A regional stance characterization method and system

Abstract

The application relates to the technical field of social media stance analysis, and discloses a regional stance characterization method and system, which comprises the following steps: extracting a user group from a social media user group; performing entity feature word and camp feature word analysis on the text published by the users in the sampled user group to obtain a first stance distribution prediction result; performing stance and personality trait double labeling on the users with high activity, using the labeled data to train a personality-stance correlation model, analyzing the personality traits of the users with low activity and inferring the stance, and statistically obtaining a second stance distribution prediction result; and performing weighted fusion on the first stance distribution prediction result and the second stance distribution prediction result to generate a final regional stance characterization. The application realizes accurate and full quantitative insight into the regional stance by establishing a quantitative mapping relationship among user behavior, personality traits and potential stance, and accurately sampling according to real world population distribution.

Description

Technical Field

[0001] This invention relates to the field of social media stance analysis technology, specifically to a method and system for representing regional stances. Background Technology

[0002] Accurate and dynamic quantitative characterization of public opinion distribution within a specific geographical area is fundamental to developing effective strategies. Traditional methods relying on sampling surveys and questionnaires are not only costly and time-consuming to implement, but also susceptible to sampling frame bias and social desirability bias, leading to distorted results.

[0003] As social media becomes a crucial arena for public opinion expression, leveraging its massive content for stance analysis has emerged as a new technological approach. However, existing technologies largely focus on keyword matching and sentiment analysis of explicit statements. These methods face two major bottlenecks: First, a significant "digital divide" exists between online users and the real-world population in terms of age, region, and education, leading to severe sample representativeness bias when directly analyzing online data. Second, many users do not directly express their opinions, making their stances undetectable by current technologies, creating analytical blind spots. Therefore, a new technological framework is urgently needed that can penetrate data appearances, align online and offline distributions, and uncover implicit stances.

[0004] Current social media-based stance analysis techniques have significant shortcomings in terms of data representativeness, analytical coverage, and model depth, specifically manifested in the following ways:

[0005] First, the sample representativeness is severely insufficient. Existing methods typically analyze social media data collected in full or randomly, ignoring the spatiotemporal heterogeneity and structural biases between online user profiles and the actual population profiles of the target region. This biased approach makes it difficult for the analytical conclusions to accurately reflect the true distribution of attitudes in the region.

[0006] Second, the scope of stance identification is too narrow. The technology relies heavily on explicit text posted by users that contains clear entity or issue keywords. For users who never or rarely express relevant opinions publicly, existing technologies are largely ineffective, causing the analysis model to miss major groups in society, thus casting doubt on the reliability of the conclusions.

[0007] Third, the analytical models are too narrow in scope and lack in-depth correlation mining. Most methods remain at a superficial level of text sentiment analysis, failing to establish a correlation model between users' deep psychological traits (such as personality) and their underlying stances. This prevents the models from inferring users' inclinations on specific issues from their daily behavior, limiting the depth and breadth of the analysis. Summary of the Invention

[0008] To address the aforementioned technical problems, this invention provides a method and system for representing regional stances. By establishing a quantitative mapping relationship between user behavior, personality traits, and potential stances, and by conducting precise sampling based on real-world population distribution, it achieves accurate and comprehensive quantitative insights into regional stances.

[0009] Specifically, this invention combines macro-demographic characteristics of regional groups with micro-level network behavior data to construct a representative sample space that accurately reflects the overall structure, addressing the problem of traditional network analysis being divorced from reality. This invention mines and utilizes users' inherent personality traits to effectively infer the stances of low-activity users who have not explicitly expressed opinions, constructing a cross-domain transfer learning model from user behavior to their potential stances. This addresses the problem of being unable to detect the stances of low-activity users and overcomes survivor bias in network data. This invention effectively integrates direct stance prediction based on macro-demographics with indirect stance inference based on micro-level individual personality traits to generate a comprehensive and accurate profile of regional stance distribution.

[0010] To solve the above-mentioned technical problems, the present invention adopts the following technical solution:

[0011] In a first aspect, the present invention provides a method for characterizing a regional position, comprising:

[0012] Based on macro-population profile data of the target region, a sample user group that is consistent with the total user group in terms of multidimensional demographic characteristics is extracted from the social media user group.

[0013] Using a large language model, entity feature words and camp feature words are analyzed on the text posted by users in the sampled user group to identify the users' final stance and statistically obtain the first stance distribution prediction results.

[0014] Based on psychological personality theory, a large language model is used to label users with activity levels above a set condition with both stance and personality traits. The labeled data is used to train a personality-stance association model, and then the personality traits of users with activity levels below the set condition are analyzed and their stances are inferred. The predicted results of the second stance distribution are obtained statistically.

[0015] The prediction results of the first and second position distributions are dynamically weighted and fused using non-negative least squares modeling to generate the final regional position representation.

[0016] In one embodiment, the step of extracting a sample user group from the social media user group based on the macro-demographic profile data of the target region, which is consistent with the total user group in terms of multidimensional demographic characteristics, specifically includes:

[0017] Obtain macro-level demographic data for the target region, assuming the total user group is... ,Include Each user Having multidimensional statistical eigenvectors ,in, On behalf of users The Statistical characteristics The total number of statistical characteristics; the social media user group is denoted as ,from A subset of samples is selected from the data, denoted as the sampled user group. The number of users in the sampled user group is , making Distribution on the set key statistical features and Consistent.

[0018] In one embodiment, the step of using a large language model to analyze entity feature words and camp feature words in the text posted by users in a sampled user group to identify the user's final stance specifically includes:

[0019] Collect sample user groups users in Text posted on social media And define entity feature words related to the issue to be analyzed. and faction characteristic words ;

[0020] Utilizing large language models The system analyzes the text posted by each user and determines their stance based on entity feature words and faction feature words, respectively.

[0021] ;

[0022] ;

[0023] By integrating entity positions and faction stance To obtain user final position :

[0024] ;

[0025] For fusion operation.

[0026] In one embodiment, the statistics yield a first position distribution prediction result, specifically including:

[0027] Statistical sampling user group The final stance of each user, and the predicted first stance distribution. It is a sampled user group Frequency distribution vector of each user's position:

[0028] ;

[0029] , , These represent the number of users who support, oppose, and are neutral, respectively. This indicates the number of users in the sampled user group.

[0030] In one embodiment, the method of using a large language model to label users with activity levels exceeding a set threshold based on psychological personality theory includes:

[0031] Within a social media user base, a user is considered highly active if their posted texts contain specific keywords and the number of such posts is greater than or equal to a set threshold. For each highly active user... Utilizing large language models At the same time, indicate the corresponding stance. And the Big Five personality score ,Right now:

[0032] ;

[0033] These represent the scores for openness, conscientiousness, extraversion, agreeableness, and emotional stability in the Big Five personality traits. for The published text, This indicates a prompt related to stance detection. This indicates information related to the Big Five personality traits score.

[0034] In one embodiment, training the personality-position association model using labeled data specifically includes:

[0035] Highly active users Using the Big Five personality traits score as a feature and stance as a label, labeled data is constructed. An XGBoost classification model is trained as a personality-position association model to learn the quantitative relationship between personality traits and specific positions. The prediction function of the XGBoost classification model is represented as the sum of multiple decision trees.

[0036] ;

[0037] Indicates to The final prediction result made from the position, This represents the k-th decision tree in the XGBoost classification model. express The parameters, This represents the total number of decision trees in the XGBoost classification model.

[0038] In one embodiment, the analysis of personality traits and inference of stances of users with activity levels below a set condition, and the statistical acquisition of a second stance distribution prediction result, specifically includes:

[0039] In social media user groups, if a user's posted text does not contain specific keywords or the number of posted texts is less than a set value, then the user is considered a low-activity user.

[0040] We used a large language model to analyze the text posted by inactive users and obtained their corresponding Big Five personality traits scores. The Big Five personality traits scores of inactive users are input into a trained personality-position association model to predict the positions of inactive users. :

[0041] ;

[0042] Second position distribution prediction results It is the frequency distribution vector of the stances of all users in the social media user group.

[0043] In one embodiment, the step of dynamically weighting and fusing the prediction results of the first and second position distributions using non-negative least squares modeling to generate the final regional position representation specifically includes:

[0044] set up Given a position distribution vector from a trusted data source, model the following problem as a nonnegative least squares optimization problem: find a set of nonnegative weight vectors. This makes the weighted combination of the first position distribution prediction results and the second position distribution prediction results close to... ; They represent The weights of the first position distribution prediction results and the second position distribution prediction results in the equation;

[0045] The weight vector obtained after solving the nonnegative least squares optimization problem Used to calculate the final regional position representation :

[0046] .

[0047] The prediction results are for the first position distribution. The prediction results are for the second position distribution. represents the weights of the optimized first position distribution prediction result and the second position distribution prediction result, respectively.

[0048] In one embodiment, the nonnegative least squares optimization problem is specifically:

[0049] ;

[0050] ;

[0051] in, This represents the L2 norm.

[0052] In a second aspect, the present invention provides a computer system including a memory and a processor, wherein the memory stores a computer program, and the processor executes the computer program to implement the steps of the method of any embodiment of the first aspect.

[0053] Compared with the prior art, the beneficial technical effects of the present invention are:

[0054] 1. Significant improvements in comprehensiveness and accuracy were achieved. Precise feature sampling ensured that the demographic structure of the analyzed sample was consistent with the real-world population, resolving the problem of traditional network analysis being divorced from reality. Simultaneously, by inferring the stances of low-activity users through personality trait models, the "survivorship bias" of traditional network analysis was overcome, resulting in a more comprehensive and realistic representation of stances.

[0055] 2. Strong robustness and good complementarity. This invention integrates two logically different analytical paths: one top-down (from macro-demographic structure to stance) and the other bottom-up (from individual character to stance). The two can verify and complement each other, avoiding the limitations of a single analytical perspective and greatly enhancing the stability of the results.

[0056] 3. Achieved deep insights. This invention can not only predict "what" (position distribution) but also partially explain "why" (personality association). By constructing a correlation model between personality and position, it provides a new perspective for understanding the behavioral preferences of groups with different personality traits, and has higher application value. Attached Figure Description

[0057] Figure 1 This is a flowchart of the method in an embodiment of the present invention;

[0058] Figure 2 This is a schematic diagram of the overall framework in an embodiment of the present invention. Detailed Implementation

[0059] A preferred embodiment of the present invention will now be described in detail with reference to the accompanying drawings.

[0060] like Figure 1 As shown, a regional position characterization method of the present invention includes the following steps:

[0061] S1, based on macro-population profile data of the target region, extracts a sampled user group from the social media user group that is consistent with the total user group in terms of multidimensional demographic characteristics distribution;

[0062] S2 uses a large language model to analyze entity feature words and camp feature words in the text posted by users in the sampled user group, identifies the user's final stance, and statistically obtains the first stance distribution prediction results.

[0063] S3, based on psychological personality theory, uses a large language model to label users with activity levels above a set condition with both stance and personality traits. The labeled data is used to train a personality-stance association model, and then the personality traits of users with activity levels below the set condition are analyzed and their stances are inferred. The second stance distribution prediction results are obtained statistically.

[0064] S4 uses non-negative least squares modeling to dynamically weight and fuse the prediction results of the first and second position distributions to generate the final regional position representation.

[0065] This invention proposes a comprehensive regional position representation framework. By implementing two different but complementary analytical models in parallel and then weighted and fusing the prediction results of the two models, it achieves accurate prediction of the position distribution in a specific region. The overall framework is as follows: Figure 2 As shown.

[0066] 1. Stance prediction based on precise feature sampling.

[0067] This invention combines macro-level population profile data and micro-level social media data, and through precise sampling and large-scale language model analysis, directly predicts the stance of the sample group.

[0068] Specifically, acquire macro-level demographic data for the target region, including multi-dimensional statistical characteristics such as age, gender, geographic distribution, education level, and occupational distribution. Let the total user group be... ,Include Each user Having multidimensional statistical eigenvectors ,in Representing the Several statistical characteristics (such as age, gender, geographical distribution, education level, occupational distribution, etc.). Meanwhile, the social media user group is denoted as... The present invention is from A subset of samples is selected from the data, denoted as the sampled user group. The sample size is , making Distribution of key demographic characteristics and Highly consistent.

[0069] Collect sample user groups Chinese users Text posted on social media Define characteristic words for entities (such as products, brands, public figures, slogans, etc.) related to the topic to be analyzed. and faction characteristic words .

[0070] Using a large language model (LLM), the text posted by each user is analyzed through carefully designed prompts, based on entity feature words. and faction characteristic words Assess their stance:

[0071] ;

[0072] .

[0073] By integrating entity positions and faction stance To arrive at each user's final stance. This fusion can be achieved using a simple majority voting method:

[0074] .

[0075] First position distribution prediction results It is a sampled user group Frequency distribution vector of each user's position:

[0076] ;

[0077] , , These represent the number of users who support, oppose, and are neutral, respectively. This indicates the number of users in the sampled user group.

[0078] 2. Inferences based on personality traits.

[0079] This section utilizes the Big Five personality theory in psychology to establish a personality-position association model to predict the positions of users who do not directly express opinions online. The Big Five personality theory is an important tool in psychology used to describe and understand human personality. It includes five main dimensions: Openness to Experience, Conscientiousness, Extraversion, Agreeableness, and Neuroticism, each representing a series of related personality traits.

[0080] Based on keyword matching and limitations on the number of texts published, social media user groups Divided into two categories: This is a highly active user group; they post a lot of texts that are relevant to the topic. This is a low-activity user group; the texts they post are either irrelevant to the topic or are few in number. For Each user in Utilizing large language models At the same time, indicate its position. And the Big Five personality score ,Right now:

[0081] ;

[0082] These represent the scores for openness, conscientiousness, extraversion, agreeableness, and emotional stability in the Big Five personality traits. for The published text, This indicates a prompt related to stance detection. This indicates information related to the Big Five personality traits score.

[0083] Stance detection related prompts Example:

[0084] I will provide some posts, all from the same user:

[0085] {Post A},

[0086] {Post B},

[0087] …

[0088] Question: What is this user's stance on the {"specific issue"}? Please choose one answer from: support, opposition, or neutral.

[0089] Example of Big Five personality score-related prompts:

[0090] I will provide some posts, all from the same user:

[0091] {Post A},

[0092] {Post B},

[0093] …

[0094] Question: Based on all of this user's tweets, please determine and complete the Big Five personality traits assessment questionnaire for this user, and calculate the questionnaire results.

[0095] Annotated data is constructed using the Big Five personality traits scores of highly active users as features and their stances as labels. Train an XGBoost classification model As a personality-position association model, the XGBoost classification model learns the quantitative relationship between personality traits and specific positions. The prediction function of the XGBoost classification model can be represented as the sum of multiple decision trees:

[0096] ;

[0097] Indicates to The final prediction result made from the position, This represents the k-th decision tree in the XGBoost classification model. express The parameters, This represents the total number of decision trees in the XGBoost classification model.

[0098] For inactive users, a large language model is also used to analyze their posted text to obtain their Big Five personality score. The Big Five personality traits scores of inactive users are input into a trained personality-position association model to predict the positions of these inactive users. :

[0099] ;

[0100] Second position distribution prediction results It is the frequency distribution vector of the stances of all users in the social media user group.

[0101] 3. Fusion of prediction results from the two models.

[0102] The prediction results from the two modules are dynamically fused using a weighted model to output the final regional stance representation. The detailed steps are as follows:

[0103] set up The position distribution vector (as the true value) comes from a trusted data source. This invention seeks a set of non-negative weight vectors. This makes the weighted combination of the two prediction results as close as possible. This problem can be modeled as a nonnegative least squares optimization problem:

[0104] ;

[0105] ;

[0106] in This represents the L2 norm (Euclidean distance).

[0107] The weights obtained after solving the above optimization problem Used to calculate the final regional position characterization result. :

[0108] .

[0109] The terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the invention. The terms “comprising,” “including,” etc., as used herein indicate the presence of the stated features, steps, operations, and / or components, but do not exclude the presence or addition of one or more other features, steps, operations, or components.

[0110] It should be understood that although the steps in the flowcharts of the accompanying drawings are shown sequentially as indicated by the arrows, these steps are not necessarily executed in the order indicated by the arrows. Unless explicitly stated herein, there is no strict order restriction on the execution of these steps, and they can be executed in other orders. Moreover, at least some of the steps in the flowcharts of the accompanying drawings may include multiple steps or stages, which are not necessarily completed at the same time, but may be executed at different times, and the execution order of these steps or stages is not necessarily sequential, but may be performed alternately or in turn with other steps or at least some of the steps or stages of other steps.

[0111] In one embodiment, the present invention provides a computer system, which may be a server. The computer system includes a processor, memory, and a network interface connected via a system bus. The processor provides computing and control capabilities. The memory includes a non-volatile storage medium and internal memory. The non-volatile storage medium stores an operating system, computer programs, and a database. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage medium. The database stores data used in the methods described above. The network interface communicates with external terminals via a network connection. The computer program is executed by the processor to implement the methods described above.

[0112] The technical features of the above embodiments can be combined in any way. For the sake of brevity, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.

[0113] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the invention can be implemented in other specific forms without departing from its spirit or essential characteristics. Therefore, the embodiments should be considered in all respects as exemplary and non-limiting, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be included within the present invention, and no reference numerals in the claims should be construed as limiting the scope of the claims.

[0114] Furthermore, it should be understood that although this specification describes embodiments, not every embodiment contains only one independent technical solution. This narrative style is merely for clarity. Those skilled in the art should consider the specification as a whole, and the technical solutions in each embodiment can also be appropriately combined to form other embodiments that can be understood by those skilled in the art.

Claims

1. A method for representing regional positions, characterized in that, The method comprises the following steps: extracting a sample user group from a social media user group based on target area macro population portrait data, which is consistent with the total user group in multi-dimensional demographic feature distribution; using a large language model to analyze entity feature words and camp feature words in the text published by users in the sample user group, identifying the final stance of the users, and obtaining the first stance distribution prediction result; based on the psychological personality theory, using a large language model to perform dual labeling of users with high activity on stance and personality traits, training a personality-stance correlation model using the labeled data, and then analyzing the personality traits of users with low activity and inferring the stance, obtaining the second stance distribution prediction result; dynamically weighting and fusing the first stance distribution prediction result and the second stance distribution prediction result through non-negative least squares modeling to generate the final regional stance representation.

2. The regional stand characterization method of claim 1, wherein, The method comprises the following steps: Obtain macroscopic population portrait data of the target region, let the total user group be , containing users, each user has a multi-dimensional statistical feature vector , wherein, represents the first statistical feature of the user , and is the total number of statistical features; the social media user group is denoted as , a sample subset is extracted from , denoted as the sampled user group , the number of users in the sampled user group is , so that the distribution on the set key statistical features is consistent with .

3. The regional stand characterization method of claim 1, wherein, The method comprises the following steps: Collecting a sample user group Users in the sample user group Text posted on social media and defining entity features and camp features related to the topic to be analyzed and camp features ; Utilizing large language models and prompts, the text posted by each user is analyzed to determine the corresponding stance based on entity keywords and camp keywords, respectively: ； ； By fusing entity stance and camp stance , the final stance of the user is derived: ； is a fusion operation.

4. The regional stand characterization method of claim 3, wherein, The method comprises the following steps: statistical sampling of a user population the first stance distribution prediction result is the sampled user population the frequency distribution vector of stances of the users ； , , respectively represent the number of users with a stance of support, opposition, and neutral, represent the number of users in the sample user population.

5. The regional stand characterization method of claim 1, wherein, The method comprises the following steps: In a social media user group, if a user posts text containing a specific keyword and the number of posts is greater than or equal to a set value, the user is a high-activity user; for each high-activity user , a large language model is used to simultaneously label the corresponding stance and Big Five personality scores, that is: ； respectively represent openness score, conscientiousness score, extraversion score, agreeableness score and emotional stability score in the Big Five personality, to the published text, representing a stance detection related hint, representing a Big Five personality score related hint.

6. The regional stand characterization method of claim 5, wherein, The method comprises the following steps: Highly active users Using the Big Five personality traits score as a feature and stance as a label, labeled data is constructed. An XGBoost classification model is trained as a personality-position association model to learn the quantitative relationship between personality traits and specific positions. The prediction function of the XGBoost classification model is represented as the sum of multiple decision trees. ； final prediction result made on the stance of kth decision tree in the XGBoost classification model, parameters of total number of decision trees in the XGBoost classification model.​​ 7. The regional stand characterization method of claim 1, wherein, The method comprises the following steps: The method comprises the following steps: analyzing text posted by low-active users using a large language model to obtain corresponding big five personality scores ; inputting the big five personality scores of the low-active users into a personality-stance correlation model that is trained to completion to predict stances of the low-active users : ； Second stance distribution prediction result is a frequency distribution vector of stances of all users in the social media user group.

8. The regional stand characterization method of claim 1, wherein, The method comprises the following steps: Let be the stance distribution vector from the trusted data source, model the following problem as a non-negative least squares optimization problem: seek a set of non-negative weight vectors such that the weighted combination of the first stance distribution prediction and the second stance distribution prediction is close to ; denote the weights for the first stance distribution prediction and the second stance distribution prediction in , respectively. weight vector obtained after solving a non-negative least squares optimization problem , for computing a final regional stance representation : ； is a first stance distribution prediction result, is a second stance distribution prediction result, respectively represent the weights of the optimized first and second stance distribution prediction results.

9. The regional stand characterization method of claim 8, wherein, The method comprises the following steps: ； ； wherein denotes the L2 norm.

10. A computer system comprising a memory and a processor, said memory storing a computer program, characterized in that, The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method comprises the following steps: The method

Images