Commodity explanation card pushing method and device, equipment and storage medium

Abstract

The application discloses a commodity explanation card pushing method, device and equipment and a storage medium. The method comprises the following steps: acquiring cyclic live video data, and generating structured identification data corresponding to a video playing time axis; performing semantic identification processing based on the structured identification data and commodity information associated with live content, and determining a commodity explanation time interval corresponding to a commodity explanation event; constructing commodity explanation time sequence data based on the commodity explanation time interval; in a cyclic live playing process, acquiring a current playing progress, and matching the current playing progress with the commodity explanation time sequence data to determine a corresponding target commodity identifier; and pushing a corresponding commodity explanation card to a live room front end based on the target commodity identifier. The application realizes automatic correspondence and automatic pushing of the commodity explanation card and the cyclic live playing content, improves the accuracy of commodity explanation card pushing, and improves the timeliness of commodity information display and the efficiency of user acquisition of commodity information in a cyclic live scene.

Description

Technical Field

[0001] This invention relates to the field of computer technology, and in particular to a method, apparatus, device, and storage medium for pushing product information cards. Background Technology

[0002] With the development of internet live streaming technology, loop live streaming, due to its ability to repeatedly play pre-recorded video content, has been widely used in e-commerce live streaming scenarios. It offers advantages such as long-term unattended operation, content reuse, and low operating costs. Loop live streaming typically provides viewing services to users who enter the live stream at different times by continuously playing pre-recorded live videos.

[0003] Existing live streaming systems primarily implement functions such as looping video content and displaying product lists. When a user enters the live stream, due to the randomness of the entry time, the user can usually only watch the currently playing video content and cannot quickly obtain information about the product being explained. To view product information, users typically need to actively open the product list to search, resulting in a long path to obtain product information and impacting the user's information acquisition efficiency.

[0004] To improve the display of product information, some looping live streaming scenarios use product explanation cards for supplementary display. Current methods for pushing product explanation cards typically rely on manual operation or pre-configured fixed push times. When the live video content is adjusted, the playback rhythm changes, or the live video is replaced, the pre-configured push times can easily deviate from the actual playback content. This results in the product explanation cards not corresponding to the currently playing content, affecting the display of product information and increasing live streaming operation costs.

[0005] Therefore, existing technologies struggle to balance the accuracy and automation of product description card delivery.

[0006] It should be noted that the information disclosed in the background section above is only used to enhance the understanding of the background of the present invention, and therefore may include information that does not constitute prior art known to those skilled in the art. Summary of the Invention

[0007] In view of the problems in the prior art, the purpose of this invention is to provide a method, apparatus, device and storage medium for pushing product explanation cards, so as to solve the problem that the prior art is unable to balance the accuracy and automation of pushing product explanation cards.

[0008] The first aspect of this disclosure provides a method for pushing product explanation cards, which includes: Acquire looping live video data and generate structured recognition data corresponding to the video playback timeline based on the looping live video data; Based on structured recognition data and product information associated with the live stream content, semantic recognition processing is performed to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data. Product explanation time series data is constructed based on the product explanation time interval. The product explanation time series data includes at least the product identifier and the corresponding explanation start time and end time. During the looped live stream playback, the current playback progress is obtained, and the current playback progress is matched with the product explanation time sequence data to determine the target product identifier corresponding to the current playback progress; based on the target product identifier, the corresponding product explanation card is pushed to the front end of the live stream.

[0009] In some implementations, before performing semantic recognition processing based on structured recognition data and product information associated with the live stream content, the product explanation card push method further includes: Extract the video file identifiers corresponding to the looping live video data, and search the preset video recognition record table based on the video file identifiers to determine whether there is already generated product explanation time sequence data in the looping live video data; if it is determined that there is no already generated product explanation time sequence data, then perform semantic recognition processing based on the structured recognition data and the product information associated with the live content.

[0010] In some implementations, structured recognition data corresponding to the video playback timeline is generated based on the looping live video data, including: Extract the audio content stream of the looping live video data; submit the audio content stream to the asynchronous speech recognition engine for segmented recognition, and output text fragments containing timestamp information as structured recognition data. Each record in the structured recognition data includes at least: text content, fragment start timestamp, and fragment end timestamp.

[0011] In some implementations, semantic recognition processing is performed based on structured recognition data and product information associated with the live stream content to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data, including: The structured recognition data and product information are input into an inference engine containing at least two large language models. The contextual semantic analysis is performed in parallel by the at least two large language models to obtain the inference judgment results output by the at least two large language models respectively. When the inference judgment results output by the at least two large language models meet the preset consistency consensus conditions, the corresponding product explanation event is determined to be valid, and the corresponding product explanation time interval and corresponding product identifier are extracted based on the inference judgment results.

[0012] In some implementations, product explanation time-series data is constructed based on the product explanation time interval, including: The time intervals for explaining adjacent products are merged. If the same product identifier appears consecutively in multiple adjacent time intervals for explaining products, the multiple time intervals for explaining products are merged into a single consecutive time interval node. The start time of the merged time interval is taken from the start time of the first product time interval, and the end time is taken from the end time of the last product time interval.

[0013] In some implementations, constructing product explanation time-series data based on product explanation time intervals also includes: Call the short-segment noise filtering operator to calculate the duration of each product explanation time interval before or after merging; when the duration is lower than the preset time threshold, delete the corresponding product explanation time interval.

[0014] In some implementations, the target product identifier corresponding to the current playback progress is determined by matching the current playback progress with the product explanation time sequence data, including: The system captures the actual playback relative timestamp of the live stream as the current playback progress and compares the actual playback relative timestamp with the start and end times of the product explanation time sequence data. When the actual playback relative timestamp is detected to be within the explanation time interval corresponding to any product identifier, the corresponding product identifier is extracted as the target product identifier.

[0015] In some implementations, the product information card delivery method further includes: After pushing the corresponding product explanation card to the front end of the live broadcast room based on the target product identifier, the subsequent product explanation cards are pushed through a delayed message chain scheduling mechanism based on the relative time difference between the current playback progress and the start time of the next product explanation in the product explanation time sequence data.

[0016] In some implementations, a delayed message chaining scheduling mechanism is used to drive the subsequent product description cards, including: Calculate the time difference between the current playback progress and the start time of the next product's explanation; deliver the task message containing the next product push instruction as a delayed message to the message queue, and set the delayed trigger time of the delayed message to the time difference; when the delayed message reaches the trigger node, control the message queue consumer to execute the card push, and continue to calculate the push schedule for the next product to form a chained push process.

[0017] In some implementations, the product description card push method further includes the following during the cyclical push process: Continuously monitor the playback stream heartbeat in the live room to obtain the latest actual playback progress; when the latest actual playback progress does not match the expected trigger time of the current delayed message, automatically discard the delayed message that is waiting in the message queue to trigger the chain reset of the message queue.

[0018] A second aspect of this disclosure provides a product information card delivery device, comprising: The structured recognition module acquires looping live video data and generates structured recognition data corresponding to the video playback timeline based on the looping live video data. The semantic recognition module performs semantic recognition processing based on structured recognition data and product information associated with the live stream content to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data. The time sequence construction module constructs product explanation time sequence data based on the product explanation time interval. The product explanation time sequence data includes at least the product identifier and the corresponding explanation start time and end time. The playback matching module obtains the current playback progress during the looped live broadcast and matches it with the product explanation time sequence data to determine the target product identifier corresponding to the current playback progress. The card push module pushes corresponding product explanation cards to the front end of the live broadcast room based on the target product identifier.

[0019] A third aspect of this disclosure provides an electronic device, characterized in that it includes: a processor; and a memory storing executable instructions of the processor; wherein the processor is configured to execute the product explanation card push method of any of the above embodiments by executing the executable instructions.

[0020] The fourth aspect of this disclosure provides a computer-readable storage medium for storing a program that, when executed, implements the product explanation card push method of any of the above embodiments.

[0021] The fifth aspect of this disclosure provides a computer program product having a computer program stored thereon, characterized in that, when the computer program is executed by a processor, it implements the product explanation card push method of any of the above embodiments.

[0022] The product information card push method, apparatus, device, and storage medium disclosed herein have the following advantages: Through the above steps, this embodiment first generates structured recognition data corresponding to the video playback timeline based on the looped live video data, and identifies the product explanation time interval corresponding to the product explanation event based on the structured recognition data; further, it constructs product explanation time sequence data based on the product explanation time interval, so that the product explanation content in the looped live video establishes a corresponding time sequence relationship; then, during the looped live playback, it matches the current playback progress with the product explanation time sequence data, determines the target product identifier corresponding to the currently played content, and pushes the corresponding product explanation card accordingly, thereby forming an automatic product explanation card push process based on the playback content.

[0023] Therefore, compared with the existing technology that mainly relies on manual operation or fixed push time to display product explanation cards, the present invention establishes a correspondence between the looped live video playback content and the product explanation, and determines the corresponding target product identifier based on the current playback progress. This achieves automatic correspondence and automatic push of product explanation cards and looped live playback content, which can improve the accuracy of product explanation card push, reduce manual intervention, and improve the timeliness of product information display and the efficiency of users obtaining product information in looped live scenarios.

[0024] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and are not intended to limit this disclosure. Attached Figure Description

[0025] Other features, objects, and advantages of the invention will become more apparent from the following detailed description of non-limiting embodiments with reference to the accompanying drawings.

[0026] Figure 1 A flowchart illustrating a method for pushing product information cards as provided in this disclosure; Figure 2 A flowchart illustrating another product information card delivery method provided in this disclosure; Figure 3 This is a structural block diagram of a product explanation card pushing device according to an embodiment of the present disclosure; Figure 4 This is a schematic diagram of the hardware structure of an electronic device suitable for implementing a product explanation card push method according to an embodiment of the present disclosure. Detailed Implementation

[0027] To make the technical solution, the technical problem solved, and the technical effects of the present invention clearer, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the described embodiments are merely exemplary embodiments of the present invention and are not intended to limit the scope of protection of the present invention. Those skilled in the art can make various modifications and variations to the embodiments without departing from the spirit and scope of the present invention, and all such modifications and variations should be considered to fall within the scope of protection of the present invention.

[0028] Techniques, methods, and apparatus known to those skilled in the art are not discussed in detail where appropriate, but should be considered part of this specification. In all examples shown herein, any specific numerical values ​​or parameters should be interpreted as exemplary only and not as limiting the invention.

[0029] Figure 1 This document presents a flowchart illustrating a method for pushing product explanation cards according to an embodiment of the present disclosure. The method can be implemented by a product explanation card push system, which can be deployed on a server, cloud computing platform, or other computing device with data processing capabilities. Through data interaction with a live streaming platform, it achieves automatic pushing of product explanation cards.

[0030] The automatic product description card push method provided in this disclosure can be applied to looping live streaming scenarios. During looping live streaming, the live video plays in a loop according to preset playback content, and users can enter the live streaming room to watch the live content at any time. The product description card push system can automatically determine the product corresponding to the currently playing content based on the playback content of the looping live video, and push the corresponding product description card to the front end of the live streaming room.

[0031] like Figure 1 As shown, the methods for pushing product description cards include, but are not limited to, the following steps: Step 110: Obtain looping live video data and generate structured recognition data corresponding to the video playback timeline based on the looping live video data; Step 120: Based on the structured recognition data and the product information associated with the live broadcast content, perform semantic recognition processing to determine the product explanation time interval corresponding to the product explanation event in the looped live broadcast video data; Step 130: Construct product explanation time series data based on the product explanation time interval. The product explanation time series data shall include at least the product identifier and the corresponding explanation start time and end time. Step 140: During the looped live broadcast, obtain the current playback progress, and match the current playback progress with the product explanation time sequence data to determine the target product identifier corresponding to the current playback progress; Step 150: Based on the target product identifier, push the corresponding product description card to the front end of the live broadcast room.

[0032] In this embodiment of the disclosure, a corresponding product explanation card is obtained based on the determined target product identifier and pushed to the front end of the live broadcast room, so that the currently played content and the product explanation card are in correspondence, so that users can obtain information about the currently explained product while watching the looped live broadcast.

[0033] Through the above steps, this embodiment first generates structured recognition data corresponding to the video playback timeline based on the looped live video data, and identifies the product explanation time interval corresponding to the product explanation event based on the structured recognition data; further, it constructs product explanation time sequence data based on the product explanation time interval, so that the product explanation content in the looped live video establishes a corresponding time sequence relationship; then, during the looped live playback, it matches the current playback progress with the product explanation time sequence data, determines the target product identifier corresponding to the currently played content, and pushes the corresponding product explanation card accordingly, thereby forming an automatic product explanation card push process based on the playback content.

[0034] Therefore, compared with the existing technology that mainly relies on manual operation or fixed push time to display product explanation cards, the present invention establishes a correspondence between the looped live video playback content and the product explanation, and determines the corresponding target product identifier based on the current playback progress. This achieves automatic correspondence and automatic push of product explanation cards and looped live playback content, which can improve the accuracy of product explanation card push, reduce manual intervention, and improve the timeliness of product information display and the efficiency of users obtaining product information in looped live scenarios.

[0035] In this embodiment of the disclosure, in step 110, looping live video data is acquired, and structured recognition data corresponding to the video playback timeline is generated based on the looping live video data, including but not limited to the following processing steps: First, acquire looping live video data. Looping live video data can include looping live video files and their corresponding playback content information, which is used to represent the set of video data that is continuously played in a looping live streaming scenario.

[0036] Furthermore, audio content stream extraction is performed on the looping live video data to separate the corresponding audio content stream from the looping live video data. The audio content stream is used to represent the speech information content that changes over time in the looping live video.

[0037] Then, the audio content stream is submitted to the asynchronous speech recognition engine for segmentation and recognition processing, so as to perform time-based segmentation and recognition analysis on the audio content stream and output text segments corresponding to the video playback timeline.

[0038] The asynchronous speech recognition engine is used to perform non-real-time or near-real-time speech recognition processing on audio content streams and organizes the recognition results in a structured manner according to time sequence, thereby generating structured recognition data corresponding to the video playback timeline.

[0039] In this embodiment of the disclosure, the structured recognition data is represented in the form of text fragments, wherein each text fragment corresponds to a time interval on the video playback timeline and includes at least the following fields: The text content is used to represent the speech recognition results within this time interval; The segment start timestamp is used to indicate the starting position of the text content on the video playback timeline; The segment end timestamp is used to indicate the end position of the text content on the video playback timeline.

[0040] Through the above processing method, the looping live video data is converted into structured recognition data that is strictly aligned with the video playback timeline, thereby realizing the mapping relationship between video content and time-series text information.

[0041] In this embodiment of the disclosure, in step 120, semantic recognition processing is performed based on structured recognition data and product information associated with the live stream content to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data, including the following processing steps: First, obtain the structured recognition data generated in step 110, and obtain the product information corresponding to the looped live video data. The product information includes product identifier, product name, and keyword information used to describe product characteristics, which is used to represent the set of target products involved in the live content.

[0042] Furthermore, the structured recognition data is parsed line by line to obtain the text content corresponding to each text segment and its start and end timestamps on the video playback timeline, thereby obtaining text sequence information aligned with the video playback timeline.

[0043] Then, the structured recognition data is correlated with the product information to identify the semantic expression content related to the target product in the text sequence based on the semantic correspondence between the text content and the product information.

[0044] When multiple text fragments within a continuous time interval are detected to be semantically related to the same product information, the continuous time interval is determined as the product explanation event for the corresponding product; when a change is detected in the semantically related object corresponding to a text fragment, the time position corresponding to the change is taken as the boundary node of the product explanation event.

[0045] Based on the semantic recognition results above, the product explanation event is mapped to the corresponding product explanation time interval. The product explanation time interval is used to represent the start and end times of the explanation of the corresponding product in the looped live video.

[0046] Through the above processing method, a semantic correspondence is established between structured recognition data and product information, and semantic changes are mapped to interval divisions in the time dimension, thereby realizing the automatic recognition and time interval expression of product explanation events in looped live video.

[0047] In this embodiment of the disclosure, before performing semantic recognition processing based on structured recognition data and product information associated with the live stream content, a preprocessing step may be included to improve processing efficiency and avoid redundant calculations: First, extract the video file identifiers corresponding to the looping live video data. The video file identifiers are used to uniquely identify the current looping live video data.

[0048] Furthermore, based on the video file identifier, a preset video recognition record table is retrieved to obtain the corresponding historical processing record information. The video recognition record table is used to record whether the product explanation time sequence data of each loop of live video data has been generated and its storage status.

[0049] Then, based on the search results, determine whether there is already generated product explanation time-series data in the looped live video data: If product explanation timeline data already exists, directly obtain the product explanation timeline data for matching processing in subsequent playback stages; If no product explanation timeline data has been generated, the semantic recognition process will be initiated. Based on the structured recognition data and the product information associated with the live stream content, semantic recognition processing will be performed to generate product explanation events and corresponding product explanation time intervals.

[0050] The above processing method can determine the processing status of looping live video based on the video file identifier and avoid repeatedly performing semantic recognition calculations on already processed videos, thereby reducing computing resource consumption and improving processing efficiency.

[0051] In this embodiment of the disclosure, step 120, semantic recognition processing, can also be implemented by a semantic reasoning engine that includes at least two large language models.

[0052] Specifically, structured recognition data and product information are input into the semantic reasoning engine, where at least two large language models perform parallel semantic analysis on the input data.

[0053] Each language model outputs inference and judgment results based on the same input. The inference and judgment results are used to characterize whether there is product explanation behavior and corresponding product identification information within the corresponding time interval.

[0054] Furthermore, a consistency judgment is made on the reasoning results output by at least two large language models. When the preset consistency consensus condition is met, the product explanation event is determined to be valid.

[0055] In this embodiment of the disclosure, the consensus condition is used to characterize that the judgment results of multiple large language models on the same product identifier and the corresponding product explanation time interval are consistent.

[0056] When a product explanation event is established, the corresponding product explanation time interval and product identifier are determined based on the consistent reasoning results and output for subsequent processing.

[0057] By adopting the above processing method, a parallel independent reasoning mechanism of multiple large language models is introduced, and the stability and accuracy of semantic recognition results are improved through consensus judgment, thereby improving the reliability of determining the product explanation time interval.

[0058] In this embodiment of the disclosure, step 130, which constructs product explanation time-series data based on the product explanation time interval, includes at least the product identifier and the corresponding explanation start and end times, and includes the following processing steps: First, obtain the product explanation time intervals determined in step 120 and their corresponding product identification information, and then organize each product explanation time interval in a structured manner based on the product identification to form initial product explanation time sequence data.

[0059] Furthermore, adjacent time intervals are merged for product explanation time intervals: when the same product identifier appears consecutively in multiple adjacent product explanation time intervals, the multiple product explanation time intervals are merged into a single consecutive explanation time sequence node. The start time of the merged time interval is taken from the start time of the first product explanation time interval, and the end time is taken from the end time of the last product explanation time interval.

[0060] Through the above merging process, multiple discrete time intervals corresponding to the same product during continuous explanation are unified into a continuous time series expression, thereby improving the continuity and consistency of the product explanation time series data.

[0061] Furthermore, after generating the merged product explanation time-series data, short-segment noise filtering is performed on the product explanation time intervals or merged explanation time-series nodes: Specifically, the duration of each product explanation time interval or explanation time sequence node is calculated. When the duration is lower than the preset time threshold, the corresponding product explanation time interval is determined to be an invalid noise interval and is deleted from the product explanation time sequence data.

[0062] By filtering as described above, abnormal explanation segments with excessively short durations are removed, thereby improving the stability and reliability of the product explanation time-series data.

[0063] Finally, the product explanation time intervals after merging and noise filtering are structured and stored to generate product explanation time series data. The product explanation time series data is used to represent the start time, end time and product identification information of each product in the looped live video.

[0064] In this embodiment of the disclosure, step 140, which involves acquiring the current playback progress during the looped live broadcast and matching it with the product explanation timeline data to determine the target product identifier corresponding to the current playback progress, includes the following processing steps: First, the actual playback relative timestamp of the looping live video is obtained in real time. The actual playback relative timestamp is used to represent the corresponding time point of the current playback position in the video playback timeline and is used as the current playback progress.

[0065] Furthermore, the actual playback relative timestamp is matched with the product explanation time intervals in the product explanation time sequence data to determine whether the actual playback relative timestamp falls within any product explanation time interval.

[0066] When the actual playback timestamp is detected to be within a certain product explanation time interval, the product identifier corresponding to that product explanation time interval is determined as the target product identifier.

[0067] In this embodiment of the disclosure, before the product description time series data is used for interval matching, a structure optimization process can be performed to improve matching stability. The structure optimization process includes: The time intervals for explaining adjacent products are merged. When the same product identifier appears consecutively in multiple adjacent time intervals for explaining products, the multiple time intervals for explaining products are merged into a single consecutive time interval node. The start time of the merged product is taken from the start time of the first time interval, and the end time is taken from the end time of the last time interval. Furthermore, short-segment noise filtering is performed on the merged explanation time sequence nodes. When the duration of a node is lower than a preset time threshold, the corresponding time interval is deleted from the product explanation time sequence data.

[0068] By using the above processing methods, the product explanation timing data is optimized for continuity and noise is removed before the playback progress is matched, thereby improving the stability and accuracy of target product identifier matching and enhancing the reliability of product explanation card triggering.

[0069] In this embodiment, the processing of looped live streaming includes two stages: triggering the start of the broadcast and driving the playback. During the start of the broadcast, the product explanation card push process is initialized. During the playback stage, the real-time playback progress and product explanation timing data are continuously matched to achieve dynamic updates of the product explanation cards.

[0070] In this embodiment of the disclosure, step 150, which involves pushing the corresponding product description card to the front end of the live stream based on the target product identifier, includes the following processing steps: First, the target product identifier determined in step 140 is received, and a preset product information database is queried based on the target product identifier to obtain product description card data corresponding to the target product identifier. The product description card data includes at least the product name, product image information, and product display link information.

[0071] Furthermore, the product description card data is encapsulated into a push message format that can be recognized by the front end. The push message is used to represent the display information of the target product in the current live broadcast state.

[0072] Then, the push message is sent to the front-end display channel of the live broadcast room to trigger the front-end interface to display the corresponding product explanation card, so that users can obtain the corresponding product information in real time based on the currently playing content while watching the looping live video.

[0073] In this embodiment, the timing of pushing product explanation cards corresponds to the current playback progress and product explanation timing data. When a change in the target product identifier is detected, the pushed product explanation card is updated accordingly to achieve dynamic synchronization between product information and video playback content.

[0074] By employing the above processing methods, the content displayed on the front end of the live stream can maintain real-time consistency with the actual product explanation process in the looped live stream video, thereby improving the efficiency of users in obtaining product information and their viewing experience, and enhancing the accuracy and timeliness of product information display.

[0075] In this embodiment of the disclosure, after the product explanation card corresponding to the target product identifier is pushed, a chain scheduling process for subsequent product explanation cards can also be performed.

[0076] Specifically, after pushing the corresponding product explanation card to the front end of the live broadcast room based on the target product identifier, the system calculates the scheduling basis for subsequent pushes based on the time interval between the current playback progress and the start time of the next product explanation in the product explanation time sequence data, and drives the push of subsequent product explanation cards through a delayed message chain scheduling mechanism.

[0077] The above methods enable the push of product explanation cards to be continuously scheduled based on the product explanation time sequence data, thereby achieving automated continuous push in the scenario of looped live streaming.

[0078] In this embodiment of the disclosure, the subsequent product description card push is driven by a delayed message chain scheduling mechanism, including the following processing steps: First, calculate the time difference between the current playback progress and the start time of the next product explanation in the product explanation time sequence data. The time difference is used to represent the time interval between the current playback state and the next product explanation node.

[0079] Furthermore, a task message containing the next product push instruction is generated based on the time difference, and the task message is delivered to the message queue system as a delayed message. At the same time, the trigger time of the delayed message is set to the delay duration corresponding to the time difference.

[0080] When the delayed message meets the triggering condition, the message queue consumer consumes and processes the task message and executes the push operation of the corresponding product explanation card.

[0081] Furthermore, after completing the push for the current node, the time difference corresponding to the next product explanation node is calculated based on the product explanation time sequence data, and a new delayed message is generated, thus forming a continuous chain-like push process.

[0082] By using the above processing method, the product description card push process is expanded from a single trigger to a chain execution structure based on delayed messages, thereby improving the continuity and automation of push scheduling.

[0083] In this embodiment of the disclosure, during the cyclic push process, the delayed message chain scheduling process can also be dynamically calibrated to improve the consistency between the push and the actual playback progress.

[0084] Specifically, the system continuously monitors the playback stream heartbeat information in the live stream to obtain the latest actual playback progress. The playback stream heartbeat information is used to reflect the real-time changes in the current live stream playback status.

[0085] When the latest actual playback progress is detected to be mismatched with the expected trigger time corresponding to the current delayed message, it is determined that the current chain scheduling has deviated, and the delayed messages in the waiting state in the message queue are automatically discarded.

[0086] Furthermore, after discarding the current delayed message, the system recalculates the next product explanation node in the product explanation timing data based on the latest playback progress and regenerates the delayed message to trigger the reset execution of the chained scheduling process.

[0087] By employing the above processing methods, the delayed message chain scheduling mechanism can adaptively calibrate according to the dynamic changes in the live broadcast progress, thereby improving the consistency and accuracy between the product explanation card push and the actual broadcast content.

[0088] Figure 2 The flowchart illustrates a specific implementation of the product explanation card push method provided by this disclosure. The product explanation card push method in a loop live streaming scenario includes three stages: video preprocessing stage, product explanation timing construction stage, and product card automatic push stage.

[0089] In this embodiment of the disclosure, the video preprocessing stage is used to identify the status of looping live video data and generate structured data. Specifically, it includes the following processing steps: Step 201: Obtain looping live video data. The looping live video data includes at least the video file identifier, the live room identifier, and the video duration information.

[0090] Step 202: Query the video recognition record table based on the video file identifier to determine whether the product explanation time sequence data has been generated from the looped live video data.

[0091] Step 203: When it is determined that the product explanation timing data corresponding to the video file identifier already exists, the stored product explanation timing data is directly reused to end the preprocessing process.

[0092] Step 204: When it is determined that the product explanation time sequence data does not exist, mark the loop live video data as pending processing.

[0093] Step 205: Extract the audio content stream of the looping live video data.

[0094] Step 206: Create a speech recognition task based on the audio content stream and perform segmented recognition processing through an asynchronous speech recognition engine.

[0095] Step 207: Generate structured recognition data corresponding to the video playback timeline. The structured recognition data includes text content, segment start timestamp, and segment end timestamp.

[0096] The above processing method enables the conversion of looping live video data into structured timeline text data.

[0097] In this embodiment, the product explanation time sequence construction stage is used to generate product explanation time sequence data based on structured recognition data and product information. Specifically, the product explanation time sequence construction stage includes the following processing steps: Step 208: Obtain the product information corresponding to the looping live video data. The product information includes the product identifier, product name, and product feature keywords.

[0098] Step 209: Input the structured recognition data and product information into a semantic reasoning engine containing at least two large language models, and have the semantic analysis and processing performed in parallel by the at least two large language models.

[0099] Step 210: Perform a consistency check on the outputs of at least two large language models; Step 211: When the consensus conditions are not met, no record corresponding to the product explanation event is generated.

[0100] Step 212: When the consensus condition is met, generate a product explanation event and determine the corresponding product identifier and product explanation time range.

[0101] Step 213: Perform a merging process on adjacent product explanation time intervals. When the same product identifier appears consecutively in adjacent time intervals, the corresponding time intervals are merged into a continuous explanation time sequence node. The start time of the merged time interval is taken from the start time of the first time interval, and the end time is taken from the end time of the last time interval.

[0102] Step 214: Perform short segment filtering on the merged product explanation time intervals or explanation sequence nodes. When the duration is lower than the preset threshold, delete the corresponding time interval from the product explanation sequence data.

[0103] Step 215: Persistently store the processed product explanation timeline data to the product explanation timeline library.

[0104] The above processing method enables a structured temporal representation of product explanation events in a looping live video.

[0105] In this embodiment, the automatic product card push stage is used to dynamically push cards during the cyclical live broadcast based on product explanation time-series data. Specifically, the automatic product card push stage includes the following processing steps: Step 216: After the live stream starts, consume the live stream start message event to trigger the product card push process.

[0106] Step 217: Read the product explanation timing data corresponding to the current live broadcast room from the product explanation timing library.

[0107] Step 218: Determine whether the current actual playback progress is within the explanation time range of any product.

[0108] If the result is "yes", then proceed to step 219: push the product description card with the corresponding product identifier to the front end of the live broadcast room, and then proceed to step 220.

[0109] If the result is "no", proceed directly to step 220: calculate the push time of the next product description card.

[0110] Step 221: Deliver the delayed scheduling task as a delayed message to the message queue.

[0111] Step 222: When the delayed message meets the triggering condition, the delayed message is triggered.

[0112] Step 223: Determine whether the current live stream has been turned off.

[0113] If the result is "yes", the process ends.

[0114] If the result is "no", then proceed to step 224: automatically push the corresponding product explanation card to the front end of the live broadcast room.

[0115] After the card is pushed in step 224, the chain loop for the next push begins, and the process jumps back to step 220 to continue calculating the next push time point, thus forming a chain push process.

[0116] Through the above processing method, product explanation cards can be continuously and automatically pushed based on the product explanation time sequence data, and remain dynamically synchronized with the looped live broadcast playback process. This implementation method achieves automatic recognition and continuous push of product explanation cards in looped live broadcast scenarios through video preprocessing, product explanation time sequence construction, and chained message scheduling based on time sequence data.

[0117] In this disclosure, a product information card pusher device is also provided. For example... Figure 3 As shown, the product explanation card push device 300 can be deployed in a live streaming management server, a live streaming operation server, a cloud computing platform, an edge computing node, or other computing devices capable of performing cyclical live streaming services. The product explanation card push device 300 can be implemented through software, hardware, or a combination of software and hardware.

[0118] like Figure 3 As shown, the product explanation card push device 300 includes: a structured recognition module 310, a semantic recognition module 320, a time sequence construction module 330, a playback matching module 340, and a card push module 350.

[0119] The structured recognition module 310 is used to acquire looped live video data and generate structured recognition data corresponding to the video playback timeline based on the looped live video data.

[0120] The semantic recognition module 320, connected to the structured recognition module 310, is used to perform semantic recognition processing based on structured recognition data and product information associated with the live broadcast content, so as to determine the product explanation time interval corresponding to the product explanation event in the looped live broadcast video data.

[0121] The time sequence construction module 330 is connected to the semantic recognition module 320 and is used to construct product explanation time sequence data based on the product explanation time interval. The product explanation time sequence data includes at least the product identifier and the corresponding explanation start time and end time.

[0122] The playback matching module 340, connected to the timing construction module 330, is used to obtain the current playback progress during the looped live playback process, and match the current playback progress with the product explanation timing data to determine the target product identifier corresponding to the current playback progress.

[0123] The card push module 350 is connected to the playback matching module 340 and is used to push the corresponding product explanation card to the front end of the live broadcast room based on the target product identifier.

[0124] In this embodiment of the disclosure, the structured recognition module 310 can call the speech recognition service to recognize the audio content in the looped live video data and generate structured recognition data containing timestamp information.

[0125] The semantic recognition module 320 can call the semantic reasoning engine to identify product explanation events based on structured recognition data and product information, and output the product explanation time interval; the time sequence construction module 330 can merge, filter and organize the product explanation time interval to generate product explanation time sequence data.

[0126] The playback matching module 340 can perform interval matching based on the current playback progress of the looping live video and the product explanation time sequence data to determine the corresponding target product identifier. The card push module 350 can obtain the corresponding product explanation card data based on the target product identifier and send push messages to the front end of the live room to realize the dynamic display of the product explanation card.

[0127] Furthermore, in some embodiments, the product explanation card pusher 300 may also include a scheduling control module 360.

[0128] The scheduling control module 360 ​​is connected to the card push module 350. After the current product explanation card is pushed, it generates a delayed scheduling task based on the relative time difference between the current playback progress and the start time of the next product explanation in the product explanation time sequence data, and drives the chain push of subsequent product explanation cards through a message queue.

[0129] In some implementations, the scheduling control module 360 ​​can also continuously monitor the live stream playback status. When it detects a mismatch between the actual playback progress and the expected trigger time corresponding to the delayed scheduling task, the untriggered delayed scheduling task is discarded, and a new delayed scheduling task is generated based on the latest playback progress to maintain consistency between the product explanation card push and the looping live stream playback progress.

[0130] Through the coordinated operation of the above modules, the product explanation card push device 300 can realize product explanation recognition in looped live video, construction of product explanation time sequence data, playback progress matching, and automatic push of product explanation cards. It can also realize the continuous scheduling of subsequent product explanation cards based on changes in playback progress, thereby improving the timeliness and accuracy of product information display in looped live video scenarios.

[0131] It should be noted that the structured recognition module 310, semantic recognition module 320, temporal construction module 330, playback matching module 340 and card push module 350 mentioned above correspond to the implementation process of steps 110 to 150 in the method embodiment. The functions, processing flow and application scenarios implemented by them can be found in the corresponding method embodiment, and will not be repeated here.

[0132] Those skilled in the art will understand that the various embodiments of this disclosure can be implemented as systems, methods, apparatuses, electronic devices, or computer program products. Therefore, the various modules, units, apparatuses, or platforms in this disclosure can be implemented in a completely hardware manner, a completely software manner (including firmware, executable program instructions stored in memory, microcode, etc.), or a hardware-software co-implementation manner, and used to perform corresponding data processing operations.

[0133] Based on the same inventive concept, this disclosure also provides an electronic device, which includes: a processor; and a memory for storing executable instructions of the processor; wherein the processor is configured to perform the method described above by executing the executable instructions. Since the principle by which this electronic device embodiment solves the problem is similar to that of the above method embodiments, the implementation of this electronic device embodiment can refer to the implementation of the above method embodiments, and repeated details will not be described again.

[0134] Based on the same inventive concept, this disclosure also provides an electronic device, which is described below with reference to... Figure 4This invention describes an electronic device 400 according to an embodiment of the present disclosure. The electronic device 400 includes a processor (processing unit) 410 and a memory (storage unit) 420 for storing executable instructions of the processor. In specific industrial applications, the electronic device 400 can be embodied as a live streaming operation control host, a live streaming media processing server, or a central computing unit integrated into a live streaming automation control system.

[0135] like Figure 4 As shown, the electronic device 400 is presented in the form of a general-purpose computing device. The components of the electronic device 400 may include, but are not limited to: at least one processing unit 410, at least one storage unit 420, a bus 430 connecting different system components (including storage unit 420 and processing unit 410), a display unit 440, etc.

[0136] The storage unit stores program code, which can be executed by the processing unit 410 to perform the steps described in the "Exemplary Methods" section of this specification according to various exemplary embodiments of this disclosure. For example, the processing unit 410 can perform the above-described steps. Figure 1 The steps of the method shown.

[0137] Storage unit 420 may include readable media in the form of volatile storage units, such as random access memory (RAM) 421 and / or cache memory 422, and may further include read-only memory (ROM) 423.

[0138] Storage unit 420 may also include a program / utility 424 having a set (at least one) of program modules 425, including but not limited to: an operating system, one or more application programs, other program modules, and program data, each or some combination of these examples may include an implementation of a network environment.

[0139] Bus 430 can represent one or more of several bus structures, including a memory bus or memory controller, peripheral bus, graphics acceleration port, processing unit, or a local bus using any of the various bus structures. Electronic device 400 can also communicate with one or more external devices 470 (e.g., keyboard, pointing device, Bluetooth device, etc.), and with one or more devices that enable user interaction with electronic device 400, and / or with any device that enables electronic device 400 to communicate with one or more other computing devices (e.g., router, modem, etc.). This communication can be performed through input / output (I / O) interface 450. Furthermore, electronic device 400 can also communicate with one or more networks (e.g., local area network (LAN), wide area network (WAN), and / or public networks, such as the Internet) via network adapter 460. As shown, network adapter 460 communicates with other modules of electronic device 400 via bus 430. It should be understood that, although not shown in the figure, other hardware and / or software modules may be used in conjunction with electronic device 400, including but not limited to: microcode, device drivers, redundant processing units, external disk drive arrays, RAID systems, tape drives, and data backup storage systems.

[0140] In some embodiments, this disclosure also provides a computer program product, which includes a computer program that, when executed by a processor, implements the above-described product explanation card push method.

[0141] In some embodiments, this disclosure also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the above-described product explanation card push method.

[0142] Computer-readable storage media can be readable signal media or readable storage media. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), USB flash drive, portable hard disk, optical storage device, magnetic storage device, or any suitable combination of the foregoing.

[0143] Furthermore, computer-readable storage media may also include data signals propagated in baseband or as part of a carrier wave, carrying readable program code. Such propagated data signals may take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof.

[0144] A readable signal medium can be any readable medium other than a readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. Optionally, program code contained on a computer-readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, optical fiber, RF, etc., or any suitable combination thereof.

[0145] In practical implementation, program code for performing the operations of this disclosure can be written in any combination of one or more programming languages. These programming languages ​​include object-oriented programming languages—such as Java and C++—as well as conventional procedural programming languages—such as C or similar languages. The program code can execute entirely on the user's computing device, partially on the user's computing device, as a standalone software package, partially on the user's computing device and partially on a remote computing device, or entirely on a remote computing device or server. In cases involving remote computing devices, the remote computing device can be connected to the user's computing device via any type of network, including a local area network (LAN) or a wide area network (WAN), or it can be connected to an external computing device.

[0146] It should be noted that although several modules or units for the device used to perform actions have been mentioned in the detailed description above, this division is not mandatory. In fact, according to embodiments of this disclosure, the features and functions of two or more modules or units described above can be embodied in one module or unit. Conversely, the features and functions of one module or unit described above can be further divided and embodied by multiple modules or units.

[0147] Furthermore, although the steps of the method in this disclosure are described in a specific order in the accompanying drawings, this does not require or imply that the steps must be performed in that specific order, or that all the steps shown must be performed to achieve the desired result. Additional or alternative steps may be omitted, multiple steps may be combined into one step, and / or a step may be broken down into multiple steps.

[0148] Other embodiments of this disclosure will readily occur to those skilled in the art upon consideration of the specification and practice of the invention disclosed herein. This disclosure is intended to cover any variations, uses, or adaptations of this disclosure that follow the general principles of this disclosure and include common knowledge or customary techniques in the art not disclosed herein. The specification and examples are to be considered exemplary only, and the true scope and spirit of this disclosure are indicated by the appended claims.

Claims

1. A method for pushing product explanation cards, characterized in that, include: Acquire looping live video data, and generate structured recognition data corresponding to the video playback timeline based on the looping live video data; Based on the structured recognition data and the product information associated with the live stream content, semantic recognition processing is performed to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data. Based on the product explanation time interval, product explanation time sequence data is constructed, which includes at least the product identifier and the corresponding explanation start and end times; During the looped live broadcast, the current playback progress is obtained, and the current playback progress is matched with the product explanation time sequence data to determine the target product identifier corresponding to the current playback progress. Based on the target product identifier, a corresponding product description card is pushed to the front end of the live stream.

2. The product explanation card push method according to claim 1, characterized in that, Before performing semantic recognition processing based on the structured recognition data and product information associated with the live stream content, the product explanation card push method further includes: Extract the video file identifier corresponding to the looping live video data, and search the preset video recognition record table based on the video file identifier to determine whether the looping live video data has generated product explanation time sequence data; If it is determined that the generated product explanation timeline data does not exist, semantic recognition processing is performed based on the structured recognition data and the product information associated with the live broadcast content.

3. The product explanation card push method according to claim 1, characterized in that, The generation of structured recognition data corresponding to the video playback timeline based on the looped live video data includes: Extract the audio content stream of the looping live video data; The audio content stream is submitted to an asynchronous speech recognition engine for segmented recognition, and text fragments containing timestamp information are output as the structured recognition data. Each record in the structured recognition data includes at least: text content, fragment start timestamp, and fragment end timestamp.

4. The product explanation card push method according to claim 1, characterized in that, The step of performing semantic recognition processing based on the structured recognition data and product information associated with the live stream content to determine the product explanation time interval corresponding to the product explanation event in the looped live stream video data includes: The structured recognition data and the product information are input into an inference engine containing at least two large language models. The contextual semantic analysis is performed in parallel by the at least two large language models to obtain the inference judgment results output by the at least two large language models respectively. When the inference judgment results output by the at least two large language models meet the preset consistency consensus conditions, the corresponding product explanation event is determined to be valid, and the corresponding product explanation time interval and corresponding product identifier are extracted based on the inference judgment results.

5. The product explanation card push method according to claim 1, characterized in that, The step of constructing product explanation time-series data based on the product explanation time interval includes: The adjacent product explanation time intervals are merged. If the same product identifier appears consecutively in multiple adjacent product explanation time intervals, the multiple product explanation time intervals are merged into a continuous explanation time sequence node. The start time of the merged product explanation time interval is taken from the start time of the first product explanation time interval, and the end time is taken from the end time of the last product explanation time interval.

6. The product explanation card push method according to claim 5, characterized in that, The process of constructing product explanation time-series data based on the product explanation time interval also includes: Call the short-segment noise filtering operator to calculate the duration of each product explanation time interval before or after merging; When the duration is less than a preset time threshold, the corresponding product explanation time interval is deleted.

7. The product explanation card push method according to claim 1, characterized in that, The step of matching the current playback progress with the product description timeline data to determine the target product identifier corresponding to the current playback progress includes: The actual playback relative timestamp of the live broadcast room is captured as the current playback progress, and the actual playback relative timestamp is compared with the start and end times of the product explanation time sequence data. When it is detected that the actual playback relative timestamp is within the explanation time interval corresponding to any product identifier, the corresponding product identifier is extracted as the target product identifier.

8. The product explanation card push method according to claim 1, characterized in that, The method for pushing product description cards also includes: After pushing the corresponding product explanation card to the front end of the live broadcast room based on the target product identifier, the subsequent product explanation cards are driven to be pushed through a delayed message chain scheduling mechanism according to the relative time difference between the current playback progress and the start time of the next product explanation in the product explanation time sequence data.

9. The product explanation card push method according to claim 8, characterized in that, The method of driving subsequent product description cards through a delayed message chain scheduling mechanism includes: Calculate the time difference between the current playback progress and the start time of the next product's explanation; The task message containing the next product push instruction is delivered to the message queue as a delayed message, and the delayed trigger time of the delayed message is set to the time waiting difference; When the delayed message reaches the trigger node, the control message queue consumer executes the card push and continues to calculate the push schedule for the next product to form a chained push process.

10. The product explanation card push method according to claim 9, characterized in that, During the cyclical push process, the product explanation card push method also includes: Continuously monitor the playback stream heartbeat of the live stream to obtain the latest actual playback progress; When the latest actual playback progress does not match the expected trigger time of the current delayed message, the delayed messages waiting in the message queue are automatically discarded to trigger a chain reset of the message queue.

11. A product information card pusher device, characterized in that, include: The structured recognition module acquires looped live video data and generates structured recognition data corresponding to the video playback timeline based on the looped live video data. The semantic recognition module performs semantic recognition processing based on the structured recognition data and product information associated with the live broadcast content to determine the product explanation time interval corresponding to the product explanation event in the looped live broadcast video data. The time sequence construction module constructs product explanation time sequence data based on the product explanation time interval. The product explanation time sequence data includes at least the product identifier and the corresponding explanation start time and end time. The playback matching module acquires the current playback progress during the looped live broadcast and matches the current playback progress with the product explanation time sequence data to determine the target product identifier corresponding to the current playback progress. The card push module pushes the corresponding product explanation card to the front end of the live broadcast room based on the target product identifier.

12. An electronic device, characterized in that, include: processor; as well as A memory in which executable instructions of the processor are stored; The processor is configured to execute the product explanation card push method according to any one of claims 1 to 10 by executing the executable instructions.

13. A computer-readable storage medium for storing a program, characterized in that, When the program is executed, it implements the product explanation card push method according to any one of claims 1 to 10.

14. A computer program product having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the product explanation card push method according to any one of claims 1 to 10.

Images