Live streaming interaction method and apparatus, electronic device, and computer-readable storage medium
Patent Information
- Application Number
- PCT/CN2025/096067
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-02-28
- Filing Date
- 2025-05-20
- Publication Date
- 2026-09-03
Smart Images
Figure CN2025096067_03092026_PF_FP_ABST
Abstract
Description
Live interactive methods, devices, electronic devices and computer-readable storage media
[0001] Cross-reference of related applications
[0002] This disclosure claims priority to Chinese Patent Application No. 2025102364299, filed on February 28, 2025, entitled "Live Streaming Interaction Method, Apparatus, Electronic Device and Computer-Readable Storage Medium", the entire contents of which are incorporated herein by reference. Technical Field
[0003] This disclosure relates to the field of computer technology, and more specifically, to a live interactive method, apparatus, electronic device, and computer-readable storage medium. Background Technology
[0004] With the development of the live streaming industry, functional live streaming rooms and live commerce have been successfully applied in various industries. This type of live streaming often allows the host to interact with the viewers, enabling some users to achieve certain conversions, such as following the host or purchasing goods.
[0005] Related technologies can analyze users' conversion intentions and rank users based on their conversion intentions. However, this method can only classify users and cannot guide the live streaming interaction process to promote user conversion, thus resulting in poor live streaming performance. Summary of the Invention
[0006] In view of this, the purpose of this disclosure is to provide a live streaming interactive method, apparatus, electronic device, and computer-readable storage medium to promote user conversion and improve live streaming performance.
[0007] To achieve the above objectives, the technical solutions adopted in the embodiments of this disclosure are as follows:
[0008] Firstly, this disclosure provides a live interactive method, the method comprising:
[0009] Get the first interaction parameters for each user watching the current live stream;
[0010] The first interaction parameters corresponding to each of the viewers are input into a pre-trained user value model for processing to obtain the conversion value corresponding to each of the viewers; the conversion value represents the promoting effect of the interaction behavior between the anchor and the viewers on the conversion of the viewers.
[0011] Based on the conversion value corresponding to each of the viewing users, determine the first target user whose conversion value reaches a preset value from among the multiple viewing users;
[0012] For each of the first target users, an interaction strategy corresponding to the first target user is determined through a pre-trained response model, and the interaction strategy is sent to the anchor so that the anchor can interact with the first target user according to the interaction strategy.
[0013] In an optional implementation, obtaining the first interaction parameters corresponding to each viewing user in the current live stream includes:
[0014] Obtain the live streaming parameters in the current live stream; the live streaming parameters include user comments from multiple viewers and live streaming content from multiple broadcasters.
[0015] Based on each user comment and its corresponding timestamp, each streamer's live stream content and its corresponding timestamp, the first interaction parameter corresponding to each viewing user is determined.
[0016] In an optional implementation, the step of inputting the first interaction parameters corresponding to each of the viewing users into a pre-trained user value model for processing to obtain the conversion value corresponding to each of the viewing users includes:
[0017] For each of the viewing users, the first interaction parameters corresponding to the viewing user are input into the user value model for processing to obtain the first conversion probability and the second conversion probability corresponding to the viewing user.
[0018] Wherein, the first conversion probability represents the conversion probability of the viewing user when there is an interactive behavior between the broadcaster and the viewing user, and the second conversion probability represents the conversion probability of the viewing user when there is no interactive behavior between the user and the viewing user;
[0019] The difference between the first conversion probability and the second conversion probability is determined as the conversion value of the viewing user.
[0020] In an optional implementation, determining the interaction strategy corresponding to each of the first target users through a pre-trained response model includes:
[0021] The response order for each of the first target users is determined based on the conversion value corresponding to each of the first target users.
[0022] According to the reply order of each of the first target users, the users to be replied to are determined in sequence, and the first interaction parameters corresponding to the users to be replied to are input into the reply model for processing to obtain the interaction strategy corresponding to the users to be replied to.
[0023] In an optional implementation, the method further includes:
[0024] If the first target user does not exist, at least one second target user is determined from the multiple viewers based on the conversion value of each viewer. The first interaction parameter corresponding to the second target user is input into the response model for processing to obtain the interaction strategy corresponding to the second target user.
[0025] In an optional implementation, the response model is trained through the following steps:
[0026] Acquire secondary interaction parameters and industry knowledge data between multiple converted users and the streamer;
[0027] According to a preset ratio, an interactive parameter to be processed is randomly determined from multiple second interactive parameters, and the interactive parameter to be processed is truncated to obtain the truncated interactive parameter to be processed.
[0028] The second interaction parameter, the truncated interaction parameter to be processed, and the industry knowledge data are input into a preset initial response model for training to obtain a trained response model.
[0029] In an optional implementation, the user value model is trained through the following steps:
[0030] Obtain third-party interaction parameters corresponding to multiple viewing users;
[0031] Based on the conversion and interaction of each viewing user, training labels are added to the third interaction parameters corresponding to each viewing user.
[0032] The third interaction parameter with the added training label is input into the pre-set initial user value model for training, and a trained user value model is obtained.
[0033] Secondly, this disclosure provides a live interactive device, the device comprising:
[0034] The acquisition module is configured to acquire the first interaction parameters corresponding to each viewing user in the current live stream.
[0035] The processing module is configured to input the first interaction parameters corresponding to each of the viewers into a pre-trained user value model for processing, and obtain the conversion value corresponding to each of the viewers; the conversion value represents the promoting effect of the interaction behavior between the anchor and the viewers on the conversion of the viewers.
[0036] The determination module is configured to determine, from among the multiple viewers, a first target user whose conversion value reaches a preset value based on the conversion value corresponding to each of the viewers;
[0037] The determining module is further configured to determine the interaction strategy corresponding to each first target user through a pre-trained response model, and send the interaction strategy to the anchor so that the anchor can interact with the first target user according to the interaction strategy.
[0038] Thirdly, this disclosure provides an electronic device including a processor and a memory, the memory storing a computer program executable by the processor, the processor being able to execute the computer program to implement the method described in any of the foregoing embodiments.
[0039] Fourthly, this disclosure provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described in any of the foregoing embodiments.
[0040] The live streaming interaction method, apparatus, electronic device, and computer-readable storage medium provided in this disclosure can determine the conversion value of each viewing user by using the first interaction parameters corresponding to each viewing user and a pre-trained user value model. Since the conversion value can characterize the promotional effect of the interaction behavior between the anchor and the viewing user on the conversion of the viewing user, the first target user who should be interacted with firstly should be determined according to the conversion value. The interaction strategy corresponding to each first target user is determined by the pre-trained response model and sent to the anchor. In this way, the anchor can interact with each first target user according to the interaction strategy corresponding to each first target user firstly, thereby promoting user conversion and improving the live streaming effect. Attached Figure Description
[0041] To more clearly illustrate the technical solutions of the embodiments of this disclosure, the accompanying drawings used in the embodiments will be briefly described below. It should be understood that the following drawings only show some embodiments of this disclosure and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.
[0042] Figure 1 shows a block diagram of an electronic device provided in an embodiment of the present disclosure;
[0043] Figure 2 shows a flowchart of a live interactive method provided in an embodiment of this disclosure;
[0044] Figure 3 shows a schematic diagram of the user value model;
[0045] Figure 4 shows a schematic diagram of the conversion value calculation;
[0046] Figure 5 shows a functional block diagram of a live interactive device provided in an embodiment of this disclosure.
[0047] Icons: 100 - Memory; 110 - Processor; 120 - Communication module; 200 - Acquisition module; 210 - Processing module; 220 - Determination module. Detailed Implementation
[0048] The technical solutions of the embodiments 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. The components of the embodiments of this disclosure described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.
[0049] Therefore, the following detailed description of the embodiments of this disclosure provided in the accompanying drawings is not intended to limit the scope of the claimed disclosure, but merely to illustrate selected embodiments of the disclosure. All other embodiments obtained by those skilled in the art based on the embodiments of this disclosure without inventive effort are within the scope of this disclosure.
[0050] It should be noted that relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes the element.
[0051] Currently, with the development of the live streaming industry, functional live streaming rooms and live commerce have been successfully applied in various industries. This type of live streaming often allows the host to interact with the viewers, enabling some users to achieve certain conversions, such as following the host or purchasing goods.
[0052] Current live streaming assistant software is often only configured to optimize the live streaming process to improve the quality of the live streaming content. However, when facing a large number of users, the streamer or live streaming team still needs to manually select some users to reply to.
[0053] While some algorithms in related technologies can analyze users' conversion intentions and rank them based on those intentions, this approach can only classify users and cannot determine the causal impact of the streamer's interactive behavior on user conversion. Therefore, it often fails to guide the live streaming interaction process and thus cannot promote user conversion, resulting in poor live streaming performance.
[0054] Based on this, the present disclosure provides a live interactive method, apparatus, electronic device, and computer-readable storage medium to solve the aforementioned problems. The live interactive method, apparatus, electronic device, and computer-readable storage medium provided in the present disclosure will now be described in conjunction with the accompanying drawings.
[0055] Figure 1 is a block diagram of an electronic device provided in an embodiment of this disclosure. Referring to Figure 1, the electronic device includes a memory 100, a processor 110, and a communication module 120. The memory 100, processor 110, and communication module 120 are electrically connected to each other directly or indirectly to realize data transmission or interaction. For example, these components can be electrically connected to each other through one or more communication buses or signal lines.
[0056] The memory 100 is configured to store computer programs or data that can be executed by a processor. The memory 100 may be, but is not limited to, random access memory (RAM), read-only memory (ROM), programmable read-only memory (PROM), erasable programmable read-only memory (EPROM), electrically erasable programmable read-only memory (EEPROM), etc.
[0057] The processor 110 is configured to read / write data or computer programs stored in memory and perform corresponding functions.
[0058] The communication module 120 is configured to establish a communication connection between the electronic device and other communication terminals via a network, and is configured to send and receive data via a network.
[0059] It should be understood that the structure shown in Figure 1 is only a schematic diagram of the electronic device. The electronic device may include more or fewer components than shown in Figure 1, or have a different configuration than shown in Figure 1. The components shown in Figure 1 can be implemented using hardware, software, or a combination thereof.
[0060] The following describes the live interactive method provided in this disclosure embodiment, using the electronic device in Figure 1 as the execution subject and in conjunction with the flowchart. Specifically, Figure 2 is a flowchart of a live interactive method provided in this disclosure embodiment. Referring to Figure 2, the method includes:
[0061] Step S20: Obtain the first interaction parameters corresponding to each viewing user in the current live stream.
[0062] Optionally, the first interactive parameter can be a combination of conversations corresponding to the viewing user, including the questions asked by the viewing user from the moment they enter the live broadcast room and the corresponding replies from the host.
[0063] Optionally, the electronic device can begin acquiring and processing the first interaction parameters corresponding to each viewing user in the current live stream in real time after a preset duration has elapsed since the start of the live stream.
[0064] Step S21: Input the first interaction parameters corresponding to each viewing user into the pre-trained user value model for processing to obtain the conversion value corresponding to each viewing user.
[0065] Among them, the conversion value represents the positive impact of the interaction between the streamer and the viewers on the conversion of viewers.
[0066] In this embodiment, the conversion value can reflect the causal impact of the anchor's interactive behavior on user conversion. That is, the conversion value can be used to determine whether the anchor's interactive methods for the viewing user can promote the user to convert, such as purchasing goods, following the anchor, etc.
[0067] Optionally, different live streaming rooms may have different conversion goals. For example, some live streaming rooms may want to gain user attention, some may want to make user purchases, and some may want users to leave some sales leads. Therefore, the user value model can be set according to the specific conversion goals of the live streaming room in order to obtain the conversion value corresponding to the relevant conversion goals.
[0068] Step S22: Based on the conversion value of each viewing user, determine the first target user whose conversion value reaches the preset value from multiple viewing users.
[0069] Optionally, viewers can generally be divided into four categories: users who dislike the rhetoric, non-target users, high-value users, and users who want to interact and convert.
[0070] Among them, the "Averse to Scripts" user group refers to users who resist the host's proactive interactive guidance, leading to a decrease in their trust in the live stream's professionalism. However, these users can be converted naturally if the host does not actively interact with the audience. The "Non-Target User Group" refers to users who only watch the live stream and cannot be converted regardless of whether the host interacts with them. The "High-Value User Group" refers to users who will be converted regardless of whether the host prioritizes answering their questions or actively initiates interaction. The "Interactive Conversion Group" refers to users who need to be prioritized in response or interaction to be converted; otherwise, they will leave the live stream.
[0071] Optionally, the primary target users refer to viewers whose conversion is significantly boosted by the streamer's interactive behavior, which is the aforementioned interactive conversion group.
[0072] Understandably, for these types of users, if the host does not prioritize answering their questions or actively guide them, the conversion rate of these users is relatively low. Conversely, if the host prioritizes answering their questions or actively guides them, these users are more likely to convert. Therefore, it is necessary to identify these primary target users from among multiple viewers and prioritize their interaction.
[0073] Optionally, the preset value can be set according to the actual application situation, and this disclosure does not impose too many restrictions on it. For example, if the number of people in the live broadcast room is small, only 500 people, a smaller preset value can be set to ensure that 50% of the viewers are the primary target users; if the number of people in the live broadcast room is large, such as 100,000 people, a higher preset value can be set to ensure that 0.5% of the viewers are the primary target users.
[0074] Step S23: For each first target user, determine the corresponding interaction strategy for the first target user through a pre-trained response model, and send the interaction strategy to the host so that the host can interact with the first target user according to the interaction strategy.
[0075] Alternatively, the response model can be a large language model.
[0076] Optionally, the interactive strategy can be the response to an unanswered question posed by the first target user (i.e., predicting the host's reply), or it can be guiding the discussion on topics that the first target user is interested in (i.e., predicting the user's next question), such as asking guiding questions that the first target user is interested in, or introducing content that the first target user is interested in.
[0077] For example, if a primary target user asks a question, the electronic device can generate at least one answer corresponding to the question through the response model and send it to the host, who can then select an answer to interact with the primary target user in the live broadcast room. If a primary target user has not yet asked a question, a guiding question (i.e., the user's next question) can be generated based on the content that the primary target user is interested in.
[0078] Specifically, in response to the host's reply, the host can reply to the first target user to interact with the first target user. In response to the user's next question, the host can proactively ask the first target user if they want to ask that question to interact with the first target user.
[0079] The live streaming interaction method provided in this embodiment can determine the conversion value of each viewing user by using the first interaction parameters corresponding to each viewing user and a pre-trained user value model. Since the conversion value can characterize the promotional effect of the interaction behavior between the anchor and the viewing user on the conversion of the viewing user, the first target user who should be interacted with firstly can be determined according to the conversion value. The interaction strategy corresponding to each first target user can be determined by the pre-trained response model and sent to the anchor. In this way, the anchor can interact with each first target user according to the interaction strategy corresponding to each first target user, thereby promoting user conversion and improving the live streaming effect.
[0080] Optionally, for ease of application, the response model and user value model can be trained first.
[0081] In this embodiment, a corresponding response model and user value model can be trained for each live broadcast room, or a general response model and user value model can be trained for multiple live broadcast rooms. The specific settings can be adjusted according to the actual application situation, and this disclosure does not impose too many restrictions on this.
[0082] The following section provides a possible approach for training the response model.
[0083] Specifically, the electronic device can acquire multiple second interaction parameters between converted users and the streamer, as well as industry knowledge data. It can randomly select the interaction parameters to be processed from the multiple second interaction parameters according to a preset ratio, truncate the interaction parameters to be processed, and obtain the truncated interaction parameters to be processed. Then, the second interaction parameters, the truncated interaction parameters to be processed, and the industry knowledge data are input into a preset initial response model for training to obtain a trained response model.
[0084] Optionally, "converted users" refers to users who have achieved the relevant conversion goals, such as those who have purchased relevant products or followed the streamer.
[0085] Optionally, the second interaction parameter may include all interaction parameters from the time the converted user enters the live stream until the conversion, i.e., user comments or questions from the converted user and the host's replies. The industry knowledge data may include dialogue records between industry experts and users, and may also include industry-related documents, etc.
[0086] For example, for car live streaming, industry knowledge data could be the dialogue between automotive industry experts and users. This data can improve the performance of the response model in terms of industry logic reasoning. If a user asks, "I live in Northeast China, can you recommend a car?", the response could be, "Winter in Northeast China is quite cold. If you only use it in the city and there are charging stations, consider an electric car. Otherwise, I suggest buying a gasoline car with a large engine."
[0087] Optionally, to ensure that the response model can predict questions that users might be interested in so as to proactively ask questions to the corresponding users, such as predicting the user's next question and predicting the effective response content of the anchor, the electronic device can also truncate the second interaction parameters. In this embodiment, the electronic device can randomly select some of the multiple second interaction parameters as interaction parameters to be processed, and divide the interaction parameters to be processed into first interaction parameters to be processed and second interaction parameters to be processed. The first interaction parameters to be processed are truncated according to the anchor's response, and the second interaction parameters to be processed are randomly truncated, thereby obtaining the truncated interaction parameters to be processed.
[0088] In one example, if the host's reply is represented as A, and the user's question or comment is represented as Q, then the second interaction parameter can be represented as Q->A->Q->…->Q->(A) (that is, the user completes the conversion after the host replies (A)), the first interaction parameter to be processed can be represented as Q->A->Q->…->A->(Q) (used to predict the user's next question (Q)), and the second interaction parameter to be processed can be represented as Q->A->Q->…->Q->(A) (the second interaction parameter is randomly truncated and configured to predict the content of the host's reply (A)).
[0089] Optionally, the preset ratio can be set according to the actual application. In one possible implementation, the second interactive parameter, the first interactive parameter to be processed, the second interactive parameter to be processed, and the industry knowledge data can be mixed in a ratio of 2:1:2:1.
[0090] In one possible implementation, the initial response model could be 1000 Questions 2.5 14B. Furthermore, the electronic device could perform LoRa tuning on this response model using supervised tuning. The tuning hardware could employ an Nvidia A800*12, 500GB RAM training cluster, with a preset number of training epochs, such as 3 epochs.
[0091] Optionally, after the response model has been trained, it can be deployed according to the actual application requirements.
[0092] It should be noted that the responses generated by large models in related technologies are often not natural enough, and they often fail to combine the user's interaction context and the live broadcast context to generate personalized responses. Therefore, they cannot solve user problems and guide users to complete conversion while engaging in targeted interaction with users.
[0093] The live streaming interaction method provided in this embodiment can input the second interaction parameters of converted users and industry knowledge data into an initial response model for training. Therefore, the trained response model can generate suitable interaction strategies by combining product expertise and dialogue history under the corresponding style and tone. Furthermore, by extracting a portion of the second interaction parameters and inputting them into the initial response model for training, the trained response model can also possess the ability to proactively ask questions, guide users, and provide high-quality responses. In this way, personalized responses can be generated during the current live stream, addressing user issues while engaging in targeted interaction and guiding users to complete the conversion process.
[0094] The following section provides a possible approach to training the user value model.
[0095] Specifically, the electronic device can acquire third interaction parameters corresponding to multiple viewing users, add training labels to the third interaction parameters corresponding to each viewing user based on the conversion and interaction status of each viewing user, and then input the third interaction parameters with added training labels into the pre-set initial user value model for training to obtain the trained user value model.
[0096] Optionally, due to the nature of live streaming, electronic devices cannot simultaneously collect data on whether a viewer has completed a conversion, both during interactive and non-interactive sessions. Therefore, interaction parameters between the streamer and the user can be collected during the live stream. Specifically, for converted users, all interaction parameters from their entry into the live stream to the conversion time can be retained; for non-converted users, all interaction parameters throughout the entire live stream can be retained.
[0097] Understandably, this third interaction parameter may include all interaction parameters of converted users from the moment they enter the live stream until the conversion time, as well as all interaction parameters of non-converted users throughout the entire live stream.
[0098] Optionally, conversion status refers to whether viewers have made a conversion, and interaction status refers to whether viewers have interacted with the streamer.
[0099] In one possible implementation, the interaction status can be represented as T and the conversion status as Y. Then, training labels can be added to the third interaction parameter corresponding to each viewing user based on the conversion status and interaction status of each viewing user, including (T=1, Y=1), (T=1, Y=0), (T=0, Y=1), and (T=0, Y=0).
[0100] Wherein, (T=1, Y=1) indicates that there is interaction between the viewing user and the broadcaster and the user is converted, (T=1, Y=0) indicates that there is interaction between the viewing user and the broadcaster and the user is not converted, (T=0, Y=1) indicates that there is no interaction between the viewing user and the broadcaster and the user is converted, and (T=0, Y=0) indicates that there is no interaction between the viewing user and the broadcaster and the user is not converted.
[0101] Understandably, the absence of interactive behavior refers to situations where viewers leave interactive parameters by asking questions or making comments, but the host does not respond to them.
[0102] In one possible implementation, considering that in actual live streaming scenarios, most viewers are users who do not interact and do not convert, the amount of data for the third interaction parameter (T=0, Y=0) may be large. In order to ensure the model training effect, the electronic device can randomly sample the third interaction parameter (T=0, Y=0) and select a portion of the third interaction parameter (T=0, Y=0) to input into the initial user value model for training.
[0103] Optionally, the initial user value model can be an Uplift+Transformer structure.
[0104] Optionally, considering that there may be multiple conversion goals, such as attention, purchase, and sales leads, multi-task training can be adopted, that is, multiple training labels can be set for the third interaction parameter, such as Y1 representing purchase, Y2 representing attention, etc. Based on this, the user value model can output values corresponding to multiple conversion goals.
[0105] Optionally, after the user value model is trained, it needs to be deployed. At this time, different conversion targets can be selected for different live streaming rooms.
[0106] In one example, Figure 3 is a schematic diagram of the user value model. Referring to Figure 3, the electronic device can input multiple third interaction parameters with added training labels into the embedding layer to generate sentence embedding vectors. The sentence embedding vectors are then input into the Transformer layer for processing. The sentence embedding vectors are context-encoded and output word vectors containing context information. Then, the vectors from different sources are concatenated through the concatenation layer to output a high-dimensional vector with comprehensive information. Finally, the concatenated vectors are nonlinearly transformed through the fully connected layer to output the final prediction result.
[0107] Optionally, considering that there are many live streaming parameters in the actual live streaming process, the electronic device first obtains the live streaming parameters in the current live stream, cleans the data, and then obtains the first interactive parameters.
[0108] In this embodiment, the live streaming parameters may include user comments from multiple viewers and live streaming content from multiple broadcasters.
[0109] Optionally, the live stream content may include the streamer's spoken announcements and comments from viewers.
[0110] Optionally, the electronic device can determine the first interaction parameter corresponding to each viewing user based on each user's message and the timestamp corresponding to the user's message, each streamer's live broadcast content and the timestamp corresponding to the streamer's live broadcast content.
[0111] In this embodiment, the electronic device can clean and integrate the live streaming parameters to obtain the session combination corresponding to each viewing user, namely the first interaction parameter.
[0112] The following provides a possible approach to inputting the first interaction parameters of each viewing user into a pre-trained user value model for processing, thereby obtaining the conversion value for each viewing user.
[0113] Specifically, the electronic device can input the first interaction parameter corresponding to each viewing user into the user value model for processing, obtain the first conversion probability and the second conversion probability corresponding to the viewing user, and determine the conversion value of the viewing user as the difference between the first conversion probability and the second conversion probability.
[0114] The first conversion probability represents the conversion probability of a viewer when there is interaction between the streamer and the viewer, while the second conversion probability represents the conversion probability of a viewer when there is no interaction between the viewer and the viewer.
[0115] Optionally, the user value model can process the first interaction parameter to output the conversion probability corresponding to the viewing user.
[0116] In this embodiment, in order to determine the positive impact of the anchor's interactive behavior on the conversion of viewers, two prediction processes can be performed using a user value model. Different interaction variables are set for the two prediction processes, thereby outputting a first conversion probability and a second conversion probability, respectively.
[0117] Understandably, the interaction variable is a binary variable, referring to whether there is any interaction between the streamer and the viewers.
[0118] In one example, please refer to Figure 4, which illustrates the conversion value calculation. The electronic device can input the first interaction parameter of the viewing user into the user value model and set the interaction variable T to 1, thereby obtaining a first conversion probability of conversion variable Y=1 for the viewing user when T=1. In addition, the electronic device also needs to set the interaction variable T to 0, thereby obtaining a second conversion probability of conversion variable Y=1 for the viewing user when T=0. The difference between the two probabilities is then determined as the conversion value corresponding to the viewing user.
[0119] Optionally, in order to further improve the conversion rate of viewers, the host needs to interact with the first target users with high conversion rates. Therefore, when determining the interaction strategy for the first target users, the electronic device can first sort the first target users.
[0120] The following provides a possible implementation method for determining the corresponding interaction strategy for each primary target user using a pre-trained response model.
[0121] Specifically, the electronic device can determine the response order for each first target user based on the conversion value corresponding to each first target user, determine the users to be responded to in turn according to the response order corresponding to each first target user, input the first interaction parameters corresponding to the users to be responded to into the response model for processing, and obtain the interaction strategy corresponding to the users to be responded to.
[0122] Optionally, the electronic device can sort the first target users according to their conversion values from high to low, thereby determining the response order for each first target user.
[0123] In this embodiment, if the user value model has only one conversion goal, it can be sorted directly according to the conversion value corresponding to the conversion goal. However, if the user value model has multiple conversion goals, such as both purchase and attention, the electronic device can perform mixed sorting of the first target users according to the preset recommendation algorithm and the conversion value corresponding to each conversion goal.
[0124] The conversion goals can include, for example, following, purchasing, or being interested. The mixed sorting refers to the weighted average obtained by taking a weighted average based on the weight of each conversion goal and sorting them in descending order.
[0125] In one possible implementation, the preset recommendation algorithm can be a click-through rate (CTR) algorithm.
[0126] Optionally, in order to ensure the effectiveness of the generated interactive strategy, the electronic device can also input the output of the user value model, the interactive goal, the duration of the last record in the first interactive parameter of the first target user from the current time, the duration of the first target user's entry into the live broadcast room, etc. into the response model.
[0127] The output of the user value model refers to the conversion probability of the first target user. The interaction goal can be to guide the user to pay attention, guide the user to purchase, or guide the user to leave sales leads, etc.
[0128] Understandably, this interaction strategy needs to be relevant to the interaction goal, such as prompting users with shopping information, guiding users to pay attention, analyzing the products that users are interested in, etc.
[0129] Optionally, in order to ensure the processing accuracy of the response model, it is also necessary to iteratively train the response model based on the interaction parameters generated in the actual application process.
[0130] In this embodiment, in order to improve the richness and diversity of interactive strategies, electronic devices can not only generate interactive strategies through response models and send them to the broadcaster, but also encourage the broadcaster to respond to users on their own in order to explore more and better response methods.
[0131] In one possible implementation, the interaction ratios of the model and the host can be set according to the actual application. In one example, the model interaction ratio can be 80%, and the host interaction ratio can be 20%. That is, the electronic device can input 80% of the first target user's first interaction parameters into the response model for processing, and send 20% of the first target user's first interaction parameters to the host, who can then respond to the first target user's questions or initiate topic guidance.
[0132] Optionally, considering the possibility that there may be no first target user, the electronic device can determine at least one second target user from multiple viewers based on the conversion value of each viewer, input the first interaction parameter corresponding to the second target user into the response model for processing, and obtain the interaction strategy corresponding to the second target user.
[0133] In this embodiment, if there is no conversion value greater than a preset value, the electronic device can find at least one second target user with a larger conversion value among multiple viewing users and generate an interaction strategy corresponding to the second target user.
[0134] In one possible implementation, the electronic device can lower a preset value and select a second target user from multiple viewing users according to the lowered preset value.
[0135] To perform the corresponding steps in the above embodiments and various possible methods, an implementation of a live interactive device is given below. Optionally, please refer to Figure 5, which is a functional block diagram of a live interactive device provided in this embodiment. It should be noted that the basic principle and technical effects of the live interactive device provided in this embodiment are the same as those in the above embodiments. For the sake of brevity, any parts not mentioned in this embodiment can be referred to the corresponding content in the above embodiments. The live interactive device includes: an acquisition module 200, a processing module 210, and a determination module 220.
[0136] The acquisition module 200 is configured to acquire the first interaction parameters corresponding to each viewing user in the current live stream.
[0137] Understandably, the acquisition module can be configured to perform the above step S20.
[0138] The processing module 210 is configured to input the first interaction parameters corresponding to each viewing user into a pre-trained user value model for processing, and obtain the conversion value corresponding to each viewing user; the conversion value represents the promoting effect of the interaction behavior between the anchor and the viewing user on the conversion of the viewing user.
[0139] Understandably, the processing module 210 can be configured to perform the above step S21.
[0140] The determining module 220 is configured to determine the first target user whose conversion value reaches a preset value from multiple viewers based on the conversion value corresponding to each viewer.
[0141] Understandably, the determining module 220 can be configured to perform the above step S22.
[0142] The determining module 220 is also configured to determine the corresponding interaction strategy for each first target user through a pre-trained response model, and send the interaction strategy to the anchor so that the anchor can interact with the first target user according to the interaction strategy.
[0143] Understandably, the determining module 220 can be configured to perform the above step S23.
[0144] Optionally, the acquisition module 200 is further configured to acquire live streaming parameters in the current live stream; the live streaming parameters include user comments from multiple viewers and live streaming content from multiple broadcasters; and the first interaction parameters corresponding to each viewer are determined based on each user comment and the timestamp corresponding to the user comment, each broadcaster's live streaming content and the timestamp corresponding to the broadcaster's live streaming content.
[0145] Optionally, the processing module 210 is further configured to input the first interaction parameter corresponding to each viewing user into the user value model for processing, thereby obtaining the first conversion probability and the second conversion probability corresponding to the viewing user; wherein, the first conversion probability represents the conversion probability of the viewing user when there is an interaction between the anchor and the viewing user, and the second conversion probability represents the conversion probability of the viewing user when there is no interaction between the user and the viewing user; the difference between the first conversion probability and the second conversion probability is determined as the conversion value of the viewing user.
[0146] Optionally, the determining module 220 is further configured to determine the response order for each first target user based on the conversion value corresponding to each first target user; determine the users to be responded to in sequence according to the response order corresponding to each first target user; input the first interaction parameters corresponding to the users to be responded to into the response model for processing; and obtain the interaction strategy corresponding to the users to be responded to.
[0147] Optionally, the determining module 220 is further configured to, if there is no first target user, determine at least one second target user from multiple viewers based on the conversion value of each viewer, input the first interaction parameter corresponding to the second target user into the response model for processing, and obtain the interaction strategy corresponding to the second target user.
[0148] Optionally, the live streaming interaction model also includes a model training module, which is configured to acquire second interaction parameters between multiple converted users and the streamer, as well as industry knowledge data; randomly select interaction parameters to be processed from the multiple second interaction parameters according to a preset ratio, truncate the interaction parameters to be processed, and obtain truncated interaction parameters to be processed; input the second interaction parameters, the truncated interaction parameters to be processed, and the industry knowledge data into a preset initial response model for training, and obtain a trained response model.
[0149] Optionally, the model training module is also configured to obtain third interaction parameters corresponding to multiple viewing users; add training labels to the third interaction parameters corresponding to each viewing user based on the conversion and interaction status of each viewing user; and input the third interaction parameters with added training labels into a pre-set initial user value model for training to obtain a trained user value model.
[0150] Optionally, the above-mentioned modules can be stored in the memory shown in Figure 1 in the form of software or firmware, or embedded in the operating system (OS) of the electronic device, and can be executed by the processor in Figure 1. Meanwhile, the data, program code, etc., required to execute the above-mentioned modules can be stored in the memory.
[0151] This disclosure also provides a computer-readable storage medium storing a computer program thereon, which, when executed by a processor, can implement the live interactive method provided in this disclosure.
[0152] In the several embodiments provided in this disclosure, it should be understood that the disclosed apparatus and methods can also be implemented in other ways. The apparatus embodiments described above are merely illustrative; for example, the flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of apparatus, methods, and computer program products according to various embodiments of this disclosure. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code, which contains one or more executable instructions configured to perform a specified logical function. It should also be noted that in some alternative implementations, the functions marked in the blocks may occur in a different order than those marked in the drawings. For example, two consecutive blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram and / or flowchart, and combinations of blocks in block diagrams and / or flowcharts, can be implemented using a dedicated hardware-based system that performs the specified function or action, or using a combination of dedicated hardware and computer instructions.
[0153] In addition, the functional modules in the various embodiments of this disclosure can be integrated together to form an independent part, or each module can exist independently, or two or more modules can be integrated to form an independent part.
[0154] If the aforementioned functions are implemented as software functional modules and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this disclosure, in essence, or the part that contributes to related technologies, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this disclosure. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.
[0155] The above description is merely an optional embodiment of this disclosure and is not intended to limit this disclosure. Various modifications and variations can be made to this disclosure by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this disclosure should be included within the scope of protection of this disclosure. Industrial applicability
[0156] By adopting the above scheme, the conversion value of each viewing user is determined through the first interaction parameter corresponding to each viewing user and the pre-trained user value model. This conversion value represents the promotional effect of the interaction behavior between the anchor and the viewing user on the conversion of that viewing user. Based on the conversion value corresponding to each viewing user, the first target user whose conversion value reaches the preset value is determined from multiple viewing users. The interaction strategy corresponding to each first target user is determined through the pre-trained response model and sent to the anchor. In this way, the anchor can prioritize interacting with each first target user according to the interaction strategy corresponding to each first target user, thereby promoting user conversion and improving the live broadcast effect.
Claims
1. A live streaming interactive method, characterized in that, The method includes: Get the first interaction parameters for each user watching the current live stream; The first interaction parameters corresponding to each of the viewers are input into a pre-trained user value model for processing to obtain the conversion value corresponding to each of the viewers; the conversion value represents the promoting effect of the interaction behavior between the anchor and the viewers on the conversion of the viewers. Based on the conversion value corresponding to each of the viewing users, determine the first target user whose conversion value reaches a preset value from among the multiple viewing users; For each of the first target users, an interaction strategy corresponding to the first target user is determined through a pre-trained response model, and the interaction strategy is sent to the anchor so that the anchor can interact with the first target user according to the interaction strategy.
2. The method according to claim 1, characterized in that, The step of obtaining the first interaction parameters corresponding to each viewing user in the current live stream includes: Obtain the live streaming parameters in the current live stream; the live streaming parameters include user comments from multiple viewers and live streaming content from multiple broadcasters. Based on each user comment and its corresponding timestamp, each streamer's live stream content and its corresponding timestamp, the first interaction parameter corresponding to each viewing user is determined.
3. The method according to claim 1 or 2, characterized in that, The step of inputting the first interaction parameters corresponding to each of the viewing users into a pre-trained user value model for processing to obtain the conversion value corresponding to each of the viewing users includes: For each of the viewing users, the first interaction parameters corresponding to the viewing user are input into the user value model for processing to obtain the first conversion probability and the second conversion probability corresponding to the viewing user. Wherein, the first conversion probability represents the conversion probability of the viewing user when there is an interactive behavior between the broadcaster and the viewing user, and the second conversion probability represents the conversion probability of the viewing user when there is no interactive behavior between the user and the viewing user; The difference between the first conversion probability and the second conversion probability is determined as the conversion value of the viewing user.
4. The method according to any one of claims 1-3, characterized in that, The step of determining the interaction strategy for each of the first target users using a pre-trained response model includes: The response order for each of the first target users is determined based on the conversion value corresponding to each of the first target users. According to the reply order of each of the first target users, the users to be replied to are determined in sequence, and the first interaction parameters corresponding to the users to be replied to are input into the reply model for processing to obtain the interaction strategy corresponding to the users to be replied to.
5. The method according to any one of claims 1-4, characterized in that, The method further includes: If the first target user does not exist, at least one second target user is determined from the multiple viewers based on the conversion value of each viewer. The first interaction parameter corresponding to the second target user is input into the response model for processing to obtain the interaction strategy corresponding to the second target user.
6. The method according to any one of claims 1-5, characterized in that, The response model is trained through the following steps: Acquire secondary interaction parameters and industry knowledge data between multiple converted users and the streamer; According to a preset ratio, an interactive parameter to be processed is randomly determined from multiple second interactive parameters, and the interactive parameter to be processed is truncated to obtain the truncated interactive parameter to be processed. The second interaction parameter, the truncated interaction parameter to be processed, and the industry knowledge data are input into a preset initial response model for training to obtain a trained response model.
7. The method according to any one of claims 1-6, characterized in that, The user value model is trained through the following steps: Obtain third-party interaction parameters corresponding to multiple viewing users; Based on the conversion and interaction of each viewing user, training labels are added to the third interaction parameters corresponding to each viewing user. The third interaction parameter with the added training label is input into the pre-set initial user value model for training, and a trained user value model is obtained.
8. The method according to any one of claims 1-7, characterized in that, The user value module has at least one conversion target; the step of inputting the first interaction parameters corresponding to each viewing user into a pre-trained user value model for processing to obtain the conversion value corresponding to each viewing user includes: The first interaction parameters corresponding to each of the viewing users are input into a pre-trained user value model for processing to obtain the conversion value corresponding to each of the viewing users under the at least one conversion target.
9. The method according to claim 8, characterized in that, The conversion target is one; the step of determining the interaction strategy corresponding to each first target user through a pre-trained response model includes: Based on the conversion value corresponding to each first target user under the aforementioned conversion goal, sort each first target user and determine the response order corresponding to each first target user; Based on the reply order of each of the first target users, the users to be replied to are determined sequentially; The first interaction parameter corresponding to the user to be replied to is input into the reply model to obtain the interaction strategy corresponding to the user to be replied to.
10. The method according to claim 8 or 9, characterized in that, The conversion targets are multiple; the step of determining the interaction strategy corresponding to each first target user through a pre-trained response model includes: The weighted average value corresponding to each first target user under the multiple conversion goals is obtained by taking a weighted average value. Based on the weighted average value corresponding to each first target user, the first target users are sorted in a mixed order to determine the response order corresponding to each first target user. Based on the reply order of each of the first target users, the users to be replied to are determined sequentially; The first interaction parameter corresponding to the user to be replied to is input into the reply model to obtain the interaction strategy corresponding to the user to be replied to.
11. The method according to any one of claims 1-10, characterized in that, The interaction strategy includes responses to unanswered questions raised by the first target user, or guidance on topics of interest to the first target user.
12. The method according to any one of claims 2-11, characterized in that, The live stream content includes: the broadcast content delivered by the broadcaster and the broadcaster's comments on the user's messages.
13. A live interactive device, characterized in that, The device includes: The acquisition module is configured to acquire the first interaction parameters corresponding to each viewing user in the current live stream. The processing module is configured to input the first interaction parameters corresponding to each of the viewers into a pre-trained user value model for processing, and obtain the conversion value corresponding to each of the viewers; the conversion value represents the promoting effect of the interaction behavior between the anchor and the viewers on the conversion of the viewers. The determination module is configured to determine, from among the multiple viewers, a first target user whose conversion value reaches a preset value based on the conversion value corresponding to each of the viewers; The determining module is further configured to determine the interaction strategy corresponding to each first target user through a pre-trained response model, and send the interaction strategy to the anchor so that the anchor can interact with the first target user according to the interaction strategy.
14. An electronic device, characterized in that, It includes a processor and a memory, the memory storing a computer program executable by the processor, the processor being able to execute the computer program to implement the method of any one of claims 1-12.
15. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by a processor, it implements the method described in any one of claims 1-12.