Picture-based viewpoint analysis monitoring method and device, equipment and medium

Abstract

The invention relates to a picture-based viewpoint analysis monitoring method, device and equipment and a medium, and relates to the field of viewpoint monitoring, and the method comprises the steps: obtaining a picture work in a first preset time period, and obtaining work data; determining a target picture work based on the work data; performing character recognition and feature recognition on each target picture work to obtain text information and a first label of an object; determining a second label of the text information, and judging whether a suspicious picture work exists or not based on the first label and the second label; if yes, acquiring a propagation path of the suspicious picture works, and determining a target node from the propagation path; obtaining to-be-selected picture works in a second preset time period, and determining a work similarity based on the first label, the second label, each to-be-selected picture work and the suspicious picture works; and determining a viewpoint score of the suspicious picture work based on the work similarity and the target node. The method has the effects of improving picture viewpoint monitoring and timely controlling viewpoints.

Description

Technical Field

[0001] The present application relates to the field of opinion monitoring, and in particular to a method, device, equipment and medium for image-based opinion analysis and monitoring. Background Art

[0002] On social media platforms, people can publish their own works in a variety of forms, such as text, pictures or videos. The rapid development and popularization of the Internet has greatly increased the speed and scope of information dissemination. Therefore, the information released by the platform needs to be reviewed and monitored to avoid the spread of illegal or bad information and cause adverse effects on society.

[0003] Video works are not easy to edit, and banned words are easy to monitor in text works. However, picture works, including text and pictures, are easy to edit and publish. As there are more and more picture works, the amount of information in them is also increasing. When there is suspicious content in picture works that affects the spread of opinions, the development of the Internet today causes picture works to spread quickly, making it difficult to control them in a timely manner. Summary of the invention

[0004] In order to timely control the viewpoints of a picture work, the present application provides a method, device, equipment and medium for analyzing and monitoring viewpoints based on a picture.

[0005] In a first aspect, the present application provides a method for analyzing and monitoring opinions based on images, which adopts the following technical solutions: A method for monitoring opinion analysis based on a picture, comprising: Acquire image works within a first preset time period, and acquire work data of each image work within the first preset time period, wherein the work data includes the number of likes, the number of reposts, and the number of comments; Determine a target image work based on the work data; Performing text recognition on each target image work to obtain text information of each target image work, and performing feature recognition on each target image work to obtain a first label of an object in each target image work; Determine a second tag of the text information, and determine whether there is a suspicious picture work based on the first tag and the second tag, wherein the suspicious picture work is a picture work containing banned words in the first tag and / or the second tag; If there is a suspicious image work, the propagation path of the suspicious image work is obtained, and a target node is determined from the propagation path, and the user attention amount at the target node reaches a preset threshold; Acquire the image works to be selected within a second preset time period, and determine the work similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work; The opinion score of the suspicious image work is determined based on the work similarity and the target node.

[0006] By adopting the above technical scheme, picture works within a first preset time period are obtained, and work data of each picture work within the first preset time period is obtained, including the number of likes, the number of reposts and the number of comments. The number of likes represents the user's recognition and liking of the picture work, the number of reposts represents the user's evaluation and recognition of the picture work, and the number of comments represents the user's communication needs. Therefore, the work data can reflect people's attention to the picture work. The target picture work is determined based on the work data, and the target picture work with high comprehensive attention in likes, reposts and comments is selected. Text recognition is performed on each target picture work to obtain text information of each target picture work. Feature recognition is performed on each target picture work to obtain a first label of the object in each target picture work, and a second label of the text information is determined. The text image contained in the picture work can convey information or express emotions. The text image can reflect the information direction of the picture. Based on the first label and the second label, it is judged whether there is a suspicious picture work with banned words in the first label and / or the second label. If there is a suspicious picture work, Then, the propagation path of the suspicious image work is obtained, which can reflect the exposure degree of the suspicious image work, and the target node is determined from the propagation path. The attention amount of the user at the target node reaches the preset threshold. When there is a user with high attention amount forwarding the suspicious image work, it means that the suspicious image work will be read by more users. If the suspicious image work has opinions, rapid diffusion will cause adverse social impact. Therefore, some users will steal the suspicious image work for secondary creation and re-publish it. Therefore, the candidate image works within the second preset time period are obtained, and the similarity between each candidate image work and the suspicious image work is determined based on the first label, the second label, each candidate image work and the suspicious image work. The proportion of the suspicious image work that has been re-created and published can be determined from the candidate image works. The opinion score of the suspicious image work is determined based on the work similarity and the target node. The opinion score can reflect the propagation range and influence of the suspicious image work. The opinion score is determined to improve the image opinion monitoring ability, so that the opinion can be controlled in time at the first time when it grows.

[0007] In another possible implementation manner, determining the second tag of the text information includes: Segmenting the text information according to word classes to obtain segmentation results, wherein two or more consecutive characters that do not belong to the word classes are rearranged and combined before segmentation; Extracting keywords from the word segmentation results, wherein the keywords are nouns, verbs, and adjectives; A second label of the text information is determined based on the quantity of each keyword.

[0008] In another possible implementation, the determining the work similarity between each candidate image work and the suspicious image work based on the first label, the second label, each candidate image work and the suspicious image work includes: Determine a first candidate label and a second candidate label for each candidate image work; Determine a first ratio of a first number of identical values ​​between the first label and the first label to be selected to the number of the first label to be selected, a second ratio of a second number of identical values ​​between the second label and the second label to be selected to the number of the second label to be selected, and a third similarity between each of the image works to be selected and the suspicious image work; The work similarity between each of the candidate picture works and the suspicious picture work is determined based on the first ratio, the second ratio, the third similarity and the respective first coefficients.

[0009] In another possible implementation, determining the opinion score of the suspicious image work based on the work similarity and the target node includes: Determine the candidate image works whose similarity to the works reaches a preset similarity threshold, and count the third number thereof; Obtain the total number of views of the suspicious image work at the target node, the total number of attentions of users at the target node, and the number of target nodes; The opinion score of the suspicious image work is determined based on the third quantity, the total number of views, the total number of attentions and the respective target coefficients.

[0010] In another possible implementation, the method further includes: Performing image recognition on the image work to determine whether it contains a preset object; If included, determining a first area ratio of the preset object in the picture work; Acquire a second area of ​​the preset object within a preset range of the image, and determine a second area ratio of the second area in the preset range; Determine a second score for the picture work based on the first area ratio, the second area ratio, and respective second coefficients; A target picture work is determined based on the second score.

[0011] In another possible implementation, the method further includes: When the word segmentation result includes a banned word, but the keyword does not include a banned word, it is determined that both the banned word and the keyword are second labels.

[0012] In another possible implementation, the method further includes: If the opinion score exceeds a preset opinion score threshold, the suspicious image work and the candidate image works whose work similarity reaches a preset similarity threshold are hidden, and prompt information is output to their respective corresponding users.

[0013] In a second aspect, the present application provides a viewpoint analysis monitoring device based on images, which adopts the following technical solution: A viewpoint analysis monitoring device based on a picture, comprising: A first acquisition module, used to acquire pictures within a first preset time period, and acquire work data of each picture work within the first preset time period, wherein the work data includes the number of likes, the number of reposts, and the number of comments; A first determination module, used to determine a target image work based on the work data; A recognition module is used to perform text recognition on each target image work to obtain text information of each target image work, perform feature recognition on each target image work to obtain a first label of an object in each target image work; a judgment module, configured to determine a second tag of the text information, and judge whether there is a suspicious picture work based on the first tag and the second tag, wherein the suspicious picture work is a picture work containing a banned word in the first tag and / or the second tag; The second acquisition module is used to acquire the propagation path of the suspicious image work if there is any suspicious image work, and determine the target node from the propagation path, and the user's attention amount at the target node reaches a preset threshold; A second determination module is used to obtain the image works to be selected within a second preset time period, and determine the similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work; The third determination module is used to determine the opinion score of the suspicious image work based on the work similarity and the target node.

[0014] By adopting the above technical solution, the first acquisition module obtains the picture works within the first preset time period, and obtains the work data of each picture work within the first preset time period, including the number of likes, the number of reposts and the number of comments. The number of likes represents the user's recognition and liking of the picture work, the number of reposts represents the user's evaluation and recognition of the picture work, and the number of comments represents the user's communication needs. Therefore, the work data can reflect people's attention to the picture works. The first determination module determines the target picture works based on the work data, and selects the target picture works with high comprehensive attention in likes, reposts and comments. The recognition module performs text recognition on each target picture work to obtain the text information of each target picture work, performs feature recognition on each target picture work, and obtains the first label of the object in each target picture work. The judgment module determines the second label of the text information. The text image contained in the picture work can convey information or express emotions. The text image can reflect the information direction of the picture, and based on the first label and the second label, it is judged whether there is a suspicious picture work with banned words in the first label and / or the second label. If there is a suspicious picture work , then the second acquisition module obtains the propagation path of the suspicious image work, which can reflect the exposure degree of the suspicious image work, and determines the target node from the propagation path. The attention of the user at the target node reaches the preset threshold. When there are users with high attention forwarding the suspicious image work, it means that the suspicious image work will be read by more users. If the suspicious image work has opinions, rapid spread will cause adverse social impact. Therefore, there will be some users who steal the suspicious image work for secondary creation and re-publish it. Therefore, the second determination module obtains the image works to be selected within the second preset time period, and determines the similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work. The proportion of the suspicious image work that has been re-created and published can be determined from the image works to be selected. The third determination module determines the opinion score of the suspicious image work based on the work similarity and the target node. The opinion score can reflect the dissemination range and influence of the suspicious image work. The determination of the opinion score is to improve the image opinion monitoring ability, so that the opinion can be controlled in time at the first time when it grows.

[0015] In another possible implementation, when determining the second tag of the text information, the judgment module is specifically configured to: Segmenting the text information according to word classes to obtain segmentation results, wherein two or more consecutive characters that do not belong to the word classes are rearranged and combined before segmentation; Extracting keywords from the word segmentation results, wherein the keywords are nouns, verbs, and adjectives; A second label of the text information is determined based on the quantity of each keyword.

[0016] In another possible implementation, when the second determination module determines the work similarity between each to-be-selected picture work and the suspicious picture work based on the first label, the second label, each to-be-selected picture work and the suspicious picture work, it is specifically used to: Determine a first candidate label and a second candidate label for each candidate image work; Determine a first ratio of a first number of identical values ​​between the first label and the first label to be selected to the number of the first label to be selected, a second ratio of a second number of identical values ​​between the second label and the second label to be selected to the number of the second label to be selected, and a third similarity between each of the image works to be selected and the suspicious image work; The work similarity between each of the candidate picture works and the suspicious picture work is determined based on the first ratio, the second ratio, the third similarity and the respective first coefficients.

[0017] In another possible implementation, when the third determination module determines the opinion score of the suspicious image work based on the work similarity and the target node, it is specifically used to: Determine the candidate image works whose similarity to the works reaches a preset similarity threshold, and count the third number thereof; Obtain the total number of views of the suspicious image work at the target node, the total number of attentions of users at the target node, and the number of target nodes; The opinion score of the suspicious image work is determined based on the third quantity, the total number of views, the total number of attentions and the respective target coefficients.

[0018] In another possible implementation, the system further includes: An image recognition module is used to perform image recognition on the image work to determine whether it contains a preset object; A proportion determination module, used for determining a first area proportion of the preset object in the picture work, if included; An area acquisition module, used for acquiring a second area of ​​the preset object within a preset range of the image, and determining a second area ratio of the second area within the preset range; A score determination module, configured to determine a second score of the picture work based on the first area ratio, the second area ratio, and respective second coefficients; The fourth determination module is used to determine the target image work based on the second score.

[0019] In another possible implementation, the system further includes: The fifth determination module is used to determine that both the banned word and the keyword are second labels when the word segmentation result includes a banned word but the keyword does not include a banned word.

[0020] In another possible implementation, the system further includes: The output module is used to hide the suspicious image works and the candidate image works whose similarity reaches the preset similarity threshold if the opinion score exceeds the preset opinion score threshold, and output prompt information to their respective corresponding users.

[0021] In a third aspect, the present application provides an electronic device, which adopts the following technical solution: An electronic device, comprising: at least one processor; Memory; At least one application, wherein at least one application is stored in a memory and configured to be executed by at least one processor, and at least one is configured to: execute a picture-based opinion analysis monitoring method shown in any possible implementation of the first aspect.

[0022] In a fourth aspect, the present application provides a computer-readable storage medium, which adopts the following technical solution: A computer-readable storage medium, when the computer program is executed in a computer, enables the computer to execute the image-based opinion analysis and monitoring method described in any one of the first aspects.

[0023] In summary, the present application includes at least one of the following beneficial technical effects: Acquire pictures within a first preset time period, and acquire work data of each picture work within the first preset time period, including the number of likes, the number of reposts, and the number of comments. The number of likes represents the user's recognition and liking of the picture work, the number of reposts represents the user's evaluation and recognition of the picture work, and the number of comments represents the user's communication needs. Therefore, the work data can reflect people's attention to the picture work. Determine the target picture work based on the work data, select the target picture work with a high comprehensive attention of likes, reposts, and comments, perform text recognition on each target picture work, obtain the text information of each target picture work, perform feature recognition on each target picture work, obtain the first label of the object in each target picture work, determine the second label of the text information, the text image contained in the picture work can convey information or express emotions, and the text image can reflect the information direction of the picture, and judge whether there is a suspicious picture work with banned words in the first label and / or the second label based on the first label and the second label. If there is a suspicious picture work, obtain the suspicious The propagation path of the picture work can reflect the exposure degree of the suspicious picture work, and determine the target node from the propagation path. The attention of the user at the target node reaches the preset threshold. When there is a user with high attention forwarding the suspicious picture work, it means that the suspicious picture work will be read by more users. If the suspicious picture work has opinions, rapid spread will cause adverse social impact. Therefore, some users will steal the suspicious picture work for secondary creation and re-publish it. Therefore, the candidate picture works within the second preset time period are obtained, and the similarity between each candidate picture work and the suspicious picture work is determined based on the first label, the second label, each candidate picture work and the suspicious picture work. The proportion of the suspicious picture works that have been re-created and published can be determined from the candidate picture works. The opinion score of the suspicious picture work is determined based on the work similarity and the target node. The opinion score can reflect the dissemination range and influence of the suspicious picture work. The determination of the opinion score is to improve the image opinion monitoring ability, so that the opinion can be controlled in time at the first time when it grows. BRIEF DESCRIPTION OF THE DRAWINGS

[0024] Figure 1 It is a flow chart of a method for opinion analysis and monitoring based on images according to an embodiment of the present application.

[0025] Figure 2 It is a structural schematic diagram of a picture-based opinion analysis monitoring device according to an embodiment of the present application.

[0026] Figure 3 It is a structural schematic diagram of an electronic device according to an embodiment of the present application. DETAILED DESCRIPTION

[0027] The present application is further described in detail below in conjunction with the accompanying drawings.

[0028] After reading this specification, those skilled in the art may make non-creative modifications to this embodiment as needed, but such modifications are protected by patent law as long as they are within the scope of the claims of this application.

[0029] 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 creative work are within the scope of protection of this application.

[0030] In addition, the term "and / or" in this article is only a description of the association relationship of associated objects, indicating that there can be three relationships. For example, A and / or B can represent: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " in this article, unless otherwise specified, generally means that the associated objects before and after are in an "or" relationship.

[0031] The embodiments of the present application are further described in detail below in conjunction with the drawings in the specification.

[0032] The embodiment of the present application provides a method for analyzing and monitoring opinions based on images, which is executed by an electronic device, which can be a server or a terminal device, wherein the server can be an independent physical server, a server cluster or a distributed system composed of multiple physical servers, or a cloud server that provides cloud computing services. The terminal device can be a smart phone, a tablet computer, a laptop computer, a desktop computer, etc., but is not limited thereto. The terminal device and the server can be directly or indirectly connected via wired or wireless communication, which is not limited in the embodiment of the present application. Figure 1 As shown, the method includes step S101, step S102, step S103, step S104, step S105, step S106 and step S107, wherein: S101, obtaining picture works within a first preset time period, and obtaining work data of each picture work within the first preset time period.

[0033] Among them, the work data includes the number of likes, reposts and comments.

[0034] For the embodiment of the present application, the electronic device obtains the picture works published on the major social media platforms within the first preset time period through the Internet or a server, and obtains the work data of each picture work within the first preset time period, including the number of likes, the number of reposts and the number of comments. The likes reflect the user's recognition and love of the picture work, the reposts reflect the user's evaluation and recognition of the picture work, and the comments reflect the user's communication needs, thereby reflecting the attention paid to the picture work.

[0035] S102, determining a target picture work based on the work data.

[0036] For the embodiment of the present application, the work data reflects people's attention to the picture work, which is reflected in the number of likes, reposts and comments. The number of likes represents the popularity of the picture work, the number of reposts represents the influence of the content, and the number of comments can represent the society's emotions and attitudes towards the picture work. Therefore, the electronic device calculates the work score based on the different weights among the three, and selects the picture work whose score reaches the preset work score threshold as the target picture work.

[0037] S103, performing text recognition on each target image work to obtain text information of each target image work, performing feature recognition on each target image work to obtain a first label of an object in each target image work.

[0038] For the embodiment of the present application, the electronic device inputs each target image work into the trained network model for text recognition and feature recognition, and obtains the text information of each target image work and the first label of the object in each target image work. The text image contained in the image work can convey information or express emotions, and the text image can reflect the information direction of the image. The network model can be a convolutional neural network model, a recurrent neural network model, or other network models.

[0039] S104, determining a second tag of the text information, and judging whether there is any suspicious picture work based on the first tag and the second tag.

[0040] Among them, the suspicious picture works are picture works with banned words in the first label and / or the second label.

[0041] For the embodiment of the present application, the second tag in the text information can reflect the core information of the text. The electronic device determines the second tag by extracting keywords from the text information, and judges whether there are suspicious works based on whether there are banned words in the first tag and / or the second tag. The presence of banned words will cause adverse social impacts.

[0042] S105: If there is any suspicious image work, the propagation path of the suspicious image work is obtained, and the target node is determined from the propagation path.

[0043] Among them, the attention amount of users at the target node reaches a preset threshold.

[0044] For the embodiment of the present application, if the electronic device determines that there is a suspicious picture work, it obtains the propagation path of the suspicious picture work, obtains information such as the attention amount and forwarding time of the user who forwards the suspicious picture work through the Internet or a server, and when the attention amount of a user reaches a preset attention amount threshold, the forwarding time of this user is determined as the target node. The determination of the target node represents a turning point in the propagation of the suspicious picture work. After the user with high attention amount forwards it, the exposure of the suspicious picture work will be greatly improved. If the suspicious picture work contains opinions, rapid spread will cause adverse social impacts. The suspicious picture work may be forwarded by multiple users with high attention amount, so there will be multiple target nodes.

[0045] S106, obtaining the image works to be selected within the second preset time period, and determining the work similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work.

[0046] For the embodiment of the present application, after the picture work is published, if high traffic is generated, some users may steal the picture work for secondary creation and publish it again. Therefore, the electronic device obtains the selected picture work within the second preset time period through the Internet or the server. Assuming that the first preset time period is 8:00-12:00, the second preset time period can be set to 10:00-14:00.

[0047] The electronic device obtains the first label and the second label of the candidate image work, and determines the similarity between each candidate image work and the suspicious image work based on the first label, the second label, each candidate image work and the suspicious image work. If the similarity is high, the spread of the suspicious image work may lead to the generation of the candidate image work. Therefore, it also reflects that if there is an opinion in the suspicious image work, the more candidate image works with high similarity are there, the more they are spread, and the greater the adverse impact is.

[0048] S107, determining the opinion score of the suspicious image work based on the work similarity and the target node.

[0049] For the embodiment of the present application, the playback volume of the suspicious picture works and the amount of user attention at the target node are both important factors affecting the dissemination influence of the suspicious picture works. The electronic device determines the opinion score of the suspicious picture works based on the similarity of the works and the target node. The opinion score can reflect the dissemination range and influence of the suspicious picture works. The higher the opinion score, the more disputes the suspicious picture works may cause. The suspicious picture works are screened out through the work data and text image content of the picture works, the dissemination path of the suspicious picture works is determined, and the target node is determined. The similarity of similar picture works in two time periods is compared to determine the similarity of the works. The opinion score of the suspicious picture works is determined based on the similarity and the target node, so as to improve the ability to monitor the opinions of the pictures, so that the opinions can be controlled in time at the first time when they are generated.

[0050] In a possible implementation of the embodiment of the present application, determining the second tag of the text information in step S104 specifically includes step S1041 (not shown in the figure), step S1042 (not shown in the figure), and step S1043 (not shown in the figure), wherein: Step S1041, segmenting the text information according to word classes to obtain segmentation results.

[0051] Among them, two or more consecutive characters that do not belong to word classes are rearranged and combined before word segmentation.

[0052] For the embodiment of the present application, the electronic device segments text information according to word classes such as nouns, verbs or adverbs. In order to circumvent the platform's monitoring, some users change the order of sensitive words and publish them. Therefore, it is necessary to rearrange and combine two or more consecutive words that do not belong to word classes before segmenting them to obtain the segmentation results.

[0053] Step S1042, extracting keywords from the word segmentation results.

[0054] Among them, the keywords are nouns, verbs and adjectives.

[0055] For the embodiment of the present application, the electronic device extracts some frequently appearing keywords from the word segmentation results. The keywords reflect the core content of the text information. The keywords are generally nouns, verbs or adjectives.

[0056] Step S1043: determining a second label of the text information based on the quantity of each keyword.

[0057] For the embodiment of the present application, there may be multiple keywords, which are high-frequency words or high-frequency characters appearing in the text information. The electronic device determines the keywords of the text information and determines the keywords as the second tag of the text information.

[0058] In a possible implementation of the embodiment of the present application, in step S106, the similarity between each candidate image work and the suspicious image work is determined based on the first label, the second label, each candidate image work and the suspicious image work, specifically including step S1061 (not shown in the figure), step S1062 (not shown in the figure) and step S1063 (not shown in the figure), wherein: Step S1061, determining a first candidate label and a second candidate label for each candidate image work.

[0059] For the embodiment of the present application, the electronic device determines the first candidate label and the second candidate label of each candidate image work by performing text recognition and feature recognition on the candidate image work. The first candidate label reflects the content of the object in the candidate image work, and the second candidate label reflects the core content of the text information.

[0060] Step S1062, determine a first ratio of a first number of first labels that are identical to the first candidate label to the number of first candidate labels, a second ratio of a second number of second labels that are identical to the second candidate label to the number of second candidate labels, and a third similarity between each candidate image work and the suspicious image work.

[0061] For the embodiment of the present application, the electronic device compares the first label and the first candidate label to determine the same first quantity, and determines a first ratio between the first quantity and the first candidate label quantity, and compares the second label and the second candidate label to determine the same second quantity, and determines a second ratio between the second quantity and the second candidate label quantity. The larger the first ratio, the greater the number of objects in the candidate picture work that are identical or significantly similar to those in the suspicious picture work, and the larger the second ratio, the more similar the core content of the text information in the candidate picture work and the suspicious picture work.

[0062] The similarity can be calculated by the structural similarity measure (SSIM). It can also be calculated by cosine similarity. The change of words in the text information will seriously affect the similarity between the two. Therefore, the electronic device first removes the text information from the candidate image works and the suspicious image works, and then represents the remaining candidate image works and suspicious image works containing images as vectors. The third similarity between the candidate image works and the suspicious image works is characterized by calculating the cosine distance between the vectors. The similarity can also be calculated by histogram, and the similarity can also be calculated by other methods, which are not limited here.

[0063] Step S1063, determining the work similarity between each candidate image work and the suspicious image work based on the first ratio, the second ratio, the third similarity and the respective first coefficients.

[0064] For the embodiment of the present application, the first ratio and the second ratio represent the degree of similarity between the works in the selected picture works and the suspicious picture works that have the same or most of the content. The third similarity represents the degree of change between the images in the selected picture works and the suspicious picture works. The higher the third similarity, the less adjustment is made to the image in the selected picture works compared with the suspicious picture works. Therefore, the electronic device determines a first score based on the first ratio, the second ratio, the third similarity and the respective first coefficients to determine the work similarity between each selected picture work and the suspicious picture work. The work similarity represents the degree of similarity between the selected picture work and the suspicious picture work. The higher the work similarity, the higher the possibility that the selected picture work is a re-creation of the suspicious picture work, and the higher the similarity between the selected picture work and the suspicious picture work.

[0065] In a possible implementation of the embodiment of the present application, in step S107, the opinion score of the suspicious image work is determined based on the work similarity and the target node, specifically including step S1071 (not shown in the figure), step S1072 (not shown in the figure) and step S1073 (not shown in the figure), wherein: Step S1071, determining the candidate image works whose work similarity reaches a preset similarity threshold, and counting the third number thereof.

[0066] For the embodiment of the present application, the electronic device determines the candidate image works whose similarity reaches a preset similarity threshold, and counts their third number. The more candidate image works with high similarity, the more suspicious image works have been re-created, and the adverse events or problems reflected by the suspicious image works have attracted social attention.

[0067] Step S1072, obtaining the total number of views of the suspicious image works at the target node, the number of attentions of users at the target node, and the number of target nodes.

[0068] For the embodiment of the present application, the electronic device obtains the total playback volume of the suspicious image work at the target node and the total attention volume of the users at the target node through the Internet or the server. The total playback volume represents the degree of attention of the users to the work, and the total attention volume represents the dissemination ability of the work. The more the total playback volume and the total attention volume, the greater the degree of spread of the suspicious image work may be, resulting in adverse social impact.

[0069] Step S1073, determining the opinion score of the suspicious image work based on the third quantity, the total number of views, the total number of attentions, the number of target nodes, and the respective target coefficients.

[0070] For the embodiment of the present application, the larger the third number is, the greater the possibility of bringing adverse effects to the society; the larger the total number of playbacks is, the higher the degree of attention to the work is; the larger the total number of attentions is, the stronger the dissemination ability of the work is; therefore, the electronic device determines the opinion score of the suspicious picture work based on the third number, the total number of playbacks, the total number of attentions, the number of target nodes and their respective target coefficients; assuming that the third number is 150,000, the total number of playbacks is 2 million times, the total number of attentions is 300,000, the number of target nodes is 30, and the target coefficients of the third number, total number of playbacks, total number of attentions and target node number are 0.6, 0.08, 0.5 and 0.8 respectively; therefore, the electronic device calculates the opinion score as 15×0.6+200×0.08+30×0.5+30×0.8=64; and the opinion score ultimately represents the size of the opinion risk of the suspicious picture work.

[0071] In a possible implementation manner of the embodiment of the present application, the method further includes step 1, step 2, step 3, step 4 and step 5, wherein: Step 1: Perform image recognition on the picture to determine whether it contains a preset object.

[0072] For the embodiment of the present application, the electronic device performs image recognition on the picture work, identifies the objects contained in the picture work, and determines whether it contains preset objects, such as banned facial images or other images that may have adverse effects, etc.

[0073] Step 2: If included, determine the first area ratio of the preset object in the picture work.

[0074] For the embodiment of the present application, the presence of a preset object indicates that the picture work may aggravate the negative emotions and dissatisfaction of the society, and the user ignores social responsibility by publishing such a picture work. If the preset object is included, the first area ratio of the preset object in the picture work is determined. The first area ratio represents the range of the preset object in the picture work. The larger the first area ratio, the easier it is for people to pay attention to the preset object when seeing the picture.

[0075] Step three, obtaining a second area of ​​the preset object within a preset range of the image, and determining a second area ratio of the second area within the preset range.

[0076] For the embodiment of the present application, the electronic device obtains the second area of ​​the preset object within the preset range of the picture, and determines the proportion of the second area in the preset range. For example, the picture can be divided into a nine-square grid, and the intersection or the center is more attractive. The preset range can be adjusted according to actual needs. The larger the proportion of the second area, the larger the area of ​​the preset object in the preset range, which is more attractive.

[0077] Step 4: Determine a second score for the image based on the first area ratio, the second area ratio, and their respective second coefficients.

[0078] For the embodiment of the present application, the first area ratio and the second area ratio are both important factors affecting the second score of the picture work. The first area ratio represents the range of the preset object in the picture work, and the second area ratio represents the ratio of the preset object within the preset range. The hidden electronic device determines the second score of the picture work based on the first area ratio, the second area ratio and their respective second coefficients. Assuming that the first area ratio is 0.2 and the second area ratio is 0.6, the second coefficients of the first area ratio and the second area ratio are 0.4 and 0.7 respectively, and therefore the second score is determined to be 0.2×0.4+0.6×0.7=0.5.

[0079] Step 5: Determine the target image work based on the second score.

[0080] For the embodiment of the present application, the electronic device determines the target picture work based on the second score. When the second score reaches the preset score threshold, it means that the preset object in the picture work is sufficiently attractive to people, so the picture work is determined as the target picture work.

[0081] In a possible implementation manner of the embodiment of the present application, the method further includes step 6, wherein step 6 may be performed after step S1043, wherein: Step 6: When the word segmentation result includes a banned word but the keyword does not include a banned word, it is determined that both the banned word and the keyword are the second label.

[0082] For the embodiment of the present application, there may be multiple keywords in the word segmentation results, all of which are high-frequency words, but the word segmentation results may also include banned words. Although banned words appear less frequently, there may be users who deliberately use this situation to weaken the banned words in the eyes of users. The inclusion of banned words in the word segmentation results will also have a certain impact. Therefore, the electronic device needs to determine that the banned words and keywords are both second labels.

[0083] In a possible implementation manner of the embodiment of the present application, the method further includes step seven, wherein step seven may be performed after step S107, wherein: Step 7: If the opinion score exceeds a preset opinion score threshold, the suspicious image works and the candidate image works whose work similarity reaches a preset similarity threshold are hidden, and prompt information is output to their respective corresponding users.

[0084] For the embodiment of the present application, the opinion score reflects the dissemination range and influence of the suspicious image work. If the electronic device determines that the opinion score exceeds the preset opinion score threshold, it means that the dissemination range will be wider and the influence will be greater. In this case, it is necessary to hide the suspicious image work and the candidate image works whose similarity reaches the preset similarity threshold, cut them off from the source, reduce the spread, and the electronic device must also output a prompt message to its respective corresponding users, such as "The work published at XX time is suspected of violating the regulations and has been hidden. Please pay attention to the subsequent works published."

[0085] The above embodiment introduces a picture-based opinion analysis monitoring method from the perspective of method flow, and the following embodiment introduces a picture-based opinion analysis monitoring device from the perspective of a virtual module or a virtual unit. For details, please refer to the following embodiment.

[0086] The present application embodiment provides a viewpoint analysis monitoring device 20 based on a picture, such as Figure 2 As shown, the image-based opinion analysis monitoring device 20 may specifically include: The first acquisition module 201 is used to acquire the picture works within a first preset time period, and acquire the work data of each picture work within the first preset time period, the work data including the number of likes, the number of reposts and the number of comments; A first determination module 202, used to determine a target image work based on the work data; The recognition module 203 is used to perform text recognition on each target image work to obtain text information of each target image work, perform feature recognition on each target image work to obtain a first label of an object in each target image work; The judgment module 204 is used to determine the second label of the text information, and judge whether there is a suspicious picture work based on the first label and the second label, and the suspicious picture work is a picture work with banned words in the first label and / or the second label; The second acquisition module 205 is used to acquire the propagation path of the suspicious image work if there is any suspicious image work, and determine the target node from the propagation path, and the user attention amount at the target node reaches a preset threshold; The second determination module 206 is used to obtain the image works to be selected within the second preset time period, and determine the work similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work; The third determination module 207 is used to determine the opinion score of the suspicious image work based on the work similarity and the target node.

[0087] The embodiment of the present application discloses a device 20 for analyzing opinions based on pictures, wherein a first acquisition module 201 acquires picture works within a first preset time period, and acquires work data of each picture work within the first preset time period, including the number of likes, the number of reposts, and the number of comments. The number of likes represents the user's recognition and liking of the picture work, the number of reposts represents the user's evaluation and recognition of the picture work, and the number of comments represents the user's communication needs. Therefore, the work data can reflect people's attention to the picture work. A first determination module 202 determines the target picture work based on the work data, and selects the target picture work with a high comprehensive attention of likes, reposts, and comments. The recognition module 203 performs text recognition on each target picture work to obtain text information of each target picture work, performs feature recognition on each target picture work, and obtains a first label of an object in each target picture work. The judgment module 204 determines a second label of the text information. The text image contained in the picture work can convey information or express emotions. The text image can reflect the information direction of the picture, and judges whether there is a suspicious picture work with banned words in the first label and / or the second label based on the first label and the second label. If there is a suspicious picture work, the second acquisition module 205 acquires the propagation path of the suspicious picture work, which can reflect the exposure degree of the suspicious picture work, and determines the target node from the propagation path. The attention amount of the user at the target node reaches the preset threshold. When there is a user with high attention amount forwarding the suspicious picture work, it means that the suspicious picture work will be read by more users. If there is an opinion in the suspicious picture work, the rapid spread will cause adverse social impact. Therefore, there will be some users who steal the suspicious picture work for secondary creation and re-publish it. Therefore, the second determination module 206 acquires the picture works to be selected within the second preset time period, and determines the work similarity between each picture work to be selected and the suspicious picture work based on the first label, the second label, each picture work to be selected and the suspicious picture work. The proportion of the suspicious picture work that has been re-created and published can be determined from the picture works to be selected. The third determination module 207 determines the opinion score of the suspicious picture work based on the work similarity and the target node. The opinion score can reflect the dissemination scope and influence of the suspicious picture work. The determination of the opinion score is to improve the image opinion monitoring ability, so that the opinion can be controlled in time at the first time when it grows.

[0088] In a possible implementation of the embodiment of the present application, when determining the second tag of the text information, the judgment module 204 is specifically configured to: Segment the text information according to word classes to obtain segmentation results, wherein two or more consecutive characters that do not belong to word classes are rearranged and combined before segmentation; Extract keywords from the word segmentation results, the keywords are nouns, verbs and adjectives; A second label of the text information is determined based on the quantity of each keyword.

[0089] In a possible implementation of the embodiment of the present application, when the second determination module 206 determines the similarity between each candidate image work and the suspicious image work based on the first label, the second label, each candidate image work and the suspicious image work, it is specifically used to: Determine a first candidate label and a second candidate label for each candidate image work; Determine a first ratio of a first number of first labels that are identical to the first candidate labels to the number of first candidate labels, a second ratio of a second number of second labels that are identical to the second candidate labels to the number of second candidate labels, and a third similarity between each candidate image work and the suspicious image work; The work similarity between each candidate image work and the suspicious image work is determined based on the first ratio, the second ratio, the third similarity and the respective first coefficients.

[0090] In a possible implementation of the embodiment of the present application, when the third determination module 207 determines the opinion score of the suspicious image work based on the work similarity and the target node, it is specifically used to: Determine the candidate image works whose work similarity reaches a preset similarity threshold, and count their third number; Obtain the playback volume of suspicious images at the target node and the attention volume of users at the target node; The opinion score of the suspicious image work is determined based on the third quantity, the number of views, the number of attentions, and the respective target coefficients.

[0091] In a possible implementation of the embodiment of the present application, the system 20 further includes: An image recognition module is used to perform image recognition on a picture work to determine whether it contains a preset object; A proportion determination module, used to determine a first area proportion of the preset object in the picture work if included; An area acquisition module, used to acquire a second area of ​​a preset object within a preset range of the image, and determine a second area ratio of the second area within the preset range; A score determination module, used to determine a second score of the picture work based on the first area ratio, the second area ratio and the respective second coefficients; The fourth determination module is used to determine the target image work based on the second score.

[0092] In a possible implementation of the embodiment of the present application, the system 20 further includes: The fifth determination module is used to determine that both the banned word and the keyword are the second label when the word segmentation result includes the banned word but the keyword does not include the banned word.

[0093] In a possible implementation of the embodiment of the present application, the system 20 further includes: The output module is used to hide the suspicious image works and the candidate image works whose work similarity reaches the preset similarity threshold if the opinion score exceeds the preset opinion score threshold, and output prompt information to their respective corresponding users.

[0094] An electronic device is provided in an embodiment of the present application, such as Figure 3 As shown, Figure 3 The electronic device 30 shown includes: a processor 301 and a memory 303. The processor 301 and the memory 303 are connected, such as through a bus 302. Optionally, the electronic device 30 may also include a transceiver 304. It should be noted that in actual applications, the transceiver 304 is not limited to one, and the structure of the electronic device 30 does not constitute a limitation on the embodiments of the present application.

[0095] Processor 301 may be a CPU (Central Processing Unit), a general-purpose processor, a DSP (Digital Signal Processor), an ASIC (Application Specific Integrated Circuit), an FPGA (Field Programmable Gate Array), or other programmable logic devices, transistor logic devices, hardware components, or any combination thereof. It may implement or execute various exemplary logic blocks, modules, and circuits described in conjunction with the disclosure of this application. Processor 301 may also be a combination that implements computing functions, such as a combination of one or more microprocessors, a combination of a DSP and a microprocessor, and the like.

[0096] The bus 302 may include a path to transmit information between the above components. The bus 302 may be a PCI (Peripheral Component Interconnect) bus or an EISA (Extended Industry Standard Architecture) bus. The bus 302 may be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 3 Only one thick line is used in the diagram, but it does not mean that there is only one bus or only one type of bus.

[0097] The memory 303 may be a ROM (Read Only Memory) or other types of static storage devices that can store static information and instructions, a RAM (Random Access Memory) or other types of dynamic storage devices that can store information and instructions, or an EEPROM (Electrically Erasable Programmable Read Only Memory), a CD-ROM (Compact Disc Read Only Memory) or other optical disk storage, optical disk storage (including compressed optical disk, laser disk, optical disk, digital versatile disk, Blu-ray disk, etc.), a magnetic disk storage medium or other magnetic storage device, or any other medium that can be used to carry or store the desired program code in the form of instructions or data structures and can be accessed by a computer, but is not limited thereto.

[0098] The memory 303 is used to store the application code for executing the solution of the present application, and the execution is controlled by the processor 301. The processor 301 is used to execute the application code stored in the memory 303 to implement the contents shown in the above method embodiment.

[0099] The electronic devices include, but are not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (personal digital assistants), PADs (tablet computers), PMPs (portable multimedia players), vehicle-mounted terminals (such as vehicle-mounted navigation terminals), and fixed terminals such as digital TVs, desktop computers, etc. It can also be a server, etc. Figure 3 The electronic device shown is merely an example and should not bring any limitation to the functions and scope of use of the embodiments of the present application.

[0100] The embodiment of the present application provides a computer-readable storage medium, which stores a computer program. When the computer-readable storage medium is run on a computer, the computer can execute the corresponding content in the above method embodiment. Compared with the related art, the embodiment of the present application obtains the picture works within the first preset time period, and obtains the work data of each picture work within the first preset time period, including the number of likes, the number of reposts and the number of comments. The number of likes represents the user's recognition and liking of the picture work, the number of reposts represents the user's evaluation and recognition of the picture work, and the number of comments represents the user's communication needs. Therefore, the work data can reflect people's attention to the picture work, determine the target picture work based on the work data, select the target picture work with high comprehensive attention of likes, reposts and comments, perform text recognition on each target picture work, obtain the text information of each target picture work, perform feature recognition on each target picture work, obtain the first label of the object in each target picture work, determine the second label of the text information, the text image contained in the picture work can convey information or express emotions, and the text image can reflect the information direction of the picture, and determine whether there is a suspicious picture work with banned words in the first label and / or the second label based on the first label and the second label. If there is a suspicious picture work If there is a suspicious image work, the propagation path of the suspicious image work is obtained, and the propagation path can reflect the exposure degree of the suspicious image work, and the target node is determined from the propagation path. The attention of the user at the target node reaches the preset threshold. When there is a user with high attention forwarding the suspicious image work, it means that the suspicious image work will be read by more users. If the suspicious image work has opinions, rapid diffusion will cause adverse social impact. Therefore, some users will steal the suspicious image work for secondary creation and re-publish it. Therefore, the candidate image works within the second preset time period are obtained, and the similarity between each candidate image work and the suspicious image work is determined based on the first label, the second label, each candidate image work and the suspicious image work. The proportion of the suspicious image work that has been re-created and published can be determined from the candidate image works. The opinion score of the suspicious image work is determined based on the work similarity and the target node. The opinion score can reflect the dissemination range and influence of the suspicious image work. The determination of the opinion score is to improve the image opinion monitoring ability, so that the opinion can be controlled in time at the first time when it grows.

[0101] It should be understood that, although the steps in the flowchart of the accompanying drawings are displayed in sequence as indicated by the arrows, these steps are not necessarily executed in sequence in the order indicated by the arrows. Unless otherwise specified herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least a part of the steps in the flowchart of the accompanying drawings may include multiple sub-steps or multiple stages, and these sub-steps or stages are not necessarily executed at the same time, but can be executed at different times, and their execution order is not necessarily sequential, but can be executed in turn or alternately with other steps or at least a part of the sub-steps or stages of other steps.

[0102] The above description is only a partial implementation method of the present application. It should be pointed out that for ordinary technicians in this technical field, several improvements and modifications can be made without departing from the principles of the present application. These improvements and modifications should also be regarded as the scope of protection of the present application.

Claims

1. A method for monitoring opinion analysis based on pictures, characterized in that: include: Acquire image works within a first preset time period, and acquire work data of each image work within the first preset time period, wherein the work data includes the number of likes, the number of reposts, and the number of comments; Determine a target image work based on the work data; Performing text recognition on each target image work to obtain text information of each target image work, and performing feature recognition on each target image work to obtain a first label of an object in each target image work; Determine a second tag of the text information, and determine whether there is a suspicious picture work based on the first tag and the second tag, wherein the suspicious picture work is a picture work containing banned words in the first tag and / or the second tag; If there is a suspicious image work, the propagation path of the suspicious image work is obtained, and a target node is determined from the propagation path, and the user attention amount at the target node reaches a preset threshold; Acquire the image works to be selected within a second preset time period, and determine the work similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work; The opinion score of the suspicious image work is determined based on the work similarity and the target node.

2. The method for image-based opinion analysis and monitoring according to claim 1, characterized in that: The determining the second tag of the text information includes: Segmenting the text information according to word classes to obtain segmentation results, wherein two or more consecutive characters that do not belong to the word classes are rearranged and combined before segmentation; Extracting keywords from the word segmentation results, wherein the keywords are nouns, verbs, and adjectives; A second label of the text information is determined based on the quantity of each keyword.

3. The method for image-based opinion analysis and monitoring according to claim 1, characterized in that: The determining the work similarity between each to-be-selected picture work and the suspicious picture work based on the first label, the second label, each to-be-selected picture work and the suspicious picture work comprises: Determine a first candidate label and a second candidate label for each candidate image work; Determine a first ratio of a first number of identical values ​​between the first label and the first label to be selected to the number of the first label to be selected, a second ratio of a second number of identical values ​​between the second label and the second label to be selected to the number of the second label to be selected, and a third similarity between each of the image works to be selected and the suspicious image work; The work similarity between each of the candidate picture works and the suspicious picture work is determined based on the first ratio, the second ratio, the third similarity and the respective first coefficients.

4. The method for image-based opinion analysis and monitoring according to claim 1, characterized in that: The determining the opinion score of the suspicious image work based on the work similarity and the target node includes: Determine the candidate image works whose similarity to the works reaches a preset similarity threshold, and count the third number thereof; Obtain the total number of views of the suspicious image work at the target node, the total number of attentions of users at the target node, and the number of target nodes; The opinion score of the suspicious image work is determined based on the third quantity, the total playback volume, the total attention volume, the number of target nodes and the respective target coefficients.

5. The method for image-based opinion analysis and monitoring according to claim 1, characterized in that: The method further comprises: Performing image recognition on the image work to determine whether it contains a preset object; If included, determining a first area ratio of the preset object in the picture work; Acquire a second area of ​​the preset object within a preset range of the image, and determine a second area ratio of the second area in the preset range; Determine a second score for the picture work based on the first area ratio, the second area ratio, and respective second coefficients; A target picture work is determined based on the second score.

6. The method for image-based opinion analysis and monitoring according to claim 2, characterized in that: The method further comprises: When the word segmentation result includes a banned word, but the keyword does not include a banned word, it is determined that both the banned word and the keyword are second labels.

7. The method for image-based opinion analysis and monitoring according to claim 1, characterized in that: The method further comprises: If the opinion score exceeds a preset opinion score threshold, the suspicious image work and the candidate image works whose work similarity reaches a preset similarity threshold are hidden, and prompt information is output to their respective corresponding users.

8. A viewpoint analysis monitoring device based on pictures, characterized in that: include: A first acquisition module, used to acquire pictures within a first preset time period, and acquire work data of each picture work within the first preset time period, wherein the work data includes the number of likes, the number of reposts, and the number of comments; A first determination module, used to determine a target image work based on the work data; A recognition module is used to perform text recognition on each target image work to obtain text information of each target image work, perform feature recognition on each target image work to obtain a first label of an object in each target image work; a judgment module, configured to determine a second tag of the text information, and judge whether there is a suspicious picture work based on the first tag and the second tag, wherein the suspicious picture work is a picture work containing a banned word in the first tag and / or the second tag; The second acquisition module is used to acquire the propagation path of the suspicious image work if there is any suspicious image work, and determine the target node from the propagation path, and the user's attention amount at the target node reaches a preset threshold; A second determination module is used to obtain the image works to be selected within a second preset time period, and determine the similarity between each image work to be selected and the suspicious image work based on the first label, the second label, each image work to be selected and the suspicious image work; The third determination module is used to determine the opinion score of the suspicious image work based on the work similarity and the target node.

9. An electronic device, characterized in that: It includes: at least one processor; Memory; At least one application, wherein the at least one application is stored in the memory and configured to be executed by the at least one processor, and the at least one application is used to execute a picture-based opinion analysis monitoring method according to any one of claims 1 to 7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the computer program is executed in a computer, the computer is caused to execute the image-based opinion analysis monitoring method according to any one of claims 1 to 7.

Images