A room violation auditing method, device, equipment, storage medium and product

Abstract

Embodiments of the present application provide a room violation auditing method and device, equipment, storage medium and product. The technical solution provided by the embodiments of the present application determines the online black list user in the room to be audited according to the black list, determines the suspected room from the room to be audited according to the online black list user, obtains the room audio information of the suspected room, determines the suspected audio from the room audio information of each suspected room, and performs violation auditing on the suspected room based on the suspected audio. The suspected room is determined by preliminary screening of the room to be audited by the online black list user, and secondary screening is performed according to the room audio information of the suspected room to determine the suspected audio. The suspected room can be accurately audited based on the suspected audio. The suspected audio that may exist violation is determined from the voice information of the live room, which effectively improves the auditing accuracy of the room and improves the room auditing effect.

Description

Technical Field

[0001] This application relates to the field of live streaming technology, and in particular to a method, apparatus, equipment, storage medium, and product for reviewing room violations. Background Technology

[0002] Live streaming rooms and voice chat rooms are currently the mainstream entertainment methods on social applications. However, some live streaming rooms often have some violations. In order to ensure the user's live streaming experience, it is necessary to review and restrict the rooms that violate the rules.

[0003] Traditional methods for reviewing rule-breaking rooms typically use a review model to analyze audio, image, and text information generated within the room to determine if it violates regulations. These methods generally employ a classification model to categorize the input information and determine room compliance based on the classification results. However, these methods usually only review single data entries from a single user, resulting in poor room-wide review effectiveness. Summary of the Invention

[0004] This application provides a method, apparatus, device, storage medium, and product for room violation review, to solve the technical problem in related technologies where room review is poor when reviewing single data generated by a single user, thereby effectively improving the accuracy and effectiveness of room review.

[0005] In a first aspect, embodiments of this application provide a method for reviewing room violations, including:

[0006] Based on the blacklist, identify online blacklist users in the rooms to be reviewed, and based on the online blacklist users, identify suspected rooms from the rooms to be reviewed;

[0007] Obtain the room audio information of the suspected room, and determine the suspected audio from the room audio information;

[0008] The room in question was reviewed for violations based on the suspected audio.

[0009] In a second aspect, embodiments of this application provide a room violation verification device, including a room determination module, an audio determination module, and a room verification module, wherein:

[0010] The room determination module is configured to determine online blacklist users in the room to be reviewed based on the blacklist list, and to determine the suspected room from the room to be reviewed based on the online blacklist users;

[0011] The audio determination module is configured to acquire room audio information of the suspected room and determine the suspected audio from the room audio information;

[0012] The room review module is configured to review the suspected room for violations based on the suspected audio.

[0013] In a third aspect, embodiments of this application provide a room violation review device, including: a memory and one or more processors;

[0014] The memory is used to store one or more programs;

[0015] When the one or more programs are executed by the one or more processors, the one or more processors implement the room violation review method as described in the first aspect.

[0016] In a fourth aspect, embodiments of this application provide a non-volatile storage medium for storing computer-executable instructions, which, when executed by a computer processor, are used to perform the room violation review method as described in the first aspect.

[0017] In a fifth aspect, embodiments of this application provide a computer program product comprising a computer program stored in a computer-readable storage medium, wherein at least one processor of the device reads from the computer-readable storage medium and executes the computer program, causing the device to perform the room violation auditing method as described in the first aspect.

[0018] This application embodiment determines online blacklist users in the rooms to be reviewed based on a blacklist list, identifies suspected rooms from the rooms to be reviewed based on the online blacklist users, obtains the room audio information of the suspected rooms, identifies suspicious audio from the room audio information of each suspected room, and conducts violation review on the suspected rooms based on the suspicious audio. By initially screening the rooms to be reviewed through online blacklist users to identify suspected rooms, and then conducting a secondary screening based on the room audio information of the suspected rooms to identify suspicious audio, violation review of suspected rooms can be accurately conducted based on the suspicious audio. By identifying potentially violating suspicious audio from the voice information of the live room, the accuracy of room review is effectively improved, and the room review effect is enhanced. Attached Figure Description

[0019] Figure 1 This is a flowchart of a room violation review method provided in an embodiment of this application;

[0020] Figure 2 This is a schematic diagram of a blacklist generation process provided in an embodiment of this application;

[0021] Figure 3 This is a flowchart of another room violation review method provided in the embodiments of this application;

[0022] Figure 4This is a schematic diagram of an audio suspect score determination process provided in an embodiment of this application;

[0023] Figure 5 This is a schematic diagram of a room violation review device provided in an embodiment of this application;

[0024] Figure 6 This is a schematic diagram of a room violation review device provided in an embodiment of this application. Detailed Implementation

[0025] To make the objectives, technical solutions, and advantages of this application clearer, specific embodiments of this application will be described in further detail below with reference to the accompanying drawings. It should be understood that the specific embodiments described herein are merely for explaining this application and not for limiting it. It should also be noted that, for ease of description, only the parts relevant to this application are shown in the drawings, not all of them. Before discussing exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flowcharts. Although the flowcharts describe operations (or steps) as sequential processes, many of these operations can be performed in parallel, concurrently, or simultaneously. Furthermore, the order of the operations can be rearranged. The process can be terminated when its operation is completed, but additional steps not included in the drawings may also be present. The above processes can correspond to methods, functions, procedures, subroutines, subroutines, etc.

[0026] The room violation review method provided in this application can be applied to live streaming scenarios, such as voice live streaming and video live streaming. It aims to initially screen rooms to be reviewed by using online blacklisted users to identify suspected rooms, and then further screen and identify suspicious audio based on the audio information of these suspected rooms. This allows for accurate violation review of suspected rooms based on the suspicious audio, effectively improving the accuracy and effectiveness of room review. Traditional room violation review schemes generally rely on manual inspections by administrators or automatic detection by artificial intelligence models. Manual inspections typically involve administrators manually reviewing the live content of each room, which is inefficient. Automatic detection by artificial intelligence models generally uses classification models to judge violations based on audio, images, and text in each live room. However, artificial intelligence models can only review single data points generated by a single user, limiting the scope of violations. Furthermore, artificial intelligence models need to review each live room sequentially, resulting in low efficiency. Moreover, online violations are often hidden within a vast number of live streaming rooms or voice chat rooms, and communication between violating users is often subtle, making it difficult to effectively combat violations through manual or automated methods. Based on this, an embodiment of the present application provides a method for room violation review to solve the technical problem that the existing room violation review scheme has poor violation review effect.

[0027] Figure 1 A flowchart of a room violation review method provided in this application embodiment is given. The room violation review method provided in this application embodiment can be executed by a room violation review device, which can be implemented by hardware and / or software and integrated into a room violation review equipment.

[0028] The following description uses a room violation verification device as an example to illustrate the method of verifying room violations. (Reference) Figure 1 The methods used to verify violations in this room include:

[0029] S110: Based on the blacklist, identify online blacklisted users in the rooms to be reviewed, and based on the online blacklisted users, identify suspected rooms from the rooms to be reviewed.

[0030] The blacklist provided in this solution contains multiple blacklisted users. Blacklisted users can be understood as users who engage in violations within a set time period, users who frequently enter or leave a live stream room that is deemed to be in violation, or the streamers of a live stream room that is deemed to be in violation.

[0031] It should be explained that blacklisted users are more likely to enter live streaming rooms that may violate regulations. Live streaming rooms that may violate regulations can be initially screened based on the number of blacklisted users online in the room, thus achieving a preliminary review of suspected rooms.

[0032] For example, the online users in each room awaiting review are identified, and the online blacklist users in each room are determined based on the blacklist. The online blacklist users provided in this solution can be understood as online users recorded in the blacklist who have entered the room awaiting review. Further, after identifying the online blacklist users in the rooms awaiting review, suspected rooms are identified from each room based on these online blacklist users. The suspected rooms provided in this solution can be understood as live streaming rooms with potential violations; that is, suspected rooms may or may not have violated regulations, requiring further review to determine if violations have occurred.

[0033] In one embodiment, identifying a suspected room from each room awaiting review based on online blacklist users can be done by determining the number of online blacklist users in the room awaiting review. For example, if the number of online blacklist users in a room awaiting review reaches a set threshold, the room awaiting review can be identified as a suspected room. In another embodiment, identifying a suspected room from each room awaiting review based on online blacklist users can also be done by determining the frequency of appearance and / or duration of stay of online blacklist users in the room awaiting review. For example, if the frequency of appearance and / or duration of stay of online blacklist users in a room awaiting review reaches a set frequency threshold and / or duration threshold, the room awaiting review can be identified as a suspected room.

[0034] In one possible embodiment, the blacklist provided by this solution can be determined based on users in rooms deemed to have violated the rules. Based on this, as... Figure 2 As shown in the provided diagram of a blacklist generation process, the room violation review method provided in this solution may further include, before determining the online blacklist users in the room to be reviewed based on the blacklist, the following:

[0035] S101: Identify candidate users who will enter the room that violates the rules, and determine the frequency of these candidate users entering the room that violates the rules.

[0036] S102: Determine blacklisted users based on their frequency of occurrence and a set third threshold, and determine the blacklist list based on the blacklisted users.

[0037] For example, this solution obtains a list of rooms with violations. This list contains multiple rooms confirmed to have engaged in violations. The violation rooms can be identified through manual review, user reports, or automated detection. Each room in the list is iterated over. For each room, users who entered the room are identified and designated as candidate users. These candidate users can be understood as users who are potential candidates for a blacklist.

[0038] Furthermore, the frequency of each candidate user entering a violation room can be determined. For example, the frequency of a candidate user's appearance can be increased each time they enter a violation room. Optionally, the frequency of each candidate user's appearance can be calculated separately for each violation room, or the frequency of each candidate user's appearance in all violation rooms can be recorded uniformly.

[0039] For each candidate user, the frequency of their appearance is compared with a set third threshold. When the frequency reaches the third threshold, the candidate user is added to the blacklist, and an initial blacklist is created based on the identified blacklist users. In one embodiment, the blacklist can be updated in real time based on users who frequently enter or stay in live streams deemed to be in violation (i.e., violation rooms), improving the accuracy of room violation review.

[0040] This solution accurately identifies blacklisted users based on the frequency of appearance of candidate users entering rooms that violate regulations, thus generating a blacklist list. Based on the blacklist list, suspected rooms can be accurately identified, improving the efficiency and accuracy of room violation review.

[0041] S120: Obtain the room audio information of the suspect room and identify the suspect audio from the room audio information.

[0042] For example, the audio information of each suspected room is obtained within a set time period. The room audio information provided by this solution can be the audio information of each user in the suspected room (such as the host, the user who takes the microphone, the host who connects with the host, etc.), that is, the various audio information that appears in the suspected room.

[0043] Furthermore, after obtaining the room audio information of the suspected room, the suspected audio is identified from the room audio information. It should be explained that the suspected audio provided in this solution can be understood as audio information that may contain violations; that is, the suspected audio may or may not contain violations, and further review is required to determine whether there are any violations.

[0044] Optionally, a trained neural network model can be used to determine whether the room audio information is suspicious, or the violation can be judged based on the text information corresponding to the room audio information, to determine whether there is any illegal text information in the text information, and to determine the room audio information as suspicious if illegal text information is found.

[0045] S130: Conduct violation review of suspected rooms based on suspected audio.

[0046] For example, after identifying suspicious audio, the suspected room is reviewed for violations based on the suspicious audio to determine whether the suspected room is a violation room. For instance, it can be determined whether the suspicious audio contains inappropriate voice, and if inappropriate voice is found, the suspected room is identified as a violation room.

[0047] Optionally, suspected audio can be sent to a pre-trained violation review model. The model can then accurately determine whether a room is in violation based on the audio. This model only needs to review rooms containing suspected audio, eliminating the need for reviewing every single live stream, thus significantly improving review efficiency. Alternatively, suspected rooms can be submitted to a manual review process based on the suspected audio. Manual review determines whether the audio contains inappropriate sounds and whether the room is in violation. This method also only requires manual review of rooms containing suspected audio, further improving review efficiency.

[0048] The above-mentioned method involves identifying online blacklisted users in rooms awaiting review based on a blacklist, identifying suspected rooms from these blacklisted users, obtaining the room audio information of these suspected rooms, identifying suspicious audio from the audio information of each suspected room, and conducting violation review on these suspected rooms based on the suspicious audio. This method uses online blacklisted users to initially screen rooms awaiting review and then conducts a secondary screening based on the room audio information of these suspected rooms to identify suspicious audio. This allows for accurate violation review of suspected rooms based on the suspicious audio, effectively improving the accuracy and effectiveness of room review by identifying potentially violating audio from the voice information of live streaming rooms.

[0049] Based on the above embodiments, Figure 3 A flowchart of another room violation review method provided in this application embodiment is given, which is a concretization of the above-described room violation review method. (Reference) Figure 3 The methods used to verify violations in this room include:

[0050] S210: Identify the online users in each room to be reviewed, and based on the blacklist, determine the online blacklist users in each room to be reviewed, as well as the user weight of each online blacklist user.

[0051] In addition to recording multiple blacklisted users, the blacklist provided by this solution also records the user weight corresponding to each blacklisted user. The higher the user weight of a blacklisted user, the higher the likelihood that they will frequently enter or linger in a live stream and engage in rule violations.

[0052] In one embodiment, the user weight provided by this solution can be determined by the frequency of blacklisted users appearing in violation rooms. For example, when determining whether a user is a blacklisted user based on the frequency of a user entering a violation room, if the user is determined to be a blacklisted user, the frequency of the blacklisted user's appearance in each violation room is used as the user weight and recorded in the blacklist list.

[0053] For example, the system identifies online users in each room awaiting review, determines whether these online users are blacklisted users recorded in the blacklist list, identifies blacklisted users in the room awaiting review as online blacklisted users, and determines the user weight of these online blacklisted users.

[0054] S220: Determine the room suspicion score for each room to be reviewed based on the number of online users and user weights.

[0055] For example, for each room to be reviewed, the number of online users in each room is determined, and a room suspicion score is determined based on the number of online users and the user weights corresponding to the online blacklisted users in the room. It is understood that the higher the room suspicion score provided by this solution, the higher the likelihood that the room to be reviewed has engaged in violations.

[0056] In one embodiment, the fewer the number of online users in a room awaiting review and the greater the sum of the user weights corresponding to the online blacklisted users in the room awaiting review, the higher the room's suspicion score. Optionally, the room suspicion score provided by this solution can be determined based on the ratio of the sum of the user weights of each online blacklisted user in the room awaiting review to the number of online users.

[0057] For example, the room suspicion score for each room awaiting review can be determined based on the following room suspicion score calculation formula:

[0058]

[0059] in, For online users in the room awaiting review, For users on the online blacklist in the room awaiting review, The user weight corresponding to the blacklisted users. This refers to the number of online users in the room to be reviewed. This solution accurately determines the room's suspicion score and the room's suspicion level by using the ratio of the sum of the user weights of each blacklisted user in the room to the total number of online users, thus improving the accuracy and efficiency of room violation review.

[0060] S230: Identify suspicious rooms from the rooms to be reviewed based on the room suspicion score and a set first threshold.

[0061] For example, for each room to be reviewed, the room suspicion score is compared with a set first threshold. When the room suspicion score reaches the first threshold, the room to be reviewed is identified as a suspected room. This solution determines the room suspicion score of each room to be reviewed based on the number of online users and the user weight of users on the online blacklist. It then accurately identifies suspected rooms based on the comparison result between the room suspicion score and the first threshold, achieving preliminary review of live streaming rooms, narrowing down the scope of live streaming rooms requiring review, and improving the accuracy and efficiency of room violation review.

[0062] S240: Obtain the user audio information of each online user in the suspected room, and sort the user audio information to obtain the room audio information.

[0063] For example, for each suspected room, the audio information of each online user in the suspected room within a set time period is obtained. The audio information of each online user includes the audio information generated in the suspected room by the room owner (including the host and other people at the host's location), users who speak on the microphone (listeners or viewers who speak on the microphone), and hosts who connect to the microphone. The online user and time information corresponding to each segment of user audio information are marked.

[0064] Furthermore, based on the time information corresponding to each segment of the user audio information, these audio segments are sorted to reconstruct the entire dialogue within a set time period. Simultaneously, the sorted user audio segments are combined to obtain the room audio information. This solution obtains the user audio information of each online user in a suspected room and sorts it to obtain room audio information. The room audio information records the complete dialogue within a set time period, which, combined with the contextual information within that time period, can accurately determine whether any inappropriate dialogue has occurred, making it easier to detect violations and improving the effectiveness of room violation review.

[0065] S250: Determine the audio suspicion score of the room audio information, and identify the suspect audio from the room audio information based on the audio suspicion score and a set second threshold.

[0066] For example, after obtaining the audio information of each suspected room, an audio suspicion score is calculated for each room. It is understood that the higher the audio suspicion score provided by this solution, the higher the probability that the corresponding room's audio information contains inappropriate speech, and consequently, the higher the probability that the suspected room contains inappropriate behavior.

[0067] Furthermore, for each room's audio information, the voice suspicion score is compared with a set second threshold. When the voice suspicion score reaches the second threshold, the room's audio information is identified as suspicious audio. This solution accurately identifies suspicious audio based on the comparison result of the audio suspicion score and the second threshold, enabling a secondary review of live streaming rooms, further narrowing down the scope of live streaming rooms requiring review, and improving the accuracy and efficiency of room violation review.

[0068] In one possible embodiment, the audio suspicion score provided by this solution can be determined by a trained semantic model. Based on this, as... Figure 4 As shown in the schematic diagram of the audio suspicion score determination process, the room violation review method provided in this solution includes the following steps when determining the audio suspicion score of room audio information:

[0069] S251: Obtain room audio information and user information corresponding to online users in the suspected room.

[0070] S252: Input the room audio information and user information into the trained semantic model, and determine the audio suspicion score of the room audio information based on the room audio information and user information through the semantic model.

[0071] For example, the system obtains the room audio information of the suspected room (which records the sorted audio information of each online user in the suspected room within a set time period), as well as the user information corresponding to each online user in the suspected room. Optionally, the user information provided in this solution can be the online user's basic information (such as the user profile on the user's homepage), the number of violations recorded by the user, whether the user is the room owner, the user's weight in the blacklist, etc. The user information can be determined through the online user's public information, without involving the user's privacy. This enriches the factors considered in the audio suspicion score, improves the accuracy of the audio suspicion score determination, and effectively ensures the privacy and security of online users.

[0072] Furthermore, the room audio information and user information are input into the trained semantic model. Upon receiving the room audio information and user information, the semantic model determines the audio suspicion score of the room audio information based on these factors. The semantic model provided in this solution can be a multi-layer LSTM model. When determining the audio suspicion score of the room audio information based on the room audio information and user information, the semantic model first converts the room audio information into text information using an ASR (Automatic Speech Recognition) algorithm, and then embeds the text information into a vector representation (sentence embedding) using a BERT Embedding model. This vector information serves as the semantic representation of the room audio information. Further, the semantic representation is concatenated with the corresponding user information to obtain concatenated features, which are then input into the semantic model. The semantic model analyzes and processes the concatenated features to obtain the corresponding audio suspicion score.

[0073] This solution analyzes and processes room audio information and user information using a semantic model to accurately determine the audio suspicion score of the room audio information. Based on the audio suspicion score, it can accurately identify suspicious audio, thereby improving the accuracy and efficiency of room violation review.

[0074] In one embodiment, when a suspected audio is confirmed to be in violation after violation review, all audio from the corresponding violation room near the time point of the violation audio, as well as public data, can be added to the training set. The semantic model is iteratively updated based on the training set model, so that the semantic model can learn the latest behavioral patterns of violating users. Even if the words used by users in violation communication become increasingly obscure, it can still accurately identify audio information that may be in violation, improve the accuracy of room violation review, and achieve continuous and effective crackdown on violation behavior.

[0075] In one embodiment, when the suspected audio is confirmed to be in violation after violation review, the violating room can be added to the list of violating rooms, and the closed live rooms in the list can be deleted to obtain a new list of violating rooms. The new list of violating rooms can be used to update the user blacklist (refer to steps S101-S102). In addition, blacklisted users are monitored. If a blacklisted user does not exhibit any violations within a set time period, or if the frequency of appearance and / or duration of stay in a violating room is less than a set threshold, the blacklisted user is removed from the blacklist. This enables real-time updates to the blacklist and ensures the accuracy of room violation review.

[0076] S260: Conduct violation review of suspected rooms based on suspected audio.

[0077] The above-mentioned method involves identifying online blacklisted users in rooms awaiting review based on a blacklist, identifying suspected rooms from these blacklisted users, obtaining the room audio information of these suspected rooms, identifying suspicious audio from the audio information of each suspected room, and conducting violation review on these suspected rooms based on the suspicious audio. This method uses online blacklisted users to initially screen rooms awaiting review and then performs a secondary screening based on the room audio information of these suspected rooms to identify suspicious audio. This allows for accurate violation review of suspected rooms based on the suspicious audio, effectively improving the accuracy and effectiveness of room review. Furthermore, by determining the room suspicion score of each room awaiting review based on the number of online users and the user weight of online blacklisted users, and comparing the room suspicion score with a first threshold, the method accurately identifies suspected rooms, achieving preliminary review of live streaming rooms, narrowing down the range of live streaming rooms requiring review, and improving the accuracy and efficiency of room violation review. Furthermore, by comparing the audio suspicion score with the second threshold, the system can accurately identify suspicious audio, enabling a second review of the live streaming room, further narrowing down the scope of live streaming rooms that need to be reviewed, and improving the accuracy and efficiency of room violation review.

[0078] Figure 5 This is a schematic diagram of a room violation verification device provided in an embodiment of this application. (Reference) Figure 5 The room violation verification device includes a room determination module 51, an audio determination module 52, and a room verification module 53.

[0079] Among them, the room determination module 51 is configured to determine the online blacklist users in the room to be reviewed based on the blacklist list, and determine the suspected room from the room to be reviewed based on the online blacklist users; the audio determination module 52 is configured to obtain the room audio information of the suspected room, and determine the suspected audio from the room audio information; the room review module 53 is configured to conduct violation review on the suspected room based on the suspected audio.

[0080] The above-mentioned method involves identifying online blacklisted users in rooms awaiting review based on a blacklist, identifying suspected rooms from these blacklisted users, obtaining the room audio information of these suspected rooms, identifying suspicious audio from the audio information of each suspected room, and conducting violation review on these suspected rooms based on the suspicious audio. This method uses online blacklisted users to initially screen rooms awaiting review and then conducts a secondary screening based on the room audio information of these suspected rooms to identify suspicious audio. This allows for accurate violation review of suspected rooms based on the suspicious audio, effectively improving the accuracy and effectiveness of room review by identifying potentially violating audio from the voice information of live streaming rooms.

[0081] In one possible embodiment, the room determination module 51 is configured as follows:

[0082] Identify the online users in each room awaiting review;

[0083] Based on the blacklist, identify the online blacklist users in each room awaiting review from the online users, as well as the user weight of each online blacklist user;

[0084] The room suspicion score for each room to be reviewed is determined based on the number of online users and user weight.

[0085] Suspected rooms are identified from the rooms to be reviewed based on the room suspicion score and a set first threshold.

[0086] In one possible embodiment, the room suspicion score is determined based on the ratio of the sum of the user weights of each online blacklisted user in the room to be reviewed to the number of online users of the online user.

[0087] In one possible embodiment, when the audio determination module 52 acquires the room audio information of the suspect room, it is configured as follows:

[0088] Obtain the audio information of each online user in the suspected room;

[0089] The room audio information is obtained by sorting the user audio information.

[0090] In one possible embodiment, when the audio determination module 52 determines the suspect audio from the room audio information, it is configured as follows:

[0091] Determine the audio suspicion score of the room's audio information;

[0092] Suspicious audio is identified from room audio information based on the audio suspicion score and a set second threshold.

[0093] In one possible embodiment, the audio determination module 52 is configured to: determine the audio suspicion score of the room audio information as follows:

[0094] Obtain audio information from the room and user information corresponding to online users in the suspected room;

[0095] The room audio information and user information are input into the trained semantic model, and the semantic model determines the audio suspicion score of the room audio information based on the room audio information and user information.

[0096] In one possible embodiment, the room violation verification device further includes a frequency determination module and a list determination module, wherein:

[0097] The frequency determination module is configured to determine candidate users who enter the room that violates the rules, and to determine the frequency of the candidate users entering the room that violates the rules.

[0098] The blacklist determination module is configured to determine blacklisted users based on their frequency of occurrence and a set third threshold, and to determine the blacklist list based on the blacklisted users.

[0099] It is worth noting that in the above-mentioned embodiments of the room violation review device, the various units and modules included are only divided according to functional logic, but are not limited to the above division, as long as the corresponding functions can be achieved; in addition, the specific names of each functional unit are only for easy differentiation and are not used to limit the protection scope of the embodiments of the present invention.

[0100] This application also provides a room violation review device, which can be integrated with the room violation review apparatus provided in this application. Figure 6 This is a schematic diagram of a room violation verification device provided in an embodiment of this application. (Reference) Figure 6 The room violation review device includes: an input device 63, an output device 64, a memory 62, and one or more processors 61; the memory 62 is used to store one or more programs; when one or more programs are executed by one or more processors 61, the one or more processors 61 implement the room violation review method provided in the above embodiments. The room violation review device, equipment, and computer provided above can be used to execute the room violation review method provided in any of the above embodiments, and have corresponding functions and beneficial effects.

[0101] This application also provides a non-volatile storage medium storing computer-executable instructions, which, when executed by a computer processor, are used to perform the room violation review method provided in the above embodiments. Of course, the computer-executable instructions provided in this application are not limited to the room violation review method provided above; they can also perform related operations in the room violation review method provided in any embodiment of this application. The room violation review device, equipment, and storage medium provided in the above embodiments can execute the room violation review method provided in any embodiment of this application. Technical details not described in detail in the above embodiments can be found in the room violation review method provided in any embodiment of this application.

[0102] Based on the above embodiments, this application also provides a computer program product. The technical solution of this application, in essence or in other words, the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. The computer program product is stored in a storage medium and includes several instructions to cause a computer device, mobile terminal, or processor therein to execute all or part of the steps of the room violation review method provided in the various embodiments of this application.

Claims

1. A method of room violation review, the method comprising: The method comprises the following steps: determining online users in each room to be audited; determining online blacklist users in each of the rooms to be audited from the online users based on a blacklist list, and a user weight of each of the online blacklist users; determining a room suspicion score of each of the rooms to be audited based on a number of the online users and the user weight of each of the online blacklist users; determining a suspicious room from the rooms to be audited based on the room suspicion score and a first threshold value set; obtaining room audio information of the suspicious room, and determining suspicious audio from the room audio information; performing a rule violation audit on the suspicious room based on the suspicious audio.

2. The room violation review method of claim 1, wherein, The room suspicion score is determined based on a ratio of a sum of the user weights of each of the online blacklist users in the room to be audited to the number of the online users.

3. The room violation review method of claim 1, wherein, The step of obtaining the room audio information of the suspicious room comprises the following steps: obtaining user audio information of each of the online users in the suspicious room; sorting the user audio information to obtain the room audio information.

4. The room violation review method of claim 1, wherein, The step of determining the suspicious audio from the room audio information comprises the following steps: determining an audio suspicion score of the room audio information; determining the suspicious audio from the room audio information based on the audio suspicion score and a second threshold value set.

5. The room violation review method of claim 4, wherein, The step of determining the audio suspicion score of the room audio information comprises the following steps: obtaining the room audio information and user information corresponding to online users of the suspicious room; inputting the room audio information and the user information into a trained semantic model, and determining the audio suspicion score of the room audio information based on the room audio information and the user information by using the semantic model.

6. The room violation review method according to any one of claims 1-5, wherein, Before the step of determining the online blacklist users in the room to be audited based on the blacklist list, the method further comprises the following steps: determining candidate users entering a rule violation room, and determining a frequency of appearance of the candidate users entering the rule violation room; determining a blacklist user based on the frequency of appearance and a third threshold value set, and determining a blacklist list based on the blacklist user.

7. A room violation auditing apparatus, characterized by, The method comprises a room determining module, an audio determining module and a room auditing module, wherein: the room determining module is configured to determine online users in each room to be audited; determine online blacklist users in each of the rooms to be audited from the online users based on a blacklist list, and a user weight of each of the online blacklist users; determine a room suspicion score of each of the rooms to be audited based on a number of the online users and the user weight of each of the online blacklist users; and determine a suspicious room from the rooms to be audited based on the room suspicion score and a first threshold value set; the audio determining module is configured to obtain room audio information of the suspicious room, and determine suspicious audio from the room audio information; the room auditing module is configured to perform a rule violation audit on the suspicious room based on the suspicious audio.

8. A room violation auditing apparatus, characterized by, The method comprises the following steps: a memory and one or more processors; the memory is used to store one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement a room violation review method as claimed in any of claims 1-6.

9. A non-transitory storage medium storing computer-executable instructions, the computer-executable instructions comprising instructions for: The computer executable instructions, when executed by a computer processor, perform a room violation review method as claimed in any of claims 1-6.

10. A computer program product comprising a computer program, characterized in that, The computer program, when executed by a processor, implements a room violation review method as claimed in any of claims 1-6.

Images