives multiple recommended home furnishing videos through the first round of voice input. During the browsing process, the user says "suitable for small apartments" as a second voice input. The server recognizes the "scene filtering" intent (filtering home furnishings suitable for small apartments), combines the intent tag "home furnishing recommendations" from the first round of interaction with the product category information involved in the currently played recommended content, filters home furnishings that match the "suitable for small apartments" tag from the product library, generates a second recommended content sequence, and returns it to the user.

[0101] Through the above methods, the embodiments of this application support multi-round continuous voice interaction during sequence browsing. The generation of each round of recommendation results inherits the semantic context of the previous round of interaction, thereby achieving progressive refinement and precise adjustment of recommendation results without requiring users to repeatedly express their existing needs. This effectively avoids repeated recommendations or semantic breaks, significantly improves the intelligence of voice interaction and the accuracy of recommendation results, and provides users with a more coherent and natural conversational content recommendation experience.

[0102] Furthermore, in one embodiment, after the server performs semantic parsing on the first voice input to obtain an intent tag, it can also determine the semantic relevance between the intent tag and the initial media content currently being played by the client. This semantic relevance is used to measure the degree of correlation between the user's voice intent and the currently playing content.

[0103] As one possible approach, the server can determine the semantic relevance by calculating the semantic similarity between the intent tag and the text features such as tags, titles, and descriptions of the initial media content. This application does not limit the specific calculation method for the semantic relevance.

[0104] Based on the determination of the semantic relevance, the server executes different response strategies according to the comparison result of the semantic relevance with the preset threshold: When the semantic relevance is below a preset threshold, it indicates a low correlation between the user's voice intent and the currently playing content. The user is inclined to exit the current content to obtain entirely new recommended content unrelated to the current content. In this case, the server generates a first recommended content sequence based on the intent tag and returns this first recommended content sequence as response data to the user. Upon receiving this response data, the user executes a switching operation, switching the currently playing initial media content to playing one of the recommended content items in the first recommended content sequence.

[0105] When the semantic relevance is not lower than a preset threshold, it indicates a high correlation between the user's voice intent and the currently playing content. The user tends to seek supplementary information related to the currently playing content rather than switching to entirely new content. At this point, the server retrieves supplementary information matching the first voice input (e.g., details of the currently playing product, information about other works by the current video blogger, extended information about knowledge points related to the currently playing content, etc.) and returns this supplementary information as response data to the user. Upon receiving this response data, the user displays the supplementary information overlaid on the currently playing page, for example, as a floating layer or pop-up window on the currently playing video screen, allowing the user to obtain the necessary supplementary information without interrupting the current playback.

[0106] In the above manner, this application embodiment adaptively selects two different response methods, "switching content" or "overlaying information," based on the semantic relevance between the user's voice intent and the currently playing content. When the user's needs are not related to the current content, a brand-new recommendation sequence is provided to meet the user's cross-domain exploration needs. When the user's needs are related to the current content, supplementary information is overlaid on the current content to deepen the current viewing experience. Thus, while meeting the user's information needs at different levels, the continuity and immersion of the browsing experience are maintained to the maximum extent.

[0107] This application also provides a content recommendation method, such as... Figure 8 As shown, Figure 8 A flowchart illustrating the content recommendation method provided in this application embodiment, which can be generated by... Figure 1 The server in the system shown executes the command. For example... Figure 8 As shown, the method may include the following steps: Step 801: Receive the first voice input sent by the user terminal. The first voice input is collected during the initial media content playback process on the content playback page.

[0108] Specifically, during the initial media content playback on the user's device, in response to the user's triggering of the voice input control, the system collects the user's first voice input and sends it to the server. Correspondingly, the server receives the first voice input from the user. This first voice input reflects the user's immediate needs, such as "Recommend some shoes suitable for long-distance running," "I want to buy a black hoodie," or "What are some good hot pot restaurants nearby?"

[0109] Step 802: Semantic parsing is performed on the first voice input data to obtain intent tags. Based on the intent tags, a first recommended content sequence is generated, which contains at least two recommended content items.

[0110] Specifically, the server performs semantic parsing on the received first voice input to identify the user's intent. This process includes the following sub-steps: First, the server performs speech recognition on the initial voice input, converting the voice signal into corresponding speech text. For example, if a user says, "Recommend a few pairs of shoes suitable for long-distance running," the server uses an acoustic model and a language model to convert it into the text "Recommend a few pairs of shoes suitable for long-distance running."

[0111] Then, the server performs semantic parsing on the speech text to determine the user's intent tags. Intent tags are used to summarize the user's core needs, such as "content recommendation intent," "product search intent," "local service intent," or "information search intent." As one possible approach, the server can use a pre-trained natural language understanding model to perform semantic parsing on the speech text and output the corresponding intent tags and related slot information (such as product category, price range, geographical location, etc.).

[0112] Optionally, when the server identifies reference words (such as "this," "this model," "similar," etc.) in the speech text, the server obtains the contextual features of the initial media content at the moment the first speech input was triggered (such as the video title, tags, category information, visual feature vectors, etc.), and performs joint semantic parsing on the speech text and the aforementioned contextual features to accurately determine the intent label. This method can effectively improve the accuracy and robustness of intent recognition when the user's voice command contains referential words, by combining the context of the currently playing content.

[0113] After obtaining the intent tag, the server retrieves the raw recommendation data based on the intent tag to construct the first recommended content sequence. Different intent tags correspond to different data sources and retrieval logic. When the intent tag is content recommendation intent (such as when a user says "recommend some funny cat videos"), the server retrieves target media content (such as funny cat video) that matches the semantics of the voice text from the media content library as the raw recommendation data.

[0114] When the intent tag is a local service intent (such as a user saying "What are some good hot pot restaurants nearby with group deals"), the server obtains the user's location information and, based on this location information and the voice text, retrieves service-related media content (such as videos of hot pot restaurants) and service discount information (such as group deals and vouchers) that match the voice text from the local life service platform as raw recommendation data.

[0115] When the intent tag is a product query intent (such as a user saying "I want to buy a black hoodie"), the server retrieves the target product description and purchase entry information that match the voice text from the e-commerce product database as the raw recommendation data.

[0116] When the intent tag is an information query intent (such as a user saying "What will the weather be like in Beijing tomorrow?"), the server obtains the corresponding query results (such as a weather information summary) from the corresponding data source (such as a weather API) and the media content associated with the query results (such as a "video on what to wear in hot weather"). The information card containing the query results and the associated media content are used together as the original recommendation data.

[0117] Optionally, before acquiring the original recommendation data and generating the first recommended content sequence, the server can first determine the semantic relevance between the intent tag and the initial media content currently being played by the client. This semantic relevance is used to measure the correlation between the user's voice intent and the currently playing content. Based on the comparison result of this semantic relevance with a preset threshold, the server determines different response strategies: When the semantic relevance is below a preset threshold, it indicates that the user's voice intent is not highly relevant to the currently playing content, and the user wants to move beyond the current content to access new content. In this case, the server executes the process described above, which involves obtaining raw recommendation data based on intent tags and generating the first recommended content sequence.

[0118] When the semantic relevance is not lower than a preset threshold, it indicates that the user's voice intent is highly relevant to the currently playing content, and the user wants to obtain supplementary information related to the current content rather than switching content. At this time, the server generates supplementary information matching the first voice input (such as details of the currently playing product, extended information of the current video, etc.) and sends this supplementary information to the user's terminal so that the user's terminal can overlay and display the supplementary information on the current playback page.

[0119] Finally, the server generates a first recommended content sequence based on the acquired raw recommendation data. This first recommended content sequence is a structured sequence formed by arranging multiple raw recommendation data according to preset sorting rules (such as relevance, popularity, price, etc.), and this sequence contains at least two recommended content items.

[0120] Step 803: Send the first recommended content sequence to the user terminal so that the user terminal switches the initial media content played on the content playback page to play one of the recommended contents in the first recommended content sequence.

[0121] Specifically, the server returns the generated first recommended content sequence as response data to the user. Upon receiving this response data, the user switches the currently playing initial media content on the current content playback page to play one of the recommended contents from the first recommended content sequence (e.g., the first recommended content), thus achieving a seamless transition from passive browsing to actively seeking recommended content without interrupting the user's playback experience.

[0122] Furthermore, in one embodiment, when the user is playing a recommended item in the first recommended content sequence, the user can trigger the voice input control again to input a second voice input to express further needs. The user sends the collected second voice input to the server. After receiving the second voice input, the server performs semantic parsing on the second voice input to obtain a second intent tag. Unlike the first round of interaction, when generating the second recommended content sequence, the server not only uses the second intent tag but also combines it with the contextual information of the first recommended content sequence for joint analysis. This contextual information includes, but is not limited to: the original intent tag corresponding to the first recommended content sequence, the content characteristics of the played recommended content (such as product category, price range, or service type), and the user's browsing behavior in the first recommended content sequence (such as the position already scrolled to, the duration of stay on the current recommended content, whether the purchase entry has been clicked, etc.). The server generates the second recommended content sequence based on the second intent tag and the above contextual information and returns the second recommended content sequence to the user. The user then switches the currently playing recommended content to play a recommended item in the second recommended content sequence. This approach allows the second round of recommendations to inherit the semantic context of the first round of interaction, avoiding duplicate recommendations or semantic breaks, and achieving a more intelligent and coherent multi-round conversational recommendation experience.

[0123] In this manner, the server receives voice input from the user, performs semantic parsing to obtain intent tags, generates a recommended content sequence containing at least two items based on these tags, and returns this sequence to the user to trigger content switching. Throughout the interaction, the server and user cooperate to enable the user to proactively express immediate needs through voice interaction and obtain a sequence of recommended content for continuous switching without interrupting the current content playback experience. This improves the accuracy of content recommendations, the continuity of interaction, and the user's operational efficiency.

[0124] The foregoing has described specific embodiments of this specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recited in the claims may be performed in a different order than that shown in the embodiments and may still achieve the desired result. Furthermore, the processes depicted in the drawings do not necessarily require the specific or sequential order shown to achieve the desired result. In some embodiments, multitasking and parallel processing are possible or may be advantageous.

[0125] According to another embodiment, a content recommendation device is provided. Figure 9 A schematic block diagram of the content recommendation device according to one embodiment is shown, the device being disposed in Figure 1 The user end in the illustrated architecture. For example... Figure 9 As shown, the device 900 includes a page display unit 901 and a listening unit 902. The main functions of each component are as follows: Page display unit 901 is configured to play initial media content on the content playback page; The listening unit 902 is configured to listen to voice input on the content playback page; The page display unit is further configured to: in response to receiving a first voice input, switch the initial media content to play a recommended content in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained by semantic parsing the first voice input, and the first recommended content sequence contains at least two recommended content items.

[0126] Optionally, the page display unit 901 is also configured as follows: In response to a swipe operation on the currently displayed recommended content on the content playback page, the previous or next recommended content in the first recommended content sequence is switched.

[0127] Optionally, the page display unit 901 is also configured as follows: When playing a recommended content from the first recommended content sequence, in response to receiving a second voice input, the currently playing recommended content is switched to playing a recommended content from the second recommended content sequence; wherein, the second recommended content sequence is generated based on the intent tag obtained by semantic parsing the second voice input and the context information of the first recommended content sequence, and the second recommended content sequence contains at least two recommended content items.

[0128] Optionally, the page display unit 901 is specifically configured as follows: In response to receiving a first voice input, the system sends the first voice input to the server and receives response data returned by the server in response to the first voice input; wherein, when the semantic correlation between the intent tag and the initial media content is lower than a preset threshold, the response data is the first recommended content sequence, and when the semantic correlation between the intent tag and the initial media content is not lower than the preset threshold, the response data is additional information matching the first voice input; When the response data is the first recommended content sequence, the initial media content is switched to play one of the recommended contents in the first recommended content sequence; The method further includes: When the response data is the additional information, the additional information is overlaid on the content playback page.

[0129] Optionally, the intent tag includes at least one of the following: content recommendation intent, product query intent, local service intent, and information query intent.

[0130] Optionally, if the intent tag is the content recommendation intent, then the recommended content is media content, and the display format of the recommended content includes: The recommended content is played on the content playback page.

[0131] Optionally, if the intent tag is the product query intent, then the display format of the recommended content includes: Play a video explaining the target product on the content playback page; In addition, purchase entry information is superimposed on the playback interface of the product explanation video; wherein, the purchase entry information includes at least one of the following: product price information, purchase button, product details page link or add to cart entry.

[0132] Optionally, if the intent tag is the local service intent, then the display format of the recommended content includes: Play service-related media content on the content playback page; In addition, service discount information is overlaid on the playback interface of the media content associated with the service; wherein the service discount information includes at least one of the following: group purchase discount information, store location information, distance information, coupon information, voucher information, or discount activity information.

[0133] Optionally, if the intent tag is the information query intent, then the display format of the recommended content includes: An information card is displayed on the content playback page, the information card containing query results for the first voice input; Furthermore, on the content playback page, media content associated with the query results is arranged sequentially after the information card, and the associated media content is played in response to the user's swipe operation.

[0134] Optionally, page display unit 901 is specifically configured as follows: In response to a triggering operation on the first voice input control on the content playback page, a voice acquisition prompt message is displayed on the content playback page, and the first voice input is acquired; The voice acquisition prompts include at least one of the following: sound wave effects, volume indication, and text transcription display during the voice recognition process.

[0135] According to another embodiment, a content recommendation device is provided. Figure 10 A schematic block diagram of the content recommendation device according to one embodiment is shown, the device being disposed in Figure 1 The server side in the illustrated architecture. For example... Figure 10 As shown, the device 1000 includes: a receiving unit 1001, a generating unit 1002, and a transmitting unit 1003. The main functions of each component are as follows: The receiving unit 1001 is configured to receive a first voice input sent by the user terminal, which is collected during the initial media content playback process on the content playback page. The generation unit 1002 is configured to perform semantic parsing on the first voice input data to obtain intent tags, and generate a first recommended content sequence based on the intent tags, wherein the first recommended content sequence contains at least two recommended content items; The sending unit 1003 sends the first recommended content sequence to the user terminal, so that the user terminal switches the initial media content played on the content playback page to play one of the recommended contents in the first recommended content sequence.

[0136] Optionally, the receiving unit 1001 is further configured to receive a second voice input sent by the user terminal, the second voice input being collected during the process of the user terminal playing a recommended content from the first recommended content sequence; The generation unit 1002 is further configured to perform semantic parsing on the second voice input to obtain a second intent tag, and generate a second recommended content sequence based on the context information of the second intent tag and the first recommended content sequence, wherein the second recommended content sequence contains at least two recommended contents; The sending unit 1003 is further configured to send the second recommended content sequence to the user terminal, so that the user terminal switches the currently playing recommended content to playing one of the recommended contents in the second recommended content sequence.

[0137] Optionally, the generation unit 1002 is specifically configured as follows: The first voice input is subjected to speech recognition to obtain the voice text. Perform semantic parsing on the speech text to determine the intent tag; Based on the intent tags, original recommendation data is obtained from the corresponding content sources to form the first recommended content sequence.

[0138] Optionally, the generation unit 1002 is specifically configured as follows: Perform semantic parsing on the spoken text; When the speech text is identified to contain a reference word referring to the initial media content, the contextual features of the initial media content at the moment when the first speech input is triggered are obtained; The intent label is determined by performing joint semantic parsing on the spoken text and the contextual features.

[0139] Optionally, the generation unit 1002 is specifically configured as follows: When the intent tag is a content recommendation intent, target media content that matches the voice text is obtained from the corresponding content source and used as the original recommendation data. When the intent tag is a local service intent, the user's location information is obtained, and based on the local service intent and the user's location information, service-related media content and service offer information that match the voice text are obtained from the corresponding content source as the original recommendation data; When the intent tag is a product query intent, the target product explanation content and purchase entry information that match the voice text are obtained from the corresponding data source as the original recommendation data; When the intent tag is an information query intent, the corresponding query result and the media content associated with the query result are obtained from the corresponding data source, and the information card containing the query result and the media content associated with the query result are used as the original recommendation data.

[0140] Optionally, the generation unit 1002 is specifically configured as follows: Determine the semantic correlation between the intent tag and the initial media content currently being played by the client; When the semantic relevance is lower than a preset threshold, the first recommended content sequence is generated; The method further includes: When the semantic relevance is not lower than the preset threshold, additional information matching the first voice input is generated; The additional information is sent to the user terminal so that the user terminal can overlay the additional information on the content playback page.

[0141] The various embodiments in this specification are described in a progressive manner. Similar or identical parts between embodiments can be referred to mutually. Each embodiment focuses on describing the differences from other embodiments. In particular, for system or device embodiments, since they are basically similar to method embodiments, the description is relatively simple, and relevant parts can be referred to the description of the method embodiments. The system and device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0142] It should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, data stored, data displayed, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties. Furthermore, the collection, use and processing of the relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and corresponding operation entry points are provided for users to choose to authorize or refuse.

[0143] In addition, embodiments of this application also provide a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0144] And an electronic device, comprising: One or more processors; and A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method described in any of the foregoing method embodiments.

[0145] This application also provides a computer program product, including a computer program that, when executed by a processor, implements the steps of the method described in any of the foregoing method embodiments.

[0146] in, Figure 11 The architecture of an electronic device is illustrated, which may include a processor 1110, a video display adapter 1111, a disk drive 1112, an input / output interface 1113, a network interface 1114, and a memory 1120. The processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, and memory 1120 can communicate with each other via a communication bus 1130.

[0147] The processor 1110 can be implemented using a general-purpose CPU, microprocessor, application-specific integrated circuit (ASIC), or one or more integrated circuits to execute relevant programs and implement the technical solution provided in this application.

[0148] The memory 1120 can be implemented in the form of ROM (Read Only Memory), RAM (Random Access Memory), static storage device, dynamic storage device, etc. The memory 1120 can store the operating system 1121 for controlling the operation of the electronic device 1100, and the basic input / output system (BIOS) 1122 for controlling the low-level operations of the electronic device 1100. Additionally, it can store a web browser 1123, a data storage management system 1124, and content recommendation devices 900 / 1000, etc. The aforementioned content recommendation devices 900 / 1000 can be the application programs that specifically implement the aforementioned steps in this embodiment. In summary, when implementing the technical solution provided in this application through software or firmware, the relevant program code is stored in the memory 1120 and executed by the processor 1110.

[0149] Input / output interface 1113 is used to connect input / output modules to realize information input and output. Input / output modules can be configured as components in the device (not shown in the figure) or externally connected to the device to provide corresponding functions. Input devices may include keyboards, mice, touch screens, microphones, various sensors, etc., and output devices may include displays, speakers, vibrators, indicator lights, etc.

[0150] Network interface 1114 is used to connect a communication module (not shown in the figure) to enable communication between this device and other devices. The communication module can communicate via wired means (such as USB, Ethernet cable, etc.) or wireless means (such as mobile network, WIFI, Bluetooth, etc.).

[0151] Bus 1130 includes a pathway for transmitting information between various components of the device, such as processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, and memory 1120.

[0152] It should be noted that although the above-described device only shows the processor 1110, video display adapter 1111, disk drive 1112, input / output interface 1113, network interface 1114, memory 1120, bus 1130, etc., in specific implementations, the device may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the above-described device may only include the components necessary for implementing the solution of this application, and does not necessarily include all the components shown in the figures.

[0153] As can be seen from the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware platforms. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a computer program product. This computer program product can be stored in a storage medium, such as ROM / RAM, magnetic disk, optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in various embodiments or some parts of the embodiments of this application.

[0154] The technical solutions provided in this application have been described in detail above. Specific examples have been used to illustrate the principles and implementation methods of this application. The descriptions of the above embodiments are only for the purpose of helping to understand the methods and core ideas of this application. Furthermore, those skilled in the art will recognize that, based on the ideas of this application, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of this application.

Claims

1. A content recommendation method, characterized in that, include: Play the initial media content on the content playback page; In response to receiving a first voice input, the initial media content is switched to play one of the recommended contents in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained by semantic parsing the first voice input, and the first recommended content sequence contains at least two recommended contents.

2. The content recommendation method according to claim 1, characterized in that, The method further includes: In response to a swipe operation on the currently displayed recommended content on the content playback page, the previous or next recommended content in the first recommended content sequence is switched.

3. The content recommendation method according to claim 1, characterized in that, The method further includes: When playing a recommended content from the first recommended content sequence, in response to receiving a second voice input, the currently playing recommended content is switched to playing a recommended content from the second recommended content sequence; wherein, the second recommended content sequence is generated based on the intent tag obtained by semantic parsing the second voice input and the context information of the first recommended content sequence, and the second recommended content sequence contains at least two recommended content items.

4. The method according to claim 1, characterized in that, The step of switching the initial media content to play one of the recommended contents in the first recommended content sequence in response to receiving the first voice input includes: In response to receiving a first voice input, the system sends the first voice input to the server and receives response data returned by the server in response to the first voice input; wherein, when the semantic correlation between the intent tag and the initial media content is lower than a preset threshold, the response data is the first recommended content sequence, and when the semantic correlation between the intent tag and the initial media content is not lower than the preset threshold, the response data is additional information matching the first voice input; When the response data is the first recommended content sequence, the initial media content is switched to play one of the recommended contents in the first recommended content sequence; The method further includes: When the response data is the additional information, the additional information is overlaid on the content playback page.

5. The content recommendation method according to any one of claims 1 to 4, characterized in that, The intent tags include at least one of the following: content recommendation intent, product query intent, local service intent, and information query intent.

6. The content recommendation method according to claim 5, characterized in that, If the intent tag represents the content recommendation intent, then the recommended content is media content, and the display format of the recommended content includes: The recommended content is played on the content playback page.

7. The content recommendation method according to claim 5, characterized in that, If the intent tag represents the product query intent, then the display format of the recommended content includes: Play a video explaining the target product on the content playback page; In addition, purchase entry information is superimposed on the playback interface of the product explanation video; wherein, the purchase entry information includes at least one of the following: product price information, purchase button, product details page link or add to cart entry.

8. The content recommendation method according to claim 5, characterized in that, If the intent tag is the local service intent, then the display format of the recommended content includes: Play service-related media content on the content playback page; In addition, service discount information is overlaid on the playback interface of the media content associated with the service; wherein the service discount information includes at least one of the following: group purchase discount information, store location information, distance information, coupon information, voucher information, or discount activity information.

9. The content recommendation method according to claim 5, characterized in that, If the intent tag represents the information query intent, then the display format of the recommended content includes: An information card is displayed on the content playback page, the information card containing query results for the first voice input; Furthermore, on the content playback page, media content associated with the query results is arranged sequentially after the information card, and the associated media content is played in response to the user's swipe operation.

10. The content recommendation method according to claim 1, characterized in that, The response to receiving the first voice input includes: In response to a triggering operation on the first voice input control on the content playback page, a voice acquisition prompt message is displayed on the content playback page, and the first voice input is acquired; The voice acquisition prompts include at least one of the following: sound wave effects, volume indication, and text transcription display during the voice recognition process.

11. A content recommendation device, characterized in that, The device includes: The page display unit is configured to play initial media content on the content playback page; The monitoring unit is configured to monitor voice input on the content playback page; The page display unit is further configured to: in response to receiving a first voice input, switch the initial media content to play a recommended content in a first recommended content sequence; wherein the first recommended content sequence is generated based on intent tags obtained by semantic parsing the first voice input, and the first recommended content sequence contains at least two recommended content items.

12. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the steps of the method according to any one of claims 1 to 10.

13. An electronic device, characterized in that, include: One or more processors; as well as A memory associated with the one or more processors, the memory being used to store program instructions that, when read and executed by the one or more processors, perform the steps of the method according to any one of claims 1 to 10.

14. A computer program product, comprising a computer program, characterized in that, When executed by a processor, the computer program implements the steps of the method according to any one of claims 1 to 10.