A quality evaluation method and device for short video similarity

Abstract

The present application relates to a quality evaluation method and device for short video similarity. By collecting quality evaluation records of to-be-evaluated segments of short videos, when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is greater than a first threshold, the corresponding quality level is determined as the quality level of the short video. According to the level, the quality evaluation record library of the to-be-evaluated segments of the short video is graded to obtain a set of short video segments corresponding to the quality level, and information association filtering is performed to determine whether it belongs to a positive or non-positive association, and determine the similar plot matching interval value corresponding to the quality level. This scheme only needs to collect the current quality evaluation record of the short video, and can directly determine the quality level of the short video through the matching interval value. It can be achieved without running a complex model, is more efficient and accurate, reduces the calculation complexity, and can further reduce misidentification or missed identification.

Description

Technical Field

[0001] The present application relates to the field of video processing technology, and in particular to a quality evaluation method and device for short video similarity. Background Art

[0002] With the rapid development of computer technology, Internet technology and natural language big model technology, short videos are appearing in more and more businesses.

[0003] Currently, video content has become the mainstream of self-media, and users share their daily lives by shooting short videos. In order to obtain short videos with rich content, users can freely combine various video materials shot by shooting devices (drones, handheld gimbals, cameras or mobile phones), and edit and merge multiple video materials into one video and publish it on social networking sites.

[0004] Due to the emergence of massive short videos, it is becoming increasingly difficult for users to find short videos that meet their quality requirements among the massive short videos. Therefore, there is an urgent need for an efficient and accurate quality evaluation scheme for short video similarity. Summary of the invention

[0005] To achieve the above objectives, this application provides the following technical solutions:

[0006] According to a first aspect of the present invention, the present invention claims a method for evaluating the quality of short video similarity, comprising:

[0007] Collecting a quality evaluation record of a to-be-evaluated segment of a short video, wherein the quality evaluation record includes a plurality of evaluation indicators;

[0008] When, among the multiple evaluation indicators, the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value is greater than a first threshold, the quality level corresponding to the similar plot matching interval value is determined as the quality level of the short video.

[0009] Further, the plurality of evaluation indicators are used to indicate advertising ratio and / or performance indicators;

[0010] The quality level includes a character quality level or a script quality level, and the similar plot matching interval value corresponds to the character quality level or the script quality level;

[0011] The similar plot matching interval value is obtained by the following operation:

[0012] Based on the quality evaluation record library of the short video segments to be evaluated, grading is performed to obtain a set of short video segments corresponding to the quality level, wherein the quality evaluation record library of the short video segments to be evaluated contains the quality evaluation records of the short video segments to be evaluated obtained in advance;

[0013] Performing information association filtering on the short video clip set to determine whether the information association of the short video clip set is a positive association or a non-positive association;

[0014] According to the information association of the short video segment set, a similar plot matching interval value corresponding to the quality level is determined.

[0015] Furthermore, the information association filtering of the short video clip set specifically includes:

[0016] Dividing the quality evaluation records in the short video clip set into video frame nodes to obtain a plurality of video frame nodes;

[0017] Taking the video frame node with the largest number of video frames among the multiple video frame nodes as the mean interval of the short video segment set;

[0018] When the number of video frames of the short video segment set in the first interval is equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a positive association;

[0019] When the number of video frames of the short video segment set in the first interval is not equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a non-positive association;

[0020] The first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval.

[0021] Furthermore, the method further comprises: when, among the multiple evaluation indicators, the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value is not greater than the first threshold, determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality indicator value;

[0022] The determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality index value includes:

[0023] Of the two first short video reference quality index values, the quality level corresponding to the first short video reference quality index value closer to the quality evaluation record is determined as the quality level of the short video; or,

[0024] Among three or more first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value closest to the quality evaluation record as the quality level of the short video;

[0025] The determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality index value includes:

[0026] When the absolute value of the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determining that the quality level is the character quality level;

[0027] When the absolute value of the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, the quality level is determined to be the script quality level;

[0028] The first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level.

[0029] Furthermore, it also includes:

[0030] When the attribution degree of the quality evaluation record of the to-be-evaluated segment of the short video is greater than or equal to the third threshold, the quality evaluation record whose number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is not greater than the first threshold is added to the quality evaluation record library of the to-be-evaluated segment of the short video;

[0031] When the duration of the quality level being the script quality level is greater than or equal to the second threshold, the quality level is controlled by using one or more control strategies including the following operations: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation;

[0032] The control strategy is determined according to the script sequence of the short video, and the script sequence is determined according to the quality evaluation record and the similar plot matching interval value corresponding to the quality level;

[0033] re-determining a first script sequence of the short videos after a first period, the first period corresponding to the first script sequence;

[0034] The operations in the control strategy are selected randomly.

[0035] According to a second aspect of the present invention, the present invention claims protection for a quality evaluation device for short video similarity, comprising:

[0036] A collection module, used for collecting quality evaluation records of the short video clips to be evaluated, wherein the quality evaluation records include multiple evaluation indicators;

[0037] The determination module is used to determine the quality level corresponding to the similar plot matching interval value as the quality level of the short video when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among the multiple evaluation indexes is greater than a first threshold.

[0038] Further, the plurality of evaluation indicators are used to indicate advertising ratio and / or performance indicators;

[0039] The quality level includes a character quality level and a script quality level, and the similar plot matching interval value corresponds to the character quality level or the script quality level;

[0040] A calculation module, wherein the calculation module is used for:

[0041] Based on the quality evaluation record library of the short video segments to be evaluated, grading is performed to obtain a set of short video segments corresponding to the quality level, wherein the quality evaluation record library of the short video segments to be evaluated contains the quality evaluation records of the short video segments to be evaluated obtained in advance;

[0042] Performing information association filtering on the short video clip set to determine whether the information association of the short video clip set is a positive association or a non-positive association;

[0043] According to the information association of the short video segment set, a similar plot matching interval value corresponding to the quality level is determined.

[0044] Furthermore, the calculation module is specifically used for:

[0045] Dividing the quality evaluation records in the short video clip set into video frame nodes to obtain a plurality of video frame nodes;

[0046] Taking the video frame node with the largest number of video frames among the multiple video frame nodes as the mean interval of the short video segment set;

[0047] When the number of video frames of the short video segment set in the first interval is equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a positive association;

[0048] When the number of video frames of the short video segment set in the first interval is not equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a non-positive association;

[0049] The first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval.

[0050] Furthermore, the determination module is further configured to determine the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality indicator value when the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value among the multiple evaluation indicators is not greater than the first threshold value;

[0051] The determining module is specifically used for:

[0052] Of the two first short video reference quality index values, the quality level corresponding to the first short video reference quality index value closer to the quality evaluation record is determined as the quality level of the short video; or,

[0053] Among three or more first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value closest to the quality evaluation record as the quality level of the short video;

[0054] The determining module is specifically used for:

[0055] When the absolute value of the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determining that the quality level is the character quality level;

[0056] When the absolute value of the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, the quality level is determined to be the script quality level;

[0057] The first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level.

[0058] Furthermore, it also includes:

[0059] The information library updating module is used to: when the attribution degree of the quality evaluation record of the to-be-evaluated segment of the short video is greater than or equal to a third threshold, add the quality evaluation record of the to-be-evaluated segment whose number of the evaluation index belonging to the similar plot matching interval value is not greater than the first threshold to the quality evaluation record library of the to-be-evaluated segment of the short video;

[0060] A control module, for controlling the quality level by adopting one or more control strategies including the following operations when the duration of the quality level being the script quality level is greater than or equal to a second threshold: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation;

[0061] The control strategy is determined according to the script sequence of the short video, and the script sequence is determined according to the quality evaluation record and the similar plot matching interval value corresponding to the quality level;

[0062] The control module is further used to re-determine a first script sequence of the short video after a first period, wherein the first period corresponds to the first script sequence;

[0063] The operations in the control strategy are randomly selected by the control module.

[0064] The present application relates to a quality evaluation method and device for short video similarity. By collecting quality evaluation records of to-be-evaluated segments of short videos, when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is greater than a first threshold, the corresponding quality level is determined as the quality level of the short video. According to the level, the quality evaluation record library of the to-be-evaluated segments of the short video is graded to obtain a set of short video segments corresponding to the quality level, and information association filtering is performed to determine whether it belongs to a positive or non-positive association, and determine the similar plot matching interval value corresponding to the quality level. This scheme only needs to collect the current quality evaluation record of the short video, and can directly determine the quality level of the short video through the matching interval value. It can be achieved without running a complex model, is more efficient and accurate, reduces the calculation complexity, and can further reduce misidentification or missed identification. BRIEF DESCRIPTION OF THE DRAWINGS

[0065] Figure 1 A workflow diagram of a method for quality evaluation of short video similarity as claimed in an embodiment of the present application;

[0066] Figure 2 An exemplary structural diagram of a quality assessment device for short video similarity claimed in an embodiment of the present application;

[0067] Figure 3 An exemplary flow chart of quality level evaluation of a quality evaluation method for short video similarity claimed in an embodiment of the present application;

[0068] Figure 4 This is an exemplary flowchart of the association of filtering information for a quality assessment method for short video similarity as claimed in an embodiment of the present application. DETAILED DESCRIPTION

[0069] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only 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.

[0070] The terms "first", "second" and "third" in this application are only used for descriptive purposes and cannot be understood as indicating or implying relative importance or implicitly indicating the number of indicated technical features. Therefore, the features defined as "first", "second" and "third" can explicitly or implicitly include at least one of the features. In the description of this application, the meaning of "multiple" is at least two, such as two, three, etc., unless otherwise clearly and specifically defined. All directional indications (such as up, down, left, right, front, back...) in the embodiments of this application are only used to explain the relative position relationship, movement, etc. between the components under a certain specific posture (as shown in the accompanying drawings). If the specific posture changes, the directional indication also changes accordingly. In addition, the terms "including" and "having" and any of their variations are intended to cover non-exclusive inclusions. For example, a process, method, device, product or equipment that includes a series of steps or units is not limited to the listed steps or units, but optionally also includes steps or units that are not listed, or optionally also includes other steps or units inherent to these processes, methods, products or equipment.

[0071] Reference to "embodiments" herein means that a particular feature, structure, or characteristic described in conjunction with the embodiments may be included in at least one embodiment of the present application. The appearance of the phrase in various locations in the specification does not necessarily refer to the same embodiment, nor is it an independent or alternative embodiment that is mutually exclusive with other embodiments. It is explicitly and implicitly understood by those skilled in the art that the embodiments described herein may be combined with other embodiments.

[0072] The following takes a short video as an example to explain in detail the implementation of the embodiment of the present application.

[0073] Figure 1 The exemplary process of the method for evaluating the quality of short video similarity in the embodiment of the present application is shown. Figure 1 As shown, the quality evaluation method for short video similarity in the embodiment of the present application may include:

[0074] Step S101, collecting quality evaluation records of the to-be-evaluated segments of the short video, where the quality evaluation records include multiple evaluation indicators;

[0075] The indicator corresponding to the evaluation index can characterize a feature that affects the playback volume during the short video film and television drama short video process. The feature that affects the playback volume may include but is not limited to advertising ratio features, performance index features, language features, or any other features related to the status of the short video during the film and television drama short video process. In practical applications, the indicator can be pre-associated, selected in real time, or selected in response to the short video operation based on the needs of the specific application scenario, the status of the short video, the requirements of the evaluation accuracy, or various other factors. For example, for different categories of short videos, due to the different needs of short videos in film and television drama short videos, the indicators need to be adjusted according to the different short video performance indicators.

[0076] The evaluation index may be information obtained after pre-matching such as real-time cleaning and normalization through the index evaluation instructions collected in real time. In some examples, if the matching interval value is a probability value of [0,1], the evaluation index may be a normalized value obtained based on the index evaluation instructions collected in real time, and the value of the normalized value is between [0,1]. In this way, it is not only beneficial for the device (for example, the computing device described below) to efficiently determine whether the evaluation index belongs to the matching interval value, but also has a small number of video frames and low computational complexity. In other examples, the evaluation index may also be an integer value obtained based on the index evaluation instructions collected in real time. In this way, it is not only beneficial for efficiently determining whether the evaluation index belongs to the corresponding matching interval value, but also has low requirements on hardware performance such as matchers, which can further reduce hardware costs.

[0077] In at least some embodiments, the indicators may include but are not limited to advertising ratio indicators, performance indicators, etc. In this embodiment, the characteristics of the short video's impact on the playback volume can be comprehensively and accurately perceived through a variety of indicators, and a quality evaluation record that can reflect the comprehensive characteristics of the short video's impact on the playback volume can be obtained, thereby efficiently and accurately determining the quality level of the short video.

[0078] The advertisement ratio index may indicate the status of the short video evaluation instructions related to the playback. The advertisement ratio index evaluation instructions may be collected by a short video evaluation instruction evaluation detector. Here, the short video evaluation instruction evaluation detector may provide these evaluation instructions to the device (e.g., the computing device described below) through various wireless playback operations, or may provide these evaluation instructions to the device (e.g., the computing device described below) in the form of electrical evaluation instructions through a wired connection with the device.

[0079] The short video evaluation instruction evaluation detector may include but is not limited to a scene transition detector, a quality detector, a title detector, an advertisement detector, an end detector, etc. Among them, the scene transition detector can be used to collect scene transition evaluation instructions in real time during the short video film and television drama short video process, the quality detector can be used to collect quality evaluation instructions in real time during the short video film and television drama short video process, the end detector can be used to collect end evaluation instructions in real time during the short video film and television drama short video process, the title detector can be used to collect title evaluation instructions in real time during the short video film and television drama short video process, and the advertisement detector can be used to collect advertisement evaluation instructions in real time during the short video film and television drama short video process. Usually, you can choose between title and advertisement.

[0080] The performance index can indicate the execution status of the film and television drama action of the short video. The index evaluation instructions of the performance index can be collected by the performance index detector. The performance index detector can provide the evaluation instructions collected by itself in real time to the device (for example, the computing device below) through various wireless playback operations, and can also provide the evaluation instructions collected by itself in real time to the device (for example, the computing device below) in the form of electrical evaluation instructions through a wired connection with the device.

[0081] It is understandable that the indicator types in the embodiments of the present application are not limited to the above-mentioned performance indicators and advertising ratios, and may also include other types that can characterize the conditions that affect the playback volume of short videos, film and television dramas, and short videos, such as indicators related to voice, pupils, etc. For example, the indicators of short videos may also include language indicators, and the evaluation indicators of voice may include the frequency value and amplitude value of language indicators, etc. The indicator evaluation instructions of language may be collected by a voice detector such as a microphone array.

[0082] The quality evaluation record is a collection of various evaluation indicators at the same time. The quality evaluation record has at least two dimensions, and its dimensions correspond to the indicators one by one, which can reflect the comprehensive characteristics of the impact of the playback volume of the short video at a certain moment. For example, taking the pre-selected 6 indicators as an example, the 6 indicators include scene transition, quality, advertising (or, title), short video segment evaluation score, and short video segment duration. The corresponding quality evaluation record can have 6 dimensions, which represent the above 6 indicators respectively. The quality evaluation record at the i-th moment can be expressed as {ai, bi, ci, di, ei}, ai represents the scene transition value at the i-th moment, bi represents the quality value at the i-th moment, ci represents the advertising value or title value at the i-th moment, di represents the short video segment evaluation score value at the i-th moment, and ei represents the short video segment duration value at the i-th moment. It should be noted that the "same moment" here not only includes strict time synchronization, but also includes the situation where the acquisition cycle is less than a predetermined threshold, and the predetermined threshold can be determined by the performance indicators of each detector. In practical applications, the threshold can be an empirical value or a set value.

[0083] Step S102: when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among the multiple evaluation indexes in the quality evaluation record is greater than a first threshold, the quality level corresponding to the similar plot matching interval value is determined as the quality level of the short video.

[0084] Here, the similar plot matching interval value generally refers to a single-class matching interval value, which can be used to indicate a single quality level, and can include matching interval values ​​corresponding to each indicator. Detailed technical details can be found in the following description, which will not be repeated here.

[0085] The quality level may indicate the state of the impact on the playback volume related to the short video film and television drama. In the embodiment of the present application, the quality level may include two or more. In this way, by dividing into multiple quality levels, it is convenient for short videos to intuitively understand whether their own state is suitable for film and television drama short videos, and at the same time, it can reduce the dependence of the state evaluation process on the supervised machine learning model, effectively reducing the computational complexity while improving accuracy.

[0086] In some implementations, the quality level may include character quality level, short script quality level, medium script quality level, and long script quality level. Each short video may have four types of matching interval values ​​corresponding to the four quality levels, that is, four similar plot matching interval values, which correspond to the character quality level, short script quality level, medium script quality level, and long script quality level. In this example, the quality level of the short video can be directly determined by multiple types of matching interval values, which can reduce the computational complexity and further reduce misidentification or missed identification.

[0087] In some implementations, the quality level may include the character quality level and the script quality level, and the matching interval value of the short video may include a type of matching interval value corresponding to the character quality level and a type of matching interval value corresponding to the script quality level, that is, two similar plot matching interval values, which correspond to the character quality level and the script quality level, respectively. In this example, the quality level of the short video can be directly determined to be the to-be-matched or script quality level through two types of matching interval values, which reduces the computational complexity and the possibility of misidentification or missed identification.

[0088] In at least some embodiments, each type of matching interval value may include a matching interval value corresponding to an indicator one-to-one, that is, a similar plot matching interval value may include a matching interval value corresponding to an indicator one-to-one. Thus, the quality level of a short video can be determined by comprehensively considering various features that affect the playback volume during playback, which can further improve the accuracy of the short video quality level evaluation.

[0089] Taking 6 pre-selected indicators as an example, the 6 indicators include scene transition, quality, advertising (or, opening), short video clip evaluation score, short video clip length, and short video clip curling. Assuming that the quality level includes character quality level and script quality level, the corresponding matching interval values ​​are divided into two categories, one category of matching interval values ​​corresponds to character quality level, and the other category of matching interval values ​​corresponds to script quality level.

[0090] In step S102, if the value of the evaluation index in the quality evaluation record belongs to the matching interval value corresponding to the corresponding index in the similar plot matching interval value, the evaluation index can be considered to belong to the similar plot matching interval value. If the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value in a quality evaluation record is greater than the first threshold, the quality evaluation record can be considered to belong to the similar plot matching interval value of the short video. In this embodiment, the quality level of the short video can be determined by judging whether a specific value belongs to the corresponding value interval. The algorithm is simple and easy to implement, with low hardware cost, high efficiency and small error.

[0091] After analysis, it is found that even if the short video quality level is to be matched, some of its evaluation indicators may also belong to the similar plot matching interval value of the corresponding script quality level. Similarly, when the short video quality level is script, some of its evaluation indicators may also fall into the similar plot matching interval value of the corresponding character quality level. Therefore, in the method of the embodiment of the present application, the quality level of the short video is determined by setting the first threshold value in combination with the matching interval value, which is more in line with the actual situation and the determination record is more accurate.

[0092] In at least some embodiments, the quality evaluation method for short video similarity may further include:

[0093] Step S103, when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among the multiple evaluation indexes of the quality evaluation record is not greater than the first threshold, the quality level of the short video is determined according to the deviation between the quality evaluation record and the first short video reference quality index value.

[0094] Therefore, the quality evaluation can be used to record the deviation of the first short video reference quality index value corresponding to different quality levels to assist in the evaluation of the quality level, and further achieve real-time, efficient and accurate evaluation of the short video quality level with lower computational complexity.

[0095] The first short video reference quality index value generally refers to a short video reference quality index value corresponding to a single quality level, that is, a single type of matching interval value can correspond to a short video reference quality index value, that is, a similar plot matching interval value corresponds to a first short video reference quality index value. In some implementation operations, similar to the matching interval value, each first short video reference quality index value can correspond to a quality level, and the first short video reference quality index value has at least two values ​​corresponding to multiple indicators, that is, the first short video reference quality index value can be represented by a multi-dimensional information, and the dimension of the multi-dimensional information is equal to the number of index to be evaluated segments. Specifically, each dimension of the first short video reference quality index value corresponds to an indicator, and the value of each dimension can be used as the mean value of the corresponding indicator. Therefore, the deviation of the quality evaluation record to the short video reference quality index value can be accurately estimated by using the values ​​of each dimension of the short video reference quality index value and each evaluation indicator in the quality evaluation record of the short video to be evaluated segment. In this implementation, the quality level of the short video is determined by the deviation between multi-dimensional information, and the quality level of the short video can be judged by the comprehensive characteristics of the short video that affect the playback volume, which reduces the calculation complexity and improves the accuracy of the quality level evaluation.

[0096] In some embodiments, similar to the similar plot matching interval value, the first short video reference quality index value may include two or more, and each first short video reference quality index value corresponds to a quality level. Taking the quality level including the character quality level and the script quality level as an example, the short video reference quality index value of the short video may include a short video reference quality index value corresponding to the character quality level and a short video reference quality index value corresponding to the script quality level, that is, one or two first short video reference quality index values ​​may be set, and these two first short video reference quality index values ​​correspond to the character quality level and the script quality level, respectively, or the first short video reference quality index value may correspond to the character quality level or the script quality level. Thus, the quality level of the short video can be determined by the deviation between the quality evaluation record and the first short video reference quality index value, which can further reduce the possibility of misidentification or missed identification while reducing the computational complexity.

[0097] In some embodiments, in step S103, among the two first short video reference quality index values, the quality level corresponding to the first short video reference quality index value that is closer to the quality evaluation record is determined as the quality level of the short video. Alternatively, among three or more first short video reference quality index values, the quality level corresponding to the first short video reference quality index value that is closest to the quality evaluation record is determined as the quality level of the short video. In this embodiment, the short video reference quality index value that is closest to the quality evaluation record is obtained through the deviation between the quality evaluation record and the short video reference quality index value, and then the quality level corresponding to the short video reference quality index value that is closest to the quality evaluation record is used as the quality level of the short video, with low calculation complexity and small error.

[0098] In some embodiments, if there are multiple similar plot matching interval values, in step S103, when the number of to-be-evaluated segments of the evaluation index belonging to each similar plot matching interval value among the multiple evaluation indexes in the quality evaluation record is not greater than the first threshold, the quality level of the short video is determined according to the deviation between the quality evaluation record and the first short video reference quality index value. In this way, the quality level can be determined by preferentially using the similar plot matching interval value and the first short video reference quality index value as an auxiliary means, which is conducive to improving efficiency and accuracy.

[0099] In some implementations, the deviation between the quality evaluation record and the first short video reference quality index value may be, but is not limited to, the Euclidean deviation, the deviation, etc. In some examples, the deviation between the quality evaluation record and the first short video reference quality index value may be the deviation, which can reduce the computational cost and eliminate errors caused by taking approximate values ​​in operations such as square root.

[0100] In some implementation operations, determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality index value in step S103 may include: step a1, when the absolute value of the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determining the quality level to be the character quality level; step a2, when the absolute value of the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, determining the quality level to be the script quality level; the first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level. Thus, the deviation between the quality evaluation record and the first short video reference quality index value can be determined by a simple algorithm such as averaging and deviation, with low computational complexity and small error, which can further reduce hardware costs and improve the accuracy of quality level evaluation.

[0101] In some implementation operations, when the accuracy requirement for quality level evaluation is relatively low, after estimating the deviation of the quality evaluation record from the first short video reference quality index value, the quality level corresponding to the first short video reference quality index value with the smallest deviation can be determined as the quality level of the short video. This implementation operation has lower computational complexity and relatively low hardware overhead.

[0102] In at least some embodiments, the similar plot matching interval value and the first short video reference quality index value can be obtained synchronously through various applicable algorithms. For example, a classification or other unsupervised learning algorithm can be used to determine the matching interval value of the short video and the reference quality index value of the short video. For another example, the similar plot matching interval value and the first short video reference quality index value can be determined by information analysis and statistical related algorithms. The specific algorithm for determining the similar plot matching interval value and the first short video reference quality index value is not limited in the embodiments of the present application.

[0103] In some implementation operations, the similar plot matching interval value and the first short video reference quality index value can be obtained based on the quality evaluation record library of the short video itself, and the quality evaluation record library contains the quality evaluation records of the short video to be evaluated segments obtained in advance. In an implementation operation, the similar plot matching interval value and the first short video reference quality index value can be determined by grading the existing quality evaluation records in the short video quality evaluation record library. Specifically, the similar plot matching interval value can be obtained by the following operations: step b1, grading based on the quality evaluation record library of the short video to be evaluated segments to obtain a short video segment set corresponding to the quality level, the quality evaluation record library of the short video to be evaluated segments contains the quality evaluation records of the short video to be evaluated segments obtained in advance; step b2, information association filtering of the short video segment set to determine whether the information association of the short video segment set belongs to positive association or non-positive association; step b3, according to the information association of the short video segment set, determine the similar plot matching interval value corresponding to the quality level. In this implementation operation, the short video reference quality index value of the short video segment set corresponding to the quality level is the first short video reference quality index value corresponding to the similar plot matching interval value, which can be determined in the grading of step b1. In this way, by grading based on the short video's own quality evaluation record library to obtain the short video's similar plot matching interval value and the first short video reference quality index value, not only the similar plot matching interval value and the first short video reference quality index value are set according to the specific type of the short video, and then the quality level evaluation is personalized or customized, the accuracy of the quality level evaluation is improved, and missed identification or misidentification can be reduced.

[0104] In some implementation operations, information association filtering of a short video clip set may be achieved through the following steps: step c1, dividing the quality evaluation records in the short video clip set into video frame nodes to obtain multiple video frame nodes, each of the multiple video frame nodes having the same number of video frames; step c2, taking the video frame node with the largest number of video frames among the multiple video frame nodes as the mean interval of the short video clip set, when the number of video frames of the short video clip set in the first interval is equal to the number of video frames of the short video clip set in the second interval, the information association of the short video clip set is positive association; when the number of video frames of the short video clip set in the first interval is not equal to the number of video frames of the short video clip set in the second interval, the information association of the short video clip set is non-positive association; wherein the first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval. In this implementation, the interval video frame node division and information counting, as well as the comparison of the number of information on the left and right halves of the mean interval are used to determine whether the set of short video clips is positively correlated or non-positively correlated. This has the effects of low complexity and accurate information correlation filtering, which can further reduce hardware costs and improve the accuracy of short video quality level evaluation.

[0105] The film and television drama operation of the same short video in different periods (for example, different age groups) may be quite different. For similar situations or other situations, in at least some embodiments, the similar plot matching interval value and the first short video reference quality index value can be adjusted in real time as needed, so that the similar plot matching interval value and the first short video reference quality index value can change dynamically with the change of the short video itself, thereby improving the accuracy of the quality level evaluation of the same short video in different periods, that is, achieving the quality level evaluation that varies from time to time.

[0106] In some implementation operations, the process of adjusting the similar plot matching interval value and the first short video reference quality index value may include: updating the quality evaluation record library of the short video to be evaluated according to the newly collected quality evaluation record, and re-determining the similar plot matching interval value using the unclassified quality evaluation record in the updated quality evaluation record library. Here, the unclassified quality evaluation record refers to the quality evaluation record in the short video segment set that is not classified into each quality level, that is, the quality evaluation record of the similar plot matching interval value that does not belong to the short video. Here, the algorithm used to re-determine the similar plot matching interval value and the first short video reference quality index value is the same as the above-mentioned related algorithm, and will not be repeated. In this embodiment, the unclassified quality evaluation record is used to realize the adjustment of the similar plot matching interval value and the first short video reference quality index value, without the need to repeatedly match all the quality evaluation records of the short video, and the update of the similar plot matching interval value and the first short video reference quality index value can be completed with less calculation amount and lower calculation complexity, so as to realize the dynamic adjustment of the similar plot matching interval value and the first short video reference quality index value, that is, the quality level evaluation can be realized from time to time at a lower hardware cost.

[0107] In some implementation operations, the quality evaluation record of the to-be-evaluated segment of the short video can be stored in the quality evaluation record library of the to-be-evaluated segment of the short video in the form of a time information sequence. The quality evaluation record library can be read and written by verifying preset information. The preset information may include any information related to identifying a specific short video. The present application does not limit the type of preset information. In addition, the matching interval value of the short video and the reference quality index value of the short video can also be stored in the quality evaluation record library of the to-be-evaluated segment of the short video. In this way, the similar plot matching interval value of the short video and the reference quality index value of the first short video can be queried in real time through the preset information of the short video, and the quality evaluation record of the to-be-evaluated segment of the short video can be updated, thereby realizing the evaluation of the quality level of the short video at any time and anywhere, without being restricted by the playback status, region, equipment, short video and other usage conditions.

[0108] In the embodiment of the present application, the quality evaluation record library can be updated through various applicable operations.

[0109] In some implementation operations, each time a new quality evaluation record is collected, the new quality evaluation record can be directly added to the quality evaluation record library of the short video segment to be evaluated. This exemplary implementation operation is conducive to quickly obtaining the quality evaluation record library of the short video segment to be evaluated. This exemplary implementation operation is applicable when the attribution of the quality evaluation record of the short video segment to be evaluated is low or the quality evaluation record library of the short video segment to be evaluated is initially created.

[0110] In some implementation operations, when the attribution of the quality evaluation record of the to-be-evaluated segment of the short video is greater than or equal to the third threshold, the quality evaluation record of the to-be-evaluated segment whose number of evaluation indicators belonging to the similar plot matching interval value is not greater than the first threshold is added to the quality evaluation record library of the to-be-evaluated segment of the short video. Therefore, when the similar plot matching interval value of the short video and the reference quality indicator value of the first short video are relatively stable, the judgment strategy for adding quality evaluation records to the information library can be used to reduce invalid or duplicate information, avoid unnecessary repeated calculations, and improve the information completeness of the short video quality evaluation record library. It is possible to reduce redundant information in the quality evaluation record library, reduce the amount of information calculation and the complexity of calculation while continuously improving the short video segments.

[0111] In the above implementation, the attribution degree of the quality evaluation record refers to: the proportion of quality evaluation records belonging to the similar plot matching interval value to all quality evaluation records, and the quality evaluation record belonging to the similar plot matching interval value refers to the quality evaluation record whose number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is greater than the first threshold. In practical applications, the attribution degree of the quality evaluation record of the to-be-evaluated segment of the short video can be calculated in real time by counting the quality evaluation records belonging to the similar plot matching interval value.

[0112] In practical applications, the specific value of the third threshold can be pre-associated, dynamically determined, or selected in response to the short video operation based on one or more factors such as the needs of specific application scenarios, the requirements of evaluation accuracy, relevant regulations of accurate film and television dramas, the needs of short videos, the conditions of short videos, etc. In some embodiments, the third threshold can be an empirical value, a function value of a predefined function with the above factors as variables, etc. The third threshold can be a fixed value, an interval value, or a variable value.

[0113] In at least some embodiments, the calculation and update of the similar plot matching interval value and the first short video reference quality index value can be implemented locally on the short video evaluation terminal, or can be performed through a cloud server. In some implementation operations, the calculation and update of the similar plot matching interval value and the first short video reference quality index value can be obtained through a cloud server and provided to the short video evaluation terminal. In this implementation operation, the calculation of the similar plot matching interval value and the first short video reference quality index value is performed on the cloud server, and the evaluation of the quality level is performed locally on the short video evaluation terminal.

[0114] In at least some embodiments, the method of the embodiment of the present application may further include a quality level control step. After the short video is evaluated to be at the script quality level, the quality level of the short video may be controlled through a single-operation or multi-operation evaluation instruction, thereby timely and effectively improving the script quality level of the short video.

[0115] In some implementation operations, the above method of the embodiment of the present application may also include:

[0116] Step S104, when the duration of the script quality level reaches a second threshold, one or more control strategies including the following operations are used to control the quality level: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation.

[0117] In this implementation, the control strategy is triggered after a certain period of time, which can not only achieve timely and effective intervention in the script quality level, but also avoid false triggering of the control strategy and improve the short video experience.

[0118] In practical applications, the specific value of the second threshold can be pre-associated, dynamically determined, or selected in response to the short video operation based on one or more factors such as the needs of specific application scenarios, the requirements of evaluation accuracy, relevant regulations for accurate film and television dramas, the needs of short videos, the conditions of short videos, etc. In some embodiments, the second threshold can be an empirical value, a function value of a predefined function with the above factors as variables, etc. The second threshold can be a fixed value, an interval value, or a variable value. For example, the second threshold can be 5 seconds, that is, when the duration of the quality level of the short video being the script quality level is greater than or equal to 5 seconds, the control strategy for the quality level of the short video is triggered.

[0119] In some implementation operations, the control strategy can be determined according to the script order of the short video, and the script order of the short video can be determined according to the quality evaluation record of the to-be-evaluated segment of the short video and the similar plot matching interval value corresponding to its quality level. For different script sequences, the script order of the short video can be re-determined after different cycles. That is, the first script order of the short video is re-determined after the first cycle, and the first cycle corresponds to the first script order. In this way, the state control of different script qualities using appropriate operations can significantly improve the effect of script quality level intervention.

[0120] Here, the first script sequence generally refers to a single script sequence, and the first period generally refers to a period corresponding to or associated with the single script sequence.

[0121] In some examples, the similar plot matching interval value of the short video can be divided into multiple sub-intervals, each sub-interval corresponds to a different script sequence, and each script sequence is associated with a different control strategy. The script sequence corresponding to the sub-interval of the similar plot matching interval value of the quality evaluation record of the short video to be evaluated is the script sequence of the short video, and the control strategy corresponding to the script sequence can be used to control the quality level of the short video.

[0122] In some implementations, the operations in the control strategy can be randomly selected. That is, a single-operation control strategy or a multi-operation control strategy can be adopted, and the specific operation adopted during control can be randomly selected, which can effectively improve the tolerance of the control strategy.

[0123] The quality evaluation method for short video similarity in the embodiment of the present application collects quality evaluation records of the to-be-evaluated segments of the short video, and directly determines the quality level of the short video by judging the number of to-be-evaluated segments that belong to the matching interval values ​​among the multiple evaluation indicators of the quality evaluation records. The method has small computational complexity and is easy to implement. It can accurately, efficiently and in real time evaluate the quality level of the short video with low computational complexity, while improving the short video experience and reducing the application cost.

[0124] The quality evaluation method for short video similarity in the embodiment of the present application is highly reusable and can be used to evaluate various states of short videos, film and television dramas, and short videos. The same method and device can be reused to evaluate brightness state, fatigue state, safety state, and film and television drama in-loop state, and can also be used to assist in accurate quality evaluation and short video safety evaluation.

[0125] Figure 2 The exemplary structure of the quality evaluation device for short video similarity provided by the embodiment of the present application is shown. Figure 2 An exemplary quality evaluation device for short video similarity in the application embodiment shown may include:

[0126] A collection module 21, used for collecting quality evaluation records of the short video segments to be evaluated, where the quality evaluation records include multiple evaluation indicators;

[0127] The determination module 22 is used to determine the quality level corresponding to the similar plot matching interval value as the quality level of the short video when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among multiple evaluation indexes is greater than a first threshold.

[0128] The quality evaluation device for short video similarity in the embodiment of the present application collects quality evaluation records of the to-be-evaluated segments of the short video, and can directly determine the quality level of the short video by judging the number of to-be-evaluated segments that belong to matching interval values ​​among multiple evaluation indicators in the quality evaluation records. It has a small amount of computation and is easy to implement. It can accurately, efficiently and in real time evaluate the quality level of the short video with low computational complexity, while improving the short video experience and reducing application costs.

[0129] In some implementations, multiple evaluation indicators are used to indicate advertising ratio and / or performance indicators. In other words, the indicators corresponding to the evaluation indicators include advertising ratio indicators and / or performance indicators. Thus, the impact of the short video on the playback volume can be comprehensively and accurately perceived through multiple indicators, thereby efficiently and accurately determining the quality level of the short video.

[0130] In some implementations, the quality level includes the character quality level and the script quality level, and the similar plot matching interval value corresponds to the character quality level or the script quality level. Therefore, the quality level of the short video can be directly determined as to be matched or the script through the two types of matching interval values ​​of the short video, which reduces the computational complexity and reduces misidentification or missed identification.

[0131] In some implementations, each type of matching interval value, i.e., each similar plot matching interval value, may include a matching interval value corresponding to an indicator corresponding to the evaluation indicator. Thus, the quality level of the short video can be determined by comprehensively considering various features that affect the playback volume during the playback of the short video.

[0132] In some implementation operations, the quality assessment device for short video similarity in the embodiment of the present application may further include:

[0133] Calculation module 23. The similar plot matching interval value is determined by calculation module 23, and calculation module 23 can obtain the similar plot matching interval value by the following steps: grading based on the quality evaluation record library of the short video to be evaluated to obtain a short video segment set corresponding to the quality level, and the quality evaluation record library of the short video to be evaluated contains the quality evaluation records of the short video to be evaluated segments obtained in advance; performing information association filtering on the short video segment set to determine whether the information association of the short video segment set belongs to positive association or non-positive association; according to the information association of the short video segment set, determining the similar plot matching interval value corresponding to the quality level. In this way, the matching interval value and the short video reference quality index value are set according to the specific type of the short video, while reducing missed recognition or misrecognition.

[0134] In some implementation operations, the calculation module 23 can also associate and filter the short video clip set through the following steps: divide the quality evaluation records in the short video clip set into video frame nodes to obtain multiple video frame nodes, and the number of video frames of each video frame node in the multiple video frame nodes is the same; the video frame node with the largest number of video frames in the multiple video frame nodes is used as the mean interval of the short video clip set; when the number of video frames of the short video clip set in the first interval is equal to the number of video frames of the short video clip set in the second interval, the information association of the short video clip set is positive association; when the number of video frames of the short video clip set in the first interval is not equal to the number of video frames of the short video clip set in the second interval, the information association of the short video clip set is non-positive association; wherein the first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval. Therefore, the quality evaluation device for short video similarity in the embodiment of the present application determines whether a set of short video clips is positively correlated or non-positively correlated by dividing the interval video frame nodes and counting the information, and comparing the number of information on the left and right halves of the mean interval. It has the effects of low complexity and accurate information correlation filtering, which can further reduce hardware costs and improve the accuracy of short video quality level evaluation.

[0135] In some implementation operations, the determination module 22 can also be used to determine the quality level of the short video based on the deviation between the quality evaluation record and the first short video reference quality indicator value when the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value among the multiple evaluation indicators of the quality evaluation record is not greater than the first threshold.

[0136] In some embodiments, the first short video reference quality index value may include two or more, and each first short video reference quality index value corresponds to a quality level. The determination module 22 is specifically used to determine the quality level corresponding to the first short video reference quality index value that is closer to the quality evaluation record among the two first short video reference quality index values ​​as the quality level of the short video; or, among three or more first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value that is closest to the quality evaluation record as the quality level of the short video. Thus, the deviation of the quality evaluation record of the short video to be evaluated can assist in completing the evaluation of the short video quality level, and achieve real-time, efficient and accurate evaluation of the short video quality level with low computational complexity.

[0137] In some implementation operations, the determination module 22 can be specifically used to: when the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determine the quality level as the character quality level; when the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, determine the quality level as the script quality level; the first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level. In this way, the short video quality level evaluation can be determined based on the deviation with lower computational complexity.

[0138] In some implementation operations, the quality assessment device for short video similarity in the embodiment of the present application may further include:

[0139] The information library updating module 24 is used to add the quality evaluation records of the to-be-evaluated segments of the short video whose number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is not greater than the first threshold to the quality evaluation record library of the to-be-evaluated segments of the short video when the attribution degree of the quality evaluation record of the to-be-evaluated segments of the short video is greater than or equal to the third threshold. In this way, the redundant information entering the quality evaluation record library can be significantly reduced, and the amount of information calculation and the complexity of calculation can be reduced.

[0140] In some implementation operations, the quality assessment device for short video similarity in the embodiment of the present application may further include:

[0141] The control module 25 is used to control the quality level by using one or more control strategies including the following operations when the duration of the quality level being the script quality level is greater than or equal to the second threshold: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation. Thus, timely and effective intervention in the quality level of the short video script can be achieved, while avoiding false triggering of the control strategy.

[0142] In some implementations, the control strategy can be determined according to the script sequence of the short video, and the script sequence can be determined according to the quality evaluation record of the short video clip to be evaluated and the similar plot matching interval value corresponding to the quality level. In this way, using appropriate operations for different script qualities to perform state control can significantly improve the effect of script quality level intervention.

[0143] In some implementations, the control module 25 may also be used to re-determine the first script sequence of the short video after the first cycle, and the first cycle corresponds to the first script sequence. In this way, re-evaluating the script quality in different cycles is more in line with the actual situation of improving the quality level of the short video, and can significantly improve the effect of script quality intervention.

[0144] In some implementations, the operations in the control strategy may be randomly selected by the control module 25. In this way, the tolerance of the control strategy can be effectively improved.

[0145] Some exemplary specific implementations of the embodiments of the present application are described in detail below. The following implementations may be arbitrarily combined with each other, and the various exemplary implementation operations of each implementation may also be arbitrarily combined as needed. It should be noted that the following implementations and any combination of their exemplary implementation operations are only examples of the above-mentioned methods or devices of the embodiments of the present application, and are not intended to limit the embodiments of the present application.

[0146] Figure 3 An exemplary process of quality level evaluation in the second embodiment of the present application is shown. The exemplary process may include the following steps:

[0147] Step S301: During the playback of the short video, the short video evaluation terminal device installed on the short video collects preset information of the short video and numerical evaluation instructions of various advertising ratio indicators and performance indicators;

[0148] Step S302: the short video evaluation terminal device performs local pre-matching on various numerical evaluation instructions of the short video to obtain a quality evaluation record of the short video at the current moment;

[0149] Here, the quality evaluation record includes an evaluation index corresponding to the indicator one by one, and the evaluation index is a normalized value, which is obtained by pre-matching the numerical evaluation instruction of the corresponding indicator, and the normalized value is between [0, 1].

[0150] Step S303: the short video evaluation terminal device locally searches for a matching interval value and a short video reference quality index value that match the preset information of the short video;

[0151] Step S304, the short video evaluation terminal device determines whether the quality level of the short video at the current moment meets the high-quality script according to the matching interval value of the short video, the reference quality index value of the short video and the quality evaluation record at the current moment;

[0152] Step S305: If it meets the requirements, the short video evaluation terminal device controls the script quality level of the short video.

[0153] Step S306, determining whether the short video evaluation terminal device is currently online;

[0154] Step S307: If offline, the short video evaluation terminal device locally stores the quality evaluation record of the short video at the current moment.

[0155] Step S308, if online, the short video evaluation terminal device uploads the quality evaluation record of the short video to be evaluated and its preset information stored locally to the cloud server through its front-end communication module.

[0156] In one implementation operation, the local pre-matching in step S302 may include information pre-matching (e.g., information cleaning), unified sampling, information normalization, etc. The corresponding quality evaluation record includes evaluation indicators corresponding to the indicators one by one, and each evaluation indicator includes a normalized value, and the value of the normalized value is obtained by performing local pre-matching on the numerical evaluation instruction of the corresponding indicator, and the normalized value is between [0, 1].

[0157] In one implementation operation, step S308 may include steps corresponding to information verification and information library. Here, information verification may include verifying whether the preset information of the short video is legal (for example, whether there is a legal film and television drama certificate, etc.) or exists. The steps corresponding to the information library may include: finding a quality evaluation record library corresponding to the preset information of the short video. After finding the quality evaluation record library of the to-be-evaluated segment of the short video, the quality evaluation record of the to-be-evaluated segment of the short video may be stored in its quality evaluation record library.

[0158] In one implementation operation, the quality evaluation record library of the short video clips to be evaluated can store the quality evaluation records of the short video clips to be evaluated in the form of a time information sequence, and each record in the time information sequence contains the time information of a quality evaluation record and the normalized value of each film and television drama indicator corresponding to the quality evaluation record.

[0159] In this embodiment, multi-dimensional quality evaluation records are obtained through numerical evaluation instructions of multiple indicators, and the quality level of short videos is determined in real time based on the multi-dimensional quality evaluation records. It has low requirements on the collection environment, a small number of video frames, and low requirements on computing power and bandwidth for information matching and transmission. The short video experience is better and the cost of use is lower.

[0160] This embodiment uses a multi-dimensional information grading method, a non-classification method, which can avoid the problem of linear inseparability when the information dimension is high, and the grading record does not directly act on the classification of intermediate features such as brightness and color, but directly characterizes the quality level of the short video through the graded short video clip set. In this way, the quality level of the short video can be directly determined by using the matching interval value of the graded short video clip set and the short video reference quality index value, which reduces the calculation complexity while reducing misidentification or missed identification, thereby improving the accuracy of short video quality level evaluation.

[0161] Figure 4 An exemplary flow chart of filtering information association is shown. The exemplary flow chart of filtering information association in this embodiment may include the following steps:

[0162] Step S401, slicing all quality evaluation records in the short video clip set by equal intervals of video frame numbers to obtain multiple slices with the same number of video frames;

[0163] Step S402, quickly sorting the slices obtained by dividing the video frame nodes according to the number of video frames, and finding the slice with the largest number of video frames;

[0164] Step S403, determining whether the number of video frames of the short video segments concentrated in the first interval is equal to the number of video frames in the second interval, if yes, proceed to step S404, otherwise, jump to step S405;

[0165] Step S404, determining that the information association of the short video clip set is a positive association;

[0166] Step S405: determine that the information association of the short video clip set is a non-positive association.

[0167] If the evaluation index in the quality evaluation record has been normalized and matched in the local pre-matching process, the value of each evaluation index in the quality evaluation record will be associated between [0,1], and there is no need to perform normalization matching before step S401 in this process. If the value of the evaluation index is not a normalized value associated between [0,1], it is necessary to add a step of normalizing each quality evaluation record in the short video clip set before step S401.

[0168] In this embodiment, information association is filtered by dividing information by video frame nodes, and different operations are used to calculate matching interval values ​​for different types of information associations, thereby obtaining various types of matching interval values ​​that can accurately characterize the boundaries of various quality levels of short videos, providing an accurate basis for accurate evaluation of the quality level of short videos.

[0169] This embodiment describes in detail an exemplary implementation operation of controlling the quality level of short videos.

[0170] In some implementations, the triggering process of controlling the quality level of the short video may include: determining whether the duration of the short video at the script quality level has reached a preset second threshold, and if so, triggering the control of the quality level of the short video, otherwise there is no need to intervene in the quality level of the short video. This implementation controls the quality level of the short video only after it has been at a level for a period of time, which can avoid false triggering of the control strategy and improve the short video experience.

[0171] Generally speaking, the higher the quality of the script, the longer it takes to control the status of the short video. Therefore, in at least some embodiments, a stepped duration threshold can be used, that is, the first duration is less than the second duration, and the second duration is less than the third duration. The specific values ​​of the first duration, the second duration, and the third duration can be empirical values ​​or dynamically changing values, which can be adjusted according to the characteristics of the short video.

[0172] In some implementation operations, the process of determining the script quality can be: pre-dividing each matching interval value in a class of matching interval values ​​corresponding to the script quality level into multiple sub-intervals, and these multiple sub-intervals represent multiple levels of script quality of different levels. By judging which sub-interval of the corresponding matching interval value the mean of the evaluation index in all quality evaluation records in the current control period belongs to, the level of the current script quality of the short video can be determined. For example, each matching interval value in a class of matching interval values ​​corresponding to the script quality level is equally divided into three sub-intervals, and these three sub-intervals represent the high, medium and low levels of script quality, respectively. If the mean of the same evaluation index of all quality evaluation records in the first duration belongs to the corresponding sub-interval, then the current script quality of the short video is determined to be the level corresponding to the sub-interval.

[0173] In some implementations, the operations in various control strategies are randomly determined. That is, one or more of the following four operations are randomly selected: emotional lines, rational plot, low advertising ratio, and clear character association. By controlling the quality level of short videos through randomly selected operations, it is possible to effectively prevent the short videos from developing tolerance due to long-term use of the same or more evaluation instructions, and significantly improve the tolerance of the control strategy.

[0174] The control method of this embodiment has control resistance.

[0175] In this embodiment, when the degree of attribution of a newly generated quality evaluation record is greater than a certain value, the strategy for the quality evaluation record to enter the quality evaluation record library is changed from the original operation of directly sending the quality evaluation record into the quality evaluation record library to "only selecting points that do not have a matching interval value dimension greater than half to enter the information library". In this way, while continuously improving the information fragments of the short video status, the redundant information entering the quality evaluation record library can be reduced, thereby reducing the amount of information calculations and reducing the complexity of calculations.

[0176] In the several embodiments provided in the present application, it should be understood that the disclosed devices, apparatuses and methods can be implemented through other operations. For example, the device embodiments described above are only schematic, for example, the division of units is only a logical function division, and there may be other division operations in actual implementation, such as multiple units or components can be combined or integrated into another device, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0177] In addition, each functional unit in each embodiment of the present application can be integrated into a matching unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit. The above is only an implementation method of the present application, and does not limit the patent range of the present application. Any equivalent structure or equivalent process transformation made using the content of the specification and drawings of this application, or directly or indirectly used in other related technical fields, is also included in the patent protection range of the present application.

[0178] The specific implementation methods of the invention are described in detail above, but they are only examples, and the present application is not limited to the specific implementation methods described above. For those skilled in the art, any equivalent modification or substitution of the invention is also within the scope of the present application, and therefore, the equalization, modification, and improvement made without departing from the spirit and principle of the present application should be included in the scope of the present application.

Claims

1. A quality evaluation method for short video similarity, characterized in that: include: Collecting a quality evaluation record of a to-be-evaluated segment of a short video, wherein the quality evaluation record includes a plurality of evaluation indicators; When the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among the multiple evaluation indexes is greater than a first threshold, the quality level corresponding to the similar plot matching interval value is determined as the quality level of the short video; The multiple evaluation indicators are used to indicate advertising ratios and / or performance indicators; The quality level includes a character quality level or a script quality level, and the similar plot matching interval value corresponds to the character quality level or the script quality level; The similar plot matching interval value is obtained by the following operation: Based on the quality evaluation record library of the short video segments to be evaluated, grading is performed to obtain a set of short video segments corresponding to the quality level, wherein the quality evaluation record library of the short video segments to be evaluated contains the quality evaluation records of the short video segments to be evaluated obtained in advance; Performing information association filtering on the short video clip set to determine whether the information association of the short video clip set is a positive association or a non-positive association; According to the information association of the short video segment set, a similar plot matching interval value corresponding to the quality level is determined.

2. The quality evaluation method for short video similarity according to claim 1, characterized in that: The information association filtering of the short video clip set specifically includes: Dividing the quality evaluation records in the short video clip set into video frame nodes to obtain a plurality of video frame nodes; Taking the video frame node with the largest number of video frames among the multiple video frame nodes as the mean interval of the short video segment set; When the number of video frames of the short video segment set in the first interval is equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a positive association; When the number of video frames of the short video segment set in the first interval is not equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a non-positive association; The first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval.

3. The quality evaluation method for short video similarity according to claim 2 is characterized in that: The method further includes: when, among the multiple evaluation indicators, the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value is not greater than the first threshold, determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality indicator value; The determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality index value includes: Of the two first short video reference quality index values, the quality level corresponding to the first short video reference quality index value closer to the quality evaluation record is determined as the quality level of the short video; or, Among three or more first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value closest to the quality evaluation record as the quality level of the short video; The determining the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality index value includes: When the absolute value of the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determining that the quality level is the character quality level; When the absolute value of the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, the quality level is determined to be the script quality level; The first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level.

4. The quality evaluation method for short video similarity according to claim 3 is characterized in that: Also includes: When the attribution degree of the quality evaluation record of the to-be-evaluated segment of the short video is greater than or equal to the third threshold, the quality evaluation record whose number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value is not greater than the first threshold is added to the quality evaluation record library of the to-be-evaluated segment of the short video; When the duration of the quality level being the script quality level is greater than or equal to the second threshold, the quality level is controlled by using one or more control strategies including the following operations: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation; The control strategy is determined according to the script sequence of the short video, and the script sequence is determined according to the quality evaluation record and the similar plot matching interval value corresponding to the quality level; re-determining a first script sequence of the short videos after a first period, the first period corresponding to the first script sequence; The operations in the control strategy are selected randomly.

5. A quality evaluation device for short video similarity, characterized in that: include: A collection module, used for collecting quality evaluation records of the short video clips to be evaluated, wherein the quality evaluation records include multiple evaluation indicators; A determination module, configured to determine, when the number of to-be-evaluated segments of the evaluation index belonging to the similar plot matching interval value among the multiple evaluation indexes is greater than a first threshold, the quality level corresponding to the similar plot matching interval value as the quality level of the short video; The multiple evaluation indicators are used to indicate advertising ratios and / or performance indicators; The quality level includes a character quality level and a script quality level, and the similar plot matching interval value corresponds to the character quality level or the script quality level; A calculation module, wherein the calculation module is used for: Based on the quality evaluation record library of the short video segments to be evaluated, grading is performed to obtain a set of short video segments corresponding to the quality level, wherein the quality evaluation record library of the short video segments to be evaluated contains the quality evaluation records of the short video segments to be evaluated obtained in advance; Performing information association filtering on the short video clip set to determine whether the information association of the short video clip set is a positive association or a non-positive association; According to the information association of the short video segment set, a similar plot matching interval value corresponding to the quality level is determined.

6. The quality assessment device for short video similarity according to claim 5, characterized in that: The calculation module is specifically used for: Dividing the quality evaluation records in the short video clip set into video frame nodes to obtain a plurality of video frame nodes; Taking the video frame node with the largest number of video frames among the multiple video frame nodes as the mean interval of the short video segment set; When the number of video frames of the short video segment set in the first interval is equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a positive association; When the number of video frames of the short video segment set in the first interval is not equal to the number of video frames of the short video segment set in the second interval, the information association of the short video segment set is a non-positive association; The first interval and the second interval are obtained by dividing the information interval of the short video clip set according to the mean interval.

7. The device for evaluating the quality of short video similarity according to claim 6, characterized in that: The determination module is further configured to determine the quality level of the short video according to the deviation between the quality evaluation record and the first short video reference quality indicator value when the number of to-be-evaluated segments of the evaluation indicator belonging to the similar plot matching interval value among the multiple evaluation indicators is not greater than the first threshold value; The determination module is specifically used for: Of the two first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value closer to the quality evaluation record as the quality level of the short video; or, Among three or more first short video reference quality index values, determine the quality level corresponding to the first short video reference quality index value closest to the quality evaluation record as the quality level of the short video; The determination module is specifically used for: When the absolute value of the deviation between the first deviation and the second deviation is greater than the average of the first deviation and the second deviation, determining that the quality level is the character quality level; When the absolute value of the deviation between the first deviation and the second deviation is not greater than the average of the first deviation and the second deviation, the quality level is determined to be the script quality level; The first deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the character quality level, and the second deviation is the deviation between the quality evaluation record and the first short video reference quality index value corresponding to the script quality level.

8. The device for evaluating the quality of short video similarity according to claim 7, characterized in that: Also includes: The information library updating module is used to: when the attribution degree of the quality evaluation record of the to-be-evaluated segment of the short video is greater than or equal to a third threshold, add the quality evaluation record of the to-be-evaluated segment whose number of the evaluation index belonging to the similar plot matching interval value is not greater than the first threshold to the quality evaluation record library of the to-be-evaluated segment of the short video; A control module, for controlling the quality level by adopting one or more control strategies including the following operations when the duration of the quality level being the script quality level is greater than or equal to a second threshold: plot rationalization operation, dialogue emotionalization operation, advertisement low ratio operation, character association clarification operation; The control strategy is determined according to the script sequence of the short video, and the script sequence is determined according to the quality evaluation record and the similar plot matching interval value corresponding to the quality level; The control module is further used to re-determine the first script sequence of the short video after the first period, and the first period corresponds to the first script sequence; The operations in the control strategy are randomly selected by the control module.

Images