Public opinion analysis method and device, medium and equipment

By collecting and analyzing carrier data from multiple platforms and combining with the Llma3 model for public opinion analysis, the problem of insufficient analysis of a single platform in the existing technology is solved, and accurate analysis and comprehensive feedback on the target multi-platform public opinion are achieved.

CN119089046APending Publication Date: 2024-12-06SHENZHEN YAWEN INFORMATION COMMUNICATION CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202411207710.1
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-08-30
Publication Date
2024-12-06

AI Technical Summary

Technical Problem

The existing technology is limited to a single OTT streaming media platform in public opinion analysis, which cannot fully reflect consumers' attitudes towards the target object, and lacks consumer feedback on different platforms and regions, so it is impossible to conduct multi-platform public opinion analysis.

Method used

Provide a public opinion analysis method, by collecting carrier data from multiple platforms, performing pre-processing and quantitative and qualitative analysis, using the Llma3 model for language recognition and emotion score, calculating popularity index, and weighted aggregation to form both quantitative and Qualitative public opinion analysis.

Benefits of technology

The target target has achieved accurate analysis of public opinion on multiple platforms, taking into account the habit differences between different platforms and regions, providing more comprehensive public opinion feedback, and improving the accuracy and coverage of the analysis.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN119089046A_ABST
    Figure CN119089046A_ABST
Patent Text Reader

Abstract

The invention relates to the technical field of public opinion analysis, and provides a public opinion analysis method which comprises the following steps: collecting carriers of a plurality of target objects from a platform; carrying out pretreatment on the collected carrier; calculating a quantitative first public opinion index for the preprocessed carrier; calculating a qualitative second public opinion index for the preprocessed carrier; and qualitative and quantitative analysis is formed on a specific carrier. The public opinion analysis method provided by the invention is not limited to a single platform, can perform multi-platform analysis according to actual conditions, not only performs quantitative public opinion analysis on the target object, but also performs qualitative public opinion analysis, and considers habits and differences of different platform subjects. A public opinion analysis method with quantitative and qualitative analysis is formed, an overseas public opinion analysis method beyond China is accurately created, and the method has good application value.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the technical field of public opinion analysis, and in particular to a public opinion analysis method, device, medium and equipment. Background Art

[0002] Overseas entertainment content and user consumption habits are spread across various social media platforms overseas. Currently, overseas public opinion analysis is based only on feedback from a single OTT streaming media, which is not enough to explain consumers' attitudes towards the content of the target object itself. In addition, many OTT platforms only rate some copyrighted content, but the market's evaluation of the target object comes from consumers, who exist on multiple platforms, of which OTT platforms are only a small part, and it is impossible to analyze different platforms.

[0003] Moreover, the public opinion analysis of the platform is limited to the indicator values ​​officially announced by the social media platform, lacking specific explanations of public opinion and no specific feedback from consumers on entertainment content. At the same time, overseas entertainment platforms include users from many different countries and regions, but the platforms are limited by their own market audiences and only integrate single data indicators from overseas countries and regions, lacking consumer feedback information from a specific region or country. Summary of the invention

[0004] In order to solve the above-mentioned deficiencies in the prior art, the present invention provides a public opinion analysis method, comprising: S100, collecting carriers of several target objects from the platform; S200, pre-processing the collected carriers; S300, calculating a quantitative first public opinion index for the preprocessed carrier; S400, calculating a qualitative second public opinion index for the preprocessed carrier; S500, forming a qualitative and quantitative analysis of the specific carrier.

[0005] Furthermore, the specific steps of S100 are: S110, analyzing the website and front-end page of the platform; S120, configuring the search keywords of the object and the full platform search mechanism; the keywords include at least one language version; S130, using search keywords and the full platform search mechanism to obtain carriers of several target objects from the platform; S140, structured acquisition of crawling settings or calling settings of carriers of different platforms of the target object, and taking dormant measures to crawl or call the carriers.

[0006] Furthermore, in step S200, the collected carrier information and comment text are input into the Llma3 model; Llma3 is used to judge the carrier relevance and output the carrier relevance probability; and the carriers with a carrier relevance probability lower than 20% are removed.

[0007] Furthermore, during the carrier removal process, if there are more than 5 irrelevant carriers belonging to the same publishing entity, the basic information of the publishing entity corresponding to the irrelevant carriers will be input into the Llma3 model, and the Llma3 model will be used to calculate the correlation of the publishing entity for judgment, and the correlation probability of the publishing entity will be output; the publishing entity whose correlation probability is lower than 20% will be blocked.

[0008] Furthermore, the first public opinion indicator is the popularity of the target object on the entire platform. The specific steps of S300 are: S310, determining the target object, and screening out all carriers of the target object; S320, retrieve comment information of all carriers of the target object, and use the Llma3 model to perform language recognition; S330, using the specific time when each comment occurred as a popularity anchor point, determining a carrier reach index, and assigning it to each comment with the identified language; S340, calculating the popularity of the same language target object during the time when the comment occurred according to the comment information and the reach index; S350, normalizing the popularity calculated for different target objects, different platforms and different languages; S360: Calculate the carrier weights of different platforms for the same target object, perform weighted aggregation on the popularity of different platforms to obtain a popularity index, and then multiply the popularity index by a preset value to amplify it.

[0009] Further, S340, the popularity of the same language target object during the time when the comment occurred is calculated according to the comment information and the reach index, and the calculation formula is:

[0010] Among them, o is the current carrier, It is the reach indicator of the current carrier and the current language on that day. is the total number of comments on the current carrier on that day, n is the number of comments generated on that day, and t is the length of the time period during which the comments occurred; S350, normalizing the popularity calculated for different target objects on different platforms and in different languages, the specific formula is:

[0011] Among them, p is the source platform, l is the current language, is the preset quantile of the current platform, is the popularity of the target audience in the current language on the current platform, is the normalized popularity of the current language target object on the current platform; S360: Calculate the carrier weights of different platforms for the same target object, perform weighted aggregation on the popularity of different platforms to obtain a popularity index, and multiply the popularity index by a preset value to amplify it. The specific formula is:

[0012]

[0013] Where a is the number of carriers on the current platform, b is the number of carriers on the entire platform, and c is the preset value. is the carrier weight of different platforms, P is the number of all platforms, p is the current platform, l is the current language, is the current platform carrier quantity share, The popularity of target objects across all platforms. is the current platform popularity.

[0014] Furthermore, the specific steps of S400 are: S410, providing entertainment knowledge information related to the target object to train the Llma3 model; S420, retrieve the comment information of all carriers of the target object, and clean the comment information; S430, extracting entity information related to the target object and audience feedback emotions from the comment information using the Llma3 model; S440: Perform a sentiment score on the target object based on the entity information and the feedback emotion, where the sentiment score ranges from -1 to 1.

[0015] The present invention also provides a public opinion analysis device, comprising: The collection module is used to collect public opinion data of specific carriers; A processing module is used to pre-process the collected public opinion data; A quantitative analysis module, used to calculate a quantitative first public opinion index for the preprocessed public opinion data; A qualitative analysis module is used to calculate a qualitative second public opinion index based on the preprocessed public opinion data; An integrated module for qualitative and quantitative analysis of specific vectors.

[0016] The present invention also provides a computer-readable storage medium, which stores computer instructions and implements the public opinion analysis method as described in any one of the above when executed by a computer processor.

[0017] The present invention also provides an electronic device, comprising at least one processor and a memory communicatively connected to the processor, wherein the memory stores instructions executable by at least one processor, and the instructions are executed by at least one processor so that the processor executes the public opinion analysis method as described in any one of the above items. Based on the above, compared with the prior art, the public opinion analysis method provided by the present invention is not limited to a single platform, and can perform analysis on multiple platforms according to actual conditions. It not only performs quantitative public opinion analysis on the target object, but also performs qualitative public opinion analysis, and takes into account the habits and differences of different platform entities, forming a public opinion analysis method that combines quantitative and qualitative analysis, and accurately creates an overseas public opinion analysis method outside of China, which has good application value.

[0018] Other features and beneficial effects of the present invention will be described in the following description, and partly become apparent from the description, or understood by practicing the present invention. The purpose and other beneficial effects of the present invention can be realized and obtained by the structures particularly pointed out in the description, claims and drawings. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work. The positional relationships described in the drawings in the following description are based on the directions in which the components are drawn in the diagrams, unless otherwise specified.

[0020] Figure 1 A flowchart of a public opinion analysis method provided in Embodiment 1 of the present invention; Figure 2 A flowchart of step S100 of the public opinion analysis method provided in the first embodiment of the present invention; Figure 3 This is a flowchart of step S300 of the public opinion analysis method provided in the first embodiment of the present invention; Figure 4 A flowchart of step S400 of the public opinion analysis method provided in the first embodiment of the present invention; Figure 5 This is a result diagram of sentiment analysis provided in the first embodiment of the present invention; Figure 6 This is a structural block diagram of a public opinion analysis device provided in Embodiment 2 of the present invention; Figure 7 This is a structural block diagram of an electronic device provided in Embodiment 4 of the present invention. DETAILED DESCRIPTION

[0021] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, rather than all the embodiments; the technical features designed in different implementation modes of the present invention described below can be combined with each other as long as they do not conflict with each other; based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative work are within the scope of protection of the present invention.

[0022] In the description of the present invention, it should be noted that all terms (including technical terms and scientific terms) used in the present invention have the same meanings as those generally understood by ordinary technicians in the field to which the present invention belongs, and should not be understood as limitations on the present invention; it should be further understood that the terms used in the present invention should be understood to have the same meanings as these terms in the context of this specification and in the relevant fields, and should not be understood in an idealized or overly formal sense, unless explicitly defined in this invention.

[0023] Embodiment 1 The present invention provides a method for analyzing public opinion. Figure 1 As shown, including: S100, collecting carriers of several target objects from the platform.

[0024] Specifically, the carrier includes basic knowledge information about the target and public data about the target on social media that has generated a lot of discussion about Chinese entertainment content. The public data includes indicator data and audience comment text. The platform includes two types: text-based social media and video-based social media. Text-based social media such as Facebook and Instagram are mainly based on user posts, and their carriers are also specific posts posted by users; video-based social media include YouTube and Tiktok, which are mainly based on videos posted on the platform, and their carriers are also corresponding videos; it should be noted that it is not limited to platforms such as Facebook, Instagram, YouTube and Tiktok, and the platform range can be determined according to specific circumstances, and the corresponding text social media or video social media can be increased or decreased.

[0025] S200, pre-processing the collected carriers.

[0026] The second step is to pre-process and clean the collected carriers to obtain structured carrier data, and perform preliminary language analysis to obtain the language category, publishing entity, and the region where the carrier may come from.

[0027] S300, calculating a quantitative first public opinion index for the preprocessed carrier.

[0028] The first public opinion indicator is the popularity of the target object overseas, which captures the popularity and popularity of the target overseas. It should be noted that the overseas scope can be determined according to the target market, and it must be all countries and regions except China.

[0029] S400: Calculate a qualitative second public opinion index for the preprocessed carrier.

[0030] The second public opinion indicator is natural semantic content analysis including entity resolution and sentiment analysis, which captures the reasons why the target is popular or discussed overseas.

[0031] S500, forming a qualitative and quantitative analysis of the specific carrier.

[0032] Specifically, the first public opinion index and the second public opinion index obtained by the quantitative analysis module and the qualitative analysis module are integrated to obtain all public opinion analysis data of the target object.

[0033] like Figure 2 As shown, the specific steps of S100 are: S110, analyzing the website and front-end page of the platform; Specifically, the aforementioned platform collects data on domestic entertainment content, including videos, pictures, reposts, comments, etc. The data collection method is a crawler mechanism, such as using Python standard software packages for obtaining public information such as beautifulSoup / request / scrapy / selenium. For mature website pages, it is necessary to standardize the capture of dynamic content of the website through engineering methods through a large number of spare IPs, accounts and simulation of real users, and obtain the corresponding public opinion carriers and corresponding information.

[0034] Before using the crawler mechanism, first parse the URL and front-end web page of the target platform. Different platforms have different crawling mechanisms. Therefore, parsing the target platform first is conducive to the configuration of the subsequent crawling mechanism and speeding up the carrier crawling speed.

[0035] S120, configuring search keywords and a full-platform search mechanism for the target object; the keywords include at least one language version.

[0036] Specifically, search keywords are established according to the specifics of the target object, and suggestions for full-platform search mechanisms are made based on the characteristics of different platforms. If the target object is a TV series, the search keywords can be keywords such as the TV series name, the name of the male lead, the name of the female lead, etc.; if the target object is a star, it can be the star's name, the name of the TV character played, the name of the song performed, etc., which needs to be determined according to the specific situation. Keywords are searched in multiple languages ​​and can be determined based on the target market country. For example, if the key market is Indonesia, Vietnam, Thailand and other countries, Indonesian, Vietnamese and Thai should be searched in particular.

[0037] S130. Use search keywords and the full-platform search mechanism to obtain carriers of several target objects from the platform.

[0038] Specifically, after determining the search keywords, search mechanism and language, we start collecting carriers, mainly obtaining browsing data and interaction data of the carriers. Browsing data includes video playback volume or post viewing volume, and interaction data includes the number of comments or forwarding volume.

[0039] S140, structured acquisition of crawling settings or calling settings of carriers of different platforms of the target object, and taking dormant measures to crawl or call the carriers.

[0040] Specifically, the crawling settings or call settings of different platforms of the target object are structured, and dormant measures are taken to crawl or call the carrier. The dormant measures are set according to the imitation of human user behavior. The dormant measures make the crawling more like a human user rather than a robot, which is conducive to dealing with anti-crawling mechanisms.

[0041] In step S200, the collected carrier information and comment text are input into the Llma3 model; Llma3 is used to judge the carrier relevance and output the carrier relevance probability; and carriers with a carrier relevance probability lower than 20% are removed.

[0042] Specifically, after the carrier collection is completed, the Llma3 model is used to judge the carrier relevance and process irrelevant carriers. The input content is the carrier information and comment text, and the output is the carrier relevance probability. The higher the probability, the higher the possibility of relevance. Carriers with a probability below 20% are removed. This preprocessing mechanism excludes carriers in a large range.

[0043] During the carrier removal process, if there are more than 5 irrelevant carriers belonging to the same publishing entity, the basic information of the publishing entity corresponding to the irrelevant carriers will be input into the Llma3 model, and the Llma3 model will be used to calculate the correlation of the publishing entity for judgment, and the correlation probability of the publishing entity will be output; the publishing entity whose correlation probability is less than 20% will be blocked.

[0044] Specifically, in the process of using carrier relevance for exclusion, if more than 5 irrelevant carriers belong to the same publishing subject, the basic information of the publishing subject corresponding to the irrelevant carrier is input into the Llma3 model, and the Llma3 model is used to calculate the relevance of the publishing subject for judgment, and the relevant probability of the publishing subject is output; the basic information of the publishing subject input is the channel name, introduction, corresponding video title and description, and the output is the probability of whether the publishing subject is a relevant publishing subject. Publishing subjects with a probability lower than 20% will be blocked. This preprocessing mechanism performs refined screening and exclusion through publishing subjects.

[0045] Through the two preprocessing mechanisms, carrier preprocessing can be performed from two aspects: the carrier aspect and the publishing subject aspect, making the preprocessed data more accurate.

[0046] Through the above steps, the carriers are collected and preprocessed. Next, the quantitative first public opinion index is constructed based on the preprocessed carriers. Specifically, the long-term popularity of the target object is constructed based on the obtained carriers, such as Figure 3 As shown, the specific steps of S300 are: S310: Determine at least two target objects, and screen out all carriers of the target objects.

[0047] The first quantitative public opinion indicator is judged based on popularity, but a single target object cannot be judged for popularity. It can calculate a specific popularity value, but without a reference, its specific popularity cannot be determined. Therefore, at least two target objects are needed to compare and judge the popularity of the two target objects. Therefore, all carriers of at least two target objects need to be collected at the same time for subsequent popularity calculation.

[0048] S320, retrieve the comment information of all carriers of the target object, and use the Llma3 model for language recognition.

[0049] Specifically, the Llma3 model with prior knowledge is used in combination with the Python langdetect package to perform language identification on all comment information and filter out target countries of business interest, such as Thailand and Indonesia. It should be noted that although language settings are also performed in the keyword establishment stage, the comments corresponding to the corresponding language carriers filtered by keywords may not be limited to the corresponding languages. There are cases where Thai language carriers contain English, Thai and Vietnamese comments. Therefore, a secondary screening of the comment language is required here to make the data more accurate and the popularity analysis of the target countries more accurate.

[0050] S330. Using the specific time when each comment occurred as a popularity anchor point, determine the reach index of the carrier and allocate it to each comment with the identified language.

[0051] Specifically, there are gaps in the date between the comments corresponding to a carrier, so the time of each comment is used as the popularity anchor point, in days, and the carrier's reach index is determined. The reach index is the number of video views or post views. Differentiate comments in different languages ​​based on the same reach index. For example, if you need to analyze the popularity in Thailand, only Thai comments corresponding to the video or post are analyzed, and comments from other target countries are not considered.

[0052] S340, calculating the popularity of the same language target object during the time when the comment occurred according to the comment information and the reach index, the calculation formula is:

[0053] Among them, o is the current carrier, It is the reach indicator of the current carrier and the current language on that day. is the total number of comments on the current carrier on that day, n is the number of comments generated on that day, and t is the length of the time period during which the comments occurred; For example, if we analyze the popularity of celebrity A on the Thai YouTube platform from July 1 to July 2, we retrieved 4 carriers on July 1, with 7 comments on the 4 carriers, 2 comments on July 1, and a total of 2000 reach indicators. The calculation formula for celebrity A's popularity on July 1 is:

[0054] On July 2, the number of carriers related to celebrity A in Thailand is 3, the number of comments is 4, the number of comments generated on that day is 1, and the reach index is 4000. The calculation formula for celebrity A's popularity in Thailand on July 2 is:

[0055] The total popularity for two days is:

[0056] According to the above embodiment, the popularity of each day in this solution does not interfere with each other. For example, the popularity calculation on July 2 has excluded the carriers that have been calculated on July 1 and have no updated reach indicators or number of comments on July 2. And the carriers that were missed on July 1 can be completed on July 2. It should be noted that the above July 1 and July 2 are only used for example demonstration, and other dates and time spans can also be used, which are determined according to the time to be analyzed.

[0057] S350: normalize the popularity calculated for different objects, different platforms and different languages.

[0058] As mentioned above, a single target object can calculate a specific popularity value, but without a reference, its specific popularity cannot be determined, so it is necessary to determine at least two target objects for comparison and judgment of popularity values. However, the popularity between different target objects may differ by a large amount, so the popularity of different objects needs to be normalized so that the popularity values ​​of different magnitudes can be compared. The specific formula is:

[0059] Among them, p is the source platform, l is the current language, is the preset quantile of the current platform, is the popularity of the target audience in the current language on the current platform, is the normalized popularity of the target object in the current language on the current platform. The preset percentiles of the popularity data distribution of different objects on the same platform are: when the social media is text-based social media, the preset percentile is 90% of the current popularity; when the social media is video-based social media, the preset percentile is 95% of the current popularity.

[0060] S360: Calculate the carrier weights of different platforms for the same target object, perform weighted aggregation on the popularity of different platforms to obtain a popularity index, and then multiply the popularity index by a preset value to amplify it.

[0061] The specific formula is:

[0062]

[0063] Where a is the number of carriers on the current platform, b is the number of carriers on the entire platform, and c is the preset value. is the carrier weight of different platforms, P is the number of all platforms, p is the current platform, l is the current language, is the current platform carrier quantity share, The popularity of target objects across all platforms. is the current platform popularity.

[0064] Specifically, the preset value is 10000. The total amount of carriers used by different platforms is used as the weight, for example, 100 YouTube videos and 120 Facebook posts. If only the popularity of these two platforms is calculated, the weight of YouTube is 0.45 and the weight of Facebook is 0.55. The popularity of different platforms is weighted and aggregated to finally get a separate popularity index, which is then multiplied by 10000 to amplify. This makes the popularity more intuitive, sensitive and friendly, and can capture more daily changes.

[0065] The above popularity calculation method uses a natural language pre-training model, combined with carrier-directed information, to perform language recognition on the carrier's comment data. After recognition, it is used as the base of popularity, weighted according to user feedback indicators, and then normalized to make the index comparable. Finally, a single but decomposable popularity index is formed, that is, quantitative analysis. The result has become a reference for the export business of China's entertainment industry, providing a data basis for choosing which film and television works to export overseas and when to export overseas in the future.

[0066] The second public opinion indicator is the entity information of the target object, the audience's feedback emotions and emotional scores, such as Figure 4 As shown, the specific steps of S400 are: S410, provide entertainment knowledge information related to the target object to train the Llma3 model. This step can make the subsequent output more relevant to the target object and reduce extraction errors. For example, for a TV series, you can input comments related to the existing TV series for training, TV series related words, etc., and stars are similar, and the model is trained first.

[0067] S420, retrieve the comment information of all carriers of the target object, and clean the comment information. Use the Llma3 model to filter out irrelevant comments such as advertisements, personal attacks, political declarations, etc., to reduce the impact of dirty data on the model.

[0068] S430, extracting entity information related to the target object and audience feedback emotions from the comment information using the Llma3 model; Specifically, the cleaned comment text is taken as the parsing object, combined with the text information of the carrier itself as the background information, and the word frequency as the weight, and the Llma3 model is used to find the key relevant information of the target object, such as relationship, appearance, and music. The corresponding entity information is extracted according to the key relevant information. The entity information includes such as husband-wife relationship, handsomeness, song name, and the feedback emotions of the audience of the corresponding entity information are obtained, such as jealousy, praise, and good listening.

[0069] S440: Perform a sentiment score on the target object based on the entity information and the feedback emotion, where the sentiment score ranges from -1 to 1.

[0070] Specifically, a target object may have multiple entity information, and one entity information may correspond to multiple different audience reaction emotions. Entity aspects such as the appearance, plot, acting skills, etc. of the object will be used to analyze the public opinion direction of the object. The key emotional words of the comments will be used to describe the emotional score of this aspect. The emotional score ranges from -1 to 1. The higher the value, the positive emotion is, the lower the value is negative, and the value close to 0 indicates neutral emotion. It should be noted that the entity aspects are not limited to the above-mentioned aspects, and the entity aspects need to be determined according to the specific situation. In addition, there are some target objects that only have emotions but no corresponding entity aspects. For example, there is only a compliment in a comment, but this word cannot be clearly classified into a certain emotion.

[0071] Preferably, if Figure 5 As shown, the aspect and entity information and sentiment scores are displayed in the form of labels and word clouds. The labels are mainly nouns, indicating that the direction of public opinion is not divided into big and small, and only the direction that appears more than 5 times is taken. The word cloud is mainly adjectives, and the shape and size of the words indicate the number of appearances. The colors range from red, green and blue to represent the sentiment scores respectively. Red represents negative emotions, green represents positive emotions, and blue represents neutral emotions. The sentiment analysis finally forms a qualitative and quantitative analysis.

[0072] Embodiment 2 The present invention also provides a public opinion analysis device, such as Figure 6 As shown, including: The collection module is used to collect public opinion data of specific carriers; A processing module is used to pre-process the collected public opinion data; A quantitative analysis module, used to calculate a quantitative first public opinion index for the preprocessed public opinion data; A qualitative analysis module is used to calculate a qualitative second public opinion index based on the preprocessed public opinion data; The integration module is used to form a qualitative and quantitative analysis of a specific carrier. Specifically, the first public opinion index and the second public opinion index obtained by the quantitative analysis module and the qualitative analysis module are integrated and output.

[0073] Embodiment 3 The present invention also provides a computer-readable storage medium, which stores computer instructions and implements any of the above-mentioned public opinion analysis methods when executed by a computer processor.

[0074] Embodiment 4 The present invention also provides an electronic device, such as Figure 7 As shown, it includes at least one processor and a memory connected to the processor in communication, wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor to enable the processor to perform any of the public opinion analysis methods mentioned above.

[0075] In addition, those skilled in the art should understand that, although there are many problems in the prior art, each embodiment or technical solution of the present invention can be improved in only one or several aspects, without having to solve all the technical problems listed in the prior art or background technology at the same time. Those skilled in the art should understand that the content not mentioned in a claim should not be used as a limitation on the claim.

[0076] Although the terms such as platform, target object, carrier, first public opinion index, second public opinion index, popularity, etc. are used more frequently in this article, the possibility of using other terms is not excluded. These terms are used only to more conveniently describe and explain the essence of the present invention; interpreting them as any additional restrictions is contrary to the spirit of the present invention; the terms "first", "second", etc. (if any) in the description and claims of the embodiments of the present invention and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence.

[0077] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for analyzing public opinion, characterized in that: include: S100, collecting carriers of several target objects from the platform; S200, pre-processing the collected carrier; S300, calculating a quantitative first public opinion index for the preprocessed carrier; S400, calculating a qualitative second public opinion index for the preprocessed carrier; S500, performing qualitative and quantitative analysis on the specific carrier.

2. The public opinion analysis method according to claim 1, characterized in that: The specific steps of S100 are: S110, parsing the website and front-end page of the platform; S120, configuring search keywords for objects and a full-platform search mechanism; the keywords include at least one language version; S130, using the search keyword and the full-platform search mechanism to obtain a number of carriers of the target object from the platform; S140, structurally obtaining the crawling settings or calling settings of the carriers of the target objects on different platforms, and taking dormant measures to crawl or call the carriers.

3. The public opinion analysis method according to claim 2, characterized in that: In step S200, the collected carrier information and comment text are input into the Llma3 model; Llma3 is used to judge the carrier relevance and output the carrier relevance probability; the carriers with the carrier relevance probability lower than 20% are removed.

4. The public opinion analysis method according to claim 3, characterized in that: During the carrier removal process, if there are more than 5 irrelevant carriers belonging to the same publishing entity, the basic information of the publishing entity corresponding to the irrelevant carriers will be input into the Llma3 model, and the Llma3 model will be used to calculate the relevance of the publishing entity for judgment, and the relevant probability of the publishing entity will be output; the publishing entity whose relevant probability of the publishing entity is lower than 20% will be blocked.

5. The public opinion analysis method according to claim 1, characterized in that: The first public opinion indicator is the popularity of the target object on the entire platform. The specific steps of S300 are: S310, determining at least two target objects, and screening out all carriers of the target objects; S320, retrieve comment information of all carriers of the target object, and use the Llma3 model to perform language recognition; S330, using the specific time when each comment occurred as a popularity anchor point, determining the reach index of the carrier, and assigning it to each comment with the identified language; S340, calculating the popularity of the target object in the same language during the time when the comment occurred according to the comment information and the reach index; S350, normalizing the popularity calculated for different target objects on different platforms and in different languages; S360: Calculate the carrier weights of different platforms for the same target object, perform weighted aggregation on the popularity of different platforms to obtain a popularity index, and then multiply the popularity index by a preset value to amplify it.

6. The public opinion analysis method according to claim 5, characterized in that: S340, calculating the popularity of the target object in the same language during the time when the comment occurred according to the comment information and the reach index, and the calculation formula is: Among them, o is the current carrier, It is the reach indicator of the current carrier and the current language on that day. is the total number of comments on the current carrier on that day, n is the number of comments generated on that day, and t is the length of the time period during which the comments occurred; S350, normalizing the popularity calculated for different target objects on different platforms and in different languages, the specific formula is: Among them, p is the source platform, l is the current language, is the preset quantile of the current platform, is the popularity of the target object in the current language on the current platform, The normalized popularity of the target object in the current language of the current platform; S360: Calculate the carrier weights of different platforms for the same target object, perform weighted aggregation on the popularity of different platforms to obtain a popularity index, and multiply the popularity index by a preset value to amplify it. The specific formula is: Where a is the number of carriers on the current platform, b is the number of carriers on the entire platform, and c is the preset value. is the carrier weight of different platforms, P is the number of all platforms, p is the current platform, l is the current language, is the current platform carrier quantity share, The popularity of target objects across all platforms. is the current platform popularity.

7. The public opinion analysis method according to claim 1, characterized in that: The second public opinion indicator is the entity information of the target object, the audience's feedback emotion and the sentiment score, and the specific steps of S400 are: S410, providing entertainment knowledge information related to the target object to train the Llma3 model; S420, retrieving the comment information of all carriers of the target object, and cleaning the comment information; S430, extracting entity information related to the target object and audience feedback emotions from the comment information using the Llma3 model; S440: Perform a sentiment score on the target object according to the entity information and the feedback emotion, where the sentiment score ranges from -1 to 1.

8. A public opinion analysis device, characterized in that: include: The collection module is used to collect public opinion data of specific carriers; A processing module, used for preprocessing the collected public opinion data; A quantitative analysis module, used to calculate a quantitative first public opinion index for the preprocessed public opinion data; A qualitative analysis module, used for calculating a qualitative second public opinion index based on the preprocessed public opinion data; An integrated module for qualitative and quantitative analysis of specific vectors.

9. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores computer instructions, and when the computer is executed by a processor, the public opinion analysis method according to any one of claims 1 to 7 is implemented.

10. An electronic device, characterized in that: It includes at least one processor and a memory that is communicatively connected to the processor, wherein the memory stores instructions that can be executed by at least one processor, and the instructions are executed by at least one processor so that the processor executes the public opinion analysis method as described in any one of claims 1-7.