A method and device for analyzing live broadcast effect

Abstract

The present invention discloses a method and device for analyzing live broadcast effect, which relates to the field of data intelligent analysis technology, and solves the technical problems of focusing on simple headcount statistics, lacking in-depth mining and integration of multiple key indicators in live broadcast, and inaccurate analysis of live broadcast effect. The present invention not only counts headcount information such as the number of viewers, the peak number of online viewers, and the length of viewing time, but also deeply analyzes the interactive situations such as the number of bullet screens, the content, and the number of recommendations, and comprehensively outlines the overall picture of the live broadcast. At the same time, repeated bullet screens are eliminated, and the emotional tendency of the bullet screens is accurately identified by combining machine learning algorithms and emotional dictionaries, reducing redundant data interference, and making the data analysis results closer to the real thoughts of the audience. The live broadcast effect index is synthesized through the headcount relationship index and the interactive relationship index, and the advantages and disadvantages of the same type of live broadcasts are screened horizontally. Then, the live broadcast dynamic comprehensive index is calculated in combination with the bullet screen refresh frequency and the headcount change value, and the live broadcast effect is fed back in real time and dynamically.

Description

Technical Field

[0001] The present invention relates to the technical field of intelligent data analysis, and in particular to a live broadcast effect analysis method and device. Background Art

[0002] With the rapid development of Internet technology, the live broadcast industry has shown explosive growth and has become the most influential form of content dissemination and social interaction.

[0003] The patent with publication number CN118133821A discloses a live broadcast effect analysis method, device, computer equipment and storage medium, the method comprising: obtaining multiple evaluation contents of the live broadcast process and traffic distribution information of the live broadcast process, and based on the traffic distribution information, identifying the live broadcast attention degree change information of the live broadcast process; performing word segmentation processing on each of the comment contents to obtain multiple word information of each comment content, and vectorizing each of the word information through a word vector generation model to obtain each independent word vector of each comment content; extracting the sentiment tendency information of each of the independent word vectors and the attitude information of each of the independent word vectors, and based on the sentiment tendency information of each of the independent word vectors and the attitude information of each of the independent word vectors, identifying the sentiment evaluation information of each comment content through a sentiment analysis network; based on the comment time of each of the comment contents, the sentiment evaluation information of each of the comment contents, and the live broadcast attention degree change information, identifying the live broadcast effect distribution information during the live broadcast process.

[0004] However, some existing methods for evaluating the effect of live broadcast are relatively crude, focusing on simple headcount and lacking in-depth exploration and integration of multiple key indicators in live broadcast, which further leads to inaccurate analysis of the effect of live broadcast. Summary of the invention

[0005] In view of the deficiencies in the prior art, the present invention provides a live broadcast effect analysis method and device, which solves the problem that the method and device focus on simple headcount, lack of in-depth mining and integration of multiple key indicators in live broadcast, and inaccurate analysis of live broadcast effects.

[0006] To achieve the above objectives, the present invention is implemented through the following technical solutions: a live broadcast effect analysis method, the method specifically comprising the following steps:

[0007] Step 1: Obtain live broadcast information and classify the live broadcast information into indicators to obtain the number of viewers and interaction indicators;

[0008] Step 2: Analyze the obtained number of people indicators, obtain the corresponding number of people analysis results by analyzing the number of viewers, the peak number of online viewers, and the viewing time separately, and calculate the number of people relationship indicators by comprehensively analyzing the results, and generate the number of people relationship indicator results;

[0009] Step 3: Analyze the obtained interaction indicators, obtain the interaction analysis results by analyzing the number of barrages, barrage content and recommendation times in the interaction indicators, and calculate the interaction relationship indicators based on the analysis results;

[0010] Step 4: Calculate the live broadcast effect index of the live broadcast based on the obtained number relationship index and interaction relationship index, and calculate the live broadcast effect index of the same type of live broadcasts at the same time, and screen the live broadcasts based on the live broadcast effect index, then calculate the live broadcast dynamic comprehensive index, and generate live broadcast effect information based on the live broadcast dynamic comprehensive index.

[0011] As a further solution of the present invention, the number of people index in step 1 represents the number of people information during the live broadcast, and the interaction index represents the interaction of the audience during the live broadcast.

[0012] As a further solution of the present invention, the specific method of generating the number of people relationship index result in step 2 is:

[0013] The live broadcast information is obtained, and the live broadcast duration is segmented with time t as a period, and the corresponding number of viewers in the time period t is obtained. Then the number of viewers is screened to obtain the real-time number of viewers. At the same time, the real-time number of viewers in different time periods t is obtained, and the average value of the real-time viewer data in the time period t is calculated to generate the real-time number of viewers. Then the peak number of online viewers corresponding to the live broadcast duration is obtained, and the number of viewers online at the same time in the time t and the average viewing time in this interval are calculated. In this way, the average viewing time corresponding to all time periods t is calculated;

[0014] Substitute the obtained average value of real-time viewing data, peak number of online viewers and average viewing time into the formula: population relationship index = (average value of real-time viewing data + peak number of online viewers + average viewing time) ÷ total live broadcast time, calculate the population relationship index, and generate the population relationship indicator result at the same time.

[0015] As a further solution of the present invention, the specific method of analyzing the interaction index in step 3 is:

[0016] The live broadcast information is obtained, and the total number of barrages and the corresponding barrage content within the time period t are obtained at the same time, and the number of barrages is filtered according to the barrage content to obtain the real-time barrage within the time period t as i, and i=1, 2, ..., j, where j represents the number of real-time barrages, and then the content of the real-time barrage i is identified;

[0017] The duplicate barrages existing in the real-time barrage i are filtered and denoted as a, and a=1, 2, ..., b, where b represents the number of duplicate barrages. At the same time, the comment users of the duplicate barrage a are obtained, and multiple groups of duplicate barrages corresponding to the same comment user are eliminated, and the unique duplicate barrage is retained. And so on, all the duplicate barrages are filtered to obtain a single duplicate barrage, and the single duplicate barrage is analyzed.

[0018] As a further solution of the present invention, the specific method of analyzing a single repeated bullet comment in step 3 is:

[0019] The content of the single repeated barrage is obtained, and the barrage content is identified based on the machine learning algorithm. At the same time, the barrage content is matched with the sentiment dictionary to obtain the corresponding matching result, and the barrage evaluation of the single repeated barrage is calculated according to the matching result. The barrage evaluation is classified into positive evaluation, negative evaluation and neutral evaluation. At the same time, the number of barrage evaluations of different categories is obtained and recorded as C1, C2 and C3 respectively;

[0020] Similarly, all real-time barrages i are evaluated and analyzed, and the corresponding numbers of different types are obtained, recorded as L1, L2 and L3. At the same time, the corresponding number proportions of different types of barrage evaluations are calculated and recorded as Z1, Z2 and Z3 respectively. Then, they are sorted from large to small according to the number proportion, and the corresponding number of recommendations during the live broadcast duration is obtained, and the number of recommendations is classified.

[0021] As a further solution of the present invention, the specific method of classifying the number of recommendations in step 3 is:

[0022] Obtain the commenting user corresponding to the recommended times, and at the same time obtain the real-time bullet comments corresponding to the commenting user, and classify the real-time bullet comments according to the bullet comment evaluation, then select the classified bullet comment evaluation corresponding to the largest number of bullet comment evaluations as the standard to classify the recommended times, and perform the same processing on all the recommended times in this way to obtain the recommended times classification result;

[0023] At the same time, the sum of the number of recommendations and the proportion of the same type of barrage evaluations is calculated and recorded as the sum of the values, and the largest value is selected as the interactive relationship indicator.

[0024] As a further solution of the present invention, the specific method of generating the live broadcast effect information in step 4 is:

[0025] The number of people relationship index is recorded as R, the interactive relationship index is recorded as H, and the sum of the two is calculated as the live broadcast effect index Q. Similarly, the same type of live broadcast corresponding to the same period is obtained, and the live broadcast effect index Qn of the same type of live broadcast is calculated, and n represents the number of the same type of live broadcast, where n=1, 2, ..., m, and m represents the number of the same type of live broadcast. At the same time, the number of the same type of live broadcast whose live broadcast effect index Q is greater than Qn is recorded as g;

[0026] Then obtain the barrage refresh frequency corresponding to the current live broadcast within time t, and obtain the change value of the number of people corresponding to the current live broadcast at time t, and then substitute the obtained barrage refresh frequency, number change value and live broadcast effect index Q into the formula , where LDI is the comprehensive index of live broadcast dynamics, ω1, ω2 and ω3 are the weight coefficients of barrage refresh frequency, number of people change value and live broadcast effect index respectively, DRF is the barrage refresh frequency, NCV is the number of people change value, and g is the number of live broadcasts of the same type obtained by screening;

[0027] The calculated live broadcast dynamic comprehensive index is compared with the effect judgment interval, and the specific value of the effect judgment interval is set by the operator according to the historical live broadcast data, and the corresponding live broadcast effect is determined based on the effect judgment interval to generate live broadcast effect information.

[0028] A live broadcast effect analysis device, comprising: a live broadcast data collection module, a number of people data analysis module, a bullet screen information analysis module, a comprehensive analysis module and an information output module;

[0029] The live broadcast data collection module is used to collect live broadcast information, transmit the number of people in the live broadcast information to the number of people data analysis module, and transmit the interaction index to the bullet screen information analysis module;

[0030] The crowd data analysis module is used to analyze the acquired crowd index, obtain the corresponding analysis results by analyzing the number of viewers, the peak number of online viewers, and the viewing time, and calculate the crowd relationship index by comprehensively analyzing the results, and generate the crowd relationship index results, and transmit the crowd relationship index results to the comprehensive analysis module;

[0031] The barrage information analysis module is used to analyze the obtained interactive indicators, and obtain the analysis results by analyzing the number of barrages, the content of barrages and the number of recommendations in the interactive indicators. The analysis results here include the analysis results of the number of barrages, the analysis results of the content of barrages and the analysis results of the number of recommendations, and calculate the interactive relationship indicators based on the analysis results, and transmit the interactive relationship indicators to the comprehensive analysis module at the same time;

[0032] A comprehensive analysis module, which is used to calculate the live broadcast effect index of the live broadcast according to the obtained number relationship index and interaction relationship index, and at the same time calculate the live broadcast effect index of the same type of live broadcasts, and screen the live broadcasts according to the live broadcast effect index, and then calculate the live broadcast dynamic comprehensive index, and generate live broadcast effect information according to the live broadcast dynamic comprehensive index, and at the same time transmit the live broadcast effect information to the information output module;

[0033] An information output module is used to display the acquired live broadcast effect information to the corresponding operator.

[0034] The present invention provides a method and device for analyzing live broadcast effects. Compared with the prior art, it has the following beneficial effects:

[0035] The present invention innovatively subdivides the live broadcast information into headcount indicators and interaction indicators. It not only counts the headcount information such as the number of viewers, the peak number of online viewers, and the viewing time, but also deeply analyzes the interactive situations such as the number of bullet screens, content, and number of recommendations, comprehensively outlines the overall picture of the live broadcast, accurately locates the advantages and disadvantages of the live broadcast, and provides a detailed basis for subsequent optimization. The live broadcast duration is flexibly segmented, and the average real-time number of viewers and the average viewing time of the audience in each time period are calculated in real time, and the dynamic changes of the live broadcast traffic are accurately captured. At the same time, repeated bullet screens are eliminated, and the emotional tendency of the bullet screens is accurately identified by combining machine learning algorithms and emotional dictionaries. Redundant data interference is reduced, and the data analysis results are closer to the real thoughts of the audience. The live broadcast effect index is synthesized by the headcount relationship index and the interaction relationship index, and the advantages and disadvantages of the same type of live broadcasts are screened horizontally. The live broadcast dynamic comprehensive index is calculated in combination with the bullet screen refresh frequency and the headcount change value, and the live broadcast effect is fed back in real time and dynamically. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 It is a step method diagram of the present invention;

[0037] Figure 2 The figure is a flowchart showing the functional modules of the device of the present invention. DETAILED DESCRIPTION

[0038] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0039] For example, see Figure 1 The present application provides a live broadcast effect analysis method, which specifically includes the following steps:

[0040] Step 1: Acquire the live broadcast information, and classify the live broadcast information into indicators to obtain the number of people indicator and interaction indicator. The number of people indicator represents the number of people during the live broadcast, while the interaction indicator represents the interaction of the audience during the live broadcast.

[0041] Step 2: Analyze the obtained population index, and obtain the corresponding population analysis results by separately analyzing the number of viewers, the peak number of online viewers, and the viewing time. The population analysis results here include the real-time number of viewers, the peak number of online viewers, and the average viewing time. At the same time, the population analysis results are comprehensively calculated to obtain the population relationship index, and the population relationship index results are generated.

[0042] The live broadcast information is obtained, and the live broadcast duration is segmented with time t as the cycle, and the corresponding number of viewers within the time period t is obtained. Then the number of viewers is screened to obtain the real-time number of viewers, and the real-time number of viewers here is the average real-time number of viewers within the time period t with the small cycle time u as the time division point. For example, if the time period t is set to one hour, one hour is divided into two small cycle time u. For example, if the small cycle time u is twenty minutes, it is further divided into four small cycle time periods, and then the real-time number of viewers in each small cycle time period is calculated. Finally, the average real-time number of viewers in the four small cycle time periods is calculated, and the real-time number of viewers in different time periods is calculated. The real-time number of viewers in period t is obtained, and the average value of the real-time number of viewers in time period t is calculated to generate the real-time number of viewers. Then the peak number of online viewers corresponding to the live broadcast duration is obtained, and the peak number of online viewers here is expressed as the maximum value of the real-time number of online viewers in the entire live broadcast. At the same time, the number of viewers online at the same time in time t and the average viewing time in this interval are calculated. For example, in the interval of 10-15 minutes of the live broadcast, there are 50 viewers online at the same time, and the average viewing time of these viewers in this interval is 4 minutes. Then the total simultaneous online viewing time in this interval is 50×4=200 minutes. The average viewing time corresponding to all time periods t is calculated in this way;

[0043] Substitute the obtained average value of real-time viewing data, peak number of online viewers and average viewing time into the formula: Population relationship index = (average value of real-time viewing data + peak number of online viewers + average viewing time) ÷ total live broadcast time to calculate the population relationship index. The above formula only takes numerical values ​​for calculation and generates the population relationship index result at the same time.

[0044] Assume that a live broadcast of popular science is set to last 90 minutes. During the live broadcast, the backend system records the number of real-time viewers every 10 minutes. After statistics and calculations, the average number of real-time viewers is 3,000. When the live broadcast is 45 minutes in, the traffic peaks, and the number of online people at the same time reaches 5,000, that is, the peak number of online people is 5,000. After the live broadcast ends, the audience's stay time is counted through the data analysis platform, and the average viewing time is 35 minutes.

[0045] The population relationship index is calculated according to the formula: population relationship index = (3000+5000+35) ÷ 90 ≈ 90.39.

[0046] Step 3: Analyze the obtained interaction indicators. The interaction analysis results are obtained by analyzing the number of barrages, the content of barrages, and the number of recommendations in the interaction indicators. The interaction analysis results here include the analysis results of the number of barrages, the analysis results of the content of barrages, and the analysis results of the number of recommendations. The interaction relationship indicators are calculated based on the interaction analysis results.

[0047] The live broadcast information is obtained, and the total number of barrages and the corresponding barrage content within the time period t are obtained, and the number of barrages is filtered according to the barrage content, and the specific filtering method is as follows:

[0048] Get the real-time bullet comments within the time period t, record them as i, and i=1, 2, ..., j, where j represents the number of real-time bullet comments. Then identify the content of the real-time bullet comment i, and filter the duplicate bullet comments in the real-time bullet comment i, record them as a, and a=1, 2, ..., b, where b represents the number of duplicate bullet comments. At the same time, get the comment users of the duplicate bullet comments a, and remove multiple groups of duplicate bullet comments corresponding to the same comment user, retain the unique duplicate bullet comment, and so on, filter all the duplicate bullet comments to obtain a single duplicate bullet comment.

[0049] Suppose a game live broadcast lasts for 2 hours, and the time period t is set to the entire live broadcast. After the live broadcast, statistics show that a total of 1,000 real-time barrages were generated, that is, j=1000. These barrages are varied, including players' discussions on game skills and praise or complaints about the host's operations. The text recognition algorithm is used to filter out barrages with repeated content. For example, the barrage "The host's operation is too good" was posted by 20 different users. After screening, it was marked as a repeated barrage, and b=20 at this time (only taking this type of repeated barrage as an example, multiple repeated situations were actually screened out). Then checking the users who posted these barrages "The host's operation is too good" it was found that user A posted it 3 times. According to the rules, only one barrage of this content posted by user A is retained, and the other two are removed.

[0050] Then, all the obtained single repeated barrages are analyzed, the barrage content of the single repeated barrage is obtained, and the barrage content is identified based on the machine learning algorithm, and the machine learning algorithm here includes naive Bayes and support vector machine, and the barrage content is matched with the sentiment dictionary to obtain the corresponding matching result, and the sentiment dictionary here is pre-created, and the corresponding matching result represents the same vocabulary as the content corresponding to the sentiment dictionary. For example, in entertainment live broadcasts, "great" and "like it very much" are positive words, and "boring" and "dislike" are negative words. If the barrage is "this show is really boring", it can be judged that this barrage is a negative evaluation through the sentiment dictionary, and the barrage evaluation of the single repeated barrage is calculated according to the matching result, and the barrage evaluation is classified into positive evaluation, negative evaluation and neutral evaluation. At the same time, the number of barrage evaluations of different categories is obtained and recorded as C1, C2 and C3, and the number here is the number corresponding to the repeated barrage;

[0051] Similarly, all real-time bullet comments i are evaluated and analyzed, and the number of different types is obtained and recorded as L1, L2 and L3. At the same time, the number of proportions corresponding to different types of bullet comments is calculated and recorded as Z1, Z2 and Z3 respectively. Then, they are sorted from large to small according to the number proportions, and the corresponding number of recommendations during the live broadcast duration is obtained. The number of recommendations here defaults to one user can only recommend once, and the number of recommendations is classified and processed, and the specific classification processing method is:

[0052] Obtain the commenting user corresponding to the number of recommendations, and at the same time obtain the real-time bullet comments corresponding to the commenting user, and classify the real-time bullet comments according to the bullet comment evaluation, then select the classified bullet comment evaluation corresponding to the largest number of bullet comment evaluations as the standard to classify the number of recommendations, and perform the same processing on all the number of recommendations in this way to obtain the classification result of the number of recommendations, and the classification result of the number of recommendations here is specifically expressed as positive evaluation recommendation, negative evaluation recommendation or neutral evaluation recommendation;

[0053] At the same time, the sum of the number of recommendations and the proportion of the number of similar barrage evaluations is calculated and recorded as the sum of the values. At the same time, the largest value is selected as the interactive relationship index. Here, the sum of positive evaluations, negative evaluations, and neutral evaluations is calculated separately.

[0054] Step 4: Calculate the live broadcast effect index of the live broadcast based on the obtained number relationship index and interaction relationship index, and calculate the live broadcast effect index of the same type of live broadcasts at the same time, and screen the live broadcasts based on the live broadcast effect index, then calculate the live broadcast dynamic comprehensive index, and generate live broadcast effect information based on the live broadcast dynamic comprehensive index.

[0055] The number relationship index is recorded as R, the interactive relationship index is recorded as H, and the sum of the two is calculated as the live broadcast effect index Q. Similarly, the corresponding live broadcasts of the same type in the same period are obtained. For example, during the global finals of a popular MOBA game, the official event live broadcast room is broadcasting the finals live from 2 pm to 6 pm on Saturday. The same type of live broadcasts include the personal event commentary live broadcasts of well-known game anchors and the corresponding game live broadcasts of the same type. The live broadcast effect index Qn of the same type of live broadcast is calculated, and n represents the number of the same type of live broadcast, where n=1, 2, ..., m, and m represents the number of the same type of live broadcasts. At the same time, the number of the same type of live broadcasts with a live broadcast effect index Q greater than Qn is screened and recorded as g;

[0056] Then, the barrage refresh frequency corresponding to the current live broadcast within time t is obtained, and the barrage refresh frequency is expressed as the ratio of the total amount of barrage refreshes within time t to the time. At the same time, the change value of the number of people corresponding to the current live broadcast at time t is obtained. Then, the obtained barrage refresh frequency, the change value of the number of people and the live broadcast effect index Q are substituted into the formula , where LDI is the comprehensive index of live broadcast dynamics, ω1, ω2 and ω3 are the weight coefficients of barrage refresh frequency, number of people change value and live broadcast effect index respectively, DRF is the barrage refresh frequency, NCV is the number of people change value, g is the number of live broadcasts of the same type obtained by screening. For example, when a game live broadcast is broadcast in real time, the barrage refresh frequency DRF=27 (27 barrages are refreshed per second), NCV=50, the live broadcast effect index Q=0.8, g=5, and the barrage refresh frequency, number of people change value and live broadcast effect index weight coefficients are 0.3, 0.2 and 0.5 respectively. Further substituting them into the above formula, the LDI is calculated to be 0.4012;

[0057] The calculated live broadcast dynamic comprehensive index is compared with the effect judgment interval, and the specific value of the effect judgment interval is set by the operator according to the historical live broadcast data, and the corresponding live broadcast effect is determined based on the effect judgment interval to generate live broadcast effect information.

[0058] When the comprehensive index of live broadcast dynamics is between 80-100, the live broadcast effect is judged to be "excellent", which means that this live broadcast not only attracts a large number of viewers. If the index falls between 60-79, the live broadcast effect is characterized as "good", indicating that the live broadcast is performing normally, the traffic and interaction are relatively stable. Although they have not reached the extreme, they are relatively stable, and the audience retention rate is maintained at a reasonable level. Once the index is lower than 60, the live broadcast effect is judged to be "poor", and the host's commentary may lack appeal, resulting in audience loss.

[0059] Embodiment 2, as Embodiment 2 of the present invention, is implemented on the basis of Embodiment 1, and differs from Embodiment 1 in that:

[0060] The setting of time t in step 2 is different from that in embodiment 1. In this embodiment, the value of time t can be divided according to the historical live broadcast time of the corresponding live broadcast, and the division can be even. For example, if a game anchor's live broadcast duration is 8 hours a day, the setting of time t can be divided into 2 hours.

[0061] Embodiment 3, as the embodiment 3 of the present invention, focuses on combining the implementation processes of embodiment 1 and embodiment 2 for implementation.

[0062] Example 4, please refer to Figure 2 The present application provides a live broadcast effect analysis device, which includes: a live broadcast data acquisition module, a number of people data analysis module, a bullet screen information analysis module, a comprehensive analysis module and an information output module, and in combination with the attached Figure 2 It can be known that the above functional modules are electrically connected in a unidirectional manner.

[0063] The live broadcast data collection module is used to collect live broadcast information, transmit the number of people in the live broadcast information to the number of people data analysis module, and transmit the interaction index to the bullet screen information analysis module;

[0064] A crowd data analysis module, which is used to analyze the acquired crowd index, obtain corresponding analysis results by separately analyzing the number of viewers, the online peak number of viewers, and the viewing time, and the analysis results here include the real-time number of viewers, the online peak number of viewers, and the average viewing time. At the same time, the comprehensive analysis results are calculated to obtain the crowd relationship index, and the crowd relationship index result is generated. At the same time, the crowd relationship index result is transmitted to the comprehensive analysis module, and the specific analysis method here is the same as the analysis process of step 2 in embodiment 1;

[0065] A barrage information analysis module, which is used to analyze the obtained interactive indicators, and obtains analysis results by analyzing the number of barrages, the content of barrages, and the number of recommendations in the interactive indicators. The analysis results here include the analysis results of the number of barrages, the analysis results of the content of barrages, and the analysis results of the number of recommendations, and calculates the interactive relationship indicators based on the analysis results. At the same time, the interactive relationship indicators are transmitted to the comprehensive analysis module, and the specific analysis method here is the same as the analysis process of step three in embodiment one;

[0066] A comprehensive analysis module, which is used to calculate the live broadcast effect index of the live broadcast according to the obtained number relationship index and interaction relationship index, and at the same time calculate the live broadcast effect index of the same type of live broadcasts, and screen the live broadcasts according to the live broadcast effect index, and then calculate the live broadcast dynamic comprehensive index, and generate live broadcast effect information according to the live broadcast dynamic comprehensive index, and at the same time transmit the live broadcast effect information to the information output module, and the analysis method here is the same as the processing process of step 4 in embodiment 1;

[0067] An information output module is used to display the acquired live broadcast effect information to the corresponding operator.

[0068] Some of the data in the above formulas are dimensionless and numerically calculated. Meanwhile, the contents not described in detail in this specification belong to the prior art known to those skilled in the art.

[0069] The above embodiments are only used to illustrate the technical method of the present invention rather than to limit it. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical method of the present invention may be modified or replaced by equivalents without departing from the spirit and scope of the technical method of the present invention.

Claims

1. A live broadcast effect analysis method, characterized in that: The method specifically comprises the following steps: Step 1: Obtain live broadcast information and classify the live broadcast information into indicators to obtain the number of viewers and interaction indicators; Step 2: Analyze the obtained number of people indicators, obtain the corresponding number of people analysis results by analyzing the number of viewers, the peak number of online viewers, and the viewing time separately, and calculate the number of people relationship indicators by comprehensively analyzing the results, and generate the number of people relationship indicator results; Step 3: Analyze the obtained interaction indicators, obtain the interaction analysis results by analyzing the number of barrages, barrage content and recommendation times in the interaction indicators, and calculate the interaction relationship indicators based on the analysis results; Step 4: Calculate the live broadcast effect index of the live broadcast according to the obtained number relationship index and interaction relationship index, and calculate the live broadcast effect index of the same type of live broadcast at the same time, and screen the live broadcast according to the live broadcast effect index, then calculate the live broadcast dynamic comprehensive index, and generate the live broadcast effect information according to the live broadcast dynamic comprehensive index, and the specific processing method is: The number of people relationship index is recorded as R, the interactive relationship index is recorded as H, and the sum of the two is calculated as the live broadcast effect index Q. Similarly, the same type of live broadcast corresponding to the same period is obtained, and the live broadcast effect index Qn of the same type of live broadcast is calculated, and n represents the number of the same type of live broadcast, where n=1, 2, ..., m, and m represents the number of the same type of live broadcast. At the same time, the number of the same type of live broadcast whose live broadcast effect index Q is greater than Qn is recorded as g; Then obtain the barrage refresh frequency corresponding to the current live broadcast within time t, and obtain the change value of the number of people corresponding to the current live broadcast at time t, and then substitute the obtained barrage refresh frequency, number change value and live broadcast effect index Q into the formula , where LDI is the live broadcast dynamic comprehensive index, , and are the weight coefficients of the barrage refresh frequency, the change value of the number of people, and the live broadcast effect index, respectively. DRF is the barrage refresh frequency, NCV is the change value of the number of people, and g is the number of live broadcasts of the same type obtained by screening; The calculated live broadcast dynamic comprehensive index is compared with the effect judgment interval, and the specific value of the effect judgment interval is set by the operator according to the historical live broadcast data, and the corresponding live broadcast effect is determined based on the effect judgment interval to generate live broadcast effect information.

2. A live broadcast effect analysis method according to claim 1, characterized in that: The number of people indicator in step 1 represents the number of people during the live broadcast, while the interaction indicator represents the interaction of the audience during the live broadcast.

3. A live broadcast effect analysis method according to claim 1, characterized in that: The specific method of generating the number of people relationship index results in step 2 is: The live broadcast information is obtained, and the live broadcast duration is segmented with time t as a period, and the corresponding number of viewers in the time period t is obtained. Then the number of viewers is screened to obtain the real-time number of viewers. At the same time, the real-time number of viewers in different time periods t is obtained, and the average value of the real-time viewer data in the time period t is calculated to generate the real-time number of viewers. Then the peak number of online viewers corresponding to the live broadcast duration is obtained, and the number of viewers online at the same time in the time t and the average viewing time in this interval are calculated. In this way, the average viewing time corresponding to all time periods t is calculated; Substitute the obtained average value of real-time viewing data, peak number of online viewers and average viewing time into the formula: population relationship index = (average value of real-time viewing data + peak number of online viewers + average viewing time) ÷ total live broadcast time, calculate the population relationship index, and generate the population relationship indicator result at the same time.

4. A live broadcast effect analysis method according to claim 1, characterized in that: The specific method of analyzing the interaction indicators in step 3 is as follows: The live broadcast information is obtained, and the total number of barrages and the corresponding barrage content within the time period t are obtained at the same time, and the number of barrages is filtered according to the barrage content to obtain the real-time barrage within the time period t as i, and i=1, 2, ..., j, where j represents the number of real-time barrages, and then the content of the real-time barrage i is identified; The duplicate barrages existing in the real-time barrage i are filtered and denoted as a, and a=1, 2, ..., b, where b represents the number of duplicate barrages. At the same time, the comment users of the duplicate barrage a are obtained, and multiple groups of duplicate barrages corresponding to the same comment user are eliminated, and the unique duplicate barrage is retained. And so on, all the duplicate barrages are filtered to obtain a single duplicate barrage, and the single duplicate barrage is analyzed.

5. A live broadcast effect analysis method according to claim 4, characterized in that: The specific method of analyzing a single repeated bullet comment in step 3 is as follows: The content of the single repeated barrage is obtained, and the barrage content is identified based on the machine learning algorithm. At the same time, the barrage content is matched with the sentiment dictionary to obtain the corresponding matching result, and the barrage evaluation of the single repeated barrage is calculated according to the matching result. The barrage evaluation is classified into positive evaluation, negative evaluation and neutral evaluation. At the same time, the number of barrage evaluations of different categories is obtained and recorded as C1, C2 and C3 respectively; Similarly, all real-time barrages i are evaluated and analyzed, and the corresponding numbers of different types are obtained, recorded as L1, L2 and L3. At the same time, the corresponding number proportions of different types of barrage evaluations are calculated and recorded as Z1, Z2 and Z3 respectively. Then, they are sorted from large to small according to the number proportion, and the corresponding number of recommendations during the live broadcast duration is obtained, and the number of recommendations is classified.

6. A live broadcast effect analysis method according to claim 5, characterized in that: The specific method of classifying the number of recommendations in step 3 is: Obtain the commenting user corresponding to the recommended times, and at the same time obtain the real-time bullet comments corresponding to the commenting user, and classify the real-time bullet comments according to the bullet comment evaluation, then select the classified bullet comment evaluation corresponding to the largest number of bullet comment evaluations as the standard to classify the recommended times, and perform the same processing on all the recommended times in this way to obtain the recommended times classification result; At the same time, the sum of the number of recommendations and the proportion of the same type of barrage evaluations is calculated and recorded as the sum of the values, and the largest value is selected as the interactive relationship indicator.

7. A live broadcast effect analysis device, the device being used to execute a live broadcast effect analysis method according to any one of claims 1 to 6, characterized in that: The device includes: a live broadcast data acquisition module, a number of people data analysis module, a bullet screen information analysis module, a comprehensive analysis module and an information output module; The live broadcast data collection module is used to collect live broadcast information, transmit the number of people in the live broadcast information to the number of people data analysis module, and transmit the interaction index to the bullet screen information analysis module; The crowd data analysis module is used to analyze the acquired crowd index, obtain the corresponding analysis results by analyzing the number of viewers, the peak number of online viewers, and the viewing time, and calculate the crowd relationship index by comprehensively analyzing the results, and generate the crowd relationship index results, and transmit the crowd relationship index results to the comprehensive analysis module; The barrage information analysis module is used to analyze the obtained interactive indicators, and obtain the analysis results by analyzing the number of barrages, the content of barrages and the number of recommendations in the interactive indicators. The analysis results here include the analysis results of the number of barrages, the analysis results of the content of barrages and the analysis results of the number of recommendations, and calculate the interactive relationship indicators based on the analysis results, and transmit the interactive relationship indicators to the comprehensive analysis module at the same time; A comprehensive analysis module, which is used to calculate the live broadcast effect index of the live broadcast according to the obtained number relationship index and interaction relationship index, and at the same time calculate the live broadcast effect index of the same type of live broadcasts, and screen the live broadcasts according to the live broadcast effect index, and then calculate the live broadcast dynamic comprehensive index, and generate live broadcast effect information according to the live broadcast dynamic comprehensive index, and at the same time transmit the live broadcast effect information to the information output module; An information output module is used to display the acquired live broadcast effect information to the corresponding operator.

Images