Live content recommendation method and device, terminal equipment and storage medium

Abstract

The present disclosure provides a live content recommendation method and device, terminal equipment and computer storage medium, the method comprising: obtaining user search information corresponding to a time window; determining search parameters corresponding to different live projects based on the user search information; generating live content recommendation information corresponding to the time window according to the search parameters and supply state information of the corresponding live projects.

Description

Technical Field

[0001] This disclosure relates to the field of Internet technology, and in particular to a method, apparatus, terminal device, and storage medium for recommending live streaming content. Background Technology

[0002] In current live streaming systems, user search behavior data and live content supply data are typically stored in heterogeneous data spaces with independent data links. When processing content distribution, servers struggle to fine-tune the supply of live content based on real-time search data, often requiring large-scale, indiscriminate recommendation calculations. This easily leads to increased unnecessary computation, higher signaling overhead, and redundant use of data processing resources. Furthermore, the lack of an effective correlation indexing mechanism between user search requests and the real-time status of the live stream necessitates frequent responses to duplicate query requests, further exacerbating data processing pressure and response latency. Summary of the Invention

[0003] This disclosure provides a live streaming content recommendation method, apparatus, terminal device, and computer-readable storage medium to at least partially solve the aforementioned problems existing in the related technologies.

[0004] In a first aspect, this disclosure provides a method for recommending live streaming content, comprising: obtaining user search information corresponding to a time window; determining search parameters corresponding to different live streaming projects based on the user search information; and generating live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects.

[0005] Secondly, this disclosure provides a live streaming content recommendation device, comprising: an acquisition module for acquiring user search information corresponding to a time window; a determination module for determining search parameters corresponding to different live streaming projects based on the user search information; and a generation module for generating live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects.

[0006] Thirdly, this disclosure provides a terminal device, including a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements any of the above methods.

[0007] Fourthly, this disclosure provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements any of the above methods.

[0008] The live streaming content recommendation method, apparatus, terminal device, and computer-readable storage medium disclosed herein acquire user search information corresponding to a time window; determine search parameters for different live streaming projects based on the user search information; and generate live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects. In this way, determining the search parameters for different live streaming projects based on user search information within the time window and generating live streaming content recommendation information in conjunction with the supply status information of the corresponding live streaming projects helps reduce the subjective bias in content selection by broadcasters, improves the accuracy of matching live streaming content with user search needs, and helps reduce ineffective server computation and data interaction resource consumption caused by asynchronous supply and demand information. Attached Figure Description

[0009] To more clearly illustrate the technical solutions in the embodiments of this disclosure, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this disclosure. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0010] Figure 1 This is a schematic diagram of the hardware environment for a game live streaming system provided in one implementation of this disclosure. Figure 2 This is a flowchart illustrating the live content recommendation method provided in one implementation of this disclosure; Figure 3 This is a schematic diagram of the interface of the trending search list for live streaming projects provided in one implementation of this disclosure; Figure 4 This is a schematic diagram of the interface of a sub-project provided in one implementation of this disclosure; Figure 5 This is a schematic diagram of the live streaming room interface provided in one implementation of this disclosure; Figure 6 This is a schematic diagram of the interface of the live room recommendation card provided in one implementation of this disclosure; Figure 7 This is a schematic diagram of the live content recommendation device provided in one implementation of this disclosure; Figure 8 This is a schematic diagram of the hardware structure of the terminal device provided in one implementation of this disclosure. Detailed Implementation

[0011] To make the objectives, technical solutions, and advantages of the embodiments of this disclosure clearer, the technical solutions of this disclosure will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this disclosure, and not all embodiments. Based on the embodiments of this disclosure, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this disclosure.

[0012] For ease of understanding, the prior art involved in this disclosure will first be described in detail.

[0013] In current mainstream live streaming platforms (including game live streaming, e-commerce live streaming, etc.), there is a lack of real-time information exchange mechanisms between streamers and users, resulting in a significant information asymmetry between the supply of live streaming content and users' immediate needs. Before going live, streamers typically determine the direction of their live streaming content based on their accumulated experience or coarse-grained popular lists provided by the platform. These lists are usually compiled based on broad categories (such as "games" or "beauty"), failing to provide streamers with fine-grained information about user needs. Furthermore, in existing technology, the user's search function and the live streaming recommendation mechanism are two separate and independent live streaming systems. When a user enters keywords in the search box, the platform only matches the search terms based on historical content indexes or static tags, returning historical recordings or existing text and image results, without actively identifying whether there is currently ongoing live streaming content that highly matches the user's search intent. Meanwhile, when a user enters a live stream, the host cannot distinguish whether the user entered randomly or with a clear search intent, nor can they know the specific keywords the user searched for. As a result, the host cannot make targeted content responses or adjust interaction strategies to meet the user's real-time needs, which limits the audience retention rate and the depth of interaction.

[0014] To address the aforementioned issues, this disclosure proposes the following technical concept: By real-time collection and aggregation of full user search information corresponding to a time window, search parameters corresponding to different live streaming projects are extracted based on the live streaming project dimension. Combined with the current supply status information of each live streaming project (i.e., the number of live streaming rooms currently broadcasting related content), live streaming content recommendation information is generated. This recommendation information can, on the one hand, push dynamic trending search lists to the broadcaster's end, helping them perceive, in a data-driven manner, which specific live streaming projects have high search popularity but insufficient supply. On the other hand, the live streaming system simultaneously dynamically matches the user's real-time input search terms with the multi-source content feature terms of each live streaming room. During the user's typing search, a matching live streaming room recommendation card appears instantly below the search box. When a user enters a live streaming room through a recommendation card, the live streaming system generates a user intent identifier from their search keywords and transmits it to the broadcaster's interface for visualization. Through the above technical means, this disclosure constructs a two-way closed-loop traffic flow link of user search, broadcaster perception, content adjustment, real-time matching, and intent transmission, achieving dynamic and accurate matching of content supply and demand on the live streaming platform.

[0015] Based on the above, the specific details of each technical solution provided in this disclosure will be described in detail below.

[0016] First, the hardware environment required for this disclosure will be introduced.

[0017] In one optional implementation, the live streaming content recommendation method, apparatus, storage medium, and terminal equipment provided in this disclosure can be executed by a computer device, which can be a terminal or a server. The terminal can be a smartphone, tablet, laptop, smart TV, wearable smart device, smart vehicle terminal, etc., and may also include a client, which can be a live streaming client, browser client, instant messaging client, or mini-program. The server can be a standalone physical server, a server cluster consisting of multiple physical servers, or a distributed live streaming system. It can also be a cloud server providing basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0018] For example, when this disclosure is used on a terminal device, the terminal device may include a display screen and a processor. The display screen is used to present the live stream and receive instructions generated by the user (viewer / host) interacting with the live stream. The processor is used to store the live stream generated by the live streaming application, respond to instructions, and control the display of the live stream on the display screen. When the user operates the live stream through the display screen, the live stream can control the local content of the terminal device in response to the received operation instructions. The terminal device can provide the graphical user interface to the user in various ways, such as rendering the display on the terminal device's screen or presenting the graphical user interface through holographic projection.

[0019] For example, when this disclosure is run on a server, the method can be implemented and executed based on a cloud system. A cloud system refers to an operating mode based on cloud computing. A cloud system includes a server, a broadcast client, and client devices. The main body running the live streaming application and the main body presenting the live stream are separate. The storage and operation of the live streaming application are completed on the server, while the presentation of the live stream is completed on the broadcast client and client devices. The broadcast client and client devices are mainly used for receiving and sending live data and presenting the live stream. For example, the broadcast client and client devices can be display devices with data transmission capabilities located close to the user (viewer / broadcaster), such as mobile terminals, televisions, computers, PDAs, personal digital assistants, head-mounted displays, etc. However, the terminal device for processing live data is the server in the cloud. During live streaming, the user operates the broadcast client and client devices to send instructions to the server. The server controls the operation of the live stream according to the instructions, encodes and compresses the live stream data, returns it to the broadcast client and client devices via the network, and finally, the broadcast client and client devices decode and output the live stream.

[0020] For example, in conjunction with the above description, Figure 1 A live streaming system for implementing this disclosure is shown. The system may include at least one viewer device 10, at least one broadcaster device 40, at least one server 20, and a network 30. The server 20 is responsible for receiving search requests from the viewer device 10, aggregating and processing the search behavior of users across the platform in real time, generating live streaming content recommendation information, and pushing dynamic trending search lists to the broadcaster device 40 and live streaming room recommendation cards to the viewer device 10.

[0021] In the aforementioned live streaming system, users (viewers / hosts) can log in to the live streaming application using their registered live streaming accounts to participate in the live stream. During the user's participation in the live stream, the viewer's device 10 and the host's device 40 interact with each other through the server 20. The viewer's device 10 and the host's device 40 send various information to the server 20. The server 20 determines the display data in the viewer's device 10 and the host's device 40 based on the live streaming mechanism and the received information, and sends the display data to the viewer's device 10 and the host's device 40. The viewer's device 10 displays the display data sent by the server 20 to the viewer, and the host's device 40 displays the display data sent by the server 20 to the host.

[0022] In potential application scenarios, the viewer device 10 and the broadcaster device 40 in the live streaming system can communicate and connect with each other via different networks 30 and server 20. Network 30 can be a wireless network or a wired network, such as a wireless local area network (WLAN), local area network (LAN), cellular network, 2G network, 3G network, 4G network, 5G network, etc. Furthermore, the live streaming system may include at least one database for continuously storing live streaming-related information while different users are online.

[0023] It should be noted that, Figure 1 The illustrated live streaming system diagram is merely an example. The live streaming system described in this disclosure is intended to more clearly illustrate the technical solutions provided herein and does not constitute a limitation on the technical solutions provided herein. As those skilled in the art will recognize, with the evolution of live streaming systems and the emergence of new business scenarios, the technical solutions provided herein are equally applicable to similar technical problems. The hardware environment required by this disclosure has been introduced above. The following will, based on the aforementioned hardware environment, detail various embodiments of the interactive control method in games provided by this disclosure.

[0024] Please refer to Figure 2 This is a flowchart of a live content recommendation method provided in one implementation of this disclosure. As shown in the figure, the live content recommendation method includes the following steps: Step S1: Obtain user search information corresponding to the time window; The time window refers to the time unit used by the live streaming system to periodically aggregate and statistically analyze user search behavior, defining the time range boundaries of user search data participating in the search popularity calculation. In one optional implementation, the time window can be dynamically updated using a sliding window mechanism. For example, the live streaming system maintains a sliding time window of the most recent 5 minutes within a 5-minute statistical period, continuously incorporating the latest search behavior data and removing historical data outside the time range as time progresses, ensuring that the search popularity ranking always reflects the platform's latest real-time user search dynamics. In another optional implementation, the time window can also employ a fixed-period rolling window mechanism. For example, the live streaming system uses a 1-minute fixed statistical period, performing a complete batch aggregation calculation on the search data within that period every minute to generate the current period's popularity ranking results, and resetting the statistical data at the start of the next period to complete a new round of aggregation calculation.

[0025] User search information refers to the behavioral data records generated by all users of the live streaming platform performing search operations in the search box. It is the original data source for constructing the popularity ranking of live streaming content. In one optional implementation, user search information includes multiple dimensions of fields, such as: user account identifier (used to count the number of users searching), search keyword text (used for aggregating and analyzing search content), search timestamp (used to divide the time window range), search source client type (such as App, Web, etc.), and the user's current geographical location (used to support the generation of regionalized trending search lists). In another optional implementation, user search information may also include contextual data from the user's search, such as the type of live streaming room the user was watching before initiating the search, the user's historical search records, and the user's active time on the platform, to assist the live streaming system in making more accurate judgments in subsequent semantic analysis and intent understanding stages.

[0026] Step S2: Based on the user search information, determine the search parameters corresponding to different live streaming projects; In this context, "live stream project" refers to a category of live stream content within a live stream platform, categorized by content type or theme. It serves as the fundamental dimension for the live stream system to aggregate and calculate search popularity. In one optional implementation, live stream projects can be defined according to a primary category dimension. For example, in a game live stream platform, projects might include specific game names such as "Game Product A," "Game Product B," "Game Product C," and "Game Product D." In a comprehensive e-commerce live stream platform, projects might include major product categories such as "Beauty & Skincare," "Clothing & Accessories," "Food & Beverages," and "Digital Appliances." In another optional implementation, live stream projects can be further subdivided according to a secondary sub-project dimension. For instance, under the primary live stream project "Game Product A," specific sub-projects like "Hero A," "Hero B," "Match Commentary," and "Combo Tutorials" can be further subdivided, making the aggregation and analysis of search popularity more refined.

[0027] The search parameters refer to a set of numerical indicators calculated by the live streaming system based on user search information, used to quantitatively represent the current search popularity of a particular live streaming project. In an optional implementation, the search parameters may include two core indicators: the number of search users (i.e., the number of unique user accounts that searched for terms related to the live streaming project within a time window) and the search frequency (i.e., the total number of searches for terms related to the live streaming project within a time window). The number of search users reflects the breadth of the live streaming project's audience, while the search frequency reflects the intensity of the search for the live streaming project. These two indicators complement each other, together forming a comprehensive quantitative representation of the search popularity of the live streaming project. In an optional implementation, the search parameters can be further expanded by adding a dynamic indicator: the search popularity growth rate (i.e., the percentage increase in the number of search users or the search frequency within the current time window compared to the previous time window), to help broadcasters promptly capture emerging trending content.

[0028] Step S3: Based on the search parameters and the supply status information of the corresponding live streaming project, generate live streaming content recommendation information corresponding to the time window.

[0029] Supply status information refers to supply-side data obtained by the live streaming system through real-time statistics of the current live streaming rooms on the platform, used to measure the sufficiency of content supply for each live streaming project. In an optional implementation, the core indicator of supply status information is the number of live streaming rooms corresponding to each live streaming project, that is, the total number of live streaming rooms on the platform whose content matches the live streaming project at the current point in time. For example, in a game live streaming platform, the current number of live streaming rooms corresponding to the "Game Product A" project is 150, and the current number of live streaming rooms corresponding to the "Game Product B" project is 12. In an optional implementation, supply status information may also include quality weight information for live streaming rooms, by weighting and integrating indicators such as the number of online viewers and the streamer's historical content quality rating for each live streaming room to obtain the effective supply index of the live streaming project.

[0030] The live streaming content recommendation information refers to the output data packet generated by the live streaming system after integrating search parameters and supply status information, used to convey the matching relationship between the demand and supply of live streaming content to the broadcaster or viewer. In an optional implementation, the live streaming content recommendation information is configured as a trending list of live streaming projects pushed to the broadcaster. This list is sorted primarily based on the search parameters of each live streaming project, while also indicating the number of live streaming rooms corresponding to each project, allowing the broadcaster to easily identify content opportunities with high search popularity but few live streaming rooms.

[0031] In one specific application, a game streamer on a live streaming platform is considering the content direction for their broadcast today. The streamer opens the trending search list for live streaming items pushed by the system on the streamer's interface. The list displays the real-time search popularity ranking for each game item within the current time window: 1st place is "Game Product A" (3200 search users, 280 current live streams), 2nd place is "Game Product B" (2800 search users, 45 current live streams), and 3rd place is "Game Product C" (2100 search users, 18 current live streams). The streamer notices that "Game Product C" has a high search popularity, but the number of corresponding live streams is far less than the number of searchers, indicating a significant supply-demand gap. Based on their gaming expertise, the streamer chooses to focus on broadcasting "Game Product C," successfully achieving high organic search exposure and traffic in this niche area.

[0032] Through the steps described above, the live streaming content recommendation method provided in this disclosure aggregates and analyzes all user search information within a time window, determines search parameters based on live streaming projects, and generates live streaming content recommendation information by combining the supply status information of each live streaming project. This achieves real-time correlation analysis between user search behavior data and live streaming content supply data. This solution changes the traditional one-way flow of content supply and demand information on live streaming platforms, transforming the demand intent data generated by users during the search process into valuable reference signals for broadcaster content decisions. Simultaneously, it dynamically matches and recommends ongoing live streaming content with users' real-time search intent, fundamentally breaking down the information barrier between user search behavior and live streaming content supply, and improving the accuracy of content supply and demand matching on live streaming platforms.

[0033] Furthermore, in one embodiment of this disclosure, the step of determining the search parameters corresponding to different live streaming projects based on the user search information includes: The user search information is analyzed and processed to obtain the search terms corresponding to the user search information; The search terms are normalized to determine the standard terms corresponding to different search terms; Based on the first term data associated with the first standard term for different live streaming projects, the search parameters corresponding to different live streaming projects are determined.

[0034] The analysis and processing of user search information refers to the process by which the live streaming system preprocesses and semantically analyzes the collected raw user search behavior data to extract structured search term information. In an optional implementation, the analysis and processing first cleans the raw search input text, removing noise such as special symbols and HTML tags. Then, the cleaned text is segmented into meaningful word units. For example, if a user inputs "video tutorials on hero A's skills in game product A," the live streaming system will segment it into word units such as "game product A," "hero A," "skills," and "video." If a user inputs "how much does genuine SK-II cost," the system will segment it into word units such as "SK-II," "genuine," and "price." In an optional implementation, the analysis and processing may also include an intent recognition step, using a natural language processing model to analyze the semantic intent type of the user's search text, such as determining whether the search is for content discovery or product purchase. This allows for the allocation of different weights to search terms with different intent types in subsequent processing, improving the accuracy of search popularity aggregation.

[0035] Among them, search terms refer to words or phrases with independent semantics obtained after analyzing and processing user search information, and are the basic calculation unit for search popularity aggregation. In an optional implementation, search terms are divided into two categories according to semantic completeness: single-word terms (such as "Hero A" and "sunscreen") and phrase terms (such as "Hero A combo" and "sunscreen product recommendation"). When extracting search terms, the live streaming system prioritizes retaining phrase terms that fully express the search intent to improve the accuracy of subsequent normalization processing. In an optional implementation, search terms also carry corresponding contextual attributes, such as the user account ID to which the term belongs, the search timestamp of the term, and the type of the source client.

[0036] Normalization, in this context, refers to the data processing method that uses natural language processing (NLP) technology to map and merge multiple search terms with different expressions but identical or highly similar semantics into a single standard term. This aims to eliminate naturally occurring differences in expression during user searches and improve the accuracy and coverage of search popularity statistics. In one optional implementation, normalization is achieved through a thesaurus of word synonyms and a semantic similarity model. For example, if user A searches for "Hero A combo," user B searches for "Hero A combo," and user C searches for "Hero A skills," although their expressions differ, they all point to the content dimension of "Hero A - combo skills." The live streaming system uses semantic similarity calculation to identify the semantic equivalence among the three terms, unifying them into the standard term "Hero A - combo tutorial," and subsequently combining the search behaviors of the three users into the search popularity of the same standard term. In an alternative implementation, normalization can also handle abbreviations, aliases, and homophonic variations. For example, "rose quartz bracelet" and "rose quartz beaded bracelet" can be merged into a unified standard term to ensure that all expressions of the same content requirement are included in the same statistical unit.

[0037] In this context, standard terms refer to normalized terms that serve as the basic unit for search popularity statistics, determined after normalization. Each standard term corresponds to a specific live stream content theme or user demand dimension. In an optional implementation, standard terms are categorized and indexed according to the live stream project dimension. For example, "Game Product A - Hero A - Combo Tutorial" belongs to the "Game Product A" live stream project, and "Sunscreen Product - Brand A" belongs to the "Beauty and Skincare" live stream project. The live stream system pre-maintains a mapping table between live stream projects and standard terms. Through keyword matching and semantic classification models, each standard term is assigned to the most matching live stream project dimension, laying the data foundation for subsequent aggregation and calculation of search parameters by live stream project dimension.

[0038] The first standard term refers to the subset of standard terms identified by the live streaming system as having a correlation with the content of a specific live streaming project, out of the total number of standard terms. The first term data refers to the statistical summary value of the first standard term within a time window, including the number of users searching for that first standard term and the search frequency. For example, in the last 5 minutes, the first standard term "Game Product A - Hero A - Combo Tutorial" was searched by 458 different user accounts on the platform, with a total search frequency of 612. Therefore, the first term data for this first standard term would be: Number of users searching = 458, Search frequency = 612.

[0039] It should be added that, in another optional implementation, the live streaming system can also skip the semantic normalization step, directly perform synonym replacement and noise reduction on the original search keywords, and then perform matching and counting according to the preset live streaming project keyword dictionary, thereby quickly determining the search parameters with lower computational overhead, which is suitable for real-time popularity calculation in ultra-high concurrency scenarios.

[0040] In a specific application, a game live streaming platform collected 380,000 user search behavior records from across the platform in real time within a 5-minute time window at 3 PM on a certain afternoon. The live streaming system performed batch analysis on these 380,000 records, extracted search terms, and normalized them, ultimately identifying approximately 12,000 different standard terms. The live streaming system categorized the standard terms by live streaming project, identifying the first standard terms associated with "Game Product A" as "Game Product A - Hero A - New Hero" (number of search users: 8900, search frequency: 12400) and "Game Product A - Annual Tournament" (number of search users: 6200, search frequency: 8100), etc. The live streaming system aggregated the first term data of all the first standard terms associated with "Game Product A" and calculated the comprehensive search parameters for the "Game Product A" live streaming project within the current time window: a total of 42,000 search users and a total search frequency of 58,000 times, providing accurate quantitative data support for generating live streaming content recommendation information.

[0041] Through the steps described above, the live streaming content recommendation method provided in this disclosure transforms massive amounts of fragmented user search behavior data into well-organized and precisely dimensional search parameters by performing multi-level analysis and processing (text cleaning, word segmentation, normalization, and standard term generation) and refined aggregation statistics by live streaming project dimension. This eliminates the fragmentation problem of search terms caused by differences in expression habits among different users, significantly improves the coverage and accuracy of search popularity statistics, and lays a solid data foundation for generating high-quality live streaming content recommendation information in the future.

[0042] Further, in one embodiment of this disclosure, the first term data includes at least one of: the number of search users corresponding to the first standard term and the search frequency; the step of generating live content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live project includes: The different live streaming projects are sorted based on the number of search users corresponding to the first standard term and / or the search frequency; Based on the sorting of different live streaming projects and the number of live streaming rooms corresponding to each project, live streaming content recommendation information corresponding to the time window is generated.

[0043] The sorting of different live streaming projects based on the number of search users and / or search frequency refers to the process by which the live streaming system uses the accumulated search popularity value of each live streaming project within a time window as the sorting criterion, and arranges all live streaming projects in descending order of popularity to generate a live streaming project popularity ranking list. In an optional implementation, the sorting can be based solely on the number of search users. For example, the live streaming system can sort live streaming projects such as "Game Product A" (42,000 search users), "Game Product B" (38,000 search users), and "Game Product C" (31,000 search users) from highest to lowest search user count, prioritizing the live streaming projects with the widest search audience. In another optional implementation, the sorting can also be based on a weighted composite score of the number of search users and search frequency. For example, the live streaming system can calculate the composite search popularity score for each live streaming project by weighting and summing the number of search users (weight 0.6) and search frequency (weight 0.4). Introducing a weighted factor based on search frequency can effectively capture the signal of a few highly active users repeatedly searching for a hot topic, avoiding the omission of rapidly rising niche hot topics when relying solely on the number of users searching.

[0044] The number of live streaming rooms refers to the number of live streaming rooms that are currently broadcasting and match the content of a specific live streaming project on the platform at the current time. In one optional implementation, the number of live streaming rooms is counted through a live streaming room content index maintained in real time by the live streaming system. This index records the current set of content feature terms for each live streaming room. The live streaming system matches the content feature terms of the live streaming room with the keywords of the live streaming project to count the number of live streaming rooms broadcasting for each project in real time. For example, at the current time, there are 580 live streaming rooms broadcasting for the "Game Product A" project and 430 live streaming rooms broadcasting for the "Game Product B" project. In another optional implementation, the number of live streaming rooms can also be counted by quality stratification, such as separately counting the number of high-quality live streaming rooms (live streaming rooms with more than a certain threshold of online viewers) and the number of ordinary live streaming rooms, enabling streamers to more accurately assess the level of competition for high-quality content within a specific live streaming project.

[0045] The generation of live streaming content recommendation information based on the ranking of different live streaming projects and the number of live streaming rooms corresponding to each project refers to the live streaming system comprehensively associating search popularity ranking data with live streaming room supply data to generate recommended output content that simultaneously reflects both demand-side popularity and supply-side status quo information. In an optional implementation, the recommendation information is pushed to the broadcaster in the form of a live streaming project popularity ranking. Each live streaming project entry in the ranking simultaneously displays information on three dimensions: the project's search popularity ranking, the number of search users, and the current number of live streaming rooms. By comparing the ratio between the number of search users (demand) and the number of live streaming rooms (supply), the broadcaster can quickly identify the live streaming project with the highest ratio, representing a content blue ocean within the current platform. In an optional implementation, the live streaming system can also directly calculate a content opportunity index (i.e., the ratio of the number of search users to the number of live streaming rooms) for each live streaming project and use this index as an auxiliary ranking dimension to help broadcasters quickly locate the optimal content direction.

[0046] In a specific application, streamer A on a live streaming platform opens the trending search list interface on the streamer's app. The live streaming system displays the ranking of live streaming items within the current time window: 1st place "Beauty - Sunscreen" (search users: 15,600, current live streams: 89, opportunity index: 175); 2nd place "Clothing - Hanfu" (search users: 12,400, current live streams: 32, opportunity index: 388); 3rd place "Food - Low-calorie light meals" (search users: 9,800, current live streams: 11, opportunity index: 891); 4th place "Beauty - Lipstick" (search users: 18,200, current live streams: 460, opportunity index: 40). Streamer A notices that "Food - Low-calorie light meals" has a high opportunity index of 891, representing a typical blue ocean market with a severe imbalance between supply and demand. Based on this, streamer A adjusts their live streaming strategy for the day, successfully achieving high organic search exposure and traffic.

[0047] Through the steps described above, the live streaming content recommendation method provided in this disclosure comprehensively correlates user search popularity rankings (demand-side data) with the number of live streaming rooms (supply-side data) to generate live streaming content recommendation information that combines both popularity and supply-demand ratio dimensions. This enables streamers to comprehensively evaluate the content value of different live streaming projects from both absolute popularity and relative opportunity perspectives. It effectively solves the limitations of traditional popular lists that only reflect absolute popularity and ignore the current state of supply competition, helping streamers make content decisions based on more accurate data and maximize the value of limited live streaming time.

[0048] Furthermore, in one embodiment of this disclosure, the live streaming content recommendation information is configured as a live streaming project list pushed to the broadcaster's end. The live streaming project list is generated based on the sorting of different live streaming projects, and the live streaming project list displays the corresponding number of live streaming rooms associated with each live streaming project.

[0049] Among them, the live streaming project ranking list refers to the interactive interface generated by the live streaming system based on the popularity ranking of live streaming projects and pushed to the broadcaster in real time. It intuitively displays the search popularity ranking and supply status of each live streaming project in a list format.

[0050] In an optional implementation, the live stream project rankings are displayed as a card list in a fixed area (such as a sidebar or collapsible panel) of the broadcaster's interface. Each list entry corresponds to a live stream project, and the entry content includes the live stream project name, current popularity ranking (labeled with a numerical sequence), number of users searching or search popularity index, and the number of live stream rooms / live streamers currently corresponding to that project. Please refer to [reference needed]. Figure 3 This is a schematic diagram of the interface of the live streaming project hot search list provided in one implementation of this disclosure. The live streaming project hot search list is used to display the popularity information related to each live streaming project. The hot search information includes: the ranking number of each live streaming project, the name of the live streaming project, the search popularity value, and the number of current live streaming rooms, etc. Figure 3 As shown, the streamer's client displays a trending search list interface 300, which ranks the various game products 301 (i.e., live streaming projects) from 1st to 7th according to the current number of searchers 302 (i.e., search popularity). Simultaneously, the trending search list interface 300 displays the number of streamers 303 (i.e., the number of live streaming rooms) corresponding to each game product 301, determined based on the current live streaming situation on the platform. The content displayed on this trending search list interface 300 is automatically refreshed at a fixed period (e.g., every 1 minute), with animation effects showcasing the ranking changes during refresh, allowing streamers to intuitively perceive the dynamic trends in the popularity of each live streaming project.

[0051] In one optional implementation, the number of live streaming rooms displayed for each live streaming project in the live streaming project ranking list supports click-based interaction. When a streamer clicks on the number of live streaming rooms in a specific live streaming project entry, the live streaming system displays a preview of the list of currently broadcast live streaming rooms for that project. This allows streamers to not only see the overall supply quantity but also quickly understand the specific situation of competitor live streaming rooms, thereby making a more comprehensive assessment of the content competition landscape. In another optional implementation, the live streaming project ranking list can also intuitively distinguish live streaming projects with different supply and demand states using color coding. For example, live streaming project entries with an opportunity index higher than a certain threshold are marked with a green background, representing a large supply-demand gap and abundant content opportunities for that project; live streaming project entries with an opportunity index lower than a certain threshold are marked with a gray background, indicating to the streamer that the live streaming market in that direction has abundant supply and fierce competition.

[0052] In one specific application, streamer B in a gaming section checked the trending search list on the streamer's app 30 minutes before going live. They noticed that the live stream item "Game Product A - Global Finals" had consistently risen to the top spot in the last 20 minutes, with the number of search users rapidly increasing from an initial 8,000 to the current 32,000, while the corresponding number of live streams was only 28. The list marked this item with a green "Soaring Popularity" icon and a "Blue Ocean Opportunity" label. Based on this, streamer B judged that there was a strong demand from viewers for pre-tournament content and that the supply was insufficient. They then adjusted their broadcast plan, focusing on streaming the pre-tournament commentary for Game Product A.

[0053] Through the above steps, the live streaming content recommendation method provided in this disclosure presents the live streaming content recommendation information to the broadcaster in an intuitive and user-friendly live streaming project list. The list also displays the current number of live streaming rooms for each project, enabling broadcasters to obtain two key types of information—demand-side popularity ranking and supply-side competition status—through a single interface module. This significantly reduces the information acquisition cost for broadcasters when making content decisions and upgrades the broadcaster content decision-making model from guessing content based on experience to selecting content based on data.

[0054] Furthermore, in one embodiment of this disclosure, the method further includes: Identify the second standard term in the first standard term that is associated with a single live streaming project; Determine the first user search information corresponding to the second standard term from the user search information; Based on the third standard terms in the first user's search information, sub-projects for the single live streaming project are determined.

[0055] The second standard term refers to a subset of standard terms that the live streaming system further filters from all the first standard terms associated with a single live streaming project, and that directly correspond to the live streaming project name itself or its core identifier. In an optional implementation, the determination of the second standard term is based on the strong association between the standard term and the live streaming project name. For example, in all the first standard terms of the "Game Product A" live streaming project, the live streaming system will identify standard terms containing "Game Product A" or its general abbreviation (such as "Product A") (such as "Game Product A - Hero A", "Game Product A - Season Update") as second standard terms, while excluding standard terms that only contain general game terms (such as "combo tutorial", "new hero build") but are not explicitly associated with the "Game Product A" project name. In an optional implementation, the second standard term can also be automatically identified by an entity recognition model. The live streaming system uses words in the standard term that are marked as game names (or brand names) by the named entity recognition model as anchor points for determining the second standard term. Any standard term whose anchor point words match the name of the target live streaming project is identified as the second standard term.

[0056] The first user search information refers to a subset of user search behavior records selected by the live streaming system from the full user search information, where the search terms match the second standard terms. In other words, it represents the search behavior data of a specific user group who actively searched for terms related to the live streaming project. In an optional implementation, the first user search information includes all search term records of these users within the same search session (typically defined as a series of consecutive searches within 30 minutes). For example, if User A searches for "Game Product A," "Game Product A - Hero A," "Hero A Skin Price," and "Hero A Equipment Guide" in the same search session, the live streaming system will include all four search records in User A's first user search information for the "Game Product A" project, thus fully reconstructing the user's search behavior chain within the scope of that live streaming project.

[0057] The third standard term refers to the remaining standard terms from all search terms in the first user's search information, after normalization processing, excluding those directly corresponding to the live stream project name. These terms represent the segmented content preferences of the live stream project's user group outside the main project category and are key data for identifying sub-projects within the live stream project. In an optional implementation, the third standard term is obtained by analyzing the remaining terms after removing the second standard term. For example, in the first user's search information for the "Game Product A" project, after removing terms containing the "Game Product A" project name identifier, the remaining third standard terms include "Hero A" (occurring 8900 times), "Season Update" (occurring 6200 times), and "New Map Explanation" (occurring 5400 times), revealing the most popular segmented content topics among users searching for "Game Product A".

[0058] Sub-projects refer to the subdivided content dimensions identified by analyzing the distribution patterns of third-standard terms within the broader category of live-streaming projects. In one optional implementation, sub-projects are obtained through semantic clustering analysis of third-standard terms. The live-streaming system groups semantically similar third-standard terms into the same cluster, with each cluster corresponding to a sub-project. For example, under the "Game Product A" live-streaming project, the system uses semantic clustering to divide the third-standard terms into the following main sub-projects: "New Hero / Skin" (aggregating terms such as "Hero A," "New Hero," and "Skin"), "Esports / Teams" (aggregating terms such as "Season Update" and "Esports Prediction"), and "Combo Techniques / Tutorials" (aggregating terms such as "Combo" and "Equipment Guide"). The identification of sub-projects further refines the live-streaming project popularity analysis from broad category popularity to subdivided content popularity, providing streamers with more actionable content selection guidance.

[0059] Through the above steps, the live streaming content recommendation method provided in this disclosure conducts in-depth analysis of the search behavior of loyal users of a single live streaming project, identifies the current hot sub-projects of the live streaming project, and extends the live streaming content recommendation information from the level of the major category of the live streaming project to the level of the fine granularity of the sub-project. This allows the broadcaster to not only perceive which major category of live streaming is currently popular, but also to further understand which specific sub-direction within that category is most popular with users, thus achieving a deep extension of the user demand perception dimension.

[0060] Further, in one embodiment of this disclosure, the third term data includes: the third standard term includes at least one of the number of search users and the search frequency corresponding to the third standard term; the step of generating live content recommendation information corresponding to the time window based on the sorting of different live streaming projects and the number of live streaming rooms corresponding to each live streaming project includes: Based on the number of search users corresponding to the third standard term and / or the search frequency, a sub-project ranking is generated for the single live streaming project; Based on the sorting of different live streaming projects, the ranking of sub-projects of a single live streaming project, and the number of live streaming rooms corresponding to each sub-project, live streaming content recommendation information corresponding to the time window is generated.

[0061] The third-term data refers to the statistical summary of the third standard term within a time window, specifically including at least one of the number of users searching for the third standard term and the search frequency. In an optional implementation, the collection scope of the third-term data is limited to the scope of the first user's search information; that is, only the search data of users identified as the target audience of the live stream project for the third standard term are counted, thereby ensuring that the sub-project popularity ranking reflects the true segmented needs within the user group of the live stream project. The sub-project ranking refers to the ordered list obtained by the live stream system based on the third-term data, sorting each sub-project in descending order, reflecting the popularity of different sub-content directions within the same live stream project. Please refer to [reference needed]. Figure 4 This is a schematic diagram of the sub-project interface provided in one implementation method of this disclosure. As shown in the figure, when the user clicks... Figure 3 When game product C (i.e., the live streaming project) appears in the trending search list on the live streaming platform, the following will be displayed: Figure 4 The sub-project interface 400 shown is used to display various sub-projects 401 under game product C, such as: game product C - eSports live broadcast, game product C - new hero, game product C - new map commentary, game product C - past eSports analysis, game product C - new skin, etc. The above sub-projects 401 are ranked according to the current number of searchers 402 (i.e., search popularity) on the current live broadcast platform. At the same time, the sub-project interface 400 also displays the number of live broadcasters 403 (i.e., the number of live broadcast rooms) corresponding to each sub-project 401, enabling broadcasters to make more accurate live broadcast content planning decisions.

[0062] Through the above steps, the live streaming content recommendation method provided in this disclosure introduces a sub-project ranking dimension at the live streaming project level and combines it with the number of live streaming rooms corresponding to each sub-project. This deepens the precision of live streaming content recommendation information from the coarse-grained project category level to the fine-grained sub-project level, providing broadcasters with more operational and accurate content decision-making references. It effectively helps broadcasters identify and lock in the current segmented content opportunities that simultaneously meet the dual conditions of strong user search demand and relatively scarce content supply.

[0063] Furthermore, in one embodiment of this disclosure, the live streaming content recommendation information is configured as a live streaming project ranking list pushed to the broadcaster's end. The live streaming project ranking list is generated based on the sorting of different live streaming projects and the sub-project ranking of the individual live streaming project. The live streaming project ranking list displays the corresponding number of live streaming rooms for each sub-project associated with each live streaming project.

[0064] The live stream project ranking list displays the corresponding number of live stream rooms for each sub-project under each live stream project. This refers to the live stream project trending list pushed to the broadcaster's end, which, while displaying the popularity ranking of each live stream project, simultaneously presents a list of sub-project rankings for each project in a collapsed or expanded manner. Each sub-project entry in the sub-project list is associated with a corresponding number of live stream rooms. In an optional implementation, the sub-project list is nested under each live stream project entry in an expandable / collapseable tree structure. The broadcaster sees by default a list of live stream project categories sorted by popularity (e.g., ...). Figure 3 When a live stream item is clicked, that item automatically expands to display a list of its sub-items (e.g., ...). Figure 4 The system defaults to displaying the top 1-7 most popular sub-items. Each sub-item entry displays the sub-item name, its search popularity score, and the corresponding number of live stream rooms. In an optional implementation, the sub-item list can also highlight sub-items with high opportunity indices using color coding. For example, sub-items meeting the opportunity index requirements are marked with green text and an upward arrow icon, allowing broadcasters to intuitively identify the most valuable content segments without performing calculations.

[0065] Through the above steps, the live streaming content recommendation method provided in this disclosure integrates the sub-project rankings and their corresponding number of live streaming rooms into the live streaming project ranking pushed to the broadcaster. This presents the two-level supply and demand analysis results in a concise and intuitive tree-shaped ranking in one stop, allowing broadcasters to complete the complete content decision analysis from which major category is currently popular to which sub-category under that major category has the most traffic value in a single ranking view without having to switch between multiple interfaces or data dimensions.

[0066] Furthermore, in one embodiment of this disclosure, the method further includes: Based on the standard terms in the user search information, a user intent identifier is generated; The system configures the user intent identifier for the user account corresponding to the user search information, wherein the user intent identifier is used to associate the user account with the user account for display when the user account enters the corresponding live broadcast room.

[0067] The generation of user intent identifiers based on standard terms in the user's search information refers to the live streaming system extracting the search terms entered by the user in the search box, normalizing them to determine the corresponding standard terms, and then converting these standard terms into a structured intent identifier data package. This package carries the user's content demand intent expressed in a specific search behavior in the form of a concise tag. In an optional implementation, the generation of user intent identifiers is based on the triggering event of a user initiating a search and entering a live streaming room through a recommended live streaming room card recommended by the live streaming system. Specifically, when a user clicks on the recommended live streaming room card that appears below the search box to enter a live streaming room, the live streaming system extracts the text entered in the search box at the moment the user enters the live streaming room, performs normalization processing on the text, maps it to the corresponding standard terms, and generates user intent identifiers based on these standard terms. For example, if a user enters "Hero A combo techniques" in the search box and enters a live streaming room through a recommended card, the live streaming system normalizes "Hero A combo techniques" into the standard term "Game Product A - Hero A - Combo Tutorial," and generates the user's intent identifier "Hero A Combo" for this entry. In one optional implementation, the text length of the intent identifier is limited by the interface display requirements. The live streaming system intelligently truncates or refines the normalized standard terms to ensure that the text of the intent identifier is concise (usually no more than 8 Chinese characters) and can be clearly presented as a small label in the broadcaster's interface.

[0068] The configuration of the user intent identifier for the user account corresponding to the user search information refers to the live streaming system associating and binding the generated intent identifier data with the user account that triggered the search behavior, and recording it in the user account's live streaming data, making the intent identifier a visible attribute field for the user account during this live streaming session. In an optional implementation, the intent identifier configuration operation occurs the instant the user (i.e., the viewer) clicks the live streaming recommendation card and successfully loads the live streaming page. The live streaming system writes the intent identifier into the user account's session record in the live streaming room on the server side, and pushes the intent identifier data to the corresponding live streaming host's display interface via WebSocket or a similar real-time communication protocol, achieving low-latency synchronous display from the user clicking to enter the live streaming room to the host seeing the user intent identifier. In an optional implementation, the configuration operation also requires a user privacy authorization process. That is, before the user uses the function for the first time, the live streaming system clearly informs the user that the intent identifier will be displayed to the host and obtains the user's authorization. The user can choose to disable the intent identifier display function to protect their personal search privacy.

[0069] The user intent identifier is used to associate the user account with the live stream room and display it. In an optional implementation, the intent identifier is displayed as a small badge next to the corresponding user account's nickname in the viewer list interface on the broadcaster's end. Please refer to [link / reference]. Figure 5 This is a schematic diagram of the live streaming interface provided in one implementation of this disclosure. As shown in the figure, the host displays a live streaming interface 500, and the corresponding audience list 501 displays the audience accounts that have entered the current live streaming room. If an audience account in the audience list 501 entered the current live streaming room by entering search information (i.e., user search information) in the search box of the live streaming platform, the live streaming system displays a search intent label 502 next to the corresponding audience account, generated based on the search terms in the audience's search information. Examples of such intent labels include "Hero Tutorial," "Map Analysis," and "Combo Tutorial" as shown in the figure. Conversely, if an audience account in the audience list 501 entered the current live streaming room through other means, the live streaming system does not display the aforementioned search intent label 502 next to the corresponding audience account. This visual differentiation design allows the host to identify at a glance which audiences in the audience list 501 have specific content needs, thereby providing more targeted content services and interactive responses to the former.

[0070] Through the above steps, the live streaming content recommendation method provided in this disclosure transforms the keywords entered by users during the search process into structured intent identifiers, and configures them on the user's account in real time when the user enters the live streaming room for display. This achieves real-time transmission of user search intent from the viewer's end to the streamer's end, and for the first time gives the streamer the ability to directly perceive the entry intent of individual viewers. It breaks down the information barrier between the streamer and viewers with clear search intent, and effectively improves the viewer retention rate and interaction quality of the live streaming room.

[0071] Furthermore, in one embodiment of this disclosure, the method further includes: The system collects user intent identifiers for each user account within a single live stream, obtains the distribution information of different types of user intent identifiers within that live stream, and pushes this information to the corresponding broadcaster's interface for display.

[0072] The process of statistically analyzing user intent identifiers for each user account within a single live stream involves the live streaming system categorizing and statistically analyzing the intent identifiers of all currently online viewer accounts in a live stream, summarizing the frequency of different types of intent identifiers and the corresponding number of users, and generating a panoramic view of the overall intent identifier distribution for the live stream from individual intent identifiers. In an optional implementation, the statistical operation is performed at fixed time intervals (e.g., every 30 seconds). The live streaming system iterates through the list of currently online viewers in the live stream in real time, categorizing and counting all viewer accounts carrying intent identifiers according to their intent identifier types. For example, for a game live stream with 500 currently online viewers, the live streaming system displays the following on the streamer's current live stream interface: 500 online viewers, 280 of whom entered the live stream through search; among them, the user intent identifiers carried by these 280 viewers are distributed as follows: "Combo Tutorial" (120 people, accounting for 43% of tagged viewers), "Season Interpretation" (95 people, accounting for 34%), and "New Hero" (65 people, accounting for 23%).

[0073] The distribution information of different types of user intent identifiers refers to structured data that reflects the overall content demand composition of the current audience group after summarizing and statistically analyzing various intent identifiers within the live stream. In an optional implementation, the distribution information is presented in the form of percentage statistics, such as "Hero A related intents (43%), event commentary related intents (34%), combo tutorial intents (23%)", enabling the streamer to quickly grasp the proportion of different content demands among the current audience group. In an optional implementation, the distribution information may also include dynamic trend data, such as the change in the percentage of each intent identifier type compared to the previous statistical period, enabling the streamer to perceive the dynamic shift in the focus of audience demands in real time.

[0074] The process of pushing the information to the corresponding broadcaster's device for display on the broadcaster's interface refers to the live streaming system sending the statistically generated intent identifier distribution information to the broadcaster's device via a real-time communication link, and then visually displaying it on the broadcaster's interface. In an optional implementation, the intent identifier distribution information is displayed in the broadcaster's interface as a word cloud or horizontal bar chart in the data panel area. This data panel area displays the proportion of various intent identifiers using a horizontal bar chart, which is refreshed in real-time with each statistical cycle, dynamically reflecting changes in the composition of audience intent. In another optional implementation, the live streaming system can also proactively push content adjustment suggestions to the broadcaster based on the intent identifier distribution information. For example, when the proportion of a certain type of intent identifier rises rapidly in a short period, the live streaming system displays a prompt on the broadcaster's interface to help the broadcaster quickly respond to real-time changes in audience demands.

[0075] Through the above steps, the live streaming content recommendation method provided in this disclosure generates distribution information of the overall audience demand in the live streaming room by real-time aggregation and statistics of the intent identifiers of all viewers in the live streaming room, and pushes it to the broadcaster. This upgrades the broadcaster's ability to perceive audience demand from the visibility of individual viewer intents to the panoramic visibility of the overall audience demand, enabling the broadcaster to grasp the overall content demand composition of the current audience group in real time during the live streaming process, effectively improving the accuracy of content supply and demand matching, audience retention rate and interaction quality in the live streaming room.

[0076] Furthermore, in one embodiment of this disclosure, after configuring the user intent identifier to the user account, the method further includes: In response to the user account exiting the live stream, control cancel the user intent identifier configured for the user account.

[0077] In this context, user account exiting the live stream refers to a session termination event caused by a viewer account actively closing the live stream page, switching to another live stream, or due to network disconnection. This event marks the end of the current access session between the user account and the live stream. In an optional implementation, the live stream system detects user exit behavior by listening for changes in the live stream session state. When a user account sends an exit command or the live stream system detects that the user account's live stream session has timed out (e.g., no heartbeat signal for more than 60 seconds), the live stream system triggers the intent identifier cancellation configuration process. In another optional implementation, the live stream system can further distinguish between temporary offline behavior (e.g., briefly switching to the background, interruption of a call, etc.) and genuine exit (e.g., closing the live stream page). For temporary offline behavior, the live stream system can retain the intent identifier configuration state for a period of time (e.g., 5 minutes), and directly restore it when the user reactivates the live stream session, reducing the interference of frequent changes in intent identifiers caused by temporary network fluctuations on the stability of the broadcaster's display.

[0078] Specifically, canceling the user intent identifier configured for the user account refers to the process by which the live streaming system, upon detecting a user account exiting the live stream, removes the intent identifier data bound to that user account in the current live stream session from the data displayed on the broadcaster's end, and stops displaying the user's intent information in the broadcaster's viewer list and intent identifier distribution statistics. In an optional implementation, the intent identifier cancellation operation is performed immediately. After confirming the user exit event, the live streaming system deletes the user account's intent identifier from the broadcaster's real-time display dataset and triggers a partial refresh of the broadcaster's interface, causing the intent tag next to the user account in the viewer list to disappear synchronously, ensuring that the viewer intent information displayed on the broadcaster's end always matches the actual status of currently online viewers. In an optional implementation, the cancellation of the intent tag also triggers a re-statistical analysis of intent identifier distribution information. The live streaming system deducts the intent identifier count of exiting viewers from each type of statistics, recalculates the distribution ratio of each type of intent identifier, and pushes the updated distribution information to the broadcaster's data panel for refresh display.

[0079] In a specific application, a game livestream room has 320 viewers online at a certain time, of which 145 carry intent tags. One viewer, A, with the intent tag "combo tutorial," leaves the livestream room after watching for 15 minutes. The livestream system detects viewer A's departure in real time and immediately removes their intent tag from the streamer's viewer list and distribution statistics, refreshing the streamer's data panel instantly. If viewer A re-enters the same livestream room 30 minutes later by searching for the keyword "Hero A skin," the livestream system regenerates the intent tag "Hero A skin" based on the new search term and reconfigures it to the user account. The streamer then sees viewer A's latest status with the "Hero A skin" intent tag, accurately reflecting the viewer's actual search intent upon re-entering the livestream.

[0080] Through the above steps, the live content recommendation method provided in this disclosure ensures that the user intent information displayed on the broadcaster's end is consistent with the actual status of the current online audience by promptly canceling the intent identifier configuration when the user account exits the live room. This prevents the information distortion problem caused by the continuous display of the intent identifier of users who have left the live room on the broadcaster's end interface. At the same time, it supports the dynamic updating of intent identifiers when the same user re-enters the live room with different search keywords, ensuring the real-time accuracy of intent identifier information.

[0081] Furthermore, in one embodiment of this disclosure, the method further includes: During the process of inputting user search information into the target user account, standard terms corresponding to the user search information and live room feature terms of multiple live rooms corresponding to the time window are obtained; Determine a target live streaming room from the plurality of live streaming rooms based on a matching relationship between the feature entry of the live streaming room and a standard entry corresponding to the user search information; Generate live streaming room recommendation information for the target live streaming room and push the information to a viewer end corresponding to the target user account.

[0082] Wherein, the process in which the target user account inputs the user search information refers to the real-time input process in which a live viewer user types search content character by character into a search box (instead of after the live viewer user clicks a search button to submit a search request), that is, the matching calculation and recommendation processing flow of search entries and live streaming rooms is started, so as to implement pre-search recommendation. In an optional implementation, the live streaming system configures a real-time input listener for the user search box to capture a user's input event in a character-by-character triggering manner. After the length of the input content in the user search box reaches a minimum trigger threshold preset by the live streaming system (for example, 2 characters), the live streaming system starts the recommendation calculation flow, and re-performs a matching calculation every time the user enters a new character, so as to refresh recommendation results in real time. For example, when a user sequentially inputs "brit", "britain", "britain A", the live streaming system performs matching calculation for the first time when "britain" is input (triggered by a 2-character threshold), and then re-performs matching each time a new character is input, so as to ensure that the recommendation result is continuously and accurately updated along with the change of the input content. In an optional implementation, the live streaming system may also introduce an input debounce mechanism, that is, when a user inputs continuously and quickly, the live streaming system performs throttling processing on matching calculation requests (for example, triggering at most once every 200 milliseconds), so as to avoid excessive concurrent calculation requests on the server side caused by the user's rapid continuous input, and effectively control the resource consumption of the live streaming system while maintaining the real-time performance of recommendation.

[0083] Among them, the live room feature terms refer to the set of keywords continuously extracted and maintained by the live streaming system for each live room that is currently broadcasting within the current time window, describing the real-time content theme of the live room. This serves as index data for real-time matching of search terms with live rooms. In an optional implementation, the step of obtaining the live room feature terms for the multiple live rooms corresponding to the time window includes: obtaining live room information for the multiple live rooms corresponding to the time window, wherein the live room information includes at least one of the following: live room title, live room tag, live room description, and live room screen text information; performing semantic analysis processing on the live room information to generate live room feature terms corresponding to the multiple live rooms. In other words, the live streaming system first acquires various information about each live stream, including the live stream title, tags, description, and text information displayed on the live stream screen (using OCR technology to recognize text appearing in the live video in real time). Then, the system merges and deduplicates these multi-source extracted terms to generate a real-time content feature term set for that live stream. This term set is dynamically updated at a fixed interval (e.g., every 30 seconds) to ensure that the live stream feature terms always reflect the latest content status of the live stream. In an optional implementation, the live streaming system also assigns a weight value to each term, for example, terms appearing in the live stream title and tags have the highest weight, followed by terms in the live stream description, and then text recognized by OCR. This ensures that each term accurately reflects the representativeness of the live stream's content theme during search matching.

[0084] Specifically, determining the target live stream from multiple live streams based on the matching relationship between the live stream feature terms and the standard terms corresponding to the user's search information means that the live streaming system identifies a subset of live streams whose content features highly match the user's search intent by calculating the relevance between the standard terms after the user's (i.e., the live stream viewer's) current input text is normalized and the feature term sets of each live stream. In an optional implementation, the matching relationship is calculated based on keyword inclusion relationship judgment. That is, the live streaming system checks whether the feature term sets of each live stream contain terms that are semantically the same as or highly similar to the standard terms entered by the user. Live streams whose feature term sets contain matching terms are initially identified as candidate target live streams. For example, if the user's current input standard term is "game product A - hero A", the live streaming system searches the full live stream feature term index to find all live streams whose feature term sets contain keywords such as "hero A" and "game product A", and identifies these live streams as candidate target live streams. In an optional implementation, the matching relationship can also be calculated using the semantic vector similarity method. The live streaming system converts the standard terms input by the user and the feature terms of each live streaming room into semantic vector representations. The degree of semantic matching between each live streaming room and the user's search intent is determined by vector cosine similarity calculation. Thus, based on accurate keyword matching, live streaming rooms with different expressions but similar semantics are further identified.

[0085] The live stream recommendation information refers to the content generated by the live streaming system after determining the target live stream based on matching relationships, and is displayed to viewers. In an optional implementation, the live stream recommendation information is displayed as recommendation cards in a dropdown area below the user's search box. Each recommendation card may include: a thumbnail of the live stream cover (a real-time frame screenshot), the streamer's nickname, the live stream title, the current number of online viewers, and feature terms matching the user's search keywords (highlighted in the card to help users quickly understand the matching basis between the recommended live stream and their search intent). Please refer to [reference needed]. Figure 6 This is a schematic diagram of the live stream recommendation card interface provided in one implementation of this disclosure. As shown in the figure, when a live stream viewer enters the search information "Hero A" below the search box 601 in the viewer's search interface 600, the live streaming system generates a live stream recommendation card 602 based on the search information entered by the viewer. This live stream recommendation card 602 is used to suggest at least one live stream 603 that highly matches the search term "Hero A". When the player clicks the "Enter Live Stream" button 604 provided in the live stream recommendation card 602, the live streaming system controls the entry into the corresponding live stream and displays the corresponding live stream interface and live stream on the viewer's screen. This allows the viewer to quickly enter the corresponding live stream through the live stream recommendation card provided below the search box 601 without having to click the search button to perform a complete search.

[0086] Through the above steps, the live streaming content recommendation method provided in this disclosure performs real-time matching calculations between search terms and live streaming content feature terms while the user is inputting search information, and immediately displays a live streaming recommendation card with a high degree of matching below the search box. This advances the exposure time of the live streaming room from the passive waiting after the user completes the search to the process of the user actively inputting keywords, greatly shortening the operation path between the user generating content needs and finding the target live streaming room, and effectively improving the efficiency of users discovering relevant live streaming content.

[0087] Furthermore, in one embodiment of this disclosure, the step of obtaining the live room feature terms of multiple live rooms corresponding to the time window includes: Obtain the live room information of multiple live rooms corresponding to the time window. The live room information includes at least one of the following information of the multiple live rooms: live room title, live room tag, live room description and live room screen text information. Semantic analysis is performed on the live streaming room information to generate live streaming room feature terms corresponding to the multiple live streaming rooms.

[0088] The "livestream room information" refers to a set of raw text data related to the content theme of a specific livestream room, which can be used by the livestreaming system for semantic analysis. It serves as input data for extracting feature terms from the livestream room. In one optional implementation, the livestream room information includes the livestream room title, which is the name of the livestream room set by the streamer when starting the broadcast. This title is typically a concise summary of the livestream content theme, such as "Game Product A Hero A Full Analysis - Combo / Skin / Equipment Guide." The livestream room title is the most direct textual expression of the livestream content theme. In another optional implementation, the livestream room information also includes livestream room tags, which are predefined content category tags provided by the platform for the streamer. Streamers can select tags that match the content of the livestream before starting (e.g., selecting tags like "Game Guide" or "Game Product A"). These tags are precise descriptions of the livestream content category and can be directly used as high-confidence feature terms. In yet another optional implementation, the livestream room information may also include a livestream room introduction, providing richer content theme information than the title and tags.

[0089] The text information in the live stream refers to the text content appearing in the live stream frame, extracted by OCR (Optical Character Recognition) technology through real-time frame processing of the live video stream. It is a crucial dynamic source of feature terms for the live stream content. In one optional implementation, the live streaming system captures frames of the live video stream at fixed intervals (e.g., every 30 seconds), applies an OCR model to the captured video frame images for text recognition, and extracts the visible text information from the screen. Examples include character names (e.g., "Hero A"), skill names, and map names displayed on the game interface in a game live stream; and product names (e.g., "Brand A Cushion Sunscreen SPF50+") and price information displayed in an e-commerce live stream. In one optional implementation, before being included in the live stream feature term library, the text information extracted by OCR needs to undergo quality filtering to remove ambiguous text with low recognition confidence, pure numbers, short strings without semantic value, and fixed UI text unrelated to the live stream content (e.g., "Follow," "Share," etc.) to ensure that the text information included in the feature term library has actual content descriptive value.

[0090] The semantic analysis of the live stream information refers to the process by which the live stream system comprehensively utilizes natural language processing technologies such as word segmentation, keyword extraction, and semantic understanding to deeply process multi-source live stream information text and extract high-quality feature terms that accurately describe the current content theme of the live stream. In an optional implementation, the semantic analysis process includes the following main steps: First, the live stream information text from various sources is cleaned and segmented to remove stop words; then, the TF-IDF or TextRank algorithm is used to extract keywords from the segmented words, selecting the most representative words as candidate feature terms; finally, a named entity recognition model is used to identify proper nouns (such as game character names, brand names, event names, etc.) in the text, ensuring that high-value proper noun entities are accurately identified and retained. In an optional implementation, the semantic analysis process can also incorporate a knowledge base of the live stream content domain, such as a professional thesaurus including a dictionary of game character names, an e-commerce brand dictionary, and an event dictionary, to assist the model in more accurately identifying vertical domain proper nouns in the live stream information.

[0091] Through the above steps, the live streaming content recommendation method provided in this disclosure significantly improves the coverage and real-time performance of live streaming content description by collecting information from multiple sources from four dimensions: live streaming title, tags, description, and real-time screen text. It also comprehensively utilizes semantic analysis processing technology to generate multi-dimensional live streaming feature terms. This solves the problem of insufficient matching accuracy and limited coverage caused by traditional live streaming recommendation systems that rely solely on titles and tags manually set by the broadcaster for content description. This enables the live streaming system to continuously perceive and accurately index the real-time content status of the live streaming room.

[0092] Furthermore, in one embodiment of this disclosure, the step of generating live room recommendation information for the target live room includes: Based on the matching relationship between the live room feature terms of the target live room and the standard terms corresponding to the user search information, a live room ranking is generated for the target live room; Based on the sorting of the live streams, generate live stream recommendation information corresponding to the target live stream.

[0093] The live stream ranking refers to the process by which the live streaming system prioritizes and determines the order in which live streams are displayed to the user based on the matching results between the live stream feature terms of each target live stream and the corresponding standard terms of the user's search information. In an optional implementation, the live stream ranking is calculated based on the content feature matching degree index. That is, the more keywords matched and the higher the importance weight of the keywords, the higher the matching degree score of the live stream and the higher its ranking. For example, if the user enters the standard term "game product A-hero A-combo", the feature term set of target live stream A contains three matching terms: "hero A" (weight 0.9), "combo" (weight 0.8), and "game product A" (weight 0.7), with a comprehensive matching degree score of 2.4; target live stream B only contains two matching terms: "hero A" and "game product A", with a comprehensive matching degree score of 1.6; based on this, the live streaming system ranks live stream A first and live stream B second. In an optional implementation, the live stream ranking can also introduce a multi-dimensional comprehensive scoring mechanism. In addition to the content feature matching degree, factors such as the current number of online viewers in the live stream (popularity dimension) and the broadcaster's historical content quality score (quality dimension) are also considered. The final ranking is generated through a weighted comprehensive score to ensure that the live streams recommended to users not only have a high content matching degree, but also have a good viewing experience.

[0094] The process of generating live stream recommendation information corresponding to the target live stream according to the live stream ranking refers to the process by which the live streaming system extracts the necessary information fields for display from the top-ranked target live streams based on the generated live stream ranking, and assembles them into a standardized recommendation information data package to be pushed to the audience. In an optional implementation, the live stream recommendation information data package contains the display information of the first 3 (or the number configured by the live streaming system) target live streams in the ranking order. The display information fields of each live stream may include: a unique identifier for the live stream (for client-side click redirection), a screenshot URL of the live stream cover (for displaying the card thumbnail), the streamer's nickname, the live stream title, the number of current online viewers, and keyword highlighting information matching the user-input standard terms, etc.

[0095] Through the above steps, the live streaming content recommendation method provided in this disclosure generates an accurate ranking of live streaming rooms by calculating the multi-dimensional matching relationship between the feature terms of the live streaming room and the user's standard search terms. The recommended live streaming room information is then assembled in sequence and pushed to the audience for display. This ensures that the recommended live streaming rooms that users see below the search box are always the top-ranked live streaming rooms with the most matching content and the best overall quality. This effectively improves the accuracy of real-time search recommendation results and the user's click experience satisfaction, and ultimately achieves the precise diversion of user search traffic to high-quality matching live streaming rooms.

[0096] Secondly, this disclosure also provides a live streaming content recommendation device.

[0097] Please refer to Figure 7 This is a schematic diagram of the structure of a live streaming content recommendation device provided in one implementation of this disclosure. As shown in the figure, the live streaming content recommendation device 100 may include: Module 101 is used to obtain user search information corresponding to the time window; The determining module 102 is used to determine the search parameters corresponding to different live streaming projects based on the user search information.

[0098] The generation module 103 is used to generate live content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live project.

[0099] It should be noted that the above is only a brief description of the live content recommendation device 100 provided in this disclosure. The process steps and / or functions performed by the live content recommendation device 100 when it is working are largely the same as the steps and / or functions described in the various embodiments of the live content recommendation method provided in the above text. Therefore, they will not be described in detail here.

[0100] As described above, the live streaming content recommendation device 100 provided in this disclosure obtains user search information corresponding to a time window; determines search parameters for different live streaming projects based on the user search information; and generates live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects. In this way, determining the search parameters for different live streaming projects based on user search information within the time window and generating live streaming content recommendation information in conjunction with the supply status information of the corresponding live streaming projects helps reduce the subjective bias in content selection by broadcasters, improves the accuracy of matching live streaming content with user search needs, and helps reduce ineffective server computation and data interaction resource consumption caused by asynchronous supply and demand information.

[0101] Then, a terminal device is also provided.

[0102] See Figure 8The figure shows a schematic diagram of the hardware architecture of a terminal device provided in one implementation of this disclosure. As shown, the terminal device 200 includes a processor 201, a memory 202, a communication interface, and a bus 203. The processor 201, the communication interface, and the memory 202 are connected via the bus 203. The memory 202 may include high-speed random access memory (RAM) and may also include non-volatile memory, such as at least one disk storage device. Communication between this live streaming system network element and at least one other network element is achieved through at least one communication interface (which can be wired or wireless), such as the Internet, wide area network, local area network, metropolitan area network, etc. The bus 203 may be an ISA bus, PCI bus, or EISA bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 8 The symbol is represented by a single double-headed arrow, but this does not mean that there is only one bus or one type of bus.

[0103] Processor 201 may be an integrated circuit chip with signal processing capabilities. The aforementioned processor 201 can also be a general-purpose processor, including a Central Processing Unit (CPU), a Network Processor (NP), etc.; it can also be a Digital Signal Processor (DSP), an Application Specific Integrated Circuit (ASIC), a Field-Programmable Gate Array (FPGA), or other programmable logic devices, discrete gate or transistor logic devices, or discrete hardware components. A general-purpose processor can be a microprocessor or any conventional processor.

[0104] In an alternative implementation, memory 202 stores computer-executable instructions that can be executed by processor 201. Processor 201 executes these computer-executable instructions to implement the above method, specifically including the following steps: Retrieve user search information corresponding to the time window; Based on the user search information, determine the search parameters corresponding to different live streaming projects; Based on the search parameters and the supply status information of the corresponding live streaming projects, live streaming content recommendation information corresponding to the time window is generated.

[0105] It should be noted that the above is only a brief description of the terminal device 200 provided in this disclosure. The process steps and / or functions performed by the terminal device 200 when it is working are largely the same as the steps and / or functions described in the various embodiments of the live content recommendation method provided in the previous text. Therefore, they will not be described in detail here.

[0106] As described above, the terminal device 200 provided in this disclosure acquires user search information corresponding to a time window; determines search parameters for different live streaming projects based on the user search information; and generates live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects. In this way, determining the search parameters for different live streaming projects based on user search information within the time window, and generating live streaming content recommendation information in conjunction with the supply status information of the corresponding live streaming projects, helps reduce the subjective bias in content selection by broadcasters, improves the accuracy of matching live streaming content with user search needs, and helps reduce ineffective server calculations and data interaction resource consumption caused by asynchronous supply and demand information.

[0107] Finally, this disclosure also provides a computer-readable storage medium.

[0108] In an alternative implementation, the computer-readable storage medium stores computer-executable instructions that, when invoked and executed by a processor, cause the processor to implement the above method, specifically including the following steps: Retrieve user search information corresponding to the time window; Based on the user search information, determine the search parameters corresponding to different live streaming projects; Based on the search parameters and the supply status information of the corresponding live streaming projects, live streaming content recommendation information corresponding to the time window is generated.

[0109] It should be noted that the above is only a brief description of the computer-readable storage medium provided in this disclosure. The process steps and / or functions performed by the computer-readable storage medium provided in this disclosure are largely consistent with the steps and / or functions described in the various embodiments of the live content recommendation method provided in the foregoing text disclosure, so they will not be described in detail here.

[0110] As described above, the computer-readable storage medium provided in this disclosure acquires user search information corresponding to a time window; determines search parameters for different live streaming projects based on the user search information; and generates live streaming content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live streaming projects. In this way, determining the search parameters for different live streaming projects based on user search information within the time window, and generating live streaming content recommendation information in conjunction with the supply status information of the corresponding live streaming projects, helps reduce the subjective bias in content selection by broadcasters, improves the accuracy of matching live streaming content with user search needs, and helps reduce ineffective server computation and data interaction resource consumption caused by asynchronous supply and demand information.

[0111] The embodiments described above are merely illustrative of several implementations of this disclosure, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of this patent disclosure. Furthermore, the technical features of the above embodiments can be combined arbitrarily. For the sake of brevity, not all possible combinations of the technical features in the above embodiments have been described; however, as long as the combinations of these technical features do not contradict each other, they should be considered within the scope of this specification. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this disclosure, and these all fall within the protection scope of this disclosure.

Claims

1. A method for recommending live streaming content, characterized in that, include: Retrieve user search information corresponding to the time window; Based on the user search information, determine the search parameters corresponding to different live streaming projects; Based on the search parameters and the supply status information of the corresponding live streaming projects, live streaming content recommendation information corresponding to the time window is generated.

2. The method according to claim 1, characterized in that, The step of determining the search parameters corresponding to different live streaming projects based on the user search information includes: The user search information is analyzed and processed to obtain the search terms corresponding to the user search information; The search terms are normalized to determine the standard terms corresponding to different search terms; Based on the first term data associated with the first standard term for different live streaming projects, the search parameters corresponding to different live streaming projects are determined.

3. The method according to claim 2, characterized in that, The first term data includes at least one of the number of search users and the search frequency corresponding to the first standard term; the step of generating live content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live project includes: The different live streaming projects are sorted based on the number of search users corresponding to the first standard term and / or the search frequency; Based on the sorting of different live streaming projects and the number of live streaming rooms corresponding to each project, live streaming content recommendation information corresponding to the time window is generated.

4. The method according to claim 3, characterized in that, The live streaming content recommendation information is configured as a live streaming project list pushed to the broadcaster's end. The live streaming project list is generated based on the sorting of different live streaming projects, and the number of corresponding live streaming rooms is displayed for each live streaming project in the live streaming project list.

5. The method according to claim 3, characterized in that, The method further includes: Identify the second standard term in the first standard term that is associated with a single live streaming project; Determine the first user search information corresponding to the second standard term from the user search information; Based on the third standard terms in the first user's search information, sub-projects for the single live streaming project are determined.

6. The method according to claim 5, characterized in that, The third term data includes at least one of the number of search users and search frequency corresponding to the third standard term; the step of generating live content recommendation information corresponding to the time window based on the sorting of different live streaming projects and the number of live streaming rooms corresponding to each live streaming project includes: Based on the number of search users corresponding to the third standard term and / or the search frequency, a sub-project ranking is generated for the single live streaming project; Based on the sorting of different live streaming projects, the ranking of sub-projects of a single live streaming project, and the number of live streaming rooms corresponding to each sub-project, live streaming content recommendation information corresponding to the time window is generated.

7. The method according to claim 6, characterized in that, The live streaming content recommendation information is configured as a live streaming project ranking list pushed to the broadcaster's end. The live streaming project ranking list is generated based on the sorting of different live streaming projects and the sub-project ranking of a single live streaming project. The live streaming project ranking list displays the corresponding number of live streaming rooms for each sub-project associated with each live streaming project.

8. The method according to claim 2, characterized in that, The method further includes: Based on the standard terms in the user search information, a user intent identifier is generated; The system configures the user intent identifier for the user account corresponding to the user search information, wherein the user intent identifier is used to associate the user account with the user account for display when the user account enters the corresponding live broadcast room.

9. The method according to claim 8, characterized in that, The method further includes: The system collects user intent identifiers for each user account within a single live stream, obtains the distribution information of different types of user intent identifiers within that live stream, and pushes this information to the corresponding broadcaster's interface for display.

10. The method according to claim 8, characterized in that, After configuring the user intent identifier to the user account, the method further includes: In response to the user account exiting the live stream, control cancel the user intent identifier configured for the user account.

11. The method according to claim 2, characterized in that, The method further includes: During the process of inputting user search information into the target user account, standard terms corresponding to the user search information and live room feature terms of multiple live rooms corresponding to the time window are obtained; Based on the matching relationship between the feature terms of the live room and the standard terms corresponding to the user's search information, the target live room is determined from the multiple live rooms; Generate live stream recommendation information for the target live stream and push it to the viewer terminal corresponding to the target user account.

12. The method according to claim 11, characterized in that, The step of obtaining the live room feature terms of multiple live rooms corresponding to the time window includes: Obtain the live room information of multiple live rooms corresponding to the time window. The live room information includes at least one of the following information of the multiple live rooms: live room title, live room tag, live room description and live room screen text information. Semantic analysis is performed on the live streaming room information to generate live streaming room feature terms corresponding to the multiple live streaming rooms.

13. The method according to claim 12, characterized in that, The step of generating live room recommendation information for the target live room includes: Based on the matching relationship between the live room feature terms of the target live room and the standard terms corresponding to the user search information, a live room ranking is generated for the target live room; Based on the sorting of the live streams, generate live stream recommendation information corresponding to the target live stream.

14. A live streaming content recommendation device, characterized in that, include: The acquisition module is used to acquire user search information corresponding to the time window. The determination module is used to determine the search parameters corresponding to different live streaming projects based on the user search information. The generation module is used to generate live content recommendation information corresponding to the time window based on the search parameters and the supply status information of the corresponding live project.

15. A terminal device, characterized in that, It includes a processor and a memory, wherein the memory stores a computer program that, when executed by the processor, implements the method as described in any one of claims 1-13.

16. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the method as described in any one of claims 1-13.

Images