Parent-child communication reminding system, method and device

Abstract

The embodiment of the invention provides a parent-child communication reminding system, method and device. The parent-child communication reminding system comprises an acquisition module for acquiring communication audio; extracting a target audio clip corresponding to the parent-child communication user from the communication audio, and sending the target audio clip to a processing module; the processing module inputs the target audio clip into a large language model, and performs text recall on the target audio clip through a three-way emotion recall unit contained in a retrieval unit in the large language model to obtain a candidate communication text; screening a target communication text from the candidate communication texts corresponding to the three emotion recall units according to the audio attribute information of the target audio clip; processing the target communication text and the communication type information corresponding to the target communication text through a generation unit in the large language model to obtain parent-child communication reminding information; the parent-child communication reminding information is sent to the reminding module; and the reminding module feeds back parent-child communication reminding information to the parent-child communication user according to a reminding strategy matched with the communication type information.

Description

Technical Field

[0001] The embodiments of this specification relate to the field of machine learning technology, and more particularly to parent-child communication reminder systems, methods, and devices. Background Art

[0002] At present, adolescents are facing tremendous pressure from academic, interpersonal, social and family aspects, and their psychological problems are becoming more and more serious. As the most important environmental factor in the growth of adolescents, the internal parent-child communication method has an important impact on the physical and mental health and growth of adolescents. Good family communication can provide emotional support and psychological security, help adolescents cope with the pressure and challenges of adolescence, and reduce the occurrence of psychological problems such as anxiety and depression. On the contrary, a family environment with poor communication or full of conflicts is likely to increase the risk of adolescents having mental and psychological problems. However, since the way of communication is affected by the emotions and personality of the communicators, different dialogue backgrounds and different personalities may lead to ineffective communication and even bring negative effects, thus failing to effectively provide emotional support and psychological security for adolescents. The existing technology seriously lacks effective and professional guidance for parent-child communication. When parents and adolescents face challenges such as adolescence and academic pressure at the same time, most parents tend to react subconsciously based on their own experience, including excessive accusations, interference, denial, neglect and other bad communication patterns, which further undermine the parent-child relationship and the physical and mental health of adolescents. In addition, some parents may choose to study theoretical books or go to hospitals and psychological counseling institutions for consultation to get professional advice. However, parent-child communication is different from the cure of physiological diseases, which is integrated into every moment of life. However, mental health professionals can often only provide framework guidance based on the parents and children's subsequent recollections, and it is difficult to observe the real communication scene. Parents and children may spend a lot of time and energy arguing repeatedly about who is right and who is wrong, and fail to identify the key nodes in communication in time, especially the inaccurate expressions of both parties may lead to further escalation of conflicts. Therefore, providing real-time guidance and feedback in a real family environment is crucial to improving parent-child communication. Therefore, a system and method based on professional theory and effective closed-loop evaluation, guidance and feedback of daily parent-child communication is needed. Summary of the invention

[0003] In view of this, the embodiments of this specification provide a parent-child communication reminder system. One or more embodiments of this specification also relate to a parent-child communication reminder method, a parent-child communication reminder device, a computing device, a computer-readable storage medium and a computer program product to solve the technical defects existing in the prior art.

[0004] According to a first aspect of an embodiment of this specification, a parent-child communication reminder system is provided, comprising a collection module, a processing module and a reminder module, including: The acquisition module is used to collect communication audio; by performing timbre detection on the communication audio, a target audio segment corresponding to the parent-child communication user is extracted from the communication audio, and the target audio segment is sent to the processing module; The processing module is used to input the target audio segment into the large language model, perform text recall of the target audio segment in the communication text library through the three-way emotional recall unit included in the retrieval unit in the large language model, and obtain the candidate communication text corresponding to each emotional recall unit; select the target communication text from the candidate communication texts respectively corresponding to the three-way emotional recall unit according to the audio attribute information corresponding to the target audio segment; process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; and send the parent-child communication reminder information to the reminder module; The reminder module is used to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0005] Optionally, the processing module is also used to determine a pre-stored non-reminder keyword set, and extract the target keyword corresponding to the target communication text; detect whether there is a non-reminder keyword matching the target keyword in the non-reminder keyword set; if so, feedback a collection instruction to the collection module; if not, execute the step of processing the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information.

[0006] Optionally, the processing module is also used to obtain multiple historical audio clips corresponding to the parent-child communication user at preset time intervals, and determine the historical parent-child communication reminder information corresponding to each historical audio clip; obtain the communication evaluation report corresponding to the parent-child communication user by performing semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip; and send the communication evaluation report to the user terminal of the parent-child communication user.

[0007] Optionally, the acquisition module is also used to determine at least one conversation user corresponding to the communication audio, and obtain audio proportion information corresponding to each conversation user by performing audio recognition on the communication audio; and determine the parent-child communication user among the at least one conversation user according to the audio proportion information corresponding to each conversation user.

[0008] Optionally, the acquisition module is also used to determine the timbre information corresponding to the parent-child communication user based on the communication audio, and extract the initial audio segment from the communication audio according to the timbre information; input the initial audio segment into the audio processing model for audio repair processing to obtain the target audio segment corresponding to the parent-child communication user.

[0009] Optionally, the three-way emotion recall unit includes a positive emotion recall unit, a neutral emotion recall unit and a negative emotion recall unit, wherein the positive emotion recall unit, the neutral emotion recall unit and the negative emotion recall unit correspond to different text recall strategies; The processing module is also used to perform positive text recall on the target audio segment in the communication text library through the positive emotion recall unit to obtain positive candidate communication text; perform neutral text recall on the target audio segment in the communication text library through the neutral emotion recall unit to obtain neutral candidate communication text; and perform negative text recall on the target audio segment in the communication text library through the negative emotion recall unit to obtain negative candidate communication text.

[0010] Optionally, the processing module is further used to convert the target audio segment into a target audio text; determine the audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text, and determine the positive text vector of the positive communication text, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text in the communication text library; Calculating the vector distance between the audio text vector and the positive text vector by the positive emotion recall unit, and screening the positive candidate communication text from the positive communication text according to the vector distance; Calculating the semantic similarity between the audio text semantic information and the neutral text semantic information by the neutral emotion recall unit, and screening the neutral candidate communication text from the neutral communication text according to the semantic similarity; The negative emotion recall unit performs keyword matching on the audio text keywords and the negative text keywords, and screens negative candidate communication texts from the negative communication texts according to the keyword matching results.

[0011] Optionally, the processing module is also used to determine the audio attribute information corresponding to the target audio segment, determine the communication intention of the parent-child communication user based on the audio attribute information; and select the target communication text from the candidate communication texts corresponding to the three-way emotion recall units respectively according to the communication intention.

[0012] Optionally, the processing module is further used to determine the communication type information corresponding to the target communication text, and filter the target prompt words matching the target communication text from a preset prompt word set according to the communication type information; use the target prompt words to update the target communication text and the communication type information to obtain context information; input the context information into the generation unit in the large language model for processing to obtain the parent-child communication reminder information.

[0013] According to a second aspect of the embodiments of this specification, a parent-child communication reminder method is provided, comprising: Performing timbre detection on the collected communication audio, and extracting a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; The target audio segment is input into a large language model, and the target audio segment is recalled in a communication text library by three-way emotion recall units included in a retrieval unit in the large language model to obtain candidate communication texts corresponding to each emotion recall unit; Filtering a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; Processing the target communication text and the communication type information corresponding to the target communication text by a generation unit in the large language model to obtain parent-child communication reminder information; The parent-child communication reminder information is fed back to the parent-child communication user according to a reminder strategy matching the communication type information.

[0014] According to a third aspect of the embodiments of this specification, a parent-child communication reminder device is provided, comprising: A detection module is configured to perform timbre detection on the collected communication audio, and extract a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; A recall module is configured to input the target audio segment into a large language model, perform text recall of the target audio segment in a communication text library through three-way emotion recall units included in a retrieval unit in the large language model, and obtain candidate communication texts corresponding to each emotion recall unit; A screening module is configured to screen a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; a processing module configured to process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; The reminder module is configured to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0015] According to a fourth aspect of an embodiment of this specification, a computing device is provided, including: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the above-mentioned parent-child communication reminder method are implemented.

[0016] According to a fifth aspect of the embodiments of this specification, a computer-readable storage medium is provided, which stores computer-executable instructions, and when the instructions are executed by a processor, the steps of the above-mentioned parent-child communication reminder method are implemented.

[0017] According to a sixth aspect of the embodiments of this specification, a computer program product is provided, including a computer program or instructions, which implement the steps of the above-mentioned parent-child communication reminder method when executed by a processor.

[0018] The parent-child communication reminder system provided in this embodiment can make reasonable reminders in time according to the communication content during the user communication process. After the communication audio is collected by the collection module, the communication audio can be detected by timbre, and the target audio segment corresponding to the parent-child communication user can be extracted from the communication audio, so as to ensure that the target audio segment corresponding to the parent-child communication user can be analyzed by the large language model in the processing module later. After sending the target audio segment to the processing module, the processing module can input the target audio segment into the large language model, and the target audio segment can be recalled in the communication text library by the three-way emotional recall unit contained in the retrieval unit in the large language model, and the candidate communication text corresponding to each emotional recall unit can be obtained; the candidate communication text corresponding to different emotions can be obtained by the three-way recall method, and on this basis, the target communication text can be selected from the candidate communication texts corresponding to the three-way emotional recall units according to the audio attribute information corresponding to the target audio segment; the communication emotion of the current parent-child communication user can be reflected according to the target communication text, and the parent-child communication reminder information can be constructed according to the communication emotion later. At this time, the target communication text and the communication type information corresponding to the target communication text can be processed by the generation unit in the large language model to obtain parent-child communication reminder information; after the parent-child communication reminder information is sent to the reminder module, the reminder module can feedback the parent-child communication reminder information to the parent-child communication user according to the reminder strategy that matches the communication type information. In the communication process of the parent-child communication user, the communication emotion of the parent-child communication user can be quickly determined by combining the efficient processing ability and semantic analysis ability of the large language model, and then the parent-child communication reminder information matching the emotion is generated and fed back to the parent-child communication user, so that the parent-child communication user can make timely adjustments to the communication method, thereby ensuring the effectiveness of communication, especially in the scenario of adolescent psychological counseling, by timely reminding the parent-child communication user, the parent-child communication user can make a positive impact on the psychological problems of adolescents, so as to provide emotional support and psychological security for adolescents through communication. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 It is a schematic diagram of a parent-child communication reminder system provided by one embodiment of this specification; Figure 2 It is a structural diagram of a parent-child communication reminder system provided by an embodiment of this specification; Figure 3 It is a flow chart of a parent-child communication reminder method provided by an embodiment of this specification; Figure 4 It is a structural schematic diagram of a parent-child communication reminder device provided by an embodiment of this specification; Figure 5 It is a structural block diagram of a computing device provided by an embodiment of this specification. DETAILED DESCRIPTION

[0020] Many specific details are described in the following description to facilitate a full understanding of this specification. However, this specification can be implemented in many other ways than those described herein, and those skilled in the art can make similar generalizations without violating the connotation of this specification, so this specification is not limited to the specific implementation disclosed below.

[0021] The terms used in one or more embodiments of this specification are only for the purpose of describing specific embodiments, and are not intended to limit one or more embodiments of this specification. The singular forms of "a", "said" and "the" used in one or more embodiments of this specification and the appended claims are also intended to include plural forms, unless the context clearly indicates other meanings. It should also be understood that the term "and / or" used in one or more embodiments of this specification refers to and includes any or all possible combinations of one or more associated listed items.

[0022] It should be understood that although the terms first, second, etc. may be used to describe various information in one or more embodiments of this specification, this information should not be limited to these terms. These terms are only used to distinguish the same type of information from each other. For example, without departing from the scope of one or more embodiments of this specification, the first may also be referred to as the second, and similarly, the second may also be referred to as the first. Depending on the context, the word "if" as used herein may be interpreted as "at the time of" or "when" or "in response to determining".

[0023] In addition, it should be noted that the user information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in one or more embodiments of this specification are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with the relevant laws, regulations and standards of the relevant countries and regions, and provide corresponding operation entrances for users to choose to authorize or refuse.

[0024] In this specification, a parent-child communication reminder system is provided. One or more embodiments of this specification also involve a parent-child communication reminder method, a parent-child communication reminder device, a computing device, a computer-readable storage medium, and a computer program product, which are described in detail one by one in the following embodiments.

[0025] See also Figure 1As shown in the schematic diagram, the parent-child communication reminder system provided by this embodiment can make reasonable reminders in time according to the communication content during the user communication process. After the communication audio is collected by the acquisition module, the communication audio can be detected by timbre, and the target audio segment corresponding to the parent-child communication user can be extracted from the communication audio, so as to ensure that the target audio segment corresponding to the parent-child communication user can be analyzed by the large language model in the processing module later. After sending the target audio segment to the processing module, the processing module can input the target audio segment into the large language model, and the target audio segment can be recalled in the communication text library by the three-way emotional recall unit contained in the retrieval unit in the large language model, and the candidate communication text corresponding to each emotional recall unit can be obtained; the candidate communication text corresponding to different emotions can be obtained by the three-way recall method, and on this basis, the target communication text can be selected from the candidate communication texts corresponding to the three-way emotional recall units according to the audio attribute information corresponding to the target audio segment; the communication emotion of the current parent-child communication user can be reflected according to the target communication text, and the parent-child communication reminder information can be constructed according to the communication emotion later. At this time, the target communication text and the communication type information corresponding to the target communication text can be processed by the generation unit in the large language model to obtain parent-child communication reminder information; after the parent-child communication reminder information is sent to the reminder module, the reminder module can feedback the parent-child communication reminder information to the parent-child communication user according to the reminder strategy that matches the communication type information. In the communication process of the parent-child communication user, the communication emotion of the parent-child communication user can be quickly determined by combining the efficient processing ability and semantic analysis ability of the large language model, and then the parent-child communication reminder information matching the emotion is generated and fed back to the parent-child communication user, so that the parent-child communication user can make timely adjustments to the communication method, thereby ensuring the effectiveness of communication, especially in the scenario of adolescent psychological counseling, by timely reminding the parent-child communication user, the parent-child communication user can make a positive impact on the psychological problems of adolescents, so as to provide emotional support and psychological security for adolescents through communication.

[0026] See also Figure 2 , Figure 2 A schematic diagram of the structure of a parent-child communication reminder system provided according to an embodiment of the present specification is shown. The parent-child communication reminder system 200 includes a collection module 210, a processing module 220 and a reminder module 230, and is specifically implemented as follows.

[0027] The acquisition module 210 is used to collect communication audio; by performing timbre detection on the communication audio, a target audio segment corresponding to the parent-child communication user is extracted from the communication audio, and sent to the processing module; The processing module 220 is used to input the target audio segment into the large language model, and to perform text recall of the target audio segment in the communication text library through the three-way emotional recall unit included in the retrieval unit in the large language model to obtain the candidate communication text corresponding to each emotional recall unit; to filter the target communication text from the candidate communication texts respectively corresponding to the three-way emotional recall unit according to the audio attribute information corresponding to the target audio segment; to process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; and to send the parent-child communication reminder information to the reminder module; The reminder module 230 is used to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0028] The parent-child communication reminder system provided in this embodiment can be applied to any communication scenario, such as homework tutoring communication scenario, psychological counseling communication scenario, parent-child interactive communication scenario, etc., and can be used to make reasonable reminders in time according to the communication content when communicating between users, thereby improving the effectiveness of communication. This embodiment uses the parent-child communication scenario as an example to illustrate the parent-child communication reminder system. The descriptions of other scenarios can refer to the same or corresponding descriptions in this embodiment, and this embodiment will not be repeated in detail here.

[0029] Specifically, the acquisition module specifically refers to a module in the parent-child communication reminder system for collecting user communication audio, which can be any form of radio equipment, and the acquisition module can be placed at any position in the communication scene, such as a desktop, floor, bookcase, closet, etc., or it can be placed on the user, such as clamped on the collar, or placed in a clothes pocket, and this embodiment does not make any restrictions here. Correspondingly, the processing module specifically refers to a module that connects to a large language model to generate parent-child communication reminder information for communication content. Correspondingly, the reminder module specifically refers to a module that feeds back parent-child communication reminder information to the user, and can send parent-child communication reminder information to the terminal device held by the user, or can feed back parent-child communication reminder information in a set manner, such as vibration, beeping, voice broadcast, flashing lights, etc., and this embodiment does not make any restrictions here. Among them, the acquisition module, the processing module and the reminder module are connected to each other in communication.

[0030] Correspondingly, the communication audio specifically refers to the audio collected by the acquisition module for the communication dialogue between users in the current scene; the parent-child communication user specifically refers to the user to whom the parent-child communication reminder information needs to be sent at the current moment after the timbre detection is performed on the communication audio, wherein the timbre detection can be performed by presetting the template audio of the parent-child communication user, comparing the template audio with the communication audio, and determining that there is audio content in the communication audio that matches the timbre of the template audio, and then determining that the communication audio involves the speech content of the parent-child communication user. At this time, the communication audio can be processed for subsequent feedback of parent-child communication reminder information to the parent-child communication user. For example, if the parent-child communication user is the "father" in the parent-child communication scene, the communication audio between the "father" and the "child" can be processed to effectively remind the "father" and avoid the communication content from affecting the mental health of the "child". Correspondingly, the target audio segment specifically refers to the audio segment in the communication audio that matches the speech content of the parent-child communication user.

[0031] The large language model specifically refers to an LLM (Large Language Model) model including a retrieval unit and a generation unit, which is used to analyze the target audio clip of the parent-child communication user in real time, so as to generate reasonable parent-child communication reminder information and feedback it to the parent-child communication user. The retrieval unit specifically refers to a unit used to recall the communication text for the target audio clip, so as to determine the current communication emotion of the parent-child communication user from the recalled communication text. The three-way recall unit specifically refers to the three recall units contained in the retrieval unit, each recall unit corresponds to a different communication emotion, and is used to recall candidate communication texts corresponding to different communication emotions in the communication text library, and then screen the target communication text to determine the current communication emotion corresponding to the parent-child communication user. Among them, the communication emotion corresponding to the three-way recall unit can be emotional warmth emotion, overprotection / interference emotion, and punishment / rejection emotion. The communication text library specifically refers to a database constructed by pre-entering communication texts for different emotional emotions, which stores communication texts corresponding to different communication emotions. Correspondingly, the candidate communication text specifically refers to the communication text recalled by each recall unit in the communication text library for the target audio clip. Accordingly, the target communication text specifically refers to the communication text that matches the user's current communication emotion selected from the candidate communication texts. Accordingly, the communication type information is the information that characterizes the communication emotion. Accordingly, the parent-child communication reminder information specifically refers to the information fed back to the user, which can be any data type and any reminder method. Accordingly, the reminder strategy specifically refers to the strategy for feeding back parent-child communication reminder information to the parent-child communication user. Different strategies are used to feed back parent-child communication reminder information in different ways, such as audio playback, vibration, flashing lights, etc. This embodiment does not make any limitation here.

[0032] Based on this, in order to be able to make timely and reasonable reminders according to the content of communication during the user communication process, after the communication audio is collected by the acquisition module, the tone detection of the communication audio can be performed to extract the target audio segment corresponding to the parent-child communication user from the communication audio, to ensure that the target audio segment corresponding to the parent-child communication user can be analyzed later through the large language model in the processing module.

[0033] Furthermore, after sending the target audio clip to the processing module, the processing module can input the target audio clip into the large language model, and perform text recall of the target audio clip in the communication text library through the three-way emotional recall unit contained in the retrieval unit in the large language model, and obtain the candidate communication text corresponding to each emotional recall unit; the candidate communication text corresponding to different emotions is obtained by implementing the three-way recall method, and on this basis, the target communication text is screened from the candidate communication texts corresponding to the three-way emotional recall unit according to the audio attribute information corresponding to the target audio clip; the communication emotion of the current parent-child communication user can be reflected according to the target communication text, and the parent-child communication reminder information can be constructed according to the communication emotion in the future. At this time, the target communication text and the communication type information corresponding to the target communication text can be processed by the generation unit in the large language model to obtain the parent-child communication reminder information.

[0034] Furthermore, after the parent-child communication reminder information is sent to the reminder module, the reminder module can feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0035] For example, after the acquisition module placed on the desk collects the communication audio between the father and the child, it performs timbre detection on the communication audio and extracts the target audio segment corresponding to the father from the communication audio. After sending it to the processing module, the processing module inputs the target audio segment into the large language model, and the three-way recall unit contained in the retrieval unit in the large language model performs text recall on the target audio segment in the communication text library. At this time, a candidate communication texts are obtained from the recall unit feedback corresponding to the emotional warmth, b candidate communication texts are obtained from the recall unit feedback corresponding to the emotional overprotection, and c candidate communication texts are obtained from the recall unit feedback corresponding to the emotional punishment.

[0036] On the one hand, when the father's communication emotion is determined to be an emotional warmth emotion according to the audio attribute information corresponding to the target audio segment, it means that the father expressed understanding, encouragement and positive emotional expression in the conversation with the child, reflecting the father's care, support and emotional warmth for Chinese characters. Therefore, the target communication text matching the target audio segment can be determined from the a candidate communication texts, and then the target communication text and the emotional warmth emotion type information are input into the generation unit in the large language model for processing, and the parent-child communication reminder information output by the generation unit can be obtained. By sending the parent-child communication reminder information to the reminder module, the reminder module can feedback the parent-child communication reminder information to the father according to the reminder strategy matching the emotional warmth emotion, such as displaying the "like" gesture image through the display screen of the reminder module.

[0037] On the second aspect, when the father's communication emotion is determined to be an overprotective emotion according to the audio attribute information corresponding to the target audio segment, it means that the father has overprotective or controlling behavior in the conversation with the child, restricting the child's freedom and interfering with the child's independent decision-making. Therefore, the target communication text that matches the target audio segment can be determined from the b candidate communication texts, and then the target communication text and the overprotective emotion type information are input into the generation unit in the large language model for processing, and the parent-child communication reminder information output by the generation unit can be obtained. By sending the parent-child communication reminder information to the reminder module, the reminder module can feedback the parent-child communication reminder information to the father according to the reminder strategy that matches the overprotective emotion, such as reminding the father of the problem of this communication by vibrating the reminder module once every 1s.

[0038] On the third aspect, when the father's communication emotion is determined to be a punishment emotion according to the audio attribute information corresponding to the target audio clip, it means that the father's conversation with the child expresses the father's severe punishment of the child, emotional rejection or indifference, including criticism and blame of the child. Therefore, the target communication text matching the target audio clip can be determined from the c candidate communication texts, and then the target communication text and the overprotective emotion type information are input into the generation unit in the large language model for processing, and the parent-child communication reminder information output by the generation unit can be obtained. By sending the parent-child communication reminder information to the reminder module, the reminder module can feedback the parent-child communication reminder information to the father according to the reminder strategy matching the overprotective emotion, such as playing the audio content of "Can you calm down?" through the audio playback device of the reminder module.

[0039] In addition, when the communication emotion of the parent-child communication user is a non-emotional warm emotion, when the parent-child communication reminder information is fed back to the user, in addition to serving as a reminder, prompt information for improving the effectiveness of the user's communication can also be fed back to the user, such as informing the parent-child communication user that he can say "I understand you" type of words for this communication, thereby improving the effectiveness of communication. In actual applications, the reminder content and prompt content contained in the parent-child communication reminder information can be set according to actual needs, and this embodiment does not make any limitations here.

[0040] To sum up, the parent-child communication reminder system provided in this embodiment can realize that during the communication process of the parent-child communication user, the communication emotion of the parent-child communication user can be quickly determined by combining the efficient processing capability and semantic analysis capability of the large language model, and then the parent-child communication reminder information matching the emotion is generated and fed back to the parent-child communication user, so that the parent-child communication user can make timely adjustments to the communication method, thereby ensuring the effectiveness of communication, especially in the scenario of psychological counseling for adolescents, by timely reminding the parent-child communication user, the parent-child communication user can make a positive impact on the psychological problems of adolescents, so as to provide emotional support and psychological security for the adolescents through communication.

[0041] In one or more implementations of this embodiment, the processing module is also used to determine a pre-stored non-reminder keyword set, and extract the target keyword corresponding to the target communication text; detect whether there is a non-reminder keyword matching the target keyword in the non-reminder keyword set; if so, feedback a collection instruction to the collection module; if not, execute the step of processing the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information.

[0042] Specifically, the non-reminder keyword set specifically refers to a set of user-defined keywords that do not need to trigger communication reminders. It is used to facilitate users to set non-reminder keywords according to actual conditions within a specified time interval, so that communication reminders are not triggered within the time interval. For example, in a parent-child communication scenario, if the child is sick, the user can set non-reminder keywords such as "Are you feeling better?", "Uncomfortable?", "Fever", etc. When it is detected that the target communication text contains the keyword, the communication reminder may not be made. Correspondingly, the target keyword specifically refers to the keyword extracted from the target communication text. The feedback collection instruction specifically refers to an instruction sent to the sending module to stop collecting audio.

[0043] Based on this, considering that in actual communication scenarios, other factors may cause communication content to be repeated, or the communication content may not match the actual communication emotions. In this case, in order to avoid the parent-child communication reminder system from feeding back parent-child communication reminder information that does not conform to the current communication situation, after determining the target communication text that reflects the communication emotions of the parent-child communication user, the pre-stored non-reminder keyword set can be determined, and the target keywords corresponding to the target communication text can be extracted; at this time, it can be detected whether there are non-reminder keywords matching the target keywords in the non-reminder keyword set; if so, it means that the communication behavior occurring at this time belongs to the communication behavior that the user specifies does not need to be reminded, so the collection instruction can be fed back to the collection module to inform the collection module to stop collecting communication audio. If not, it means that the communication behavior occurring at this time belongs to normal communication behavior, so the generation unit in the large language model can continue to process the target communication text and the communication type information corresponding to the target communication text to obtain the parent-child communication reminder information.

[0044] Using the above example, when it is determined that the father's communication emotion is overprotective, the target communication text that matches the target audio clip is determined from the b candidate communication texts. After that, the keywords "uncomfortable", "fever", "heartache", etc. can be extracted from the target communication text, and then matched with the non-reminder keywords preset by the father. When it is determined that the non-reminder keyword matches the keyword, it means that the communication content at this time is a situation preset by the father that does not trigger a communication reminder, such as the child's recent illness. At this time, an instruction to stop collecting audio can be sent to the acquisition module to avoid consuming more computing resources.

[0045] In summary, by setting a set of non-reminder keywords, after determining the target communication text that reflects the user's communication emotions, it is possible to determine whether to trigger a reminder operation through keyword matching, thereby reducing the consumption of computing resources and facilitating users to flexibly set information reminder strategies.

[0046] In one or more implementations of this embodiment, the processing module is also used to obtain multiple historical audio clips corresponding to the parent-child communication user at preset time intervals, and determine the historical parent-child communication reminder information corresponding to each historical audio clip; obtain the communication evaluation report corresponding to the parent-child communication user by performing semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip; and send the communication evaluation report to the user terminal of the parent-child communication user.

[0047] Specifically, historical audio clips refer to the audio clips involved in each communication of the parent-child communication user. Correspondingly, historical parent-child communication reminder information refers to the parent-child communication reminder information generated by the large language model for each historical audio clip. By performing semantic analysis on historical audio clips and historical parent-child communication reminder information, the emotional state of each communication can be determined, which is then convenient for feedback to parent-child communication users, so that parent-child communication users can learn communication methods to improve communication efficiency. Among them, the communication evaluation report can include the emotional score of each communication, communication time, statistics of multiple occurrences of speech, positive / negative / neutral tendency, and recommended speech.

[0048] Based on this, in order to facilitate parent-child communication users to understand their own communication situation, the processing module can be used to record the historical audio clips and historical parent-child communication reminder information corresponding to the parent-child communication users, and obtain multiple historical audio clips corresponding to the parent-child communication users after a set time interval, and determine the historical parent-child communication reminder information corresponding to each historical audio clip, and then perform semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip to obtain the communication evaluation report corresponding to the parent-child communication user; on this basis, by sending the communication evaluation report to the user terminal of the parent-child communication user, the user can be supported to perform communication learning and other operations.

[0049] In actual applications, the set time interval can be set as needed, such as one week, one month, or n months, etc.; in addition, in addition to the above information, the communication evaluation report can also include the number and frequency of parent-child communication users triggering bad communication methods, the number of sensitive words of parent-child communication users, the number and frequency of large emotional fluctuations of parent-child communication users, etc. This embodiment does not make any limitations here.

[0050] Continuing with the above example, the system counts the father's communication content at one-month intervals and determines that the number of times the father praises the child is m1, and the number of times the father scolds the child is m2. At the same time, based on the above information, the system can recommend to the father articles, videos or books that can help him communicate effectively with his child and improve the parent-child relationship. Furthermore, based on the above content, a corresponding communication evaluation report for the father can be generated and sent to the father's mobile phone, which can support the father's reading, learning and reflection.

[0051] In addition, considering that in actual communication scenarios, the communication content of parent-child communication users may involve sensitive information, such as ID number, bank card password and other information, in order to avoid the leakage of sensitive information, when storing historical audio clips, the historical audio clips can be desensitized before storage, thereby improving data security.

[0052] In summary, by conducting statistical analysis on historical communication content at set time intervals and providing users with communication evaluation reports, it is easier for users to understand their own communication situations, thereby improving communication effectiveness and communicating with others in a better way.

[0053] In one or more implementations of this embodiment, the acquisition module is also used to determine at least one conversation user corresponding to the communication audio, and obtain audio proportion information corresponding to each conversation user by performing audio recognition on the communication audio; and determine the parent-child communication user from the at least one conversation user according to the audio proportion information corresponding to each conversation user.

[0054] Specifically, the conversation user refers to the user involved in the communication audio, and the audio proportion information refers to the value of the audio duration relative to the communication audio after counting the speaking time of each conversation user in the communication audio. The audio proportion information can be used to determine the user who mainly speaks in the communication audio, and then the user can be used as a parent-child communication user for subsequent communication analysis and reminders.

[0055] Based on this, after the collection module collects the communication audio, in order to be able to remind the main speaking user in the communication, the communication audio can be analyzed at this time, and at least one dialogue user corresponding to the communication audio can be first determined, and the audio proportion information corresponding to each dialogue user can be obtained by performing audio recognition on the communication audio; then, by comparing the audio proportion information, the dialogue user with the highest proportion can be screened as the parent-child communication user, so as to remind the parent-child communication user later.

[0056] In addition, in addition to screening parent-child communication users by calculating the audio proportion information as mentioned above, parent-child communication users can also be identified by setting different conversation voice characteristics, and the identity of the parent-child communication user can be set as needed, such as setting it to father, mother, grandfather, etc., so as to accurately identify the identity of each conversation user in the family and make reasonable conversation reminders.

[0057] In summary, by calculating the audio proportion information to determine the parent-child communication users, it can be achieved that the determined parent-child communication users are users who speak in a higher proportion in this communication, and the reminder effect on them is greater, thereby improving the user's communication reminder experience.

[0058] In one or more implementations of this embodiment, the acquisition module is also used to determine the timbre information corresponding to the parent-child communication user based on the communication audio, and extract the initial audio segment from the communication audio according to the timbre information; input the initial audio segment into the audio processing model for audio repair processing to obtain the target audio segment corresponding to the parent-child communication user.

[0059] Specifically, the timbre information refers to the timbre feature information corresponding to the parent-child communication user. Since different users correspond to different timbres, after determining the timbre information corresponding to the parent-child communication user, the audio segment corresponding to the parent-child communication user can be segmented in the communication audio according to the timbre information. Correspondingly, the audio processing model specifically refers to a model that performs audio repair processing on the initial audio segment, and can have processing capabilities such as audio splicing, silencing, and audio enhancement to ensure that the final target audio segment is the audio corresponding to the speech content of the parent-child communication user, so as to facilitate subsequent analysis and processing.

[0060] Based on this, after determining the parent-child communication user, taking into account that the communication audio may contain speech audios of multiple users, in order to facilitate the subsequent reminder operation in combination with the high-quality audio corresponding to the parent-child communication user, the timbre information corresponding to the parent-child communication user can be determined based on the communication audio. On this basis, the initial audio segment can be extracted from the communication audio according to the timbre information; the initial audio segment obtained at this time may be multiple non-continuous audio segments, and there may also be speech content from other people. On this basis, the initial audio segment can be input into the audio processing model for audio repair processing, and the audio repair capability of the audio processing model can be used to perform audio repair on the initial audio segment, thereby ensuring that the target audio segment obtained in the end can clearly and accurately express the speech content of the parent-child communication user.

[0061] Continuing with the above example, in the parent-child communication scenario, when it is determined that the collected communication audio involves the father, mother, and child, and according to the pre-set vocal features, when it is determined that the father needs to be reminded, the timbre features corresponding to the father can be determined first, and then the collected communication audio can be identified according to the timbre features to obtain three audio clips corresponding to the father. On this basis, by inputting the three audio clips into the audio processing model for audio repair processing, the three audio clips can be spliced, and the conversation content between the mother and the child in the audio can be eliminated at the same time, and then a high-quality audio clip can be obtained, which is used for subsequent analysis of the father's communication content, so as to provide the father with parent-child communication reminder information that matches the communication emotions.

[0062] In summary, by combining the audio processing model to repair the initial audio clip, the quality of the repaired audio can be guaranteed, thereby facilitating subsequent audio analysis and reminders.

[0063] In one or more implementations of this embodiment, the three-way emotion recall unit includes a positive emotion recall unit, a neutral emotion recall unit, and a negative emotion recall unit, wherein the positive emotion recall unit, the neutral emotion recall unit, and the negative emotion recall unit correspond to different text recall strategies; The processing module is also used to perform positive text recall on the target audio segment in the communication text library through the positive emotion recall unit to obtain positive candidate communication text; perform neutral text recall on the target audio segment in the communication text library through the neutral emotion recall unit to obtain neutral candidate communication text; and perform negative text recall on the target audio segment in the communication text library through the negative emotion recall unit to obtain negative candidate communication text.

[0064] Specifically, the positive emotion recall unit refers to a recall unit for recalling the communication text corresponding to the emotional warmth of the target audio clip in the communication text library; the neutral emotion recall unit refers to a recall unit for recalling the communication text corresponding to the overprotection emotion of the target audio clip in the communication text library; the negative emotion recall unit refers to a recall unit for recalling the communication text corresponding to the punishment emotion of the target audio clip in the communication text library; among them, the positive emotion recall unit, the neutral emotion recall unit and the negative emotion recall unit correspond to different text recall strategies, that is, the three-way recall units use different methods to recall the communication text corresponding to their respective emotions to ensure that the recalled text is more accurate.

[0065] Based on this, in order to successfully recall the matching communication text for the target audio clip, the candidate communication text corresponding to each emotion can be obtained by setting three recall units to recall separately, and the target communication text can be screened on this basis. Specifically, the positive text of the target audio clip can be recalled in the communication text library through the positive emotion recall unit to obtain the positive candidate communication text; the neutral text of the target audio clip can be recalled in the communication text library through the neutral emotion recall unit to obtain the neutral candidate communication text; the negative text of the target audio clip can be recalled in the communication text library through the negative emotion recall unit to obtain the negative candidate communication text. Subsequently, the target communication text can be screened from the positive candidate communication text, the neutral candidate communication text and the negative candidate communication text to complete the construction of parent-child communication reminder information.

[0066] In actual applications, in order to enable each recall unit to successfully recall the communication text corresponding to the target audio clip in the communication text library, the candidate communication texts contained in the communication text library can be stored according to the emotion type during the storage stage to avoid cross-storage of texts, which affects the recall accuracy. Alternatively, identifications corresponding to different emotion types can be added to the communication texts corresponding to each emotion type. During recall, the corresponding communication text is first hit according to the identification, and then the text recall operation is performed through each recall unit. During specific implementation, the storage method of the communication text contained in the communication text library can be set according to actual needs, and this embodiment does not make any restrictions here. Among them, the text recall processing can recall a set number of candidate communication texts according to actual needs, and the setting of this number can be set according to actual needs, and this embodiment does not make any restrictions here.

[0067] In addition, when constructing the communication text library, in addition to directly storing communication texts that can distinguish different types of emotions, it can also store communication expression information corresponding to different types of emotions. For example, for punishment emotions, it can include negative responsibility (negative words for family members appear continuously in the language in a short period of time, such as can't, no, not allowed, etc., or the child is associated with negative words such as incompetent, incompetent, disobedient, and worrying), comparison (comparing the child with a specific person and then directly denying the child, such as look at XX's family, etc.), labeling (such as you are too pampered, etc.), repetition (repeatedly repeating a specific thing, such as drinking hot water, a certain specific food, watching TV, etc.), communication tone (excessive use of rhetorical questions, multiple use of rhetorical questions to question within a certain period of time), specific sentence patterns (such as you... again, you always..., etc.), specific things (specifying things that express disgust, or physical defects of children), direct judgment (such as you are really worrying, etc.), etc. When the above words appear in the target communication audio, they will be hit by the corresponding recall unit as candidate communication text for subsequent communication emotion type analysis, so as to make reasonable parent-child communication reminder information.

[0068] In summary, by using different recall units to recall candidate communication texts corresponding to different communication emotions for target audio clips, the problem of missing communication emotion types can be avoided. On this basis, a basis is provided for subsequent screening of target communication texts, so that the communication emotions corresponding to the current parent-child communication users can be accurately hit to make reasonable communication reminders.

[0069] In one or more implementations of this embodiment, the processing module is further used to convert the target audio segment into a target audio text; determine the audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text, and determine the positive text vector of the positive communication text, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text in the communication text library; Calculating the vector distance between the audio text vector and the positive text vector by the positive emotion recall unit, and screening the positive candidate communication text from the positive communication text according to the vector distance; Calculating the semantic similarity between the audio text semantic information and the neutral text semantic information by the neutral emotion recall unit, and screening the neutral candidate communication text from the neutral communication text according to the semantic similarity; The negative emotion recall unit performs keyword matching on the audio text keywords and the negative text keywords, and screens negative candidate communication texts from the negative communication texts according to the keyword matching results.

[0070] Specifically, the target audio text specifically refers to the text recording the conversation content obtained after format conversion of the target audio segment. Correspondingly, the audio text vector specifically refers to the vector expression obtained by embedding the target audio text. The audio text semantic information specifically refers to the semantic vector expression obtained after semantic analysis of the target audio text. The audio text keywords specifically refer to the keywords extracted from the target audio text. Correspondingly, the positive text vector specifically refers to the vector expression corresponding to the positive communication text in the communication text library. The neutral text semantic information specifically refers to the vector expression of the semantics corresponding to the neutral communication text in the communication text library. The negative text keywords specifically refer to the keywords contained in the negative communication text in the communication text library.

[0071] Based on this, in order to ensure the recall accuracy when recalling candidate communication texts for target audio clips, the target audio clips can be converted into target audio texts first; the target audio texts can be aligned with the communication texts contained in the communication text library in terms of structural type. On this basis, considering that each recall unit adopts different recall strategies for text recall, such as the positive emotion recall unit uses the method of calculating vector distance for text recall, the neutral emotion recall unit uses the method of calculating semantic similarity for text recall, and the negative emotion recall unit uses the method of matching keywords for text recall. Therefore, before recalling the text, data preparation is required. The audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text can be determined, as well as the positive text vector of the positive communication text in the communication text library, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text can be determined, so as to achieve data alignment between the communication text contained in the communication text library and the target audio text in three communication emotion dimensions.

[0072] Furthermore, the vector distance between the audio text vector and the positive text vector can be calculated by the positive emotion recall unit, and the positive communication text with the shortest vector distance can be selected as the positive candidate communication text in the positive communication text according to the vector distance; the semantic similarity between the audio text semantic information and the neutral text semantic information can be calculated by the neutral emotion recall unit, and the neutral communication text with the highest similarity can be selected as the neutral candidate communication text according to the semantic similarity; and the audio text keywords and the negative text keywords can be matched by the negative emotion recall unit, and the negative communication text with the highest matching degree can be selected as the negative candidate communication text according to the keyword matching result. This facilitates the subsequent screening of the target communication text based on the candidate communication texts corresponding to the three communication emotion types.

[0073] In practical applications, a vector distance calculation method is set for the positive emotion recall unit to perform text recall. The purpose is to take into account that the communication content corresponding to the emotional warmth is more directly expressed. Using this method for text recall can improve the accuracy under the emotional warmth type. A semantic similarity calculation method is set for the neutral emotion recall unit to perform text recall. The purpose is to take into account that the communication content corresponding to the overprotective emotion is more concealed. Using this method for text recall can be completed from the semantic dimension to avoid excessive recall. A keyword matching method is set for the negative emotion recall unit to perform text recall. The purpose is to take into account that the communication content corresponding to the punishment emotion will carry sensitive keywords. Using this method for text recall can avoid errors in emotion type analysis.

[0074] In summary, by converting the target audio segment into the target audio text and then performing subsequent text recall, the structural alignment of the audio and the text contained in the communication text library can be achieved, thereby improving the recall accuracy.

[0075] In one or more implementations of this embodiment, the processing module is also used to determine the audio attribute information corresponding to the target audio segment, determine the communication intention of the parent-child communication user based on the audio attribute information; and select the target communication text from the candidate communication texts corresponding to the three-way emotional recall units respectively according to the communication intention.

[0076] Specifically, the audio attribute information refers to the corresponding frequency domain features of the audio obtained after analyzing the target audio segment. The frequency domain features can reflect the communication emotions of the parent-child communication user in the target audio segment, thereby determining the communication intention of the parent-child communication user, wherein the communication intention is the intention information corresponding to the communication emotion.

[0077] Based on this, after the candidate communication texts are recalled by the three-way emotional recall units mentioned above, in order to accurately determine the target communication text that reflects the user's communication emotions, the audio attribute information corresponding to the target audio clip can be determined, and on this basis, the communication intention of the parent-child communication user can be determined according to the audio attribute information; thereafter, the target communication text can be screened from the candidate communication texts corresponding to the three-way emotional recall units according to the communication intention, so that parent-child communication reminder information can be constructed based on the target communication text in the future.

[0078] Following the above example, after obtaining the target audio clip corresponding to the father, the target audio clip can be converted into a target audio text. Thereafter, a text vector VT1 can be constructed for the target audio text, and a semantic vector VT2 corresponding to the target audio text can be determined, and keywords 1 and 2 can be extracted from the target audio text. On this basis, the vector distance between the text vector VT1 and the text vector corresponding to the positive communication text in the communication text library is calculated by the recall unit corresponding to the emotional warmth type, and the emotional warmth communication text is screened out by sorting the vector distance. The semantic similarity between the semantic vector VT2 and the semantic vector corresponding to the neutral communication text in the communication text library is calculated by the recall unit corresponding to the overprotective emotional type, and the overprotective communication text is screened out by sorting the semantic similarity. The recall unit corresponding to the punishment emotional type matches keyword 1 and keyword 2 with the keywords corresponding to the negative communication text in the communication text library, and the punishment communication text is screened out according to the matching results.

[0079] Furthermore, after obtaining the emotional warmth communication text, overprotection communication text, and punishment communication text that match the father's communication audio, considering that the father's communication emotion is one of the emotional warmth emotion type, overprotection emotion type, and punishment emotion type, in order to accurately identify the father's current communication emotion, the frequency domain features of the audio clip can be further extracted, and the father's current communication intention can be determined through the frequency domain features. Then, according to the communication intention, the text that matches the father's communication emotion can be selected from the emotional warmth communication text, overprotection communication text, and punishment communication text as the target communication text, so that reasonable parent-child communication reminder information can be fed back to the father based on the target communication text.

[0080] In summary, by determining the communication emotions of parent-child communication users through communication intention recognition, it is possible to accurately locate the target communication text among the candidate communication texts corresponding to the multi-channel recall unit, thereby facilitating the subsequent generation of parent-child communication reminder information to reasonably remind parent-child communication users based on the target communication text.

[0081] In one or more implementations of this embodiment, the processing module is also used to determine the communication type information corresponding to the target communication text, and select the target prompt word matching the target communication text from a preset prompt word set according to the communication type information; use the target prompt word to update the target communication text and the communication type information to obtain context information; input the context information into the generation unit in the large language model for processing to obtain the parent-child communication reminder information.

[0082] Specifically, the communication type information specifically refers to the emotion type corresponding to the current communication emotion of the parent-child communication user, such as emotional warmth type, overprotection type, and punishment type. Correspondingly, the prompt word set specifically refers to a set of prompt words stored for different communication type information, and correspondingly, the target prompt word is the prompt word that matches the current communication emotion. Context information specifically refers to the information of the input generation unit obtained after updating the target communication text and communication type information using the prompt word.

[0083] Based on this, after determining the target communication text as mentioned above, the communication type information corresponding to the current user's communication emotion can be determined according to the target communication text, and on this basis, the target prompt words matching the target communication text can be screened in the preset prompt word set according to the communication type information; at this time, the target prompt words can be used to update the target communication text and the communication type information, so as to obtain context information according to the update results; thereafter, the context information can be input into the generation unit in the large language model for processing, so as to obtain parent-child communication reminder information matching the current communication emotion of the parent-child communication user according to the processing results of the generation unit, and then the parent-child communication reminder information can be fed back to the user.

[0084] Continuing with the above example, when the target communication text is an emotional warmth communication text, it means that the father's communication emotion at this time is an emotional warmth emotion type, and then the prompt word corresponding to the emotional warmth emotion type can be selected to update the target communication text, and then the updated target communication text is input into the generation unit in the large language model for processing, and parent-child communication reminder information can be obtained. The parent-child communication reminder information is a "like" gesture image displayed on the display screen of the reminder module.

[0085] When the target communication text is an overprotective communication text, it means that the father's communication emotion at this time is an overprotective emotional type, and then the prompt word corresponding to the overprotective emotional type can be selected to update the target communication text, and then the updated target communication text is input into the generation unit in the large language model for processing, and parent-child communication reminder information can be obtained. The parent-child communication reminder information is to remind the father of the problem of this communication through the reminder module at a frequency of once every 1s.

[0086] When the target communication text is a punishment communication text, it means that the father's communication emotion at this time is a punishment emotion type, and then the prompt word corresponding to the punishment emotion type can be selected to update the target communication text, and then the updated target communication text is input into the generation unit in the large language model for processing, and parent-child communication reminder information can be obtained. The parent-child communication reminder information is the audio content of "Can you calm down?" played through the audio playback device of the reminder module.

[0087] In summary, by selecting prompt words that match the communication emotion type and inputting them into the generation unit for processing, the parent-child communication reminder information output by the generation unit can match the user's current communication emotion, so that reasonable reminders can be made to improve the user's communication experience.

[0088] In one or more implementations of this embodiment, the processing module is also used to extract sample audio clips from the sample set, input the sample audio clips into the initial large language model for processing, and obtain predicted parent-child communication reminder information; input the sample audio clips, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clips into the reward model of the initial large language model for processing to obtain model training reward information; adjust the parameters of the initial large language model based on the model training reward information until the large language model that meets the training stop condition is obtained.

[0089] Specifically, the sample set specifically refers to a collection of stored sample audio clips. Correspondingly, the predicted parent-child communication reminder information specifically refers to the parent-child communication reminder information output after being input into the initial large language model for processing. Correspondingly, the reward model specifically refers to a model for scoring the prediction results of the initial large language model, which is used in the model training stage. Correspondingly, the training reward information specifically refers to the score obtained by scoring the initial large language model for this prediction processing. Correspondingly, the training stop condition specifically refers to the condition for stopping the training of the initial large language model, which includes but is not limited to the number of iterations condition, the verification set verification condition or the loss value comparison condition. In actual applications, the training stop condition can be set according to actual needs, and this embodiment does not make any limitation here.

[0090] Based on this, in order to enable the large language model to provide accurate predictions, it needs to be fully trained. During the training process, sample audio clips can be extracted from the sample set, and after the sample audio clips are input into the initial large language model for processing, the predicted parent-child communication reminder information can be obtained; then the sample audio clips, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clips can be input into the reward model of the initial large language model for processing, so as to score the current prediction ability of the initial large language model through the reward model, and then obtain the model training reward information; thereafter, the initial large language model can be adjusted based on the model training reward information, and the adjusted model can be tested. If the training stop condition is not met, new samples can be selected for training until a certain training stage, and the model that meets the training stop condition can be used as the large language model and deployed.

[0091] In summary, by fully training the large language model, the large language model can output parent-child communication reminder information with higher accuracy, thereby facilitating parent-child communication users in the communication stage and improving the rationality of communication.

[0092] In one or more implementations of this embodiment, the processing module is also used to construct reward correction information based on historical processing information of the reward model, and input the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip into the reward model for processing to obtain initial model training reward information; use the reward correction information to adjust the initial model training reward information to obtain the model training reward information.

[0093] Specifically, the historical processing information refers to the information recorded by the reward model when scoring the initial large language model, which is constructed by the input and output of the reward model. Correspondingly, the reward correction information refers to the information obtained after analyzing and sorting the historical processing information to correct the scoring results output by the reward model in the application phase, which is used to make the scoring results of the reward model more consistent with the actual scores, thereby improving the training effect of the initial large language model. Correspondingly, the initial model training reward information refers to the initial scoring results output by the reward model.

[0094] Based on this, considering that the reward model is a machine learning model, when it scores the initial large language model, the scoring result may deviate from the true score, and this deviation may affect the training effect of the model. Therefore, the reward model can be corrected after scoring. Specifically, the reward correction information can be constructed based on the historical processing information of the reward model. At the same time, the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip can be input into the reward model for processing to obtain the initial model training reward information; on this basis, the reward correction information can be used to adjust the initial model training reward information to correct the reward information to a real combination, so as to obtain the model training reward information, and then the model parameter optimization can be completed based on this information.

[0095] The parent-child communication reminder system provided in this embodiment can make reasonable reminders in time according to the communication content during the user communication process. After the communication audio is collected by the collection module, the communication audio can be detected by timbre, and the target audio segment corresponding to the parent-child communication user can be extracted from the communication audio, so as to ensure that the target audio segment corresponding to the parent-child communication user can be analyzed by the large language model in the processing module later. After sending the target audio segment to the processing module, the processing module can input the target audio segment into the large language model, and the target audio segment can be recalled in the communication text library by the three-way emotional recall unit contained in the retrieval unit in the large language model, and the candidate communication text corresponding to each emotional recall unit can be obtained; the candidate communication text corresponding to different emotions can be obtained by the three-way recall method, and on this basis, the target communication text can be selected from the candidate communication texts corresponding to the three-way emotional recall units according to the audio attribute information corresponding to the target audio segment; the communication emotion of the current parent-child communication user can be reflected according to the target communication text, and the parent-child communication reminder information can be constructed according to the communication emotion later. At this time, the target communication text and the communication type information corresponding to the target communication text can be processed by the generation unit in the large language model to obtain parent-child communication reminder information; after the parent-child communication reminder information is sent to the reminder module, the reminder module can feedback the parent-child communication reminder information to the parent-child communication user according to the reminder strategy that matches the communication type information. In the communication process of the parent-child communication user, the communication emotion of the parent-child communication user can be quickly determined by combining the efficient processing ability and semantic analysis ability of the large language model, and then the parent-child communication reminder information matching the emotion is generated and fed back to the parent-child communication user, so that the parent-child communication user can make timely adjustments to the communication method, thereby ensuring the effectiveness of communication, especially in the scenario of adolescent psychological counseling, by timely reminding the parent-child communication user, the parent-child communication user can make a positive impact on the psychological problems of adolescents, so as to provide emotional support and psychological security for adolescents through communication.

[0096] Corresponding to the above system embodiment, this specification also provides a parent-child communication reminder method embodiment, Figure 3 FIG. 1 is a flow chart showing a parent-child communication reminder method provided by an embodiment of the present specification. Figure 3 As shown, the method includes: Step S302, performing timbre detection on the collected communication audio, and extracting a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; Step S304, inputting the target audio segment into the large language model, performing text recall of the target audio segment in the communication text library through the three-way emotion recall unit included in the retrieval unit in the large language model, and obtaining candidate communication texts corresponding to each emotion recall unit; Step S306, selecting a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; Step S308, processing the target communication text and the communication type information corresponding to the target communication text by the generation unit in the large language model to obtain parent-child communication reminder information; Step S310: feeding back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0097] In an optional embodiment, a pre-stored non-reminder keyword set is determined, and the target keyword corresponding to the target communication text is extracted; it is detected whether there is a non-reminder keyword matching the target keyword in the non-reminder keyword set; if so, a collection instruction is fed back to the collection module; if not, the target communication text and the communication type information corresponding to the target communication text are processed by the generation unit in the large language model to obtain the parent-child communication reminder information.

[0098] In an optional embodiment, multiple historical audio clips corresponding to the parent-child communication user are obtained at preset time intervals, and historical parent-child communication reminder information corresponding to each historical audio clip is determined; by performing semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip, a communication evaluation report corresponding to the parent-child communication user is obtained; and the communication evaluation report is sent to the user terminal of the parent-child communication user.

[0099] In an optional embodiment, at least one conversation user corresponding to the communication audio is determined, and audio proportion information corresponding to each conversation user is obtained by performing audio recognition on the communication audio; and a parent-child communication user is determined from the at least one conversation user according to the audio proportion information corresponding to each conversation user.

[0100] In an optional embodiment, the timbre information corresponding to the parent-child communication user is determined based on the communication audio, and an initial audio segment is extracted from the communication audio according to the timbre information; the initial audio segment is input into an audio processing model for audio repair processing to obtain a target audio segment corresponding to the parent-child communication user.

[0101] In an optional embodiment, the three-way emotion recall unit includes a positive emotion recall unit, a neutral emotion recall unit and a negative emotion recall unit, wherein the positive emotion recall unit, the neutral emotion recall unit and the negative emotion recall unit correspond to different text recall strategies; the positive emotion recall unit performs positive text recall on the target audio segment in the communication text library to obtain a positive candidate communication text; the neutral emotion recall unit performs neutral text recall on the target audio segment in the communication text library to obtain a neutral candidate communication text; the negative emotion recall unit performs negative text recall on the target audio segment in the communication text library to obtain a negative candidate communication text.

[0102] In an optional embodiment, the processing module is also used to convert the target audio segment into a target audio text; determine the audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text, and determine the positive text vector of the positive communication text, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text in the communication text library; calculate the vector distance between the audio text vector and the positive text vector through the positive emotion recall unit, and screen the positive candidate communication text in the positive communication text according to the vector distance; calculate the semantic similarity between the audio text semantic information and the neutral text semantic information through the neutral emotion recall unit, and screen the neutral candidate communication text in the neutral communication text according to the semantic similarity; perform keyword matching on the audio text keywords and the negative text keywords through the negative emotion recall unit, and screen the negative candidate communication text in the negative communication text according to the keyword matching result.

[0103] In an optional embodiment, audio attribute information corresponding to the target audio segment is determined, and the communication intention of the parent-child communication user is determined based on the audio attribute information; and the target communication text is screened from the candidate communication texts corresponding to the three-way emotion recall units respectively according to the communication intention.

[0104] In an optional embodiment, communication type information corresponding to the target communication text is determined, and target prompt words matching the target communication text are screened from a preset prompt word set according to the communication type information; the target communication text and the communication type information are updated using the target prompt words to obtain context information; the context information is input into a generation unit in the large language model for processing to obtain the parent-child communication reminder information.

[0105] In an optional embodiment, a sample audio clip is extracted from a sample set, and the sample audio clip is input into an initial large language model for processing to obtain predicted parent-child communication reminder information; the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip are input into a reward model of the initial large language model for processing to obtain model training reward information; and the initial large language model is adjusted based on the model training reward information until the large language model that meets the training stop condition is obtained.

[0106] In an optional embodiment, reward correction information is constructed based on historical processing information of the reward model, and the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip are input into the reward model for processing to obtain initial model training reward information; the initial model training reward information is adjusted using the reward correction information to obtain the model training reward information.

[0107] The parent-child communication reminder method provided in this embodiment, in order to achieve timely and reasonable reminders according to the communication content during the user communication process, can be used to extract the target audio segment corresponding to the parent-child communication user from the communication audio by performing timbre detection on the communication audio after the communication audio is collected by the collection module, so as to ensure that the target audio segment corresponding to the parent-child communication user can be analyzed by the large language model in the processing module later. After sending the target audio segment to the processing module, the processing module can input the target audio segment into the large language model, and perform text recall on the target audio segment in the communication text library through the three-way emotional recall unit contained in the retrieval unit in the large language model, and obtain the candidate communication text corresponding to each emotional recall unit; the candidate communication text corresponding to different emotions is obtained by the three-way recall method, and on this basis, the target communication text is screened in the candidate communication texts corresponding to the three-way emotional recall units respectively according to the audio attribute information corresponding to the target audio segment; the communication emotion of the current parent-child communication user can be reflected according to the target communication text, and the parent-child communication reminder information can be constructed later according to the communication emotion. At this time, the target communication text and the communication type information corresponding to the target communication text can be processed by the generation unit in the large language model to obtain parent-child communication reminder information; after the parent-child communication reminder information is sent to the reminder module, the reminder module can feedback the parent-child communication reminder information to the parent-child communication user according to the reminder strategy that matches the communication type information. In the communication process of the parent-child communication user, the communication emotion of the parent-child communication user can be quickly determined by combining the efficient processing ability and semantic analysis ability of the large language model, and then the parent-child communication reminder information matching the emotion is generated and fed back to the parent-child communication user, so that the parent-child communication user can make timely adjustments to the communication method, thereby ensuring the effectiveness of communication, especially in the scenario of adolescent psychological counseling, by timely reminding the parent-child communication user, the parent-child communication user can make a positive impact on the psychological problems of adolescents, so as to provide emotional support and psychological security for adolescents through communication.

[0108] The above is a schematic scheme of a parent-child communication reminder method of this embodiment. It should be noted that the technical scheme of the parent-child communication reminder method and the technical scheme of the parent-child communication reminder system mentioned above belong to the same concept, and the details not described in detail in the technical scheme of the parent-child communication reminder method can be found in the description of the technical scheme of the parent-child communication reminder system mentioned above.

[0109] Corresponding to the above method embodiment, this specification also provides a parent-child communication reminder device embodiment, Figure 4 FIG. 2 shows a schematic diagram of a parent-child communication reminder device provided by an embodiment of the present specification. Figure 4 As shown, the device comprises: The detection module 402 is configured to perform timbre detection on the collected communication audio, and extract a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; The recall module 404 is configured to input the target audio segment into the large language model, and perform text recall of the target audio segment in the communication text library through the three-way emotion recall unit included in the retrieval unit in the large language model to obtain the candidate communication text corresponding to each emotion recall unit; A screening module 406 is configured to screen a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; The processing module 408 is configured to process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; The reminder module 410 is configured to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

[0110] In an optional embodiment, a pre-stored non-reminder keyword set is determined, and the target keyword corresponding to the target communication text is extracted; it is detected whether there is a non-reminder keyword matching the target keyword in the non-reminder keyword set; if so, a collection instruction is fed back to the collection module; if not, the target communication text and the communication type information corresponding to the target communication text are processed by the generation unit in the large language model to obtain the parent-child communication reminder information.

[0111] In an optional embodiment, multiple historical audio clips corresponding to the parent-child communication user are obtained at preset time intervals, and historical parent-child communication reminder information corresponding to each historical audio clip is determined; by performing semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip, a communication evaluation report corresponding to the parent-child communication user is obtained; and the communication evaluation report is sent to the user terminal of the parent-child communication user.

[0112] In an optional embodiment, at least one conversation user corresponding to the communication audio is determined, and audio proportion information corresponding to each conversation user is obtained by performing audio recognition on the communication audio; and a parent-child communication user is determined from the at least one conversation user according to the audio proportion information corresponding to each conversation user.

[0113] In an optional embodiment, the timbre information corresponding to the parent-child communication user is determined based on the communication audio, and an initial audio segment is extracted from the communication audio according to the timbre information; the initial audio segment is input into an audio processing model for audio repair processing to obtain a target audio segment corresponding to the parent-child communication user.

[0114] In an optional embodiment, the three-way emotion recall unit includes a positive emotion recall unit, a neutral emotion recall unit and a negative emotion recall unit, wherein the positive emotion recall unit, the neutral emotion recall unit and the negative emotion recall unit correspond to different text recall strategies; the positive emotion recall unit performs positive text recall on the target audio segment in the communication text library to obtain a positive candidate communication text; the neutral emotion recall unit performs neutral text recall on the target audio segment in the communication text library to obtain a neutral candidate communication text; the negative emotion recall unit performs negative text recall on the target audio segment in the communication text library to obtain a negative candidate communication text.

[0115] In an optional embodiment, the processing module is also used to convert the target audio segment into a target audio text; determine the audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text, and determine the positive text vector of the positive communication text, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text in the communication text library; calculate the vector distance between the audio text vector and the positive text vector through the positive emotion recall unit, and screen the positive candidate communication text in the positive communication text according to the vector distance; calculate the semantic similarity between the audio text semantic information and the neutral text semantic information through the neutral emotion recall unit, and screen the neutral candidate communication text in the neutral communication text according to the semantic similarity; perform keyword matching on the audio text keywords and the negative text keywords through the negative emotion recall unit, and screen the negative candidate communication text in the negative communication text according to the keyword matching result.

[0116] In an optional embodiment, audio attribute information corresponding to the target audio segment is determined, and the communication intention of the parent-child communication user is determined based on the audio attribute information; and the target communication text is screened from the candidate communication texts corresponding to the three-way emotion recall units respectively according to the communication intention.

[0117] In an optional embodiment, communication type information corresponding to the target communication text is determined, and target prompt words matching the target communication text are screened from a preset prompt word set according to the communication type information; the target communication text and the communication type information are updated using the target prompt words to obtain context information; the context information is input into a generation unit in the large language model for processing to obtain the parent-child communication reminder information.

[0118] In an optional embodiment, a sample audio clip is extracted from a sample set, and the sample audio clip is input into an initial large language model for processing to obtain predicted parent-child communication reminder information; the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip are input into a reward model of the initial large language model for processing to obtain model training reward information; and the initial large language model is adjusted based on the model training reward information until the large language model that meets the training stop condition is obtained.

[0119] In an optional embodiment, reward correction information is constructed based on historical processing information of the reward model, and the sample audio clip, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio clip are input into the reward model for processing to obtain initial model training reward information; the initial model training reward information is adjusted using the reward correction information to obtain the model training reward information.

[0120] The above is a schematic scheme of a parent-child communication reminder device of this embodiment. It should be noted that the technical scheme of the parent-child communication reminder device and the technical scheme of the parent-child communication reminder method mentioned above belong to the same concept, and the details not described in detail in the technical scheme of the parent-child communication reminder device can be found in the description of the technical scheme of the parent-child communication reminder method mentioned above.

[0121] Figure 5A block diagram of a computing device 500 provided according to an embodiment of the present specification is shown. The components of the computing device 500 include, but are not limited to, a memory 510 and a processor 520. The processor 520 is connected to the memory 510 via a bus 530, and a database 550 is used to store data. The computing device 500 also includes an access device 540, which enables the computing device 500 to communicate via one or more networks 560. Examples of these networks include a public switched telephone network (PSTN), a local area network (LAN), a wide area network (WAN), a personal area network (PAN), or a combination of communication networks such as the Internet. The access device 540 may include one or more of any type of wired or wireless network interface (e.g., a network interface card (NIC)), such as an IEEE802.11 wireless local area network (WLAN) wireless interface, a Worldwide Interoperability for Microwave Access (Wi-MAX) interface, an Ethernet interface, a Universal Serial Bus (USB) interface, a cellular network interface, a Bluetooth interface, and a Near Field Communication (NFC). In one embodiment of the present specification, the above components of the computing device 500 and Figure 5 Other components not shown in the figure may also be connected to each other, for example, via a bus. It should be understood that Figure 5The computing device structure block diagram shown is for illustrative purposes only and is not intended to limit the scope of this specification. Those skilled in the art may add or replace other components as needed. The computing device 500 may be any type of stationary or mobile computing device, including a mobile computer or mobile computing device (e.g., a tablet computer, a personal digital assistant, a laptop computer, a notebook computer, a netbook, etc.), a mobile phone (e.g., a smart phone), a wearable computing device (e.g., a smart watch, smart glasses, etc.) or other types of mobile devices, or a stationary computing device such as a desktop computer or a personal computer (PC, Personal Computer). The computing device 500 may also be a mobile or stationary server. Among them, the processor 520 is used to execute the following computer executable instructions, which implement the steps of the above-mentioned parent-child communication reminder method when executed by the processor. The above is a schematic scheme of a computing device of the present embodiment. It should be noted that the technical scheme of the computing device and the technical scheme of the above-mentioned parent-child communication reminder method belong to the same concept, and the details of the technical scheme of the computing device that are not described in detail can all be referred to the description of the technical scheme of the above-mentioned parent-child communication reminder method.

[0122] This specification also provides a computer-readable storage medium in one embodiment, which stores computer-executable instructions, which implement the steps of the above-mentioned parent-child communication reminder method when executed by a processor. The above is a schematic scheme of a computer-readable storage medium of this embodiment. It should be noted that the technical solution of the storage medium belongs to the same concept as the technical solution of the above-mentioned parent-child communication reminder method. For details not described in detail in the technical solution of the storage medium, please refer to the description of the technical solution of the above-mentioned parent-child communication reminder method.

[0123] This specification also provides a computer program product in one embodiment, including a computer program or instruction, which implements the steps of the above-mentioned parent-child communication reminder method when executed by a processor. The above is a schematic scheme of a computer program product of this embodiment. It should be noted that the technical solution of the computer program product and the technical solution of the above-mentioned parent-child communication reminder method belong to the same concept, and the details not described in detail in the technical solution of the computer program product can be found in the description of the technical solution of the above-mentioned parent-child communication reminder method.

[0124] The above is a description of a specific embodiment of the specification. Other embodiments are within the scope of the appended claims. In some cases, the actions or steps recorded in the claims can be performed in an order different from that in the embodiments and still achieve the desired results. In addition, the processes depicted in the drawings do not necessarily require the specific order or continuous order shown to achieve the desired results. In some embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0125] The computer instructions include computer program codes, which may be in source code form, object code form, executable files or some intermediate forms, etc. The computer-readable medium may include: any entity or device capable of carrying the computer program code, recording medium, USB flash drive, mobile hard disk, magnetic disk, optical disk, computer memory, read-only memory (ROM), random access memory (RAM), electric carrier signal, telecommunication signal and software distribution medium, etc. It should be noted that the content contained in the computer-readable medium may be appropriately increased or decreased according to the requirements of patent practice. For example, in some regions, according to patent practice, computer-readable media do not include electric carrier signals and telecommunication signals.

[0126] It should be noted that, for the above-mentioned method embodiments, for the sake of simplicity of description, they are all expressed as a series of action combinations, but those skilled in the art should be aware that the embodiments of this specification are not limited by the order of the actions described, because according to the embodiments of this specification, some steps can be performed in other orders or simultaneously. Secondly, those skilled in the art should also be aware that the embodiments described in the specification are all preferred embodiments, and the actions and modules involved are not necessarily required by the embodiments of this specification.

[0127] In the above embodiments, the description of each embodiment has its own emphasis. For parts that are not described in detail in a certain embodiment, reference can be made to the relevant descriptions of other embodiments.

[0128] The preferred embodiments of this specification disclosed above are only used to help explain this specification. The optional embodiments do not describe all the details in detail, nor do they limit the invention to the specific implementation methods described. Obviously, many modifications and changes can be made according to the content of the embodiments of this specification. This specification selects and specifically describes these embodiments in order to better explain the principles and practical applications of the embodiments of this specification, so that technicians in the relevant technical field can understand and use this specification well.

Claims

1. A parent-child communication reminder system, characterized in that: It includes acquisition module, processing module and reminder module, including: The acquisition module is used to collect communication audio; by performing timbre detection on the communication audio, a target audio segment corresponding to the parent-child communication user is extracted from the communication audio, and the target audio segment is sent to the processing module; The processing module is used to input the target audio segment into the large language model, perform text recall of the target audio segment in the communication text library through the three-way emotional recall unit included in the retrieval unit in the large language model, and obtain the candidate communication text corresponding to each emotional recall unit; select the target communication text from the candidate communication texts respectively corresponding to the three-way emotional recall unit according to the audio attribute information corresponding to the target audio segment; process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; and send the parent-child communication reminder information to the reminder module; The reminder module is used to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

2. The parent-child communication reminder system according to claim 1, characterized in that: The processing module is also used to determine a pre-stored non-reminder keyword set and extract the target keyword corresponding to the target communication text; detect whether there is a non-reminder keyword matching the target keyword in the non-reminder keyword set; if so, feed back a collection instruction to the collection module; if not, execute the step of processing the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information.

3. The parent-child communication reminder system according to claim 1, characterized in that: The processing module is also used to obtain multiple historical audio clips corresponding to the parent-child communication user at preset time intervals, and determine the historical parent-child communication reminder information corresponding to each historical audio clip; obtain the communication evaluation report corresponding to the parent-child communication user by performing semantic analysis on the historical parent-child communication reminder information corresponding to each historical audio clip; and send the communication evaluation report to the user terminal of the parent-child communication user.

4. The parent-child communication reminder system according to claim 1, characterized in that: The acquisition module is also used to determine at least one conversation user corresponding to the communication audio, and obtain audio proportion information corresponding to each conversation user by performing audio recognition on the communication audio; and determine the parent-child communication user from the at least one conversation user according to the audio proportion information corresponding to each conversation user.

5. The parent-child communication reminder system according to claim 1, characterized in that: The acquisition module is also used to determine the timbre information corresponding to the parent-child communication user based on the communication audio, extract the initial audio segment from the communication audio according to the timbre information; input the initial audio segment into the audio processing model for audio repair processing, and obtain the target audio segment corresponding to the parent-child communication user.

6. The parent-child communication reminder system according to claim 1, characterized in that: The three-way emotion recall unit includes a positive emotion recall unit, a neutral emotion recall unit and a negative emotion recall unit, wherein the positive emotion recall unit, the neutral emotion recall unit and the negative emotion recall unit correspond to different text recall strategies; The processing module is also used to perform positive text recall on the target audio segment in the communication text library through the positive emotion recall unit to obtain positive candidate communication text; perform neutral text recall on the target audio segment in the communication text library through the neutral emotion recall unit to obtain neutral candidate communication text; and perform negative text recall on the target audio segment in the communication text library through the negative emotion recall unit to obtain negative candidate communication text.

7. The parent-child communication reminder system according to claim 6, characterized in that: The processing module is further used to convert the target audio segment into a target audio text; determine the audio text vector, audio text semantic information and audio text keywords corresponding to the target audio text, and determine the positive text vector of the positive communication text, the neutral text semantic information of the neutral communication text and the negative text keywords of the negative communication text in the communication text library; Calculating the vector distance between the audio text vector and the positive text vector by the positive emotion recall unit, and screening the positive candidate communication text from the positive communication text according to the vector distance; Calculating the semantic similarity between the audio text semantic information and the neutral text semantic information by the neutral emotion recall unit, and screening the neutral candidate communication text from the neutral communication text according to the semantic similarity; The negative emotion recall unit performs keyword matching on the audio text keywords and the negative text keywords, and screens negative candidate communication texts from the negative communication texts according to the keyword matching results.

8. The parent-child communication reminder system according to any one of claims 1 to 7, characterized in that: The processing module is also used to determine the audio attribute information corresponding to the target audio segment, determine the communication intention of the parent-child communication user according to the audio attribute information; and select the target communication text from the candidate communication texts corresponding to the three-way emotion recall units respectively according to the communication intention.

9. The parent-child communication reminder system according to any one of claims 1 to 7, characterized in that: The processing module is further configured to determine communication type information corresponding to the target communication text, and select a target prompt word matching the target communication text from a preset prompt word set according to the communication type information; Using the target prompt word, the target communication text and the communication type information are updated to obtain context information; The context information is input into the generation unit in the large language model for processing to obtain the parent-child communication reminder information.

10. The parent-child communication reminder system according to any one of claims 1 to 7, characterized in that: The processing module is further used to extract a sample audio segment from the sample set, input the sample audio segment into the initial large language model for processing, and obtain predicted parent-child communication reminder information; input the sample audio segment, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio segment into the reward model of the initial large language model for processing, and obtain model training reward information; The initial large language model is adjusted based on the model training reward information until the large language model that meets the training stop condition is obtained.

11. The parent-child communication reminder system according to claim 10, characterized in that: The processing module is further used to construct reward correction information according to the historical processing information of the reward model, and input the sample audio segment, the predicted parent-child communication reminder information, and the sample parent-child communication reminder information corresponding to the sample audio segment into the reward model for processing to obtain initial model training reward information; The initial model training reward information is adjusted using the reward correction information to obtain the model training reward information.

12. A parent-child communication reminder method, characterized in that: include: Performing timbre detection on the collected communication audio, and extracting a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; The target audio segment is input into a large language model, and the target audio segment is recalled in a communication text library by three-way emotion recall units included in a retrieval unit in the large language model to obtain candidate communication texts corresponding to each emotion recall unit; Filtering a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; Processing the target communication text and the communication type information corresponding to the target communication text by a generation unit in the large language model to obtain parent-child communication reminder information; The parent-child communication reminder information is fed back to the parent-child communication user according to a reminder strategy matching the communication type information.

13. A parent-child communication reminder device, characterized in that: include: A detection module is configured to perform timbre detection on the collected communication audio, and extract a target audio segment corresponding to the parent-child communication user from the communication audio according to the timbre detection result; A recall module is configured to input the target audio segment into a large language model, perform text recall of the target audio segment in a communication text library through three-way emotion recall units included in a retrieval unit in the large language model, and obtain candidate communication texts corresponding to each emotion recall unit; A screening module is configured to screen a target communication text from the candidate communication texts respectively corresponding to the three emotion recall units according to the audio attribute information corresponding to the target audio segment; a processing module configured to process the target communication text and the communication type information corresponding to the target communication text through the generation unit in the large language model to obtain parent-child communication reminder information; The reminder module is configured to feed back the parent-child communication reminder information to the parent-child communication user according to the reminder strategy matching the communication type information.

14. A computing device, characterized in that include: Memory and processor; The memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions. When the computer-executable instructions are executed by the processor, the steps of the method according to claim 12 are implemented.

15. A computer-readable storage medium, characterized in that: It stores computer executable instructions, which implement the steps of the method according to claim 12 when executed by a processor.

16. A computer program product, characterized in that The method comprises a computer program or an instruction, which implements the steps of the method according to claim 12 when the computer program or the instruction is executed by a processor.

Images