Quantity refreshing risk identification method, apparatus and device, and storage medium

Abstract

The embodiment of the invention relates to an amount refreshing risk identification method and device, equipment and a storage medium, and the method comprises the steps: obtaining interactive text information and corresponding video information; determining interactive text features based on the interactive text information, and determining video information features based on the video information; determining the similarity between the interactive text feature and the video information feature, and taking the similarity as a similarity feature; and inputting the interactive text features, the video information features and the similarity features into an amount refreshing identification model to obtain a measurement probability corresponding to the amount refreshing risk of the interactive text release request, and determining the amount refreshing risk of the interactive text release request based on the measurement probability. According to the technical scheme provided by the embodiment of the invention, the accuracy of identifying the interactive text quantity refreshing risk can be improved, so that the interactive text quantity refreshing behavior can be accurately identified, and the network ecological environment can be better maintained.

Description

Technical Field

[0001] The embodiments of the present application relate to the field of intelligent analysis technology, and in particular, to a method, device, equipment and storage medium for identifying brush volume risks. Background Art

[0002] With the development of the Internet, there are illegal industries that may be hired by third parties for economic interests to achieve revenue through the inflated volume of interactive texts. Or in order to increase the popularity of certain content to attract advertisers, they create false traffic and popularity on the Internet and manipulate the perception of viewers to achieve certain commercial purposes. The inflated volume of interactive texts in illegal industries often uses automated tools, scripts or robots to generate large amounts of interactive texts in batches without human intervention. Content platforms and social media operators usually need to carefully analyze the inflated volume of interactive texts to identify which are real fan interactions and which may be malicious inflated volume behaviors.

[0003] In order to effectively prevent the inflated volume of interactive texts, major content platforms and social media operators analyze interactive texts to identify whether there is a risk of inflated volume of interactive texts. However, existing methods for identifying the risk of inflated volume of interactive texts do not cover certain inflated volume behaviors of interactive texts, and the identification is not very accurate. Summary of the invention

[0004] The embodiments of the present application provide a method, apparatus, device and storage medium for identifying the risk of fake volume in interactive texts, which can improve the accuracy of identifying the risk of fake volume in interactive texts, thereby accurately identifying the behavior of fake volume in interactive texts and better maintaining the network ecological environment.

[0005] In a first aspect, the present application provides a method for identifying fraud risk, including: Obtain interactive text information and corresponding video information; Determining interactive text features based on the interactive text information, and determining video information features based on the video information; Determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The interactive text features, the video information features and the similarity features are input into a brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and the brush volume risk of the interactive text publishing request is determined based on the measurement probability.

[0006] In a second aspect, an embodiment of the present application provides a device for identifying risk of volume manipulation, including: An acquisition module, used to acquire interactive text information and corresponding video information; A first determination module, configured to determine interactive text features based on the interactive text information, and to determine video information features based on the video information; A second determination module is used to determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The risk determination module is used to input the interactive text features, the video information features and the similarity features into the brush volume recognition model to obtain the measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0007] In a third aspect, an embodiment of the present application provides an electronic device, including: Memory for storing computer programs; The processor is used to execute the computer program stored in the memory, and when the computer program is executed, the method provided in the embodiment of the present application is implemented.

[0008] In a fourth aspect, an embodiment of the present application provides a computer-readable storage medium having a computer program stored thereon, and when the computer program is executed by a processor, the method provided in the embodiment of the present application is implemented.

[0009] Compared with the prior art, the above technical solution provided by the embodiment of the present application has the following advantages: the technical solution provided by the embodiment of the present application obtains interactive text information and corresponding video information, determines interactive text features through interactive text information and determines video information features through video information, inputs interactive text features, video information features and similarity features of the two into a brush volume recognition model, obtains the measured probability of brush volume risk of the corresponding interactive text publishing request, and determines the brush volume risk of the request based on the measured probability; That is, the present application converts interactive text information into interactive text features, converts video information into video information features, and measures the correlation between interactive text information and drama information by the similarity between the two, thereby determining the relevance of the interactive text information, and obtains the measurement probability of the brush volume risk by inputting the interactive text features, video information features, and the similarity features for measuring the correlation into the brush volume identification model. The brush volume risk is determined by the measurement probability, which can improve the accuracy of identifying the brush volume risk, accurately identify interactive text brush volume behavior, and better maintain the network ecological environment. For the interactive text situation in which the illegal industries in the prior art send low-quality content, the similarity features used to measure the correlation between the interactive text information and the video information are used as a judgment factor for the brush volume risk, so that the existing brush volume risk can be accurately identified, thereby identifying the brush volume behavior of the illegal industries. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate embodiments consistent with the present application and, together with the description, serve to explain the principles of the present application.

[0011] In order to more clearly illustrate the embodiments of the present application or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, for ordinary technicians in this field, other drawings can be obtained based on these drawings without paying any creative labor.

[0012] One or more embodiments are exemplarily described by pictures in the corresponding drawings, and these exemplified descriptions do not constitute limitations on the embodiments. Elements with the same reference numerals in the drawings represent similar elements, and unless otherwise stated, the figures in the drawings do not constitute proportional limitations.

[0013] Figure 1 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 2 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 3 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 4 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 5 A flow chart for determining the brushing risk of the interactive text publishing request based on the measured probability; Figure 6 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 7 A flow chart of a method for identifying the risk of volume manipulation provided in an embodiment of the present application; Figure 8 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present application. DETAILED DESCRIPTION

[0014] In order to make the purpose, technical solution and advantages of the embodiments of the present application clearer, the technical solution in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of this application.

[0015] The disclosure below provides many different embodiments or examples to realize the different structures of the present application. In order to simplify the disclosure of the present application, the parts and settings of specific examples are described below. Of course, they are only examples, and the purpose is not to limit the present application. In addition, the present application can repeat reference numbers and / or letters in different examples. This repetition is for the purpose of simplification and clarity, and does not itself indicate the relationship between the various embodiments and / or settings discussed.

[0016] In order to solve the technical problem that certain interactive text brushing behaviors cannot be covered and identified inaccurately in the prior art, the present application provides a brushing risk identification method, device, equipment and storage medium, which can improve the accuracy of identifying interactive text brushing risks, thereby accurately identifying interactive text brushing behaviors and better maintaining the network ecological environment.

[0017] Figure 1 A flow chart of a method for identifying brush volume risks is provided in an embodiment of the present application. The method can be executed by a brush volume risk identification device. The device can be implemented by software and / or hardware. The device can be configured in a video sharing platform or a live broadcast platform. The video sharing platform or the live broadcast platform is configured in an electronic device. The method can be applied to the scenario of identifying the brush volume risks of interactive texts.

[0018] like Figure 1 As shown, the technical solution provided in the embodiment of the present application includes the following steps: S110: Obtain interactive text information and corresponding video information.

[0019] In this embodiment, the video information may include episode information, introduction information of the recorded video, or theme information of the live video. Therefore, the technical solution provided in the embodiment of the present application can be applied to the interactive scene of the episode, or can also be applied to the interactive scene of the recorded video, or can also be applied to the interactive scene of the live broadcast. Among them, the episode information includes the introduction information of a TV series, including a one-sentence recommendation, a list of actors, a list of roles, a subject classification, an emotional classification, etc. The introduction information of the recorded video may include character information, location information, theme information, time information, etc. in the recorded video. The theme information of the live video may include entertainment, facts, fashion, sports, health, etc.

[0020] In this embodiment, if applied in an interactive scene of a drama, the interactive text information may be barrage information; if applied in an interactive scene of a recorded video, the interactive text information may be comment information; if applied in an interactive scene of a live broadcast, it may be comment information in the live broadcast room.

[0021] S120: Determine interactive text features based on the interactive text information, and determine video information features based on the video information.

[0022] In an embodiment of the present application, the method for determining interactive text features based on interactive text information may be to convert the interactive text information into a structured form, for example, to obtain an interactive text feature vector through a large language model, thereby obtaining interactive text features; or to use all words in the interactive text information to express the interactive text feature vector, thereby obtaining interactive text features; or it may be to extract a preset number of feature words from the interactive text information, and to represent the interactive text information using an interactive text feature vector through the preset number of feature words; wherein the preset number may be pre-set according to the weight sorting result of the feature words; or it may be other reasonable methods.

[0023] Similarly, the implementation method of determining the features of video information based on video information is the same as the implementation method of obtaining interactive text features. Specifically, the video information can be converted into a structured form, for example, a video information feature vector is obtained through a large language model to obtain video information features; or all words in the video information are used to express the video information feature vector to obtain video information features, or a set number of feature words can be extracted from the video information, and the video information is represented by a video information feature vector through a set number of feature words, wherein the set number can be pre-set according to the weight sorting result of the feature words. Among them, the weight of the feature words is determined based on the number of times the feature words appear. Or other reasonable methods can also be used.

[0024] S130: Determine the similarity between the interactive text feature and the video information feature, and use it as a similarity feature.

[0025] In an embodiment of the present application, cosine similarity can be used to calculate the similarity between interactive text features and video information features. Cosine similarity is an indicator that measures the similarity between two vectors in a direction, and is often used to calculate the similarity between text words or other types of data in fields such as natural language processing and information retrieval. Therefore, the cosine similarity between the interactive text feature vector and the video information feature vector can be calculated, and the similarity between the interactive text features and the video information features can be determined by the cosine similarity, and the cosine similarity can be used as a similarity feature. Among them, the correlation between the interactive text information and the video information can be measured by the similarity between the interactive text features and the video information features, and the relevance of the interactive text information can be judged.

[0026] S140: Input the interactive text feature, the video information feature and the similarity feature into a brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0027] Since the technical solution provided by the embodiment of the present application is applied to interactive scenarios, the characteristics of interactive scenarios are that some interactive texts appearing in specific target videos are normal content, but appearing in other videos may be potential brushing behaviors. Therefore, in interactive scenarios, the correlation between interactive text and video content is inseparable. Therefore, the correlation between interactive text and video information is measured by similarity to determine whether there is a brushing risk. Based on this, the present application inputs interactive text features, video information features, and similarity features into a trained brushing recognition model to obtain a measured probability of the brushing risk corresponding to the interactive text publishing request, and determines the brushing risk of the request through the measured probability, thereby determining whether it is a brushing behavior of an illegal industry. Among them, the interactive text publishing request contains interactive text information, account information corresponding to the interactive text information, video information, etc.

[0028] In an embodiment of the present application, before the brush volume recognition model is used, it may also include training the brush volume recognition model. The specific method is: obtain a training sample set, perform metric probability annotation on the training samples in the training sample set to obtain a labeled sample set; use the labeled sample set to train the brush volume recognition model to obtain a trained brush volume recognition model. Among them, the training samples in the training sample set may include training interactive sample features, training video information features, and similarity features of the two. Among them, the training samples can be annotated with 1 or 0 for metric probability, wherein a metric probability of 1 indicates that it is a corresponding historical high-risk interactive text publishing request or an interactive text sending request for a brush volume behavior that is not covered in history; 0 indicates a corresponding risk-free interactive text publishing request. Among them, the brush volume recognition model can be a binary classification model.

[0029] In this embodiment, the brushing risk of the interactive text release request can be determined based on the measurement probability. Specifically, the brushing risk of the interactive text release request can be determined according to the degree of the measurement probability. The higher the measurement probability, the greater the brushing risk of the interactive text release request. Alternatively, the brushing risk of the interactive text release request can be determined according to the measurement probability and the log information of the user operation data. Specifically, the risk score of the interactive text release request can be determined according to the measurement probability and the log information of the user operation data, and the brushing risk of the interactive text release request can be determined according to the risk score. Among them, the number of times the account user sends the interactive text release request, the number of times the account is switched on the same device, whether the network environment is abnormal, etc. can also affect the determination of the brushing risk. According to this information and the measurement probability, the risk score of the interactive text release request is determined, wherein a query table of risk score-brush risk can be pre-set, and the corresponding risk score can be found by querying, and the corresponding brushing risk can be queried. Or it can be based on the measurement probability and user operation. Or the measurement probability and the log information of the user operation data can be input into the risk control model to achieve it.

[0030] Among the related technologies, Natural Language Processing (NLP) technology is used to analyze interactive text information in interactive scenarios, and text analysis technology is used to identify spam, sensitive words and recurring patterns. If the interactive text information sent by the illegal industry is not low-quality content, the risk of interactive text brushing is not accurately identified, and it is impossible to identify the brushing behavior of the illegal industry.

[0031] The technical solution provided by the embodiment of the present application obtains interactive text information and corresponding video information, determines interactive text features through the interactive text information and determines video information features through the video information, inputs the interactive text features, video information features and similarity features of the two into a brush volume recognition model, obtains a measured probability of brush volume risk of the corresponding interactive text publishing request, and determines the brush volume risk level of the request based on the measured probability; That is, the present application converts interactive text information into interactive text features, converts video information into video information features, and measures the correlation between interactive text information and drama information by the similarity between the two, thereby determining the relevance of the interactive text information, and obtains the measurement probability of the brush volume risk by inputting the interactive text features, video information features, and the similarity features for measuring the correlation into the brush volume identification model. The brush volume risk is determined by the measurement probability, which can improve the accuracy of identifying the brush volume risk, accurately identify the interactive text brush volume behavior, and better maintain the network ecological environment. For the interactive text of low-quality content sent by the illegal industries in the prior art, the similarity features used to measure the correlation between the interactive text information and the video information are used as a judgment factor for the brush volume risk, so that the existing brush volume risk can be accurately identified, thereby identifying the brush volume behavior of the illegal industries.

[0032] Figure 2 This is a flow chart of a method for identifying the risk of volume manipulation provided by an embodiment of the present application. In this embodiment, the step of "obtaining the log information of user operation data and determining the target basic characteristics based on the log information of the user operation data" is added, and the step of determining the measurement probability is limited. Figure 2 As shown, the method comprises the following steps: S210: Obtain log information of user operation data, interactive text information and corresponding video information.

[0033] In this embodiment, the log information of user operation data records the time information, account information, device information, IP address information, mobile phone number information, network environment information, etc. of all users issuing interactive text requests.

[0034] S220: Determine target basic features based on the log information of the user operation data, determine interactive text features based on the interactive text information, and determine video information features based on the video information.

[0035] In one implementation of the embodiments of the present application, optionally, user attribute features of the interactive text scene are obtained from the log information of the user operation data as the target basic features; wherein the user attribute features include counting features, conflict features and time-distance features. Among them, counting features can be the number of interactive text messages posted by account users within a preset historical time, the number of mobile phone numbers, the number of accounts or the number of videos posted by a device in history, etc. Conflict features can be the inconsistency between the city of the IP address and the location of the mobile phone number, the conflict between the device source and other parameters, etc.; time-distance features can be the time difference between two interactive text publishing requests, etc. Among them, the target basic features can also include other features extracted from the log information of the user operation data. Among them, the target basic features can be N-dimensional.

[0036] S230: Determine the similarity between the interactive text feature and the video information feature, and use it as a similarity feature.

[0037] S240: Input the target basic features, the interactive text features, the video information features and the similarity features into a brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0038] In this embodiment, the target basic features obtained from the log information of the user operation data are used as one of the input items of the brush volume identification model, and some basic features in the user operation data are used as considerations for obtaining the brush volume risk of the interactive text publishing request. This can more accurately identify the brush volume risk and thus accurately identify the brush volume behavior.

[0039] In this embodiment, target basic features are added as input items of the brush volume recognition model. Before the brush volume recognition model is used, the brush volume recognition model can also be trained. Among them, the training samples in the training sample set include training interactive text features, training video information features, and similarity features of the two, as well as training target basic features. By measuring the probability of the training sample set, a labeled sample set is obtained, and the brush volume recognition model is trained using the labeled sample set to obtain a trained brush volume recognition model.

[0040] In the related art, the behavior of robots and batch operations is often identified based on methods such as the frequency of sending interactive texts and the length of time of viewing. However, if the illegal industry has a large number of accounts and sends interactive texts to multiple accounts at a low frequency to perform text brushing, it can bypass the existing brushing risk detection strategy, and the brushing risk cannot be accurately identified, and the above-mentioned brushing behavior cannot be covered. The embodiment of the present application increases the acquisition of target basic features from user operation data, specifically including counting features, conflict features, and time features, and uses the target basic features as one of the input items of the brushing identification model, thereby using some basic features in the user operation data (counting features, conflict features, and time features, etc.) as considerations for obtaining the brushing risk of interactive text publishing requests. This can cover the illegal industry's behavior of brushing text by sending interactive texts to multiple accounts at a low frequency, improve the recognition accuracy of brushing risks, and accurately identify brushing behaviors.

[0041] Figure 3 This is a flow chart of a method for identifying the risk of volume manipulation provided by an embodiment of the present application. Figure 1 Based on the above, the steps of determining interactive text features, determining video information features and determining similarity features are limited, such as Figure 3 As shown, the method comprises the following steps: S310: Obtain interactive text information and corresponding video information.

[0042] S320: Input the interactive text information into a large language model to obtain an interactive text feature vector, and input the video information into a large language model to obtain a video information feature vector.

[0043] In this embodiment, the large language model can be an open source model bge-large-zh-v1.5, or other models can also be used. The interactive text information and video information can be converted into embedding vectors through the large language model; wherein the interactive text feature vector can be a 255-dimensional vector, and the video information feature vector can be a 255-dimensional vector. Optionally, the large language model for interactive text information input and the large language model for video information input can be the same model, or they can be different models.

[0044] S330: Determine the cosine similarity between the interactive text feature vector and the video information feature vector, and use the cosine similarity as the similarity between the interactive text feature and the video information feature, and as a similarity feature.

[0045] In this embodiment, the cosine similarity is defined as follows: (1); in, is the dot product of vector A and vector B. ||A|| and ||B|| are the Euclidean norms of vector A and vector B, respectively. The interactive text feature vector can be taken as vector A and the video information feature vector as vector B. The cosine similarity between the interactive text feature vector and the video information feature vector can be calculated by the above formula (1); in theory, the closer the directions of the two vectors are, the higher their cosine similarity is. A cosine similarity value of 1 indicates that the two vectors are exactly the same, a cosine similarity value of 0 indicates that the two are orthogonal, and a cosine similarity of -1 indicates that the two are completely opposite. Among them, the greater the cosine similarity between the interactive text feature vector and the video information feature vector, the stronger the correlation between the interactive text information and the video information.

[0046] S340: Input the interactive text features in the interactive text feature vector, the interactive text features in the video information feature vector, and the similarity features into the brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0047] In this embodiment, the interactive text information and video information are processed by a large language model to obtain an interactive text feature vector and a video information feature vector, and then the cosine similarity of the two vectors is calculated, wherein the large language model usually has a better understanding of natural language and can capture complex language patterns and contextual relationships. For the large language model, the interactive text information can be used to pre-train the large language model and fine-tune the parameters of the large language model. If the interactive text information and the video information are processed by different large language models, the video information can also be used to pre-train the large language model and fine-tune the model parameters. Thus, by pre-training the large language model, the model can be made more suitable for the field of application of the embodiment of the present application. In addition, the professional knowledge base of the interactive text scene can be summarized and precipitated and provided to the large language model, so that it can process the professional terms of the interactive text more efficiently and accurately. Alternatively, the large language model can be simplified or otherwise reduced to reduce the resource consumption caused by the large language model.

[0048] Figure 4 This is a flow chart of a method for identifying the risk of volume manipulation provided by an embodiment of the present application. Figure 2 Based on the above, the steps of determining interactive text features, determining video information features and determining similarity features are limited, such as Figure 4 As shown, the method comprises the following steps: S410: Obtain log information of user operation data, interactive text information and corresponding video information.

[0049] S420: Determine target basic features based on the log information of the user operation data, input the interactive text information into a large language model to obtain an interactive text feature vector, and input the video information into a large language model to obtain a video information feature vector.

[0050] S430: Determine the cosine similarity between the interactive text feature vector and the video information feature vector, and use the cosine similarity as the similarity between the interactive text feature and the video information feature, and as a similarity feature.

[0051] S440: Input the target basic features, the interactive text features in the interactive text feature vector, the interactive text features in the video information feature vector, and the similarity features into the brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0052] Among them, the introduction of S410-S420 can refer to the introduction of the above embodiment and will not be repeated here.

[0053] exist Figure 3 and Figure 4On the basis of the illustrated embodiment, before the brush volume recognition model is used, the interactive text feature vector and the video information feature vector may be subjected to dimensionality reduction processing respectively. Specifically, principal component analysis (PCA) may be used to perform dimensionality reduction processing on the interactive text feature vector and the video information feature vector respectively. For example, the interactive text feature vector may be reduced from a 255-dimensional vector to a 64-dimensional vector, and the video information feature vector may be reduced from a 255-dimensional vector to a 64-dimensional vector. By reducing the dimensionality of the interactive text feature vector and the video information feature vector, the consumption of computing resources may be reduced when performing calculations using the brush volume recognition model.

[0054] On the basis of the above embodiment, before determining the risk level of the interactive text publishing request based on the measured probability, it may also include: determining the segmentation threshold of the brush volume recognition model based on the performance indicators of historical data. Among them, the performance indicators include recall rate, precision rate, hit rate, etc. For example, taking precision rate as an example, the historical data is simulated through the brush volume recognition model to obtain various measured probabilities, and the corresponding segmentation threshold is determined according to the business's requirements for high risk, medium risk and no risk for precision rate. Specifically, Figure 5 is a flow chart for determining the brushing risk of the interactive text publishing request based on the measured probability, such as Figure 5 As shown, the specific steps include: S510: If the measured probability is greater than or equal to the first segmentation threshold, the risk of brushing volume of the interactive text publishing request is determined to be high risk.

[0055] S520: If the measured probability is less than the first segmentation threshold and greater than or equal to the second segmentation threshold, it is determined that the brushing risk of the interactive text publishing request is medium risk.

[0056] S530: If the measured probability is less than the second segmentation threshold, it is determined that the risk of brushing the interactive text publishing request is risk-free.

[0057] S510-S530 are introduced uniformly. The first segmentation threshold can be greater than the second segmentation threshold. The first segmentation threshold Ψ1 can be 0.9, and the second segmentation threshold Ψ2 can be 0.7, or the first segmentation threshold and the second segmentation threshold can be other values, which can be adjusted according to business needs. If the brushing risk of the interactive text publishing request is high risk, it can be identified as a brushing behavior. If the brushing risk of the interactive text publishing request is medium risk, it can be verified by other means to identify whether it is a brushing behavior. If the brushing risk of the interactive text publishing request is no risk, it can be identified as a non-brushing behavior.

[0058] Based on the above embodiment, in an optional implementation manner, optionally, determining the brushing risk of the interactive text publishing request based on the measured probability includes: The brushing risk of the interactive text publishing request is determined based on the measured probability and the log information of the user operation data. Specifically, the measured probability and the log information of the user operation data can be input into the risk control model, and the brushing risk of the corresponding interactive text publishing request is output. Among them, the risk control model can be pre-trained to obtain a trained risk control model. Among them, taking into account the suspiciousness of the user's operation behavior, the brushing risk can be obtained by comprehensively considering the log information in the user's operation data and the measured probability of the brushing risk. Therefore, by comprehensively considering the log information in the user's operation data and the measured probability of the brushing risk to obtain the brushing risk, the recognition accuracy of the brushing risk can be improved, and by using the risk control model to obtain the brushing risk, the calculation speed can be improved, and the recognition accuracy can also be improved.

[0059] On the basis of the above embodiments, after determining the risk of brushing of the interactive text publishing request based on the measured probability, the technical solution provided in the embodiments of the present application may further include: if the risk of brushing of the interactive text publishing request is judged to be high risk, the corresponding interactive text is deleted, and the interactive text is not displayed on the front end; if the risk of brushing of the interactive text publishing request is judged to be medium risk, a secondary verification method is provided; if the user passes the secondary verification, the interactive text publishing request is approved, and the corresponding interactive text is stored in the interactive text library, and the interactive text is displayed on the front end; if the risk of brushing of the interactive text publishing request is judged to be no risk, the interactive text publishing request is approved, and the corresponding interactive text is stored in the interactive text library, and the interactive text is displayed on the front end.

[0060] Specifically, if the risk of brushing for the interactive text publishing request is high, the corresponding interactive text will be deleted and not displayed on the front end. This can help the platform maintain the network ecological environment and protect the interests of users. If the risk of brushing for the interactive text publishing request is medium, a secondary verification method is provided. The specific verification method can be sliding, character selection and other verification methods. If the secondary verification passes, it is considered that it is not a brushing behavior, and the interactive text can be stored and displayed on the front end to achieve interaction; if the secondary verification fails, it is considered that there is suspicion of brushing behavior, so the interactive text will be deleted and not displayed on the front end. If the risk of brushing for the interactive text publishing request is low, it is considered that it is not a brushing behavior, and the interactive text will be stored and displayed on the front end, and the statistics of parameter video popularity and other series will be calculated. Among them, when the video information is episode information, the technical solution provided by this application can also refer to Figure 6 .

[0061] In the prior art, risk detection of interactive text information can be performed by combining risk control strategy judgment and manual review. However, most cases that cannot be determined by risk control strategies are assigned to the manual review process, which greatly increases the pressure of manual review. The technical solution provided in the embodiment of the present application, when applied before manual review, can greatly reduce the pressure of manual review and reduce labor costs. In order to ensure the authenticity and quality of interactive texts, prevent illegal industries from affecting the popularity ranking of videos by brushing, or maintain a good order in the video and live broadcast environment, the present application provides the above-mentioned technical solution. The technical solution provided by the embodiment of the present application can help video sharing platforms or live broadcast platforms maintain the network ecological environment, protect the interests of users, and provide a fairer environment for video content producers, maintain the healthy development of the platform and the good experience of users.

[0062] Figure 7 is a schematic diagram of a structure of a device for identifying risk of volume manipulation provided in an embodiment of the present application, such as Figure 6 As shown, the device includes: an acquisition module 710, a first determination module 720, a second determination module 730 and a risk determination module 740.

[0063] Wherein, the acquisition module 710 is used to acquire interactive text information and corresponding video information; A first determination module 720, configured to determine interactive text features based on the interactive text information, and determine video information features based on the video information; A second determination module 730 is used to determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The risk determination module 740 is used to input the interactive text features, the video information features and the similarity features into the brush volume identification model, obtain the measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

[0064] In a possible embodiment, the device further includes a target basic feature determination module, which is used to obtain log information of user operation data and determine the target basic feature based on the log information of the user operation data; Accordingly, the risk determination module 740 is specifically configured to: The target basic features, the interactive text features, the video information features and the similarity features are input into a brush volume recognition model to obtain a measured probability of brush volume risk of a corresponding interactive text publishing request.

[0065] In a possible embodiment, determining the interactive text feature based on the interactive text information includes: inputting the interactive text information into a large language model to obtain an interactive text feature vector; Accordingly, determining the video information feature based on the video information includes: inputting the video information into a large language model to obtain a video information feature vector; Correspondingly, the determining the similarity between the interactive text feature and the video information feature includes: determining the cosine similarity between the interactive text feature vector and the video information feature vector, and using the cosine similarity as the similarity between the interactive text feature and the video information feature; Accordingly, the step of inputting the interactive text feature, the video information feature, and the similarity feature into a brush volume recognition model includes: The interactive text features in the interactive text feature vector, the interactive text features in the video information feature vector, and the similarity features are input into the brush volume recognition model.

[0066] In a possible embodiment, determining the target basic features based on the log information of the user operation data includes: The user attribute features of the interactive text scene are obtained from the log information of the user operation data as the target basic features; wherein the user attribute features include counting features, conflict features and time interval features.

[0067] In a possible embodiment, the device further includes a segmentation threshold determination module, which is used to determine the segmentation threshold of the brush volume identification model based on the performance indicator of the historical data; Accordingly, determining the brushing risk of the interactive text publishing request based on the measured probability includes: If the measured probability is greater than or equal to the first segmentation threshold, the risk of brushing the interactive text publishing request is determined to be high risk; If the measured probability is less than the first segmentation threshold and greater than or equal to the second segmentation threshold, it is determined that the brushing risk of the interactive text publishing request is medium risk; If the measured probability is less than the second segmentation threshold, it is determined that the risk of brushing the interactive text publishing request is risk-free.

[0068] In a possible embodiment, determining the brushing risk of the interactive text publishing request based on the measured probability includes: The risk of brushing volume of the interactive text publishing request is determined based on the measurement probability and the log information of the user operation data.

[0069] In a possible embodiment, the device further includes a risk processing module, which is used to: If the risk of inflating the volume of the interactive text publishing request is determined to be high, the corresponding interactive text is deleted, and the interactive text is not displayed on the front end; If the risk of brushing of the interactive text publishing request is determined to be medium risk, a secondary verification method is provided; if the user passes the secondary verification, the interactive text publishing request is approved, and the corresponding interactive text is stored in the interactive text library, and the interactive text is displayed on the front end; If the risk of brushing the interactive text publishing request is determined to be risk-free, the interactive text publishing request is approved, and the corresponding interactive text is stored in the interactive text library and displayed on the front end.

[0070] In a possible embodiment, the device further includes a dimensionality reduction processing module, which is used to: The interactive text feature vector and the video information feature vector are respectively subjected to dimensionality reduction processing.

[0071] In a possible embodiment, the device further includes a dimensionality reduction processing module, which is used to: Obtaining a training sample set, and annotating the training samples in the training sample set by metric probability to obtain an annotated sample set; The labeled sample set is used to train the brush volume recognition model to obtain a trained brush volume recognition model.

[0072] In one possible embodiment, the video information includes episode information, introduction information of a recorded video, or theme information of a live video.

[0073] like Figure 8 As shown, an embodiment of the present application provides an electronic device, including a processor 111, a communication interface 112, a memory 113 and a communication bus 114, wherein the processor 111, the communication interface 112, and the memory 113 communicate with each other through the communication bus 114. Memory 113, used for storing computer programs; In one embodiment of the present application, the processor 111 is used to execute the program stored in the memory 113 to implement the method for identifying the risk of brushing volume provided by any of the above method embodiments, including: Obtain interactive text information and corresponding video information; Determining interactive text features based on the interactive text information, and determining video information features based on the video information; Determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The interactive text features, the video information features and the similarity features are input into a brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and the brush volume risk of the interactive text publishing request is determined based on the measurement probability.

[0074] An embodiment of the present application also provides a computer-readable storage medium having a computer program stored thereon. When the computer program is executed by a processor, the steps of the method for identifying the risk of brushing volume provided in any of the aforementioned method embodiments are implemented.

[0075] The device embodiments described above are merely illustrative, wherein the units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0076] Through the description of the above implementation methods, those skilled in the art can clearly understand that each implementation method can be implemented by means of software plus a general hardware platform, and of course, by hardware. Based on this understanding, the above technical solution, in essence, or the part that contributes to the relevant technology, can be embodied in the form of a software product, which can be stored in a computer-readable storage medium, such as ROM / RAM, a disk, an optical disk, etc., including a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in each embodiment or some parts of the embodiment.

[0077] It should be understood that the terms used herein are only for the purpose of describing specific example embodiments and are not intended to be limiting. Unless the context clearly indicates otherwise, the singular forms "one", "an" and "said" as used herein may also be meant to include plural forms. The terms "include", "comprise", "contain", and "have" are inclusive, and therefore specify the existence of stated features, steps, operations, elements and / or parts, but do not exclude the existence or addition of one or more other features, steps, operations, elements, parts, and / or combinations thereof. The method steps, processes, and operations described herein are not interpreted as necessarily requiring them to be performed in the specific order described or illustrated, unless the execution order is clearly indicated. It should also be understood that additional or alternative steps may be used.

[0078] The above description is only a specific implementation of the present application, so that those skilled in the art can understand or implement the present application. Various modifications to these embodiments will be apparent to those skilled in the art, and the general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application will not be limited to the embodiments shown herein, but will conform to the widest range consistent with the principles and novel features applied for herein.

Claims

1. A method for identifying the risk of volume manipulation, characterized in that: include: Obtain interactive text information and corresponding video information; Determining interactive text features based on the interactive text information, and determining video information features based on the video information; Determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The interactive text features, the video information features and the similarity features are input into a brush volume recognition model to obtain a measurement probability of the brush volume risk of the corresponding interactive text publishing request, and the brush volume risk of the interactive text publishing request is determined based on the measurement probability.

2. The method according to claim 1, characterized in that: Also includes: Obtaining log information of user operation data, and determining target basic features based on the log information of the user operation data; Accordingly, the interactive text feature, the video information feature and the similarity feature are input into the brush volume identification model to obtain the measured probability of the brush volume risk of the corresponding interactive text publishing request, including: The target basic features, the interactive text features, the video information features and the similarity features are input into a brush volume recognition model to obtain a measured probability of brush volume risk of a corresponding interactive text publishing request.

3. The method according to claim 1, characterized in that The determining of interactive text features based on the interactive text information includes: inputting the interactive text information into a large language model to obtain an interactive text feature vector; Accordingly, determining the video information feature based on the video information includes: inputting the video information into a large language model to obtain a video information feature vector; Correspondingly, the determining the similarity between the interactive text feature and the video information feature includes: determining the cosine similarity between the interactive text feature vector and the video information feature vector, and using the cosine similarity as the similarity between the interactive text feature and the video information feature; Accordingly, the step of inputting the interactive text feature, the video information feature, and the similarity feature into a brush volume recognition model includes: The interactive text features in the interactive text feature vector, the interactive text features in the video information feature vector, and the similarity features are input into the brush volume recognition model.

4. The method according to claim 2, characterized in that: The determining of target basic features based on the log information of the user operation data includes: The user attribute features of the interactive text scene are obtained from the log information of the user operation data as the target basic features; wherein the user attribute features include counting features, conflict features and time interval features.

5. The method according to claim 1 or 2, characterized in that: Also includes: The performance indicator based on historical data determines the segmentation threshold of the brush volume identification model; Accordingly, determining the brushing risk of the interactive text publishing request based on the measured probability includes: If the measured probability is greater than or equal to the first segmentation threshold, the risk of brushing the interactive text publishing request is determined to be high risk; If the measured probability is less than the first segmentation threshold and greater than or equal to the second segmentation threshold, it is determined that the brushing risk of the interactive text publishing request is medium risk; If the measured probability is less than the second segmentation threshold, it is determined that the risk of brushing the interactive text publishing request is risk-free.

6. The method according to claim 2, characterized in that The determining the brushing risk of the interactive text publishing request based on the measured probability includes: The risk of brushing volume of the interactive text publishing request is determined based on the measurement probability and the log information of the user operation data.

7. The method according to claim 1 or 2, characterized in that: Also includes: If the risk of inflating the volume of the interactive text publishing request is determined to be high, the corresponding interactive text is deleted, and the interactive text is not displayed on the front end; If the risk of fraud in the interactive text publishing request is determined to be medium risk, a secondary verification method is provided; If the user passes the secondary verification, the interactive text publishing request is approved, the corresponding interactive text is stored in the interactive text library, and the interactive text is displayed on the front end; If the risk of brushing the interactive text publishing request is determined to be risk-free, the interactive text publishing request is approved, and the corresponding interactive text is stored in the interactive text library and displayed on the front end.

8. The method according to claim 3, characterized in that Also includes: The interactive text feature vector and the video information feature vector are respectively subjected to dimensionality reduction processing.

9. The method according to claim 1, characterized in that: Also includes: Obtaining a training sample set, and annotating the training samples in the training sample set by metric probability to obtain an annotated sample set; The labeled sample set is used to train the brush volume recognition model to obtain a trained brush volume recognition model.

10. The method according to claim 1, characterized in that The video information includes episode information, introduction information of recorded videos or theme information of live videos.

11. A device for identifying the risk of fraudulent transactions, characterized in that: include: An acquisition module, used to acquire interactive text information and corresponding video information; A first determination module, configured to determine interactive text features based on the interactive text information, and to determine video information features based on the video information; A second determination module is used to determine the similarity between the interactive text feature and the video information feature and use it as a similarity feature; The risk determination module is used to input the interactive text features, the video information features and the similarity features into the brush volume recognition model to obtain the measurement probability of the brush volume risk of the corresponding interactive text publishing request, and determine the brush volume risk of the interactive text publishing request based on the measurement probability.

12. An electronic device, characterized in that: include: Memory for storing computer programs; A processor, configured to execute a computer program stored in the memory, and when the computer program is executed, implement the method described in any one of claims 1 to 10.

13. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed by a processor, the method described in any one of claims 1 to 10 is implemented.

Images