CN116229439A

CN116229439A - Barrage identification method, device, apparatus, storage medium and program product

Info

Publication number: CN116229439A
Application number: CN202111458954.3A
Authority: CN
Inventors: 陈小帅
Original assignee: Tencent Technology Shenzhen Co Ltd
Current assignee: Tencent Technology Shenzhen Co Ltd
Priority date: 2021-12-01
Filing date: 2021-12-01
Publication date: 2023-06-06

Abstract

The application relates to the technical field of video processing, and provides a bullet screen identification method, device, equipment, storage medium and program product, which can improve the identification accuracy of a bullet screen, and comprise the following steps: acquiring a target barrage synchronously displayed with a target video segment; determining a subsequent video clip whose playing order follows the target video clip; the comparison result between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips is used as the relativity of the target barrage and the subsequent video clips on the plot; acquiring a barrage segment formed by a barrage synchronously displayed with a subsequent video segment, and taking a comparison result between semantic features of a target barrage and semantic features of the barrage segment as a semantic relativity of the target barrage and the barrage segment; and identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a dramatic barrage.

Description

Barrage identification method, device, apparatus, storage medium and program product

Technical Field

The present application relates to the field of video processing technology, and in particular, to a bullet screen recognition method, apparatus, computer device, storage medium, and computer program product.

Background

With the rapid development of computer technology and intelligent terminals, videos are deeply favored by users because of the rapid and visual information transmission, and various video applications are also in an infinite hierarchy. In some scenes, a user may post a barrage while watching a video, for example, when user Alice is watching a certain video or a video clip, the barrage may be issued for the video clip, and thereafter, when user Bob is watching the video clip, the barrage posted by user Alice will be presented to user Bob synchronously with the video clip, thereby increasing the interactivity between users.

Because of the openness of the bullet screen content, part of bullet screens published by users belongs to the dramatic bullet screens which disclose the dramas of the subsequent video clips, and the synchronous presentation of the bullet screens and the video clips can seriously influence the watching experience of users who do not watch the subsequent video clips, and is also unfavorable for the interaction between users. However, in the conventional method, the identification of the bullet screen depends only on the content of the bullet screen, and the identification accuracy is low.

Disclosure of Invention

In view of the foregoing, it is desirable to provide a barrage recognition method, apparatus, computer device, storage medium, and computer program product.

A method of barrage identification, the method comprising:

acquiring a target barrage synchronously displayed with a target video segment;

determining a subsequent video clip whose playing order follows the target video clip;

the result of comparison between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips is used as the relativity of the target barrage and the subsequent video clips on the plot;

acquiring a barrage segment formed by a barrage synchronously displayed with the subsequent video segment, and taking a comparison result between the semantic features of the target barrage and the semantic features of the barrage segment as the semantic relativity of the target barrage and the barrage segment;

and identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a transparent barrage.

A bullet screen identification device, the device comprising:

the target bullet screen acquisition module is used for acquiring a target bullet screen synchronously displayed with the target video clips;

the subsequent video segment acquisition module is used for determining the subsequent video segments with the playing sequence behind the target video segment;

The plot correlation obtaining module is used for comparing plot characteristics of the target barrage with plot characteristics of the subsequent video clips to obtain the plot correlation of the target barrage and the subsequent video clips;

the semantic relevance acquisition module is used for acquiring a barrage segment formed by the barrage synchronously displayed with the subsequent video segment, and taking a comparison result between the semantic features of the target barrage and the semantic features of the barrage segment as the semantic relevance of the target barrage and the barrage segment;

and the drama-transparent barrage identification module is used for identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a drama-transparent barrage.

In some embodiments, the drama-transparent bullet screen identification module is further configured to obtain a tendency of the target account on the release of the drama-transparent bullet screen; the target account is an account for issuing the target barrage; based on the correlation degree on the plot, the semantic correlation degree and the positive correlation relation between the trend degree and the bullet screen dramatic permeability probability, obtaining the probability that the target bullet screen is a bullet screen; and based on the probability that the target barrage is the transparent barrage, obtaining the identification result of whether the target barrage is the transparent barrage.

In some embodiments, the drama screen identification module is further configured to obtain a first weight corresponding to the relevance on the episode and the semantic relevance, and a second weight corresponding to the trend; and obtaining the probability that the target barrage is a through barrage according to the first weight, the product result of the relevance on the plot and the semantic relevance and the product result of the second weight and the tendency.

In some embodiments, the drama screen identification module is further configured to determine a release time of the target screen; determining a total number of drama shots published by the target account and having a publication time before a publication time of the target barrage, and determining a number of drama passes published by the target account and having a publication time before the publication time of the target barrage; and taking the ratio of the number of the drama-transparent backdrop to the total number as the tendency of the target account on issuing the drama-transparent backdrop.

In some embodiments, the scenario feature of the target bullet screen is a bullet screen scenario depth representation and the scenario feature of the subsequent video clip is a video clip scenario depth representation; the scenario correlation acquisition module is further used for acquiring a scenario depth representation of the target scenario and acquiring a video clip scenario depth representation of the subsequent video clip; and carrying out interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation, and taking an interactive fusion result as the relativity of the target barrage and the subsequent video clip on the scenario.

In some embodiments, the scenario correlation acquisition module is further configured to:

performing voice recognition on the dialect of the subsequent video segment to obtain a dialect text of the subsequent video segment; constructing video clip scenario depth representation of the subsequent video clip based on the text-to-text;

and/or the number of the groups of groups,

character recognition is carried out on the video frames of the follow-up video clips, and caption text of the follow-up video clips is obtained; and constructing video clip scenario depth representation of the subsequent video clip based on the subtitle text.

In some embodiments, the semantic features of the target barrage are barrage semantic depth representations and the semantic features of the barrage segment are barrage segment semantic depth representations; the semantic relativity acquisition module is also used for acquiring the barrage semantic depth representation of the target barrage and acquiring the barrage segment semantic depth representation of the barrage segment; and carrying out interaction fusion on the barrage semantic depth representation and the barrage segment semantic depth representation, and taking an interaction fusion result as the semantic relativity of the target barrage and the barrage segment.

In some embodiments, the target bullet screen acquisition module is further configured to acquire a plurality of bullet screens displayed in synchronization with the target video clip; respectively carrying out plot detection on each bullet screen in the plurality of bullet screens to obtain plot detection results for representing whether the bullet screen has plots or not; and determining the barrage with the plot from the barrages as a target barrage according to the plot detection result.

In some embodiments, the target barrage acquisition module is further configured to perform word segmentation processing on each barrage to obtain a word segmentation sequence corresponding to each barrage; acquiring the plot detection characteristics corresponding to the barrage according to each word in the word segmentation sequence and the position of each word in the word segmentation sequence; inputting the scenario detection characteristics into a pre-trained scenario detection neural network, and outputting a scenario detection result of whether the bullet screen has the scenario through the scenario detection neural network.

In some embodiments, the scenario correlation acquisition module is further configured to input scenario features of the target bullet screen and scenario features of the subsequent video clip into a pre-trained scenario correlation detection neural network for comparison; and taking the comparison result output by the plot correlation degree detection neural network as the correlation degree of the target barrage and the subsequent video clips on the plot.

In some embodiments, the apparatus further comprises a scenario correlation detection neural network training module for acquiring a plurality of sample barrages presented in synchronization with the sample video clips; determining a first sample barrage and a second sample barrage of the plurality of sample barrages; when the audience acceptance degree corresponding to the first sample barrage is higher than the audience acceptance degree corresponding to the second sample barrage, determining that the correlation degree of the first sample barrage and the sample video clip on the plot is higher than the correlation degree of the second sample barrage and the sample video clip on the plot; training the scenario correlation detection neural network based on the first sample barrage and the sample video clip being higher in scenario correlation than the second sample barrage and the sample video clip, the scenario features of the first sample barrage, the scenario features of the second sample barrage, and the scenario features of the sample video clip.

In some embodiments, the apparatus further comprises a viewer approval determination module for obtaining a praise number for the first sample barrage and a praise number for the second sample barrage; and when the praise number of the first sample barrage is larger than that of the second sample barrage, determining that the audience approval degree corresponding to the first sample barrage is higher than that corresponding to the second sample barrage.

A computer device comprising a memory storing a computer program and a processor executing the bullet screen identification method described above.

A computer-readable storage medium having stored thereon a computer program for execution by a processor of the bullet screen recognition method described above.

A computer program product comprising a computer program which when executed by a processor implements the bullet screen identification method described above.

In the bullet screen identification method, the device, the computer equipment, the storage medium and the computer program product, when the bullet screen identification is required to be carried out on the target bullet screen, after the target bullet screen synchronously displayed with the target video segment and the subsequent video segment with the playing sequence behind the target video segment are obtained, the relativity of the target bullet screen and the subsequent video segment on the plot is obtained based on the comparison between the plot characteristics of the target bullet screen and the plot characteristics of the subsequent video segment; based on the comparison between the semantic features of the target barrage and the semantic features of the barrage segments synchronously displayed with the subsequent video, obtaining the semantic relativity of the target barrage segments synchronously displayed with the subsequent video segments; and then, identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a dramatic barrage. On one hand, based on definition of the drama-transparent barrage, identification of the drama-transparent barrage is carried out not by relying on the target barrage alone but on the relativity between the target barrage and the candidate video clips, on the other hand, the relativity comprises two dimensions of plot relativity and semantic relativity, and internal relation between the target barrage and the candidate video clips is fully considered, so that accuracy of identifying the drama-transparent barrage is greatly improved.

Drawings

FIG. 1 is a diagram of an application environment for a method of bullet screen recognition in one embodiment;

FIG. 2 is a flow chart of a method for identifying a bullet screen according to one embodiment;

FIG. 3 is a flowchart of another embodiment of a method for identifying a bullet screen;

FIG. 4 is a flow chart of obtaining scenario correlation in one embodiment;

FIG. 5 is a flow diagram of obtaining semantic relatedness in some embodiments;

FIG. 6 is a flowchart of detecting whether a bullet screen is provided with a scenario in some embodiments;

FIG. 7 is a flowchart of a method for identifying a bullet screen according to an embodiment;

FIG. 8 is a flow chart of a method for identifying a bullet screen according to one embodiment;

FIG. 9 is a block diagram of an apparatus for identifying a bullet screen in one embodiment;

fig. 10 is an internal structural view of a computer device in one embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application will be further described in detail with reference to the accompanying drawings and examples. It should be understood that the specific embodiments described herein are for purposes of illustration only and are not intended to limit the present application.

Reference in the present application to "an embodiment" means that a particular feature, structure, or characteristic described in connection with the embodiment may be included in at least some embodiments of the present application. The appearances of such phrases in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments. Those of skill in the art will explicitly and implicitly understand that the embodiments described herein may be combined with other embodiments.

The bullet screen identification method provided by the application can be applied to an application environment shown in fig. 1. Wherein the terminal 110 may communicate with the server 120 through a network; the terminal 110 acquires a video from the server 120 through a network and plays the video to the user. The terminal 110 may synchronously display the barrage published for the video while playing the video to the user. Before the terminal 110 synchronously displays the barrage aiming at the video, the server 120 can firstly perform the dynamic barrage identification on the barrage synchronously displayed with the current video segment by using the barrage identification method provided by the application; if the server 120 recognizes that the bullet screen is a transparent bullet screen, the transparent bullet screen may be dynamically processed according to the viewing condition of the user on the subsequent video segment, where the dynamic processing is, for example: if the subsequent video clips corresponding to the leaked scenario of the dynamic transparent bullet screen are watched by the user, the dynamic transparent bullet screen is not shielded, and the dynamic transparent bullet screen is synchronously displayed to the user along with the current video clips; the dynamic process may also be: and if the subsequent video segments corresponding to the scenario revealed by the dynamic bullet screen are not watched by the user, shielding the dynamic bullet screen, and not displaying the dynamic bullet screen to the user along with the current video segments.

In one embodiment, the terminal 110 may also receive a barrage released by the user for the current video clip during the process of playing the video to the user, and feed back the barrage released by the user for the current video clip to the server 120. The server 120 may identify whether the bullet screen issued by the user for the current video clip is a live bullet screen by using the bullet screen identification method provided in the present application.

The terminal 110 may be, but not limited to, various desktop computers, notebook computers, smart phones, tablet computers, internet of things devices and portable wearable devices, and the internet of things devices may be smart speakers, smart televisions, smart air conditioners, smart vehicle devices, etc. The portable wearable device may be a smart watch, smart bracelet, headset, or the like. The terminal 110 may have a client installed thereon, which may be used to play video. The server 120 may be implemented as a stand-alone server or as a server cluster or cloud server composed of a plurality of servers.

In some embodiments, as shown in fig. 2, a barrage identification method is provided, which is illustrated by using the method applied to the computer device (such as the terminal 110 or the server 120) in fig. 1 as an example, and includes the following steps:

Step S201, a target barrage synchronously displayed with the target video clip is obtained.

The target barrage is the barrage to be subjected to the dynamic barrage identification. The target video clip is any video clip that divides the original video into a plurality of video clips. For example, a video clip to be played by the terminal may be taken as a target video clip, and the computer device acquires a target barrage to be displayed synchronously with the target video clip.

A barrage may be understood as a comment containing text content that is presented in synchronization with a video. In the process of watching the video, a user can issue a barrage aiming at the watched video, and the computer equipment records the issued barrage corresponding to the video. The target barrage may be any one of a plurality of barrages presented in synchronization with the target video clip.

In one embodiment, after determining a target video segment from a plurality of video segments obtained by video segmentation, the computer device obtains a plurality of barrages synchronously displayed with the target video segment, and sequentially performs the dynamic transparent barrages identification on the plurality of barrages to obtain an identification result corresponding to each barrage, wherein the barrages processed by the computer device each time are the target barrages.

Step S202, determining a subsequent video clip whose playing order is behind the target video clip.

The playing sequence of the video clips is the sequence of the video clips before and after playing in the original video, and the subsequent video clips are the video clips after the target video clips in the playing sequence. The number of subsequent video segments may be one. For example, the next video clip after the target video clip. The number of subsequent video clips may be multiple. For example, the subsequent video clips may be all video clips in the original video that are played in order after the target video clip. It will be appreciated that the more the target video clip is played in the original video, the fewer the number of subsequent video clips that follow it, the more the target video clip is played in the original video, and the greater the number of subsequent video clips that follow it.

For example, the video clips obtained by video segmentation are a video clip (1), a video clip (2), a video clip (3) and a video clip (4), if the video clip (2) is a target video clip, since the playing sequence of the video clip (3) and the video clip (4) is behind the video clip (2), the computer device can select one video clip from the video clip (3) and the video clip (4) as a subsequent video clip. The computer device may take the video segment (4) as a subsequent video segment, and detect whether the video segment (2) is a theatrical screen for the video segment (4). The computer device may also take the video segment (3) as a subsequent video segment, and detect whether the video segment (2) is a theatrical screen for the video segment (3).

In some embodiments, the subsequent video clip may also be a video clip that follows the target video clip in order of play and is associated with the presence scenario of the target video clip. The scenario described by the target barrage is typically associated with the target video clip, and if the target barrage is a scenario in which a subsequent video clip is a scenario, the target barrage is also associated with the subsequent video clip, in which case the subsequent video clip is also associated with the scenario in which the target video clip exists to some extent. Specifically, the computer device may store in advance the video clips associated with the presence of each video clip on the episode, from which the video clips whose play order is subsequent to the target video are acquired as the subsequent video clips.

In one embodiment, the computer device may further obtain historical play record data corresponding to the current user identifier, determine video clips not watched by the user according to the historical play record data, and screen video clips not watched by the current user from video clips after the target video clip in play order, so as to determine subsequent video clips.

In one embodiment, the computer device may further screen out video clips associated with the scenario in which the target video exists and not viewed by the current user from video clips that follow the target video in the play order as subsequent video clips. Therefore, the computer equipment only needs to compare the target barrage with part of the follow-up video clips, and does not need to compare with all the follow-up video clips, so that the comparison quantity is reduced, and the recognition efficiency is improved.

And step S203, the comparison result between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips is used as the relativity of the target barrage and the subsequent video clips on the plot.

The plot features of the barrage are video plot-related features embodied by the barrage. The scenario feature of the target barrage may be a scenario depth representation corresponding to the text content contained by the target barrage itself.

The plot features of the video clips are video plot-related features embodied by the video clips, and the plot features of the subsequent video clips can be plot depth representations corresponding to video text contents of the subsequent video clips.

In one embodiment, a computer device may obtain a target barrage, extract text content from the target barrage, and obtain corresponding story features based on the text content.

In one embodiment, a computer device may obtain video text content of a subsequent video clip, obtain corresponding episode features based on the video text content. By way of example, video text content may include the speech, subtitles, captions, notes, side notes or scenario introductions of video, and so forth. For example, the computer device may obtain video text content by speech recognition of the speech of the candidate video clip. For another example, the computer device may obtain video text content by optically character recognition of video frames in subsequent video clips.

The relevance of the target barrage and the subsequent video clips on the plot is the relevance of the target barrage and the subsequent video clips on the plot. The greater the relevance of the target barrage to the follow-up video clip on the plot, the more relevant the plot described by the target barrage to the plot described by the follow-up video clip.

Specifically, after obtaining the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips, the computer device compares the similarities between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips, and if the plot characteristics of the target barrage are more similar to the plot characteristics of the subsequent video clips, the relevance of the target barrage and the subsequent video clips on the plot is higher, and the relevance on the plot can be called plot relevance.

Step S204, acquiring a barrage segment formed by the barrage synchronously displayed with the subsequent video segment, and taking the comparison result between the semantic features of the target barrage and the semantic features of the barrage segment as the semantic relativity of the target barrage and the barrage segment.

A barrage segment is a paragraph of text formed from the text content contained by a plurality of barrages. Specifically, the computer device may use, as the bullet screen segment, text content included in a plurality of bullet screens synchronously displayed with the subsequent video segment, where the release time of each bullet screen included in the bullet screen segment may be earlier than the release time of the target bullet screen, or may be later than the release time of the target bullet screen. For example, the computer device may splice text content contained in each bullet screen synchronously displayed with the subsequent video clip to obtain a text paragraph as a bullet screen segment.

Semantic features of the barrage are text-semantically related features embodied by the barrage. The semantic features of the target barrage may be semantic depth representations corresponding to text content contained by the target barrage itself.

Semantic features of the barrage segment are text semantic-related features embodied by the barrage segment. The bullet screen section includes a plurality of bullet screens, each bullet screen including text content. For example, the semantic features of a bullet screen segment may be determined based on the semantic features of each bullet screen included in the bullet screen segment, e.g., the computer device may obtain a semantic depth representation corresponding to the text content of each bullet screen included in the bullet screen segment, and obtain a semantic depth representation corresponding to the bullet screen segment based on the semantic depth representation of each bullet screen. For example, the semantic features of the bullet screen segment may be determined directly based on the text segment of the bullet screen segment itself, e.g., the computer device obtains the bullet screen segment, i.e., the text segment, and directly takes the semantic depth representation corresponding to the entire text segment as the semantic depth representation corresponding to the bullet screen segment.

The semantic relativity of the target barrage and the barrage segment is the semantic relativity of the target barrage and the barrage segment. The larger the correlation degree between the target barrage and the barrage section in terms of semantics, the more similar the semantics of the target barrage and the barrage section.

Specifically, after the computer device obtains the semantic features of the target barrage and the semantic features of the barrage segment of the subsequent video segment, the semantic features of the target barrage and the semantic features of the barrage segment of the subsequent video segment are compared, if the semantic features of the target barrage are more similar to the semantic features of the barrage segment of the subsequent video segment, the higher the semantic relevance between the target barrage and the barrage segment of the subsequent video segment is, the semantic relevance can be called semantic relevance.

Step S205, based on the relativity on the plot and the relativity on the semanteme, the target barrage is identified, and the identification result of whether the target barrage is a drama transparent barrage is obtained.

Wherein, the correlation on the plot has a positive correlation with the probability that the target barrage is a through barrage, i.e. the higher the correlation on the plot, the greater the probability that the target barrage is a through barrage. The semantically relevant degree has a positive correlation with the probability that the target barrage is a through barrage, i.e. the higher the semantically relevant degree, the greater the probability that the target barrage is a through barrage.

In one embodiment, after obtaining the relevance of the target barrage and the subsequent video clip on the plot and the relevance of the target barrage segment of the target barrage and the subsequent video clip on the semantic, the computer device obtains the probability that the target barrage is a through barrage based on the positive correlation, completes the through recognition of the target barrage, and takes the probability that the target barrage is the through barrage as a recognition result. If the probability of the target barrage being the transparent barrage is greater than or equal to the probability threshold, determining that the target barrage is the transparent barrage, and if the probability of the target barrage being the transparent barrage is less than the probability threshold, determining that the target barrage is not the transparent barrage.

In the bullet screen identification method, when the target bullet screen is required to be identified, after the target bullet screen synchronously displayed with the target video segment and the subsequent video segment with the playing sequence behind the target video segment are obtained, the relativity of the target bullet screen and the subsequent video segment on the plot is obtained based on the comparison between the plot characteristics of the target bullet screen and the plot characteristics of the subsequent video segment; based on the comparison between the semantic features of the target barrage and the semantic features of the barrage segments synchronously displayed with the subsequent video, obtaining the semantic relativity of the target barrage segments synchronously displayed with the subsequent video segments; and then, identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a dramatic barrage. On one hand, based on definition of the drama-transparent barrage, identification of the drama-transparent barrage is carried out not by relying on the target barrage alone but on the relativity between the target barrage and the candidate video clips, on the other hand, the relativity comprises two dimensions of plot relativity and semantic relativity, and internal relation between the target barrage and the candidate video clips is fully considered, so that accuracy of identifying the drama-transparent barrage is greatly improved.

In some embodiments, the computer device may determine whether the target bullet screen is a through-the-air bullet screen based on the relevance of the target bullet screen to the subsequent video clip on the episode, the semantic relevance of the bullet screen segments of the target bullet screen to the subsequent video clip, and the propensity of the account that issued the target bullet screen to issue the through-the-air bullet screen.

Specifically, the computer device may obtain a tendency of the target account on the published dynamic bullet screen; the target account is an account for issuing a target barrage; based on the positive correlation between the correlation degree on the plot, the correlation degree on the semantics and the tendency degree and the bullet screen dramatic-transmission probability, the probability that the target bullet screen is the bullet screen is obtained; based on the probability that the target barrage is a through barrage, the identification result of whether the target barrage is the through barrage is obtained.

The tendency of the account on the released backdrop can be obtained by counting the ratio of the number of the history backdrop released by the account to the number of the history backdrop released by the account. The higher the ratio of the number of historical scriptures issued by the account to the number of historical scriptures issued by the account, the higher the propensity of the account to issue scriptures.

The tendency of the target account on the released drama bullet screen is in positive correlation with the probability that the target bullet screen is the drama bullet screen, namely the higher the tendency of the target account on the released drama bullet screen is, the higher the probability that the target bullet screen is the drama bullet screen is.

After the computer equipment obtains the relativity of the target barrage and the subsequent video clips on the plot, the relativity of the target barrage sections of the target barrage and the subsequent video clips on the semantic, and the tendency of the account for issuing the target barrage on the rapid-transmission barrage, the probability that the target barrage is the rapid-transmission barrage can be obtained based on the positive relativity between the relativity on the plot, the relativity on the semantic, the tendency and the barrage rapid-transmission probability, and the probability that the target barrage is the rapid-transmission barrage is taken as a recognition result. If the probability of the target barrage being the transparent barrage is greater than or equal to the probability threshold, determining that the target barrage is the transparent barrage, and if the probability of the target barrage being the transparent barrage is less than the probability threshold, determining that the target barrage is not the transparent barrage.

In this embodiment, whether the target barrage is a through barrage is identified by combining the tendency of the target account on the through barrage, the correlation of the target barrage and the subsequent video segment on the plot, and the semantic correlation of the barrage segment of the target barrage and the subsequent video segment, so as to improve the identification accuracy.

In some embodiments, the step of obtaining the propensity of the target account to issue a theatrical performance screen comprises: determining the release time of a target barrage; determining a total number of the drama shots released by the target account and having release times before release times of the target shots, and determining a number of the drama shots released by the target account and having release times before release times of the target shots; the ratio of the number of the drama-transparent backlashes to the total number is taken as the tendency of the target account on issuing the drama-transparent backlashes.

The tendency of the target account to issue the bullet screen is Pu, and the time of issuing the target bullet screen (i.e. the issuing time of the target bullet screen) is t _n Then the computer device may determine that the target account was published and that the publication time is at t _n The total number of previous barrages, and the target account was published, and the publication time is at t _n The ratio of the number of the previous drama-transparent curtains to the total number of the curtains is taken as the tendency of the target account on the release of the drama-transparent curtains, namely, the tendency pu=the number of the drama-transparent curtains/the total number of the curtains.

In this embodiment, statistical analysis is performed on the backdrop of the historical release of the account releasing the target backdrop, and the ratio of the number of backdrops released by the target account history to the total number of backdrops released by the history is used as the tendency of the target account on the backdrop releasing the backdrop, so as to improve the accuracy of the tendency.

In some embodiments, the step of obtaining the probability that the target barrage is a theatrical barrage based on positive correlations between the plot-wise relevance, the semantic relevance, and the likelihood of each being a barrage, specifically comprises: acquiring a first weight corresponding to the relevance on the plot and the relevance on the semantic, and a second weight corresponding to the tendency; and obtaining the probability that the target barrage is a drama-transparent barrage according to the product result of the first weight, the correlation degree on the plot and the correlation degree on the semanteme and the product result of the second weight and the tendency degree.

The subsequent video segments are marked as s, the probability that the target barrage is a through barrage for the subsequent video segments is marked as Pt [ s ], the relatedness of the target barrage and the subsequent video segments on the plot is marked as Pr [ s ], the relatedness of the target barrage segments synchronously displayed by the target barrage and the subsequent video segments on the plot is marked as Prd [ s ], the first weight corresponding to the relatedness Pr [ s ] on the plot and Prd [ s ] on the plot is marked as w1, the tendency of the target account on the released through barrage is marked as Pu, and the second weight corresponding to the tendency Pu is marked as w2.

Since there is a positive correlation between each of the relevance on the plot, the semantic relevance, and the trend and the barrage dramatic probability, w1>0 and w2>0, and w1+w2=1 (i.e., the sum of the first weight and the second weight is 1). w1 and w2 may be the preferred parameters found by performing the grid search on the constructed evaluation data, and the product result w1×pr [ s ] x Prd [ s ] between the first weight w1, the correlation Pr [ s ] on the plot and the correlation Prd [ s ] on the semantic, and the product result w2×pu between the second weight w2 and the inclination Pu are added, and the obtained added result is taken as the probability that the target barrage is a drama-transparent barrage, that is, pt [ s ] =w1×pr [ s ] ×prd [ s ] +w2×pu.

In this embodiment, according to the result of the product of the first weight, the correlation degree on the plot and the correlation degree on the semantics, and the result of the product of the second weight and the inclination degree, the probability that the target barrage is a drama-transparent barrage is obtained, and the identification accuracy of the drama-transparent barrage is improved.

FIG. 3 is a flow chart illustrating a method of identifying a bullet screen according to one embodiment. The method may be performed by a computer device, and with reference to fig. 3, the method essentially comprises the steps of:

step S301, a target barrage synchronously displayed with a target video clip is obtained;

step S302, determining a subsequent video clip after the target video clip in play order;

step S303, the comparison result between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips is used as the relativity of the target barrage and the subsequent video clips on the plot;

step S304, acquiring a barrage segment formed by a barrage synchronously displayed with a subsequent video segment, and taking a comparison result between semantic features of a target barrage and semantic features of the barrage segment as a semantic relativity of the target barrage and the barrage segment;

step S305, determining the total number of the drama shots released by the target account and the release time before the release time of the target shots, and determining the number of the drama transparent shots released by the target account and the release time before the release time of the target shots;

Step S306, taking the ratio of the number of the drama-transparent backlashes to the total number as the tendency of the target account on the release of the drama-transparent backlashes;

step S307, obtaining a first weight corresponding to the relevance on the plot and the relevance on the semanteme and a second weight corresponding to the tendency;

wherein the sum of the first weight and the second weight is 1;

step S308, obtaining the probability that the target barrage is a drama-transparent barrage according to the product result of the first weight, the relativity on the plot and the relativity on the semanteme and the product result of the second weight and the tendency degree.

Wherein the description of the relevant steps can be found in the embodiments mentioned above.

In the embodiment, whether the target barrage is a transparent barrage is identified by combining the tendency of the target account on the transparent barrage, the correlation of the target barrage and the subsequent video segment on the plot and the semantic correlation of the barrage segment of the target barrage and the subsequent video segment, so that the identification accuracy is improved; the statistics analysis is carried out on the backdrop of the historical release of the account for releasing the target backdrop, and the ratio of the number of the backdrop of the historical release of the target account to the total number of the backdrop of the historical release is used as the tendency of the target account on the backdrop of the release backdrop, so that the accuracy of the tendency is improved; in addition, according to the product result of the first weight, the relativity on the plot and the relativity on the semanteme and the product result of the second weight and the tendency, the probability that the target barrage is the drama-transparent barrage is obtained, and the identification accuracy of the drama-transparent barrage is improved.

In some embodiments, obtaining the relevance of the target barrage to the subsequent video clip on the plot based on a comparison between the plot features of the target barrage and the plot features of the subsequent video clip includes: acquiring a barrage scenario depth representation of a target barrage, and acquiring a video clip scenario depth representation of a subsequent video clip; and carrying out interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation, and taking an interactive fusion result as the relevance of the target barrage and the subsequent video clip on the scenario.

The plot characteristic of the target barrage is barrage plot depth representation, and the plot characteristic of the subsequent video clip is video clip plot depth representation.

Fig. 4 is a schematic flow chart of acquiring scenario correlation in one embodiment, as shown in fig. 4. Referring to fig. 4, the computer device may obtain a barrage scenario depth representation of the target barrage and a video clip scenario depth representation of the subsequent video clip through the BERT layer (Bidirectional Encoder Representation from Transformers, a neural network-based semantic representation network) from the text of the target barrage, the text of the subsequent video clip, and the like, respectively. And then, carrying out interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation through an Attention mechanism layer (Attention), and taking an interactive fusion result as the relevance of the target barrage and the subsequent video clip on the scenario.

In the mode, the accuracy of the relevance of the obtained target barrage and the subsequent video clips on the plot can be ensured by carrying out interactive fusion on barrage plot depth representation of the target barrage and video clip plot depth representation of the subsequent video clips.

In some embodiments, obtaining a video clip episode depth representation of a subsequent video clip may include the steps of: performing voice recognition on the dialect of the subsequent video segment to obtain the dialect text of the subsequent video segment; constructing video clip scenario depth representation of the subsequent video clip based on the text; and/or character recognition is carried out on the video frames of the subsequent video clips, so as to obtain caption texts of the subsequent video clips; based on the subtitle text, a video clip episode depth representation for the subsequent video clip is constructed.

Specifically, after obtaining the subsequent video segment, the computer device may perform automatic speech recognition (ASR, automatic Speech Recognition) on the dialogue in the subsequent video segment to obtain a dialogue text, and input the dialogue text into the BERT layer shown in fig. 4 to obtain a scenario depth representation of the video segment of the subsequent video segment. The computer device may also intercept a video frame (which may be understood as a frame of image) from a subsequent video segment, perform optical character recognition (OCR, optical Character Recognition) on the video frame to obtain subtitle text, and input the subtitle text into the BERT layer shown in fig. 4 to obtain a plot depth representation of the video segment of the subsequent video segment. The computer device may also input the text of the dialog and the text of the subtitle simultaneously into the BERT layer shown in fig. 4 to obtain a video clip scenario depth representation of the subsequent video clip.

In this embodiment, the text of the following video segment is obtained through voice recognition, and the subtitle text of the following video segment is obtained through character recognition, so that the scenario depth representation of the constructed video segment can reflect the scenario features of the following video segment more truly.

In some embodiments, the comparing the plot feature of the target barrage with the plot feature of the subsequent video clip as the relevance of the target barrage and the subsequent video clip on the plot includes: inputting the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips into a pre-trained plot correlation detection neural network for comparison; and taking the comparison result output by the plot correlation detection neural network as the correlation degree of the target barrage and the subsequent video clips on the plot.

The processing flow of the scenario correlation detection neural network may be as shown in fig. 4.

After the computer equipment obtains the text of the target barrage and the text of the dialogue text and the subtitle text of the subsequent video segment, the text of the target barrage and the text of the dialogue text and the subtitle text of the subsequent video segment can be respectively input into the BERT layer of the plot correlation detection neural network to obtain the barrage plot depth representation and the video segment plot depth representation output by the BERT layer. And then, the computer equipment carries out interactive fusion on the barrage plot depth representation and the video clip plot depth representation through the plot correlation detection neural network attention mechanism layer to finish comparison between the barrage plot depth representation and the video clip plot depth representation, and takes an interactive fusion result (namely a comparison result) as the correlation of the target barrage and the subsequent video clip on the plot.

In the embodiment, the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips are compared through the plot correlation detection neural network, so that the accuracy of the correlation of the target barrage and the subsequent video clips on the plot is ensured.

In some embodiments, the scenario correlation detection neural network is constructed by:

acquiring a plurality of sample barrages synchronously displayed with the sample video clips; determining a first sample barrage and a second sample barrage of the plurality of sample barrages; when the audience acceptance degree corresponding to the first sample barrage is higher than that corresponding to the second sample barrage, determining that the correlation degree of the first sample barrage and the sample video fragment on the plot is higher than that of the second sample barrage and the sample video fragment on the plot; the scenario correlation detection neural network is trained based on the first sample barrage and the sample video clip having a higher scenario correlation than the second sample barrage and the sample video clip, the scenario features of the first sample barrage, the scenario features of the second sample barrage, and the scenario features of the sample video clip.

The audience acceptance corresponding to the barrage may be indicative of audience acceptance of the barrage.

Illustrating: and in the plurality of sample barrages synchronously displayed with the sample video clips, the acceptance of the sample barrage a by the audience is higher than that of the sample barrage b by the audience, and the correlation of the sample barrage a and the sample video clips on the plot is higher than that of the sample barrage b and the sample video clips on the plot. The computer device can perform model training according to the condition characteristics of the sample barrage a, the condition characteristics of the sample barrage b and the condition characteristics of the sample video clips based on that the correlation degree of the sample barrage a and the sample video clips is higher than that of the sample barrage b and the sample video clips, so as to obtain the condition correlation degree detection neural network.

In this embodiment, based on the relative level of audience approval obtained by the sample barrage, the relative level of relevance between each sample barrage and the sample video clip on the plot is determined, so as to ensure the prediction accuracy of the constructed plot relevance detection neural network on the plot relevance.

In some embodiments, the relative level between the audience approval corresponding to the first sample barrage and the audience approval corresponding to the second sample barrage may be determined by: acquiring the praise number of the first sample barrage and the praise number of the second sample barrage; when the praise number of the first sample barrage is greater than the praise number of the second sample barrage, determining that the audience acceptance corresponding to the first sample barrage is higher than the audience acceptance corresponding to the second sample barrage.

After obtaining the praise number of the first sample barrage and the praise number of the second sample barrage, the computer device determines that the audience acceptance of the first sample barrage is higher than the audience acceptance of the second sample barrage if the praise number of the first sample barrage is greater than the praise number of the second sample barrage.

In this embodiment, the relative sizes of the praise numbers of the sample barrages determine the relative sizes of audience approval degrees obtained by the sample barrages, so as to ensure the accuracy of the relative sizes of the correlation degrees between the sample barrages and the sample video clips on the plot.

In some embodiments, the comparison result between the semantic features of the target barrage and the semantic features of the barrage segment is used as the relativity of the target barrage and the barrage segment in terms of semantics, and the method may include the following steps: acquiring a barrage semantic depth representation of semantic features representing a target barrage, and acquiring a barrage segment semantic depth representation of semantic features representing barrage segments; and carrying out interactive fusion on the barrage semantic depth representation and the barrage segment semantic depth representation, and taking an interactive fusion result as the semantic relativity of the target barrage and the barrage segment.

The semantic features of the target barrage are barrage semantic depth representations, and the semantic features of the barrage segments are barrage segment semantic depth representations.

FIG. 5 is a flow diagram of obtaining semantic relatedness in one embodiment. Referring to fig. 5, the computer device may obtain, via the BERT layer, a barrage semantic depth representation characterizing the semantic features of the target barrage and a barrage segment semantic depth representation characterizing the semantic features of the barrage segment, respectively, from the text of the target barrage, the text of the barrage segment, and the like. And then, carrying out interaction fusion on the barrage semantic depth representation and the barrage segment semantic depth representation through the attention mechanism layer, and taking an interaction fusion result as the semantic relativity of the target barrage and the barrage segment.

In this embodiment, the barrage semantic depth representation of the target barrage and the barrage segment semantic depth representation of the barrage segment are interactively fused, so as to ensure the accuracy of the obtained target barrage and barrage segment semantic relativity.

In some embodiments, obtaining a target bullet screen presented in synchronization with a target video clip includes: acquiring a plurality of barrages synchronously displayed with a target video clip; respectively carrying out plot detection on each of the plurality of barrages to obtain plot detection results representing whether the barrages have plots or not; and determining the barrage with the plot from the barrages as a target barrage according to the plot detection result.

Specifically, after obtaining a plurality of bullet curtains synchronously displayed with the target video clip, the computer device can respectively perform scenario detection on each bullet curtain to determine whether the bullet curtain has a scenario. If the plot detection result represents that the barrage has plots, the barrage is used as a target barrage to carry out subsequent drama transparent barrage identification.

In this embodiment, before the identification of the bullet screen, the detection of whether the bullet screen has a plot is performed, so as to improve the identification efficiency of the bullet screen.

In some embodiments, the scenario detection is performed on each of the plurality of barrages, which may specifically include the following steps: performing word segmentation processing on each barrage to obtain word segmentation sequences corresponding to each barrage; acquiring plot detection characteristics corresponding to the barrage according to each word in the word segmentation sequence and the position of each word in the word segmentation sequence; the scenario detection characteristics are input into a pre-trained scenario detection neural network, and whether the bullet screen has the scenario detection result is output through the scenario detection neural network.

The method comprises the steps of dividing the barrage to obtain barrage words and positions of the barrage words in the barrage, wherein the barrage words comprise barrage words and the positions of the barrage words in the barrage, and obtaining word division sequences formed by the barrage words according to the positions of the barrage words in the barrage. FIG. 6 is a flow chart of detecting whether a bullet screen has a scenario according to one embodiment. Referring to fig. 6, the positions of the bullet screen words 0, 1, 2, and m in the bullet screen are respectively position 0, position 1, position 2, and position m. The corresponding scenario features of the barrage include the barrage words in the barrage and the respective positions of the barrage words in the barrage (i.e., the position of each word in the word segmentation sequence).

The scenario detection neural network is a deep neural network for detecting whether the bullet screen has a scenario. The scenario detection neural network is obtained by performing model training on pre-constructed scenario-possessing and scenario-free barrage data.

The computer device inputs the bullet screen content into the scenario detection neural network, and the scenario detection neural network outputs the probability that the bullet screen has a scenario as a scenario detection result. When the probability of the bullet screen having the plot is greater than or equal to a preset threshold value, determining that the bullet screen has the plot, and judging that the bullet screen is transparent can be performed on the bullet screen.

In this embodiment, the situation detection neural network is used to obtain the detection result of whether the bullet screen has the situation according to each bullet screen word included in the bullet screen and the position of the bullet screen word in the bullet screen, so as to improve the detection accuracy of whether the bullet screen has the situation.

Fig. 7 is a schematic flow chart of a method for identifying a bullet screen according to an embodiment. The method may be performed by a computer device. Referring to fig. 7, the method includes:

step S701, determining whether the target barrage synchronously displayed with the target video clip has a scenario.

In step S702, it is determined whether the target barrage is a theatrical transparent barrage for a subsequent video clip.

In step S703, when the target barrage is a through barrage for a subsequent video clip, the target barrage is processed based on the viewing condition of the subsequent video clip by the user.

In one embodiment, before determining whether a target bullet screen is a theatrical transmission bullet screen, the computer device may first determine whether the target bullet screen is provided with a scenario, may perform scenario determination through a flow as shown in fig. 6, and may determine whether the bullet screen is provided with a scenario through the BERT layer according to bullet screen words in the bullet screen and positions of bullet screen words in the bullet screen.

Fig. 8 is a flow chart of a method for identifying a bullet screen according to an embodiment. The method may be performed by a computer device. Referring to fig. 8, the method includes the steps of:

step S801, a plurality of barrages synchronously displayed with a target video clip are obtained;

step S802, performing word segmentation processing on each barrage to obtain a word segmentation sequence corresponding to each barrage;

step S803, according to each word in the word segmentation sequence and the position of each word in the word segmentation sequence, acquiring the plot detection characteristics corresponding to the barrage;

step S804, inputting the scenario detection characteristics into a pre-trained scenario detection neural network, and outputting the scenario detection result of whether the bullet screen has the scenario through the scenario detection neural network;

In step S805, according to the scenario detection result, a scenario-equipped bullet screen of the plurality of bullet screens is set as a target bullet screen.

Step S806, determining the subsequent video clips with the playing sequence behind the target video clip;

step S807, acquiring a barrage scenario depth representation of a target barrage and acquiring a video clip scenario depth representation of a subsequent video clip;

step S808, performing interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation, and taking the interactive fusion result as the relativity of the target barrage and the subsequent video clip on the scenario;

step S809, acquiring a barrage segment formed by a barrage synchronously displayed with a subsequent video segment, acquiring a barrage semantic depth representation of a target barrage, and acquiring a barrage segment semantic depth representation of the barrage segment;

step S810, performing interaction fusion on the barrage semantic depth representation and the barrage segment semantic depth representation, and taking an interaction fusion result as the semantic relativity of the target barrage and the barrage segment;

step S811, determining the total number of the drama shots released by the target account and having release time before the release time of the target shots, and determining the number of the drama transparent shots released by the target account and having release time before the release time of the target shots;

In step S812, the ratio of the number of the drama screens to the total number is used as the tendency of the target account to release the drama screens.

Step S813 of acquiring a first weight corresponding to the relevance on the plot and the semantic relevance, and a second weight corresponding to the tendency;

step S814, obtaining the probability that the target barrage is a drama-transparent barrage according to the product result of the first weight, the correlation degree on the plot and the correlation degree on the semanteme and the product result of the second weight and the tendency degree;

according to the bullet screen identification method, when the target bullet screen is required to be identified, after the target bullet screen synchronously displayed with the target video segment and the subsequent video segment with the playing sequence behind the target video segment are obtained, the relativity of the target bullet screen and the subsequent video segment on the plot is obtained based on the comparison between the plot characteristics of the target bullet screen and the plot characteristics of the subsequent video segment; based on the comparison between the semantic features of the target barrage and the semantic features of the barrage segments synchronously displayed with the subsequent video, obtaining the semantic relativity of the target barrage segments synchronously displayed with the subsequent video segments; and then, identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a dramatic barrage. On one hand, based on definition of the drama-transparent barrage, identification of the drama-transparent barrage is carried out not by relying on the target barrage alone but on the relativity between the target barrage and the candidate video clips, on the other hand, the relativity comprises two dimensions of plot relativity and semantic relativity, and internal relation between the target barrage and the candidate video clips is fully considered, so that accuracy of identifying the drama-transparent barrage is greatly improved.

It should be understood that, although the steps in the flowcharts of fig. 2 to 8 are sequentially shown as indicated by arrows, these steps are not necessarily sequentially performed in the order indicated by the arrows. The steps are not strictly limited to the order of execution unless explicitly recited herein, and the steps may be executed in other orders. Moreover, at least a portion of the steps of fig. 2-8 may include steps or stages that are not necessarily performed at the same time, but may be performed at different times, nor does the order in which the steps or stages are performed necessarily occur sequentially, but may be performed alternately or alternately with other steps or at least a portion of the steps or stages in other steps. It should be noted that references to "first," "second," etc. above are used to distinguish similar objects and are not necessarily used to describe a particular order or sequence.

In some embodiments, as shown in fig. 9, there is provided a bullet screen recognition apparatus including:

the target bullet screen acquisition module 901 is used for acquiring a target bullet screen synchronously displayed with the target video clip;

A subsequent video clip obtaining module 902, configured to determine a subsequent video clip whose playing sequence follows the target video clip;

the plot correlation obtaining module 903 is configured to use a comparison result between plot characteristics of the target barrage and plot characteristics of the subsequent video segment as a plot correlation between the target barrage and the subsequent video segment;

the semantic relativity acquisition module 904 is configured to acquire a barrage segment formed by a barrage synchronously displayed with the subsequent video segment, and take a comparison result between the semantic features of the target barrage and the semantic features of the barrage segment as the semantic relativity of the target barrage and the barrage segment;

and the drama-transparent barrage identification module 905 is configured to identify the target barrage based on the relevance on the plot and the semantic relevance, and obtain an identification result of whether the target barrage is a drama-transparent barrage.

In some embodiments, the drama screen identification module 905 is further configured to obtain a tendency of the target account on issuing the drama screen; the target account is an account for issuing the target barrage; based on the correlation degree on the plot, the semantic correlation degree and the positive correlation relation between the trend degree and the bullet screen dramatic permeability probability, obtaining the probability that the target bullet screen is a bullet screen; and based on the probability that the target barrage is the transparent barrage, obtaining the identification result of whether the target barrage is the transparent barrage.

In some embodiments, the drama screen identification module 905 is further configured to obtain a first weight corresponding to the relevance on the episode and the semantic relevance, and a second weight corresponding to the trend; and obtaining the probability that the target barrage is a through barrage according to the first weight, the product result of the relevance on the plot and the semantic relevance and the product result of the second weight and the tendency.

In some embodiments, the drama screen identification module 905 is further configured to determine a release time of the target barrage; determining a total number of drama shots published by the target account and having a publication time before a publication time of the target barrage, and determining a number of drama passes published by the target account and having a publication time before the publication time of the target barrage; and taking the ratio of the number of the drama-transparent backdrop to the total number as the tendency of the target account on issuing the drama-transparent backdrop.

In some embodiments, the scenario feature of the target bullet screen is a bullet screen scenario depth representation and the scenario feature of the subsequent video clip is a video clip scenario depth representation; the scenario correlation obtaining module 903 is further configured to obtain a scenario depth representation of the target scenario, and obtain a video clip scenario depth representation of the subsequent video clip; and carrying out interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation, and taking an interactive fusion result as the relativity of the target barrage and the subsequent video clip on the scenario.

and/or character recognition is carried out on the video frames of the subsequent video segments, so that caption text of the subsequent video segments is obtained; and constructing video clip scenario depth representation of the subsequent video clip based on the subtitle text.

In some embodiments, the semantic features of the target barrage are barrage semantic depth representations and the semantic features of the barrage segment are barrage segment semantic depth representations; the semantic relativity acquisition module 904 is further configured to acquire a barrage semantic depth representation of the target barrage, and acquire a barrage segment semantic depth representation of the barrage segment; and carrying out interaction fusion on the barrage semantic depth representation and the barrage segment semantic depth representation, and taking an interaction fusion result as the semantic relativity of the target barrage and the barrage segment.

In some embodiments, the target barrage acquisition module 901 is further configured to perform word segmentation processing on each barrage to obtain a word segmentation sequence corresponding to each barrage; acquiring the plot detection characteristics corresponding to the barrage according to each word in the word segmentation sequence and the position of each word in the word segmentation sequence; inputting the scenario detection characteristics into a pre-trained scenario detection neural network, and outputting a scenario detection result of whether the bullet screen has the scenario through the scenario detection neural network.

In some embodiments, the scenario correlation obtaining module 903 is further configured to input the scenario features of the target bullet screen and the scenario features of the subsequent video clip into a pre-trained scenario correlation detection neural network for comparison; and taking the comparison result output by the plot correlation degree detection neural network as the correlation degree of the target barrage and the subsequent video clips on the plot.

In the bullet screen identification device, when the target bullet screen is required to be identified, after the target bullet screen synchronously displayed with the target video segment and the subsequent video segment with the playing sequence behind the target video segment are obtained, the relativity of the target bullet screen and the subsequent video segment on the plot is obtained based on the comparison between the plot characteristics of the target bullet screen and the plot characteristics of the subsequent video segment; based on the comparison between the semantic features of the target barrage and the semantic features of the barrage segments synchronously displayed with the subsequent video, obtaining the semantic relativity of the target barrage segments synchronously displayed with the subsequent video segments; and then, identifying the target barrage based on the relevance on the plot and the semantic relevance to obtain an identification result of whether the target barrage is a dramatic barrage. On one hand, based on definition of the drama-transparent barrage, identification of the drama-transparent barrage is carried out not by relying on the target barrage alone but on the relativity between the target barrage and the candidate video clips, on the other hand, the relativity comprises two dimensions of plot relativity and semantic relativity, and internal relation between the target barrage and the candidate video clips is fully considered, so that accuracy of identifying the drama-transparent barrage is greatly improved.

For specific limitations of the bullet screen recognition apparatus, reference is made to the above limitations of the bullet screen recognition method, and no further description is given here. The modules in the bullet screen recognition device can be all or partially implemented by software, hardware and a combination thereof. The above modules may be embedded in hardware or may be independent of a processor in the computer device, or may be stored in software in a memory in the computer device, so that the processor may call and execute operations corresponding to the above modules.

In some embodiments, a computer device is provided, which may be a terminal or a server as shown in fig. 1, and an internal structure diagram thereof may be as shown in fig. 10. The computer device includes a processor, a memory, and a communication interface connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The database of the computer device is used to store the barrage data. The network interface of the computer device is used for communicating with an external computer device through a network connection. The computer program when executed by a processor implements a barrage identification method.

It will be appreciated by those skilled in the art that the structure shown in fig. 10 is merely a block diagram of some of the structures associated with the present application and is not limiting of the computer device to which the present application may be applied, and that a particular computer device may include more or fewer components than shown, or may combine certain components, or have a different arrangement of components.

In some embodiments, a computer device is provided, comprising a memory storing a computer program and a processor implementing the steps of the method embodiments described above when the processor executes the computer program.

In some embodiments, a computer readable storage medium is provided, on which a computer program is stored which, when executed by a processor, implements the steps of the various method embodiments described above.

In some embodiments, a computer program product is provided, comprising a computer program which, when executed by a processor, implements the steps of the various method embodiments described above.

It should be noted that, the user information (including, but not limited to, user equipment information, user personal information, etc.) and the data (including, but not limited to, data for analysis, stored data, presented data, etc.) referred to in the present application are information and data authorized by the user or sufficiently authorized by each party, and the collection, use and processing of the related data are required to comply with the related laws and regulations and standards of the related countries and regions.

Those skilled in the art will appreciate that implementing all or part of the above-described methods may be accomplished by way of a computer program stored on a non-transitory computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in embodiments provided herein may include at least one of non-volatile and volatile memory. The nonvolatile Memory may include Read-Only Memory (ROM), magnetic tape, floppy disk, flash Memory, optical Memory, or the like. Volatile memory can include random access memory (Random Access Memory, RAM) or external cache memory. By way of illustration, and not limitation, RAM can be in the form of a variety of forms, such as static random access memory (Static Random Access Memory, SRAM) or dynamic random access memory (Dynamic Random Access Memory, DRAM), and the like.

The technical features of the above embodiments may be arbitrarily combined, and all possible combinations of the technical features in the above embodiments are not described for brevity of description, however, as long as there is no contradiction between the combinations of the technical features, they should be considered as the scope of the description.

The foregoing examples represent only a few embodiments of the present application, which are described in more detail and are not to be construed as limiting the scope of the invention. It should be noted that it would be apparent to those skilled in the art that various modifications and improvements could be made without departing from the spirit of the present application, which would be within the scope of the present application. Accordingly, the scope of protection of the present application is to be determined by the claims appended hereto.

Claims

1. A method of barrage identification, the method comprising:

acquiring a target barrage synchronously displayed with a target video segment;

2. The method of claim 1, wherein identifying the target barrage based on the plot-wise relevance and the semantic relevance to obtain an identification of whether the target barrage is a theatrical barrage, comprises:

acquiring the tendency of a target account on a release of a bullet screen; the target account is an account for issuing the target barrage;

based on the correlation degree on the plot, the semantic correlation degree and the positive correlation relation between the trend degree and the bullet screen dramatic permeability probability, obtaining the probability that the target bullet screen is a bullet screen;

and based on the probability that the target barrage is the transparent barrage, obtaining the identification result of whether the target barrage is the transparent barrage.

3. The method of claim 2, wherein deriving the probability that the target barrage is a theatrical barrage based on the positive correlation between the storyboard relevance, the semantic relevance, and the trend each and a barrage theatrical probability comprises:

Acquiring a first weight corresponding to the relevance on the plot and the semantic relevance, and a second weight corresponding to the tendency;

and obtaining the probability that the target barrage is a through barrage according to the first weight, the product result of the relevance on the plot and the semantic relevance and the product result of the second weight and the tendency.

4. The method of claim 2, wherein the obtaining the propensity of the target account to release the theatrical performance includes:

determining the release time of the target barrage;

determining a total number of drama shots published by the target account and having a publication time before a publication time of the target barrage, and determining a number of drama passes published by the target account and having a publication time before the publication time of the target barrage;

and taking the ratio of the number of the drama-transparent backdrop to the total number as the tendency of the target account on issuing the drama-transparent backdrop.

5. The method of claim 1, wherein the scenario feature of the target bullet screen is a bullet screen scenario depth representation and the scenario feature of the subsequent video clip is a video clip scenario depth representation;

And the comparison result between the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips is used as the relativity of the target barrage and the subsequent video clips on the plot, and comprises the following steps:

acquiring a barrage scenario depth representation of the target barrage, and acquiring a video clip scenario depth representation of the subsequent video clip;

and carrying out interactive fusion on the barrage scenario depth representation and the video clip scenario depth representation, and taking an interactive fusion result as the relativity of the target barrage and the subsequent video clip on the scenario.

6. The method of claim 5, wherein obtaining a video clip episode depth representation of the subsequent video clip comprises:

performing voice recognition on the dialect of the subsequent video segment to obtain a dialect text of the subsequent video segment; constructing video clip scenario depth representation of the subsequent video clip based on the text-to-text; and/or the number of the groups of groups,

7. The method of claim 1, wherein the semantic features of the target barrage are barrage semantic depth representations and the semantic features of the barrage segments are barrage segment semantic depth representations;

the step of using the comparison result between the semantic features of the target barrage and the semantic features of the barrage segment as the semantic relativity between the target barrage and the barrage segment comprises the following steps:

acquiring a barrage semantic depth representation of the target barrage, and acquiring a barrage section semantic depth representation of the barrage section;

and carrying out interaction fusion on the barrage semantic depth representation and the barrage segment semantic depth representation, and taking an interaction fusion result as the semantic relativity of the target barrage and the barrage segment.

8. The method of claim 1, wherein the obtaining a target bullet screen that is presented in synchronization with a target video clip comprises:

acquiring a plurality of barrages synchronously displayed with a target video clip;

respectively carrying out plot detection on each bullet screen in the plurality of bullet screens to obtain plot detection results for representing whether the bullet screen has plots or not;

and determining the barrage with the plot from the barrages as a target barrage according to the plot detection result.

9. The method of claim 8, wherein the separately plot detecting each of the plurality of curtains comprises:

performing word segmentation processing on each barrage to obtain word segmentation sequences corresponding to each barrage;

acquiring the plot detection characteristics corresponding to the barrage according to each word in the word segmentation sequence and the position of each word in the word segmentation sequence;

inputting the scenario detection characteristics into a pre-trained scenario detection neural network, and outputting a scenario detection result of whether the bullet screen has the scenario through the scenario detection neural network.

10. The method of claim 1, wherein the comparing the plot feature of the target barrage with the plot feature of the subsequent video clip as a relevance of the target barrage to the plot feature of the subsequent video clip comprises:

inputting the plot characteristics of the target barrage and the plot characteristics of the subsequent video clips into a pre-trained plot correlation detection neural network for comparison;

and taking the comparison result output by the plot correlation degree detection neural network as the correlation degree of the target barrage and the subsequent video clips on the plot.

11. The method according to claim 10, wherein the method further comprises:

acquiring a plurality of sample barrages synchronously displayed with the sample video clips;

determining a first sample barrage and a second sample barrage of the plurality of sample barrages;

when the audience acceptance degree corresponding to the first sample barrage is higher than the audience acceptance degree corresponding to the second sample barrage, determining that the correlation degree of the first sample barrage and the sample video clip on the plot is higher than the correlation degree of the second sample barrage and the sample video clip on the plot;

training the scenario correlation detection neural network based on the first sample barrage and the sample video clip being higher in scenario correlation than the second sample barrage and the sample video clip, the scenario features of the first sample barrage, the scenario features of the second sample barrage, and the scenario features of the sample video clip.

12. The method of claim 11, wherein the method further comprises:

acquiring the number of praise of the first sample barrage and the number of praise of the second sample barrage;

and when the praise number of the first sample barrage is larger than that of the second sample barrage, determining that the audience approval degree corresponding to the first sample barrage is higher than that corresponding to the second sample barrage.

13. A bullet screen identification device, said device comprising:

14. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that the processor implements the method of any one of claims 1 to 12 when executing the computer program.

15. A computer readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the method of any one of claims 1 to 12.

16. A computer program product comprising a computer program, characterized in that the computer program, when executed by a processor, implements the method of any one of claims 1 to 12.