Multi-mode public opinion monitoring intelligent analysis system and method

Abstract

The invention relates to the field of public opinion intelligent analysis, in particular to a multi-mode public opinion monitoring intelligent analysis system and method. The method comprises the following steps of obtaining public opinion source information and source time, analyzing the public opinion source information to obtain public opinion word vectors, and constructing a public opinion distribution diagram according to the public opinion word vectors and the source time; according to the real-time public opinion source information and the real-time source time, public opinion circle moving points and public opinion circle change points are obtained, a public opinion dynamic graph is obtained according to the public opinion circle change points, the public opinion circle moving points and the public opinion distribution graph, the public opinion dynamic graph is analyzed, and a focusing area and an observation area are obtained; analyzing the focus area in combination with the public opinion dynamic graph to obtain a popularity evaluation value, analyzing the popularity evaluation value to obtain a popularity factor and a negative factor, and analyzing the popularity factor and the negative factor to obtain a public opinion trend value; and analyzing the public opinion trend value to obtain a public opinion management strategy. The public opinion condition observation capability can be improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent public opinion analysis, and particularly to a multi-modal intelligent public opinion monitoring and analysis system and method. Background Art

[0002] With the rapid development of artificial intelligence technology, especially the application breakthroughs of deep learning in the fields of natural language processing (NLP), computer vision (CV), speech recognition, etc., more and more public opinion monitoring systems have begun to introduce multi-modal data analysis technology. Multi-modal analysis combines information from different data sources; public opinion monitoring refers to the tracking and analysis of the opinions, attitudes, and emotions of the general public. Especially in the digital age of rapid and fragmented information dissemination, the formation and outbreak of public opinion events have become even more rapid; public opinion monitoring is of great significance to enterprises, the media, etc. For example, emergencies, brand management, etc.

[0003] During the spread of public opinion events, there are problems of data security. The spread processes of each public opinion event itself and between public opinion events are complex. The impact of the changes in public opinion events on the spread of public opinion events, the rapid changes in the spread of public opinion events, and the delay in acquisition; the analysis of the key influencing factors in the spread process of public opinion events, how to respond to emergencies, enterprise brand monitoring, hot event analysis, etc. are the problems we need to solve. Summary of the Invention

[0004] The object of the present invention is to propose a multi-modal intelligent public opinion monitoring and analysis system and method for the problems existing in the background art.

[0005] The technical solution of the present invention: A multi-modal intelligent public opinion monitoring and analysis method includes the following steps:

[0006] S1. Obtain public opinion source information and source time, analyze the public opinion source information to obtain a public opinion word vector, and construct a public opinion distribution map according to the public opinion word vector and the source time;

[0007] S2. Obtain real-time public opinion source information and real-time source time, obtain a public opinion circular moving point and a public opinion circular change point according to the real-time public opinion source information and the real-time source time, obtain a public opinion dynamic map according to the public opinion circular change point, the public opinion circular moving point, and the public opinion distribution map, and analyze the public opinion dynamic map to obtain a focused area and an observation area;

[0008] S3. Analyze the focused area in combination with the public opinion dynamic map to obtain a heat evaluation value, analyze the heat evaluation value to obtain a heat factor and a negative factor, and analyze the heat factor and the negative factor to obtain a public opinion trend value;

[0009] S4. Analyze the public opinion trend value to obtain a public opinion management strategy.

[0010] Preferably, the process of obtaining public opinion source information and source time, analyzing the public opinion source information to obtain public opinion word vectors, and constructing a public opinion distribution map according to the public opinion word vectors and source time includes:

[0011] Obtain the location where public opinion is generated, set an information collection end according to the obtained location where public opinion is generated. The information collection end consists of several information capture points, and the information capture points are used to obtain public opinion source information and source time; the public opinion source information includes public opinion text information, public opinion pictures, and public opinion videos;

[0012] Analyze the public opinion source information to obtain diverse text information; identify the words in the diverse text information through natural language processing technology to obtain text words; analyze the text words through the GloVe algorithm to obtain public opinion word vectors;

[0013] Obtain the text sentences of the diverse text information, and obtain public opinion sentence vectors according to the public opinion word vectors of the text words in the text sentences;

[0014] Mark all public opinion sentence vectors in a rectangular coordinate system according to the public opinion sentence vectors, set a radius standard value and a standard number of points; obtain the distances between all public opinion points, and perform distribution processing on the distances between public opinion points. The distribution processing specifically refers to: analyzing the distances between each public opinion point through the standard number of points and the radius standard value to obtain a public opinion center point and points inside the public opinion circle, and obtaining a public opinion distribution area according to the public opinion center point and the radius standard value;

[0015] Convert the source time into a time stamp, denoted as the source time stamp, obtain the source time of the text words corresponding to the public opinion distribution area, construct the association relationship between the public opinion center points of the public opinion distribution area through the source time, and construct a public opinion distribution map according to the public opinion distribution area, the association relationship, and the source time stamp.

[0016] Preferably, the process of obtaining real-time public opinion source information and real-time source time, obtaining a public opinion circle moving point and a public opinion circle changing point according to the real-time public opinion source information and real-time source time, obtaining a public opinion dynamic map according to the public opinion circle changing point, the public opinion circle moving point, and the public opinion distribution map, and analyzing the public opinion dynamic map to obtain a focused area and an observation area includes:

[0017] Obtain real-time public opinion source information and real-time source time, and obtain an online public opinion distribution map through the real-time public opinion source information and real-time source time;

[0018] Analyze the online public opinion distribution map and the public opinion distribution map. Analyze the points inside the public opinion circles in each public opinion distribution area of the online public opinion distribution map and the points inside the public opinion circles in the corresponding public opinion distribution areas of the public opinion distribution map to obtain real circle moving points and empty circle moving points, and record the real circle moving points and the empty circle moving points as public opinion circle moving points;

[0019] Obtain public opinion circle variable points according to the points in the public opinion distribution areas corresponding to the public opinion circle moving points;

[0020] Construct the dynamic change relationship between the public opinion circle variable points and the corresponding public opinion circle center points in the public opinion distribution map. According to the public opinion circle variable points, real circle moving points, empty circle moving points and the dynamic change relationship, obtain the public opinion dynamic map;

[0021] Analyze the public opinion distribution map, and analyze the public opinion circle center points through the dynamic change relationship to obtain the focus area and the observation area.

[0022] Preferably, the process of analyzing the focus area in combination with the public opinion dynamic map to obtain the heat evaluation value is as follows:

[0023] According to the public opinion dynamic map and the focus area, count the number of real circle moving points and empty circle moving points in the focus area, and record them as the number of real points and the number of empty points respectively. According to the number of real points and the number of empty points, obtain the heat evaluation value R h ;

[0024]

[0025] Among them, mps1 and mps2 refer to the heat influence coefficients of real circle moving points on the focus area and the heat influence coefficients of empty circle moving points on the focus area; R h refers to the heat evaluation value of the hth focus area; ds refers to the number of real points; ks refers to the number of empty points; w refers to the value obtained by subtracting the area occupied by the empty circle moving points from the sum of the area occupied by the real circle moving points and the area occupied by the points inside the public opinion circle; S refers to the area of the public opinion distribution area.

[0026] Preferably, the process of analyzing the heat evaluation value to obtain the heat factor and the negative factor includes:

[0027] Analyze the heat evaluation value and the focus area to obtain the heat area and the negative area;

[0028] Analyze the heat area to obtain the heat factor Y f ;

[0029]

[0030] Among them, Y f refers to the fth focus heat area; ds refers to the number of real points; ns refers to the total number of points inside the public opinion circle;

[0031] Analyze the negative region to obtain the negative factor Y g ;

[0032]

[0033] Among them, Y g refers to the g-th focused negative region; ks refers to the number of empty points; ns refers to the total number of points within the public opinion circle.

[0034] Preferably, the process of analyzing the heat factor and the negative factor to obtain the public opinion trend value includes:

[0035] Analyze the heat factor and the heat region to obtain the focused heat region; analyze the negative factor and the negative region to obtain the focused negative region;

[0036] Obtain the heat similarity of the heat region according to the diverse text information corresponding to the focused heat region and the diverse text information corresponding to each heat region; obtain the negative similarity of the negative heat region according to the diverse text information corresponding to the focused negative region and the diverse text information corresponding to each negative region;

[0037] Set the spread similarity; analyze the spread similarity, the heat similarity, and the heat region to obtain the heat spread region and the negative spread region;

[0038] Obtain the public opinion trend value Qt of the focused heat region according to the heat factor of the heat spread region and the heat evaluation value of the focused heat region 热 ;

[0039]

[0040] Among them, R b represents the heat evaluation value of the focused heat region; Y i represents the heat factor of each heat spread region; c represents the total number of heat spread regions; D i represents the heat similarity of each heat spread region;

[0041] Obtain the public opinion trend value of the focused negative region according to the negative factor of the negative spread region and the heat evaluation value of the focused negative region;

[0042]

[0043] Among them, R p represents the heat evaluation value of the focused negative region; Y l represents the negative factor of each negative spread region; x represents the total number of negative spread regions; D l represents the negative similarity of each negative spread region.

[0044] Preferably, the process of analyzing the public opinion trend value to obtain the public opinion management strategy includes:

[0045] Setting the public opinion trend threshold for the focused heat area and the public opinion trend threshold for the focused negative area;

[0046] When the public opinion trend value in the focused heat area is greater than or equal to the public opinion trend threshold or the public opinion trend value in the focused negative area is greater than or equal to the public opinion trend threshold, it is recorded as public opinion level one; when the public opinion trend value in the focused heat area is less than the public opinion trend threshold, it is recorded as public opinion level two;

[0047] Sending public opinion level one and public opinion level two to relevant personnel, and the relevant personnel formulate management strategies for relevant events through public opinion level one and public opinion level two to obtain the public opinion management strategy.

[0048] The present invention also discloses a multi-modal public opinion monitoring intelligent analysis system, including a management center, and the management center is communicatively connected to a data collection module, a data monitoring and analysis module, a data processing module, and a public opinion management module:

[0049] The data collection module is used to obtain public opinion source information and source time, analyze the public opinion source information to obtain a public opinion word vector, and construct a public opinion distribution map according to the public opinion word vector and the source time;

[0050] The data monitoring and analysis module is used to obtain real-time public opinion source information and real-time source time, obtain a public opinion circular moving point and a public opinion circular change point according to the real-time public opinion source information and the real-time source time, obtain a public opinion dynamic map according to the public opinion circular change point, the public opinion circular moving point, and the public opinion distribution map, and analyze the public opinion dynamic map to obtain a focused area and an observation area;

[0051] The data processing module is used to analyze the focused area in combination with the public opinion dynamic map to obtain a heat evaluation value, analyze the heat evaluation value to obtain a heat factor and a negative factor, and analyze the heat factor and the negative factor to obtain a public opinion trend value;

[0052] The public opinion management module is used to analyze the public opinion trend value to obtain the public opinion management strategy.

[0053] Compared with the prior art, the above technical solution of the present invention has the following beneficial technical effects:

[0054] Enhance the perception ability of public opinion information through information from multiple public opinion sources; construct a public opinion distribution map with the help of public opinion word vectors, visually display the changes in public opinion information, improve the ability to observe public opinion situations, enhance the security of public opinion-related information during public opinion analysis, and help identify irrelevant generated content; construct a public opinion dynamic map with the help of public opinion circular moving points and public opinion circular change points, which helps to understand the dynamic situation and dissemination process of public opinion; through the focus area and the observation area, focus on analyzing the influence degree of the focus area of public opinion information, and improve the accuracy of public opinion analysis; combine the heat evaluation value, heat factor, and negative factor to analyze the focus area, and according to the obtained public opinion trend value, help public opinion managers understand the development trend of network public opinion, formulate targeted public opinion management strategies, and at the same time help identify potential risks or emergencies, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0055] Figure 1 It is a schematic diagram of an embodiment proposed by the present invention.

[0056] DETAILED DESCRIPTION OF THE EMBODIMENTS Embodiment 1, as Figure 1 shown, a multi-modal public opinion monitoring intelligent analysis method proposed by the present invention includes the following steps:

[0057] S1. Obtain public opinion source information and source time, analyze the public opinion source information to obtain public opinion word vectors, and construct a public opinion distribution map according to the public opinion word vectors and source time;

[0058] S2. Obtain real-time public opinion source information and real-time source time, obtain public opinion circular moving points and public opinion circular change points according to the real-time public opinion source information and real-time source time, obtain a public opinion dynamic map according to the public opinion circular change points, public opinion circular moving points, and public opinion distribution map, and analyze the public opinion dynamic map to obtain a focus area and an observation area;

[0059] S3. Analyze the focus area in combination with the public opinion dynamic map to obtain a heat evaluation value, analyze the heat evaluation value to obtain a heat factor and a negative factor, and analyze the heat factor and the negative factor to obtain a public opinion trend value;

[0060] S4. Analyze the public opinion trend value to obtain a public opinion management strategy.

[0061] It should be further noted that in the specific implementation process, the process of obtaining public opinion source information and source time, analyzing the public opinion source information to obtain public opinion word vectors, and constructing a public opinion distribution map according to the public opinion word vectors and source time is as follows:

[0062] Obtain the location where public opinion is generated, set up an information collection end according to the obtained location where public opinion is generated. The information collection end consists of several information capture points, and the information capture points are used to obtain public opinion source information and the source time. The public opinion source information includes public opinion text information, public opinion pictures, and public opinion videos. The source time refers to the time when the information capture point obtains the public opinion source information.

[0063] Analyze the public opinion source information to obtain diverse text information.

[0064] It should be further noted that in the specific implementation process, the specific process of analyzing the public opinion source information is as follows: By capturing the dialogue content in the public opinion video, converting the captured dialogue content into text information to obtain video text information; Based on the public opinion pictures, generating public opinion picture description text information through natural language generation technology to obtain picture text information; Through the source time corresponding to the public opinion source information, the public opinion text information, video text information, and picture text information corresponding to the public opinion source information with the same source time are recorded as diverse text information.

[0065] Through natural language processing technology, identify the words in the diverse text information to obtain text words, and number the text words, denoted as j, where j = 1, 2,..., n, n > 0 and n is an integer; Analyze the text words through the GloVe algorithm to obtain the public opinion word vector, denoted as [u j , v j , where u j and v j are the values obtained through the GloVe algorithm.

[0066] It should be further noted that in the specific implementation process, the specific process of analyzing the text words through the GloVe algorithm is as follows: Construct a vocabulary table of text words, record the co-occurrence frequency of each text word with other words through the vocabulary table, and construct a co-occurrence matrix; The GloVe algorithm optimizes the co-occurrence matrix to make the inner product of the public opinion word vectors closest to the logarithm of the co-occurrence probability of the text words, and optimizes the co-occurrence matrix by minimizing the loss function to obtain the public opinion word vectors.

[0067] Obtain the text sentences of the diverse text information, and number the text sentences, denoted as e, where e = 1, 2,..., m, m > 0 and m is an integer. Calculate the weighted average of the public opinion word vectors of each text word in the text sentence to obtain the public opinion sentence vector, denoted as [u e , v e , where u e is the weighted average of the u j of all text words in the text sentence numbered e, and v e is the v of all text words in the text sentence numbered e.j The weighted average value;

[0068] Let u e be the abscissa of the rectangular coordinate system, and v e be the ordinate of the rectangular coordinate system. Mark all public opinion sentence vectors in the rectangular coordinate system, denoted as public opinion point e, and set the radius standard value and the standard number of points; obtain the distances between all public opinion points, perform distribution processing on the distances between public opinion points, and obtain the public opinion distribution area;

[0069] It should be further noted that in the specific implementation process, the specific process of performing distribution processing on the distances between public opinion points is as follows: when there are a public opinion points whose distances from the public opinion point e are less than or equal to the radius standard value, and the value of a is greater than or equal to the standard number of points, then mark the public opinion point e as the public opinion center point, and mark the a public opinion points as the public opinion points inside the circle; repeat the above operation for all public opinion points until there are no changes in the public opinion center point and the public opinion points inside the circle. The area obtained by drawing a circle with the public opinion center point as the center and the radius standard value as the center radius is denoted as the public opinion distribution area;

[0070] Convert the source time into a timestamp, denoted as the source timestamp, obtain the source time of the text words corresponding to the public opinion distribution area, construct an association relationship between the public opinion center points of the public opinion distribution areas with the same source time, and mark the source timestamp into the association relationship. According to the public opinion distribution area, the association relationship, and the source timestamp, construct a public opinion distribution map;

[0071] It should be further noted that in the specific implementation process, the public opinion distribution map will continuously generate the public opinion distribution maps corresponding to each source time through the real-time acquisition of public opinion source information and the source time.

[0072] It should be further noted that in the specific implementation process, obtain the real-time public opinion source information and the real-time source time. According to the real-time public opinion source information and the real-time source time, obtain the public opinion circle moving point and the public opinion circle changing point. According to the public opinion circle changing point, the public opinion circle moving point, and the public opinion distribution map, obtain the public opinion dynamic map. The process of analyzing the public opinion dynamic map to obtain the focused area and the observation area is as follows:

[0073] Obtain the real-time public opinion source information and the real-time source time, and through the real-time public opinion source information and the real-time source time, obtain the corresponding public opinion distribution map, denoted as the online public opinion distribution map;

[0074] Analyze the online public opinion distribution map and the public opinion distribution map. Correspond the coordinates of the points inside the public opinion circles in each public opinion distribution area of the online public opinion distribution map with the coordinates of the points inside the public opinion circles in the corresponding public opinion distribution areas of the public opinion distribution map. If there is a non-overlapping situation in the coordinates of the points inside the public opinion circles of the online public opinion distribution map and the coordinates of the corresponding points inside the public opinion circles of the public opinion distribution map, the non-overlapping situation includes an increase, a decrease, and a change in the points inside the public opinion circles of the public opinion distribution map. When the points inside the public opinion circles of the public opinion distribution map increase or change, mark the corresponding increased or changed points inside the public opinion circles of the online public opinion distribution map at the corresponding positions in the public opinion distribution map, denoted as solid circle moving points; when the points inside the public opinion circles decrease, mark the decreased points inside the public opinion circles of the public opinion distribution map at the corresponding positions in the public opinion distribution map, denoted as empty circle moving points, and denote the solid circle moving points and the empty circle moving points as public opinion circle moving points;

[0075] Perform a weighted average calculation on the coordinates of all points (except the public opinion circle center points) in the public opinion distribution area corresponding to the public opinion circle moving points to obtain public opinion circle change points;

[0076] Construct the dynamic change relationship between the public opinion circle change points and the corresponding public opinion circle center points in the public opinion distribution map, and update the public opinion circle change points, solid circle moving points, empty circle moving points, and the dynamic change relationship into the public opinion distribution map to obtain the public opinion dynamic map;

[0077] Analyze the public opinion distribution map. When there is a corresponding dynamic change relationship for the public opinion circle center point, denote the public opinion distribution area corresponding to the public opinion circle center point as the focused area; when there is no corresponding dynamic change relationship for the public opinion circle center point, denote the public opinion distribution area corresponding to the public opinion circle center point as the observation area.

[0078] It should be further noted that in the specific implementation process, the process of analyzing the focused area in combination with the public opinion dynamic map to obtain the heat evaluation value, analyzing the heat evaluation value to obtain the heat factor and the negative factor, and analyzing the heat factor and the negative factor to obtain the public opinion trend value is as follows:

[0079] Combined with the public opinion dynamic map, analyze each focused area, count the number of solid circle moving points and empty circle moving points in the focused area, denoted as the number of solid points and the number of empty points respectively, and obtain the heat evaluation value R according to the number of solid points and the number of empty points h ;

[0080]

[0081] where mps1 and mps2 refer to the heat influence coefficients of the solid circle moving points on the focused area and the heat influence coefficients of the empty circle moving points on the focused area; R hRefers to the heat evaluation value of the h-th focused area; ds refers to the number of real points; ks refers to the number of empty points; w refers to the value obtained by subtracting the area occupied by the moving real circle point and the sum of the areas of the points within the public opinion circle from the area occupied by the moving empty circle point; S refers to the area of the public opinion distribution area;

[0082] Mark the heat evaluation values of each focused area at the variable points of the public opinion circle. The focused areas with heat evaluation values greater than 0 are recorded as heat areas; the focused areas with heat evaluation values less than 0 are recorded as negative areas; the focused areas with heat evaluation values equal to 0 remain as focused areas;

[0083] Analyze the heat areas to obtain the heat factor Y f ;

[0084]

[0085] Among them, Y f Refers to the f-th focused heat area; ds refers to the number of real points; ns refers to the total number of points within the public opinion circle;

[0086] Analyze the negative areas to obtain the negative factor Y g ;

[0087]

[0088] Among them, Y g Refers to the g-th focused negative area; ks refers to the number of empty points; ns refers to the total number of points within the public opinion circle;

[0089] Record the heat area corresponding to the maximum heat factor as the focused heat area; record the negative area corresponding to the maximum negative factor as the focused negative area;

[0090] Compare the diverse text information corresponding to the focused heat area with the diverse text information corresponding to each heat area to obtain the heat similarity of the heat area; compare the diverse text information corresponding to the focused negative area with the diverse text information corresponding to each negative area to obtain the negative similarity of the negative heat area;

[0091] Set the spread similarity; the spread similarity is related to the influence of the heat similarity and the negative similarity on the focused heat area and the focused negative area respectively; record the heat area corresponding to the heat similarity greater than or equal to the spread similarity as the heat spread area; record the heat area corresponding to the heat similarity greater than or equal to the spread similarity as the negative spread area;

[0092] According to the heat factor of the heat spread area and the heat evaluation value of the focused heat area, obtain the public opinion trend value Qt of the focused heat area 热 ;

[0093]

[0094] Among them, R b represents the heat evaluation value of the focused heat area; Y i represents the heat factor of each heat-affected area; c represents the total number of heat-affected areas; D i represents the heat similarity of each heat-affected area;

[0095] According to the negative factor of the negatively affected area and the heat evaluation value of the focused negative area, obtain the public opinion trend value of the focused negative area;

[0096]

[0097] Among them, R p represents the heat evaluation value of the focused negative area; Y l represents the negative factor of each negatively affected area; x represents the total number of negatively affected areas; D l represents the negative similarity of each negatively affected area.

[0098] It should be further noted that in the specific implementation process, the process of analyzing the public opinion trend value to obtain the public opinion management strategy is as follows:

[0099] Set the public opinion trend threshold of the focused heat area and the public opinion trend threshold of the focused negative area;

[0100] When the public opinion trend value of the focused heat area is greater than or equal to the public opinion trend threshold or the public opinion trend value of the focused negative area is greater than or equal to the public opinion trend threshold, it means that the public opinion risk of the heat event reflected by the focused heat area is high or the public opinion risk of the negative event in the focused negative area is high, denoted as public opinion level one; when the public opinion trend value of the focused heat area is less than the public opinion trend threshold, it means that the public opinion risk of the heat event reflected by the focused heat area is low or the public opinion risk of the negative event in the focused negative area is low, denoted as public opinion level two;

[0101] Send the public opinion level one and the public opinion level two to the relevant personnel, and the relevant personnel formulate the management strategy for the relevant event through the public opinion level one and the public opinion level two, denoted as the public opinion management strategy.

[0102] Embodiment 2. In a multi-modal public opinion monitoring intelligent analysis system proposed by the present invention, which is applied to a multi-modal public opinion monitoring intelligent analysis method described in Embodiment 1, specifically includes a management center, and the management center is communicatively connected to a data collection module, a data monitoring and analysis module, a data processing module, and a public opinion management module:

[0103] The data acquisition module is used to obtain the public opinion source information and the source time, analyze the public opinion source information to obtain the public opinion word vectors, and construct a public opinion distribution map according to the public opinion word vectors and the source time;

[0104] The data monitoring and analysis module is used to obtain the real-time public opinion source information and the real-time source time, obtain the public opinion circular moving points and the public opinion circular changing points according to the real-time public opinion source information and the real-time source time, obtain the public opinion dynamic map according to the public opinion circular changing points, the public opinion circular moving points and the public opinion distribution map, and analyze the public opinion dynamic map to obtain the focused area and the observed area;

[0105] The data processing module is used to analyze the focused area in combination with the public opinion dynamic map to obtain a heat evaluation value, analyze the heat evaluation value to obtain a heat factor and a negative factor, and analyze the heat factor and the negative factor to obtain a public opinion trend value;

[0106] The public opinion management module is used to analyze the public opinion trend value to obtain a public opinion management strategy.

[0107] The embodiments of the present invention have been described in detail above with reference to the accompanying drawings. However, the present invention is not limited thereto. Various changes can be made without departing from the spirit of the present invention within the knowledge of those skilled in the art of the relevant technology.

Claims

1. A multi-modal public opinion monitoring intelligent analysis method, characterized in that It includes the following steps: S1. Obtain the public opinion source information and the source time, analyze the public opinion source information to obtain the public opinion word vectors, and construct a public opinion distribution map according to the public opinion word vectors and the source time; S2. Obtain the real-time public opinion source information and the real-time source time, obtain the public opinion circular moving points and the public opinion circular changing points according to the real-time public opinion source information and the real-time source time, obtain the public opinion dynamic map according to the public opinion circular changing points, the public opinion circular moving points and the public opinion distribution map, and analyze the public opinion dynamic map to obtain the focused area and the observation area; S3. Analyze the focused area in combination with the public opinion dynamic map to obtain the heat evaluation value, analyze the heat evaluation value to obtain the heat factor and the negative factor, and analyze the heat factor and the negative factor to obtain the public opinion trend value; S4. Analyze the public opinion trend value to obtain the public opinion management strategy.

2. The multimodal public opinion monitoring intelligent analysis method according to claim 1, wherein The process of obtaining the public opinion source information and the source time, analyzing the public opinion source information to obtain the public opinion word vectors, and constructing a public opinion distribution map according to the public opinion word vectors and the source time includes: Obtain the location where the public opinion is generated, set up an information collection end according to the obtained location where the public opinion is generated, and the information collection end is composed of several information capture points, and the information capture points are used to obtain the public opinion source information and the source time; the public opinion source information includes public opinion text information, public opinion pictures and public opinion videos; Analyze the public opinion source information to obtain diverse text information; identify the words in the diverse text information through natural language processing technology to obtain text words; analyze the text words through the GloVe algorithm to obtain the public opinion word vectors; Obtain the text sentences of the diverse text information, and obtain the public opinion sentence vectors according to the public opinion word vectors of each text word in the text sentences; Mark all the public opinion sentence vectors in the rectangular coordinate system according to the public opinion sentence vectors, and set the radius standard value and the standard number of points; obtain the distances between all the public opinion points, and perform distribution processing on the distances between the public opinion points. The distribution processing specifically refers to: analyze the distances between each public opinion point through the standard number of points and the radius standard value to obtain the public opinion center point and the points inside the public opinion circle, and obtain the public opinion distribution area according to the public opinion center point and the radius standard value; Convert the source time into a time stamp, denoted as the source time stamp, obtain the source time of the text words corresponding to the public opinion distribution area, construct the correlation relationship between the public opinion center points of the public opinion distribution area through the source time, and construct the public opinion distribution map according to the public opinion distribution area, the correlation relationship and the source time stamp.

3. The multimodal public opinion monitoring intelligent analysis method according to claim 2, characterized in that, The process of obtaining the real-time public opinion source information and the real-time source time, obtaining the public opinion circular moving points and the public opinion circular changing points according to the real-time public opinion source information and the real-time source time, obtaining the public opinion dynamic map according to the public opinion circular changing points, the public opinion circular moving points and the public opinion distribution map, and analyzing the public opinion dynamic map to obtain the focused area and the observation area includes: Obtain the real-time public opinion source information and the real-time source time, and obtain the online public opinion distribution map through the real-time public opinion source information and the real-time source time; Analyze the online public opinion distribution map and the public opinion distribution map. Analyze the points inside the public opinion circles in each public opinion distribution area of the online public opinion distribution map and the points inside the public opinion circles in the corresponding public opinion distribution areas of the public opinion distribution map to obtain real-circle moving points and empty-circle moving points, and record the real-circle moving points and empty-circle moving points as public opinion circle moving points; Obtain public opinion circle variable points based on the points in the public opinion distribution areas corresponding to the public opinion circle moving points; Construct the dynamic change relationship between the public opinion circle variable points and the corresponding public opinion circle center points in the public opinion distribution map. Based on the public opinion circle variable points, real-circle moving points, empty-circle moving points, and the dynamic change relationship, obtain the public opinion dynamic map; Analyze the public opinion distribution map and analyze the public opinion circle center points through the dynamic change relationship to obtain the focus area and the observation area.

4. A multi-modal public opinion monitoring intelligent analysis method according to claim 3, characterized in that The process of analyzing the focus area in combination with the public opinion dynamic map to obtain the heat evaluation value is as follows: According to the public opinion dynamics graph and the focused area, count the number of solid circle moving points and hollow circle moving points within the focused area, which are respectively recorded as the number of solid points and the number of hollow points, and obtain the heat evaluation value R according to the number of solid points and the number of hollow points h ; Among them, mps1 and mps2 refer to the heat influence coefficients of the real circle moving points on the focused area and the heat influence coefficients of the empty circle moving points on the focused area; R h refers to the heat evaluation value of the h-th focused area; ds refers to the number of real points; ks refers to the number of empty points; w refers to the value obtained by subtracting the area occupied by the empty circle moving points from the sum of the area occupied by the real circle moving points and the area occupied by the points within the public opinion circle; S refers to the area of the public opinion distribution area.

5. A multi-modal public opinion monitoring intelligent analysis method according to claim 4, characterized in that, The process of analyzing the heat evaluation value to obtain the heat factor and the negative factor includes: Analyze the heat evaluation value and the focus area to obtain the heat area and the negative area; Analyze the heat region to obtain the heat factor Y f ; Among them, Y f refers to the f-th focused popularity area; ds refers to the number of real points; ns refers to the total number of points within the public opinion circle; Analyze the negative area to obtain the negative factor Y g ; Among them, Y g refers to the g-th focused negative region; ks refers to the number of empty points; ns refers to the total number of points within the public opinion circle.

6. A multimodal public opinion monitoring intelligent analysis method according to claim 1 or 5, characterized in that, The process of analyzing the heat factor and the negative factor to obtain the public opinion trend value includes: Analyze the heat factor and the heat area to obtain the focused heat area; analyze the negative factor and the negative area to obtain the focused negative area; Obtain the heat similarity of the heat area according to the diverse text information corresponding to the focused heat area and the diverse text information corresponding to each heat area; obtain the negative similarity of the negative heat area according to the diverse text information corresponding to the focused negative area and the diverse text information corresponding to each negative area; Set the spread similarity; analyze the spread similarity, heat similarity, and heat area to obtain the heat spread area and the negative spread area; Obtain the public opinion trend value Qt of the focused heat area based on the heat factor of the heat-affected area and the heat evaluation value of the focused heat area 热 ; Among them, R b represents the heat evaluation value of the focused heat area; Y i represents the heat factor of each heat-affected area; c represents the total number of heat-affected areas; D i represents the heat similarity of each heat-affected area; Obtain the public opinion trend value of the focused negative area according to the negative factor of the negative spread area and the heat evaluation value of the focused negative area; Among them, R p represents the heat evaluation value focusing on the negative area; Y l represents the negative factors of each negatively affected area; x represents the total number of negatively affected areas; D l represents the negative similarity of each negatively affected area.

7. A multi-modal public opinion monitoring intelligent analysis method according to claim 6, characterized in that, The process of analyzing the public opinion trend value to obtain the public opinion management strategy includes: Set the public opinion trend threshold of the focused heat area and the public opinion trend threshold of the focused negative area; When the public opinion trend value of the focused heat area is greater than or equal to the public opinion trend threshold or the public opinion trend value of the focused negative area is greater than or equal to the public opinion trend threshold, it is recorded as public opinion level one; when the public opinion trend value of the focused heat area is less than the public opinion trend threshold, it is recorded as public opinion level two; Send public opinion level one and public opinion level two to relevant personnel. The relevant personnel formulate management strategies for relevant events through public opinion level one and public opinion level two to obtain the public opinion management strategy.

8. A multi-modal public opinion monitoring intelligent analysis system, specifically applied to a multi-modal public opinion monitoring intelligent analysis method according to any one of claims 1 to 7, including a management center, characterized in that, The management center is communicatively connected to a data collection module, a data monitoring and analysis module, a data processing module, and a public opinion management module: The data collection module is used to obtain public opinion source information and source time, analyze the public opinion source information to obtain public opinion word vectors, and construct a public opinion distribution map according to the public opinion word vectors and the source time; The data monitoring and analysis module is used to obtain real-time public opinion source information and real-time source time, obtain the public opinion circular moving points and public opinion circular change points according to the real-time public opinion source information and real-time source time, obtain the public opinion dynamic map according to the public opinion circular change points, public opinion circular moving points and public opinion distribution map, analyze the public opinion dynamic map, and obtain the focused area and the observation area; The data processing module is used to analyze the focused area in combination with the public opinion dynamic map to obtain a heat evaluation value, analyze the heat evaluation value to obtain a heat factor and a negative factor, and analyze the heat factor and the negative factor to obtain a public opinion trend value; The public opinion management module is used to analyze the public opinion trend value to obtain a public opinion management strategy.

Images