Quick group chat voice reply method and device, storage medium and computer equipment

Abstract

According to the group chat voice quick reply method and device, the storage medium and the computer equipment provided by the invention, after the output voice of the user in the group chat room is received, the reply text carrying the metadata is generated for the user, and the reply text is added to the text priority queue; when queue updating is detected each time, a reply priority is given to each reply text in the queue based on the metadata, so that the intelligence of a reply sequence is improved; and then marking the reply text with the highest priority in the queue as a to-be-processed text, so that a generation bit of the audio reply queue pulls the to-be-processed text when being idle, further converting the to-be-processed text voice into a reply audio through the generation bit, and when detecting that a playing bit of the audio reply queue is null, playing the reply audio through the generation bit. The reply audio in the generation bit is sent to the playing bit to be played; through a double-buffering mechanism of the generation bit and the playing bit, audio generation and playing can be decoupled and processed in parallel, so that the speed of voice reply is quickly improved.

Description

Methods, devices, storage media, and computer equipment for quick voice replies in group chats Technical Field

[0001] This application relates to the field of computer technology, and in particular to a method, apparatus, storage medium, and computer device for quick voice reply in group chat. Background Technology

[0002] With the rapid development of artificial intelligence technology, AI is being applied more and more widely in the field of social entertainment. Various voice group chat platforms are gradually introducing AI, enabling it to chat and interact with real users in real time within the room, thus making communication more diverse and interesting, and bringing users a novel and unique experience.

[0003] Currently, in the field of AI voice interaction, the technology is relatively mature for one-on-one dialogue scenarios, and AI can communicate with users relatively smoothly. However, when faced with multiple real users speaking simultaneously, engaging in lively discussions and taking turns speaking, existing AI typically responds sequentially. As time accumulates, the AI ​​may still be responding to past conversations, resulting in significant delays in responses and consequently affecting the user experience in group voice chats. Summary of the Invention

[0004] The purpose of this application is to at least address one of the aforementioned technical deficiencies, particularly the technical deficiency in existing technologies where AI responses are severely delayed in voice group chat scenarios, thereby affecting the user experience in voice group chats.

[0005] This application provides a method for quickly replying to voice messages in group chats, the method comprising:

[0006] The system receives voice output from users in a group chat room in real time, and after generating a reply text carrying metadata for the voice output, adds the reply text to a text priority queue.

[0007] After detecting an update to the text priority queue, the metadata of each reply text in the text priority queue is comprehensively analyzed based on the priority allocation strategy to obtain the reply priority of each reply text.

[0008] The reply text with the highest reply priority in the text priority queue is marked as the text to be processed, so that the generation position of the preset audio reply queue can pull the text to be processed when it is idle;

[0009] The text to be processed is converted into speech using the generated bit to obtain the response audio. When the playback bit of the audio response queue is detected to be empty, the response audio in the generated bit is sent to the playback bit for playback.

[0010] Optionally, generating a response text carrying metadata for the output speech includes:

[0011] The output speech is converted into output text using a speech recognition model, and the user information corresponding to the output speech is obtained.

[0012] The output text and the user information are input into a preset large language model, so that the large language model performs intent understanding and response generation on the output text based on the user information, and obtains response text carrying metadata; the metadata includes the text generation time and user information.

[0013] Optionally, the step of comprehensively analyzing the metadata of each reply text in the text priority queue based on the priority allocation strategy to obtain the reply priority of each reply text includes:

[0014] For each reply text, obtain the user information and text generation time from the metadata of that reply text; wherein, the user information includes membership level, consumption level, and interaction level;

[0015] The response waiting time is calculated based on the current system time and the text generation time, and the user contribution value is calculated based on the membership level, the consumption level, and the interaction level.

[0016] Obtain the balance weight of the group chat room, and determine the reply priority of the reply text based on the reply waiting time, the user contribution value, and the balance weight.

[0017] Optionally, the expression for calculating the response priority includes:

[0018]

[0019] In the formula, Indicates weight priority; Indicates the response waiting time; Indicates the balancing weights; This represents the user's contribution value.

[0020] Optionally, the step of converting the text to be processed into speech using the generated bits to obtain the response audio includes:

[0021] When the generating bit receives the text to be processed, it performs speech conversion on the text to be processed through the speech synthesis model built into the generating bit to obtain the response audio.

[0022] Optionally, the method further includes:

[0023] The reply waiting time for each reply text in the text priority queue is calculated periodically by using the cleanup task set in the text priority queue.

[0024] Reply texts whose reply waiting time exceeds a preset threshold are removed from the text priority queue.

[0025] Optionally, the method further includes:

[0026] If any error is detected in the output speech during the response text generation process, or if any error is detected in the text to be processed during the response audio generation process, then the output speech or the text to be processed will be discarded.

[0027] This application also provides a group chat voice quick reply device, including:

[0028] The voice receiving module is used to receive the voice output of users in the group chat room in real time, and after generating a reply text carrying metadata for the output voice, add the reply text to the text priority queue.

[0029] The priority allocation module is used to perform comprehensive analysis on the metadata of each reply text in the text priority queue based on the priority allocation strategy after detecting an update of the text priority queue, so as to obtain the reply priority of each reply text.

[0030] The text retrieval module is used to mark the reply text with the highest reply priority in the text priority queue as the text to be processed, so that the generation position of the preset audio reply queue can retrieve the text to be processed when it is idle;

[0031] The voice response module is used to convert the text to be processed into speech through the generation bit to obtain the response audio, and when it is detected that the playback bit of the audio response queue is empty, the response audio in the generation bit is sent to the playback bit for playback.

[0032] This application also provides a storage medium storing computer-readable instructions, which, when executed by one or more processors, cause the one or more processors to perform the steps of the group chat voice quick reply method as described in any of the above embodiments.

[0033] This application also provides a computer device, including: one or more processors, and memory;

[0034] The memory stores computer-readable instructions, which, when executed by the one or more processors, perform the steps of the group chat voice quick reply method as described in any of the above embodiments.

[0035] As can be seen from the above technical solutions, the embodiments of this application have the following advantages:

[0036] The group chat voice rapid reply method, device, storage medium, and computer equipment provided in this application can receive the output voice of users in a group chat room in real time, ensuring that no user's communication information is missed. Then, it can generate reply text carrying metadata for the received output voice and add the reply text to a text priority queue, providing a data basis for subsequent voice replies. After detecting an update to the text priority queue, it can comprehensively analyze the metadata of each reply text in the text priority queue based on a priority allocation strategy, considering factors such as the timeliness of the topic and the user's activity level, and assign a corresponding reply priority to each reply text. Furthermore, the intelligence of the response order can be improved. Next, the highest-priority response text in the text priority queue can be marked as the text to be processed, so that the generation bit of the preset audio response queue can pull the text to be processed when idle, thus achieving ordered response text processing. During response text processing, the text to be processed can first be converted into speech through the generation bit to obtain the response audio. When the playback bit of the audio response queue is detected to be empty, the response audio in the generation bit is sent to the playback bit for playback. Through the double buffering mechanism of the generation bit and the playback bit, this application can decouple and parallelize audio generation and playback, thereby rapidly improving the speed of voice response. Attached Figure Description

[0037] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 is a flowchart illustrating a method for quickly replying to voice messages in a group chat, as provided in an embodiment of this application.

[0039] Figure 2 is a flowchart illustrating a response priority calculation process provided in an embodiment of this application;

[0040] Figure 3 is a structural schematic diagram of a group chat voice quick reply device provided in an embodiment of this application;

[0041] Figure 4 is a schematic diagram of the internal structure of a computer device provided in an embodiment of this application. Detailed Implementation

[0042] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0043] Currently, in the field of AI voice interaction, the technology is relatively mature for one-on-one dialogue scenarios, and AI can communicate with users relatively smoothly. However, when faced with multiple real users speaking simultaneously, engaging in lively discussions and taking turns speaking, existing AI typically responds sequentially. As time accumulates, the AI ​​may still be responding to past conversations, resulting in significant delays in responses and consequently affecting the user experience in group voice chats.

[0044] In one embodiment, as shown in FIG1, FIG1 is a flowchart illustrating a method for quickly replying to voice messages in a group chat according to an embodiment of this application; this application also provides a method for quickly replying to voice messages in a group chat, specifically including the following:

[0045] S110: Receives voice output from users in the group chat room in real time, and after generating a reply text carrying metadata for the voice output, adds the reply text to the text priority queue.

[0046] In this step, within the group chat room, the computer device can receive the voice output from users in the room in real time, ensuring that no user's communication information is missed. Then, it can generate a reply text carrying metadata for the received voice output and add the reply text to the text priority queue, providing a data basis for subsequent voice replies.

[0047] Specifically, the computer device can continuously monitor and receive the output voice messages sent by each user in the AI ​​group chat room in real time. Upon receiving each output voice message, the computer device can perform speech recognition and semantic analysis on the output voice message, and then generate a reply text that matches the content of the output voice message. In addition, during the reply text generation process, the computer device can also add corresponding metadata to it to characterize information such as the source user, reply time, and semantic topic of the reply text. Subsequently, the computer device can write the reply text carrying metadata into a preset text priority queue. Through this text priority queue, this application can uniformly cache, dynamically sort, and output the reply texts of different output voice messages, thereby avoiding resource conflicts and chaos caused by processing multiple reply texts simultaneously.

[0048] S120: After detecting an update to the text priority queue, perform a comprehensive analysis of the metadata of each reply text in the text priority queue based on the priority allocation strategy to obtain the reply priority of each reply text.

[0049] In this step, whenever an update to the text priority queue is detected, the computer device can perform a comprehensive analysis of the metadata of each reply text in the text priority queue based on a priority allocation strategy, such as the timeliness of the topic and the user's activity level, and assign a corresponding reply priority to each reply text, thereby improving the intelligence of the reply order.

[0050] Specifically, when a new reply text is added to the text priority queue, or when a reply text is extracted from the text priority queue for processing, the computer device can trigger a priority reallocation mechanism. Based on a preset priority allocation strategy, the computer device comprehensively analyzes and processes the metadata carried by each reply text in the text priority queue. During this analysis, the computer device can dynamically evaluate each reply text by considering factors such as the timeliness of the reply text's generation, the real-time activity level of the user speaking in the room, and its relevance to the group chat topic, and assign a corresponding reply priority to each reply text accordingly.

[0051] Understandably, by continuously adjusting and optimizing the priority of reply texts in the text priority queue, computer devices can prioritize reply texts that are more timely and have greater discussion value when processing replies in the future, thereby avoiding delayed responses to outdated topics and improving the intelligence and flexibility of AI reply order.

[0052] S130: Mark the reply text with the highest reply priority in the text priority queue as the text to be processed, so that the generation bit of the preset audio reply queue can pull the text to be processed when idle.

[0053] In this step, after determining the response priority of each response text in the text priority queue through step S120, the computer device can mark the response text with the highest response priority in the text priority queue as the text to be processed, so that the generation position of the preset audio response queue can pull the text to be processed when idle, thereby realizing the orderly processing of response text.

[0054] Specifically, after each round of priority evaluation, the computer device can mark the highest-priority response text in the text priority queue as the text to be processed, serving as the target object for the audio response queue in the next round of voice response generation. This audio response queue employs a double-buffering mechanism, including a generation bit and a playback bit. The generation bit can be used to generate audio from the text to be processed. Therefore, when the generation bit is idle, the text to be processed can be retrieved from the text priority queue, triggering the audio response process, thereby achieving ordered and controllable response text processing.

[0055] S140: The generated bit performs speech conversion on the text to be processed to obtain the response audio, and when the playback bit of the audio response queue is detected to be empty, the response audio in the generated bit is sent to the playback bit for playback.

[0056] In this step, after the text to be processed is retrieved from the text priority queue through the generation bit of the audio response queue in step S130, the computing device can first perform speech conversion on the text to be processed through the generation bit to obtain the response audio. When it is detected that the playback bit of the audio response queue is empty, the response audio in the generation bit is sent to the playback bit for playback. Through the double buffering mechanism of the generation bit and the playback bit, this application can decouple and process audio generation and playback in parallel, thereby rapidly improving the voice response rate.

[0057] Specifically, after the audio response queue triggers the audio response process, the computer device can first use the generation bit in the audio response queue to perform speech conversion processing on the text to be processed, thereby converting the text content into the corresponding audio stream, and temporarily storing the generated response audio in the generation bit. During this process, the computer device can continuously monitor the status of the playback bit in the audio response queue. When it detects that the playback bit is in an idle state, it can send the synthesized response audio from the generation bit to the playback bit, and the playback bit controls the actual output and playback of the response audio.

[0058] Understandably, the generation and playback bits in the audio response queue operate independently, forming a double-buffered processing mechanism that decouples and parallelizes the audio generation and playback processes. While the playback bit is playing the response audio, the generation bit can process the next text to be processed in parallel. When the response audio finishes playing, the playback bit is cleared, and the newly generated response audio in the generation bit can be transferred to the playback bit. The generation bit can then retrieve the text to be processed from the text priority queue again for processing. This cycle continues, allowing this application to generate the next response audio segment before the previous one has finished playing, thus significantly reducing the waiting time for voice responses.

[0059] In the above embodiments, in a group chat room, the voice output of users in the room can be received in real time to ensure that no user's communication information is missed. Then, a reply text carrying metadata can be generated for the received voice output, and the reply text is added to the text priority queue to provide a data basis for subsequent voice replies. After detecting an update to the text priority queue, the metadata of each reply text in the text priority queue can be comprehensively analyzed based on a priority allocation strategy, such as the timeliness of the topic and the user's activity level, to assign a corresponding reply priority to each reply text, thereby improving the intelligence of the reply order. Then, the reply text with the highest reply priority in the text priority queue can be marked as the text to be processed, so that the generation position of the preset audio reply queue can pull the text to be processed when idle, thereby realizing the orderly processing of reply text. When processing reply text, the text to be processed can be converted into speech through the generation position to obtain reply audio. When the playback position of the audio reply queue is detected to be empty, the reply audio in the generation position is sent to the playback position for playback. Through the double buffering mechanism of the generation position and the playback position, this application can decouple and parallelize the audio generation and playback, thereby rapidly improving the voice reply rate.

[0060] In one embodiment, the process of generating response text carrying metadata for the output speech in step S110 may include:

[0061] S111: Convert the output speech into output text using a speech recognition model, and obtain the user information corresponding to the output speech.

[0062] S112: Input the output text and user information into the preset large language model, so that the large language model can perform intent understanding and response generation on the output text based on the user information, and obtain response text carrying metadata; the metadata includes the text generation time and user information.

[0063] In this embodiment, after receiving the output speech, the computer device can convert the output speech into output text through a speech recognition model and obtain the user information corresponding to the output speech. Then, the output text and user information can be input into a preset large language model so that the large language model can perform intent understanding and response generation based on the user information to obtain response text carrying metadata.

[0064] Specifically, the computer device can invoke a preset speech recognition model to process the output speech, converting continuous speech signals into corresponding output text. Simultaneously, it can retrieve user information corresponding to the output speech from the session management module, representing the specific user identity of the speech source and their interaction status in the group chat room. Subsequently, the computer device can input the output text and user information together into a preset large language model. This allows the large language model to understand the semantic content while combining user information to perform intent recognition and contextual analysis on the output text, generating a reply text that matches the group chat topic. This reply text also includes corresponding metadata, such as text generation time and user information, providing a reliable basis for subsequent priority evaluation and queue scheduling.

[0065] In one embodiment, as shown in Figure 2, which is a flowchart illustrating a response priority calculation process provided by an embodiment of this application, step S120, which involves comprehensively analyzing the metadata of each response text in the text priority queue based on a priority allocation strategy to obtain the response priority of each response text, may include:

[0066] S121: For each reply text, obtain the user information and text generation time from the metadata of that reply text; wherein, the user information includes membership level, consumption level and interaction level.

[0067] S122: Calculate the response waiting time based on the current system time and the text generation time, and calculate the user contribution value based on the membership level, consumption level, and interaction level.

[0068] S123: Obtain the balance weight of the group chat room, and determine the reply priority of the reply text based on the reply waiting time, user contribution value, and balance weight.

[0069] In this embodiment, for each reply text in the text priority queue, the computer device can obtain user information and text generation time from the metadata of the reply text. The user information includes membership level, consumption level, and interaction level. Then, the reply waiting time can be calculated based on the current system time and the text generation time, and the user contribution value can be calculated based on the membership level, consumption level, and interaction level. Furthermore, the computer device can also obtain the balancing weight of the group chat room, and thus determine the reply priority of the reply text based on the reply waiting time, user contribution value, and balancing weight.

[0070] Specifically, the metadata of the reply text mainly consists of user information and the text generation time. The user information includes the corresponding user's membership level, consumption level, and interaction level on the platform, primarily reflecting the user's overall participation value in the group chat room. Therefore, based on the time difference between the text generation time and the current system time, the computer device can calculate the reply waiting time corresponding to the reply text, representing the length of time the reply text remains in the queue. Simultaneously, the computer device can also comprehensively calculate the user contribution value based on the various levels of information in the user information, according to preset mapping rules or weighted calculation models, thereby quantifying the activity level and value weight of different users in the group chat room. Subsequently, the computer device can obtain the balance weight corresponding to the current group chat room and, using the reply waiting time, user contribution value, and balance weight as input parameters, comprehensively evaluate the reply priority of the reply text.

[0071] This balancing weight can be used to adjust the influence ratio between response waiting time and user contribution value in the priority calculation process. For example, in a group chat room that focuses on entertainment and interaction and emphasizes a lively atmosphere, this application can adjust the balancing weight to favor response waiting time; in a group chat room that is dominated by a streamer or has a strong paid attribute, this application can adjust the balancing weight to favor user contribution value.

[0072] Understandably, by introducing time and contribution factors into the calculation of response priority and using balanced weights for dynamic adjustment, the bias caused by responding solely based on time sequence or a single user attribute can be effectively avoided. This can improve the rationality and intelligence of AI responses, and enhance user satisfaction and participation in group chat interactions.

[0073] In one embodiment, the expression for calculating the response priority in step S120 may include:

[0074]

[0075] In the formula, Indicates weight priority; Indicates the response waiting time; Indicates the balancing weights; This represents the user's contribution value.

[0076] In this embodiment, the core objective of the calculation expression is to determine the order of AI responses by dynamically balancing the real-time nature of replies and the importance of users in scenarios involving simultaneous interaction by multiple users. Specifically, the user contribution value can be smoothed using the logarithmic function log10 to prevent users with larger contribution values ​​from overwhelmingly influencing priority, thereby maintaining the rationality and fairness of the scheduling. It should be noted that in this application, the smaller the response priority value, the higher the priority.

[0077] In one embodiment, the process of generating bits to convert the text to be processed into speech and obtaining the response audio in step S140 may include:

[0078] S141: When the generation bit receives the text to be processed, it performs speech conversion on the text to be processed through the speech synthesis model built into the generation bit to obtain the response audio.

[0079] In this embodiment, after the audio response queue pulls the text to be processed selected and marked by the text priority queue into the generation position, the computer device can call the speech synthesis model built into the generation position to perform speech conversion processing on the text to be processed, mapping the text-based response content into the corresponding speech signal, thereby generating a response audio that can be directly used for playback.

[0080] Specifically, during the speech synthesis process, the speech synthesis model can combine preset pronunciation rules, intonation parameters, and speech rate control strategies to process the response audio in a natural way, so as to ensure the clarity and coherence of the generated speech, thereby improving the user's auditory experience.

[0081] In one embodiment, the method may further include:

[0082] S150: The reply waiting time for each reply text in the text priority queue is calculated periodically through the cleanup task set in the text priority queue.

[0083] S160: Remove reply texts from the text priority queue whose reply waiting time exceeds a preset threshold.

[0084] In this embodiment, a cleanup task is set in the text priority queue. Through this cleanup task, the computer device can periodically calculate the reply waiting time of each reply text in the text priority queue, and then clean up reply texts whose reply waiting time exceeds a preset threshold, so as to avoid a large number of outdated reply texts accumulating in the text priority queue.

[0085] Understandably, the cleanup tasks in the text priority queue are automatically triggered and executed by the computer device according to a preset time period. During task execution, the computer device can traverse each reply text in the text priority queue and calculate the reply waiting time for each reply text based on the current system time and the text generation time, thus reflecting the duration of the reply text's continuous stay in the queue. Furthermore, the computer device can compare each reply waiting time with a preset waiting time threshold. If the comparison result shows that the reply waiting time of a reply text exceeds the preset threshold, it indicates that the reply text no longer has the timeliness or interactive value for further reply. Therefore, the computer device can remove it from the text priority queue, avoiding delayed responses from AI to topics that have ended or clearly shifted, and also reducing ineffective computation and resource waste.

[0086] In one embodiment, the method may further include:

[0087] S170: If any error is detected in the output speech during the response text generation process, or if any error is detected in the text to be processed during the response audio generation process, the output speech or the text to be processed will be discarded.

[0088] In this embodiment, when the computer device is processing and responding to the output voice in the group chat room, if it detects any output voice failure, semantic parsing abnormality, or large language model generation error during the response text generation process, or if it detects an abnormality in the response audio generation process when performing speech synthesis processing on the selected text to be processed, the computer device can immediately trigger the abnormality handling mechanism and directly discard the corresponding output voice or text to be processed to avoid affecting queue scheduling and causing lag in AI voice response.

[0089] The group chat voice quick reply device provided in the embodiments of this application is described below. The group chat voice quick reply device described below can be referred to in correspondence with the group chat voice quick reply method described above.

[0090] In one embodiment, as shown in FIG3, FIG3 is a structural schematic diagram of a group chat voice quick reply device provided in an embodiment of the present application; the present application also provides a group chat voice quick reply device, including a voice receiving module 210, a priority allocation module 220, a text retrieval module 230, and a voice reply module 340, specifically including the following:

[0091] The voice receiving module 210 is used to receive the voice output of users in the group chat room in real time, and after generating a reply text carrying metadata for the output voice, add the reply text to the text priority queue.

[0092] The priority allocation module 220 is used to perform comprehensive analysis on the metadata of each reply text in the text priority queue based on the priority allocation strategy after detecting an update in the text priority queue, so as to obtain the reply priority of each reply text.

[0093] The text retrieval module 230 is used to mark the reply text with the highest reply priority in the text priority queue as the text to be processed, so that the generation position of the preset audio reply queue can retrieve the text to be processed when idle.

[0094] The voice response module 340 is used to convert the text to be processed into speech by generating a bit to obtain the response audio, and when it detects that the playback bit of the audio response queue is empty, it sends the response audio in the generating bit to the playback bit for playback.

[0095] In the above embodiments, in a group chat room, the voice output of users in the room can be received in real time to ensure that no user's communication information is missed. Then, a reply text carrying metadata can be generated for the received voice output, and the reply text is added to the text priority queue to provide a data basis for subsequent voice replies. After detecting an update to the text priority queue, the metadata of each reply text in the text priority queue can be comprehensively analyzed based on a priority allocation strategy, such as the timeliness of the topic and the user's activity level, to assign a corresponding reply priority to each reply text, thereby improving the intelligence of the reply order. Then, the reply text with the highest reply priority in the text priority queue can be marked as the text to be processed, so that the generation position of the preset audio reply queue can pull the text to be processed when idle, thereby realizing the orderly processing of reply text. When processing reply text, the text to be processed can be converted into speech through the generation position to obtain reply audio. When the playback position of the audio reply queue is detected to be empty, the reply audio in the generation position is sent to the playback position for playback. Through the double buffering mechanism of the generation position and the playback position, this application can decouple and parallelize the audio generation and playback, thereby rapidly improving the voice reply rate.

[0096] In one embodiment, the voice receiving module 210 may include:

[0097] The speech recognition submodule is used to convert the output speech into output text through a speech recognition model, and to obtain the user information corresponding to the output speech.

[0098] The text generation submodule is used to input the output text and user information into a preset large language model, so that the large language model can perform intent understanding and response generation based on the user information to obtain response text carrying metadata, including text generation time and user information.

[0099] In one embodiment, the priority allocation module 220 may include:

[0100] The metadata acquisition submodule is used to retrieve user information and text generation time from the metadata of each reply text; the user information includes membership level, consumption level, and interaction level.

[0101] The parameter calculation submodule is used to calculate the response waiting time based on the current system time and the text generation time, and to calculate the user contribution value based on the membership level, consumption level, and interaction level.

[0102] The priority calculation submodule is used to obtain the balance weight of the group chat room and determine the reply priority of the reply text based on the reply waiting time, user contribution value and balance weight.

[0103] In one embodiment, the expression for calculating the response priority in the priority allocation module 220 may include:

[0104]

[0105] In the formula, Indicates weight priority; Indicates the response waiting time; Indicates the balancing weights; This represents the user's contribution value.

[0106] In one embodiment, the voice response module 340 may include:

[0107] The speech conversion submodule is used to convert the text to be processed into speech using the speech synthesis model built into the generation unit after the generation unit receives the text to be processed, so as to obtain the response audio.

[0108] In one embodiment, the apparatus may further include:

[0109] The task triggering module is used to periodically calculate the response waiting time for each reply text in the text priority queue based on the cleanup tasks set in the text priority queue.

[0110] The queue cleanup module is used to remove reply texts from the text priority queue whose reply waiting time exceeds a preset threshold.

[0111] In one embodiment, the apparatus may further include:

[0112] The data discarding module is used to discard the output speech or the text to be processed if any error is detected in the output speech during the generation of the response text, or if any error is detected in the generation of the response audio.

[0113] In one embodiment, this application also provides a storage medium storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the group chat voice quick reply method as described in any of the above embodiments.

[0114] In one embodiment, this application also provides a computer device storing computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the group chat voice quick reply method as described in any of the above embodiments.

[0115] Schematic, as shown in FIG4, FIG4 is a schematic diagram of the internal structure of a computer device 300 provided in an embodiment of this application. The computer device 300 can be provided as a server. Referring to FIG4, the computer device 300 includes a processing component 302, which further includes one or more processors, and memory resources represented by memory 301 for storing instructions executable by the processing component 302, such as application programs. The application programs stored in memory 301 may include one or more modules, each corresponding to a set of instructions. Furthermore, the processing component 302 is configured to execute instructions to perform the group chat voice quick reply method of any of the above embodiments.

[0116] The computer device 300 may also include a power supply component 303 configured to perform power management of the computer device 300, a wired or wireless network interface 304 configured to connect the computer device 300 to a network, and an input / output (I / O) interface 305. The computer device 300 may operate on an operating system stored in memory 301, such as Windows Server™, Mac OS X™, Unix™, Linux™, Free BSD™, or similar.

[0117] Those skilled in the art will understand that the structure shown in Figure 4 is merely a block diagram of a portion of the structure related to the present application and does not constitute a limitation on the computer device to which the present application is applied. Specific computer devices may include more or fewer components than those shown in the figure, or may combine certain components, or may have different component arrangements.

[0118] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only 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 a process, method, article, or apparatus. Without further limitations, 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 said element.

[0119] The various embodiments in this specification are described in a progressive manner. Each embodiment focuses on the differences from other embodiments. The various embodiments can be combined as needed, and the same or similar parts can be referred to each other.

[0120] The above description of the disclosed embodiments enables those skilled in the art to make or use this application. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of this application. Therefore, this application is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for quickly replying to voice messages in group chats, characterized in that, The method includes: receiving voice output from users in a group chat room in real time, and after generating reply text carrying metadata for the voice output, adding the reply text to a text priority queue; after detecting an update to the text priority queue, performing comprehensive analysis on the metadata of each reply text in the text priority queue based on a priority allocation strategy to obtain the reply priority of each reply text; marking the reply text with the highest reply priority in the text priority queue as a text to be processed, so that the generation position of a preset audio reply queue can retrieve the text to be processed when idle; performing speech conversion on the text to be processed through the generation position to obtain reply audio, and sending the reply audio from the generation position to the playback position for playback when the playback position of the audio reply queue is detected to be empty.

2. The method for quickly replying to group chat voice messages according to claim 1, characterized in that, The step of generating a response text carrying metadata for the output speech includes: converting the output speech into output text using a speech recognition model, and obtaining user information corresponding to the output speech; inputting the output text and the user information into a preset large language model, so that the large language model performs intent understanding and response generation on the output text based on the user information, thereby obtaining a response text carrying metadata; the metadata includes the text generation time and user information.

3. The method for quickly replying to group chat voice messages according to claim 1, characterized in that, The step of comprehensively analyzing the metadata of each reply text in the text priority queue based on the priority allocation strategy to obtain the reply priority of each reply text includes: for each reply text, obtaining the user information and text generation time from the metadata of the reply text; wherein, the user information includes membership level, consumption level, and interaction level; calculating the reply waiting time based on the current system time and the text generation time, and calculating the user contribution value based on the membership level, the consumption level, and the interaction level; obtaining the balance weight of the group chat room, and determining the reply priority of the reply text based on the reply waiting time, the user contribution value, and the balance weight.

4. The method for quickly replying to group chat voice messages according to claim 3, characterized in that, The expression for calculating the response priority includes: In the formula, Indicates weight priority; Indicates the response waiting time; Indicates the balancing weights; This represents the user's contribution value.

5. The method for quick reply to group chat voice messages according to claim 1, characterized in that, The step of converting the text to be processed into speech using the generation bit to obtain a response audio includes: when the generation bit receives the text to be processed, it converts the text to speech using a speech synthesis model built into the generation bit to obtain a response audio.

6. The method for quickly replying to group chat voice messages according to claim 1, characterized in that, The method further includes: periodically calculating the reply waiting time for each reply text in the text priority queue through a cleanup task set in the text priority queue; and removing reply texts in the text priority queue whose reply waiting time exceeds a preset threshold.

7. The method for quickly replying to group chat voice messages according to claim 1, characterized in that, The method further includes: if any output speech is detected to have an error during the generation of the response text, or if any text to be processed is detected to have an error during the generation of the response audio, then the output speech or the text to be processed is discarded.

8. A group chat voice rapid reply device, characterized in that, include: The voice receiving module is used to receive the voice output of users in the group chat room in real time, and after generating a reply text carrying metadata for the output voice, add the reply text to the text priority queue. The priority allocation module is used to perform comprehensive analysis on the metadata of each reply text in the text priority queue based on the priority allocation strategy after detecting an update of the text priority queue, so as to obtain the reply priority of each reply text; the text retrieval module is used to mark the reply text with the highest reply priority in the text priority queue as the text to be processed, so that the generation position of the preset audio reply queue can retrieve the text to be processed when it is idle. The voice response module is used to convert the text to be processed into speech through the generation bit to obtain the response audio, and when it is detected that the playback bit of the audio response queue is empty, the response audio in the generation bit is sent to the playback bit for playback.

9. A storage medium, characterized in that: The storage medium stores computer-readable instructions that, when executed by one or more processors, cause the one or more processors to perform the steps of the group chat voice quick reply method as described in any one of claims 1 to 7.

10. A computer device, characterized in that, include: One or more processors, and a memory; the memory stores computer-readable instructions that, when executed by the one or more processors, perform the steps of the group chat voice quick reply method as described in any one of claims 1 to 7.

Images