A method and system for determining attributes and weights of public event response decisions

Abstract

The present invention discloses a method and system for determining attributes and weights of public event response decisions, the method comprising: using a large language model to obtain a model-recommended attribute set and weights according to the current state of the public event and public opinion; extracting the attribute set of public concern from the public opinion of the current public event, and calculating the public's preference measurement, emotional urgency and topic popularity for each attribute, and comprehensively obtaining the public opinion weight of the attribute of public concern; fusing the model recommendation and the attribute set of public concern to obtain the standard attribute set of the current public event response decision; and simultaneously synthesizing the model-recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set of the current public event response decision. The present invention can generate a comprehensive list including key decision attributes, and introduce public opinion into the dynamic adjustment process of attribute weights to generate response decisions that better reflect the public's immediate needs and concerns.

Description

Technical Field

[0001] The present invention belongs to the technical field of data processing for emergency decision-making, and in particular relates to a method and system for determining attributes and weights of public event response decisions. Background Art

[0002] In the process of responding to major public emergencies, the determination of decision attribute sets and their weights is the core link in formulating effective emergency strategies. The attribute set includes all event characteristics that may affect the development of the event and its response measures. The attribute weights reflect the relative importance of different event characteristics in affecting the development of the event and response measures, which is crucial for the rational allocation of resources and the setting of priorities.

[0003] However, existing technologies still have certain limitations in identifying the attributes and determining the weights of major public emergencies. Traditional methods mostly rely on manual experience or data analysis. [1,2] ,This method is easily limited by subjective judgment and data processing capabilities, and it is difficult to quickly and accurately grasp the multiple key attributes of an event and their relative importance in the early stages of the event.,First, the construction of traditional attribute sets often relies on manual screening and expert judgment. [3] ,This method is not only inefficient when dealing with massive information and rapidly changing events, but it is also prone to missing key attributes or including irrelevant attributes, which affects the accuracy of decision making.,Secondly, the determination of attribute weights is usually based on historical data or ,expert experience [4,5] ,These methods are difficult to capture the complex correlation of event attributes and the ,temporal dynamics of weights, which may lead to that the weight allocation may ,not match the actual situation, thus affecting the efficiency and ,effectiveness of emergency response. Summary of the invention

[0004] The present invention provides a method and system for determining the attributes and weights of public event response decisions, which can generate a comprehensive list including key decision attributes, and introduce public opinions into the dynamic adjustment process of attribute weights to generate response decisions that better reflect the public's immediate needs and concerns.

[0005] In order to achieve the above technical objectives, the present invention adopts the following technical solutions:

[0006] A method for determining attributes and weights of public event response decisions, including:

[0007] Step 1: Input the current state of public events, public opinions, and decision-making experience of historical public events into the large language model. By learning the decision-making experience of historical public events, a list of attributes for the current public event response decision is generated, which is recorded as the model recommendation attribute set.

[0008] Step 2: Input the model suggested attribute set, the current state of the public event, and public opinion into the large language model to generate the weight of each attribute in the model suggested attribute set, which is recorded as the model suggested weight;

[0009] Step 3, obtaining public opinions on current public events and extracting attribute sets of public concern;

[0010] Step 4: For each attribute that the public is concerned about: on the one hand, the public's preference measurement and emotional urgency are obtained through preference opinion evaluation and emotional opinion analysis; on the other hand, the topic popularity is calculated based on the frequency of public opinion expression; then the public's preference measurement, emotional urgency and topic popularity are combined to obtain the public opinion weight of the attribute that the public is concerned about;

[0011] Step 5: Fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for the current public event response decision-making; at the same time, combine the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for the current public event response decision-making.

[0012] Furthermore, when using the large language model to generate an attribute list, the input and output of the model are carried out in a question-and-answer format: first, the question setting criteria and question input template are clarified; second, the public event response decision is imported into the large language model as a scenario, and the decision-making role of the large language model is clarified to guide the large language model to refer to the decision-making experience of historical public events and output the attribute list in accordance with the specified answer specification requirements.

[0013] Furthermore, when extracting the attribute set of public concern, the obtained public opinion text is first segmented by combining the question-specific dictionary and the stop dictionary, and then the word vector and sentence vector are constructed based on the word2vec technology. Finally, the attribute set of public concern is obtained by LDA topic clustering.

[0014] Furthermore, the public's preference opinions are evaluated on the attributes of public concern to obtain the public's preference measurement, including:

[0015] (1) Establishing a six-level granularity language term set LTSs and its preference dictionary and probabilistic language term set. The language term set includes six-level granularity language terms: super, most, extraordinarily, more, slightly, and not at all; the preference dictionary is used to store the correspondence between the public's preference opinion expressions and the terms in the language term set;

[0016] (2) Use the preference dictionary to normalize the preference opinion expression of the public's concerned attributes into language terms of various levels of granularity, and then calculate the probability of each language term of each level of granularity of the public's concerned attributes to obtain the preference opinion expression of the probabilistic language term set of the public's opinion;

[0017] (3) Aggregate the probabilistic language term sets of each public concern attribute to express the preference opinions and obtain the public concern attribute Preference measure :

[0018]

[0019] In the formula, For properties of public concern The total frequency of preference opinions expressed, The number of attributes that are of public concern; is the intermediate variable, , is the probability of the k-th level granularity language term, is the weight vector of the k-th level granularity language term in the language term set LTS.

[0020] Furthermore, sentiment analysis is performed on the attributes that the public is concerned about to obtain the public's emotional urgency, including:

[0021] (1) Based on the sentiment polarity of each sentiment word provided by the sentiment dictionary, the sentiment polarity of public opinion sentiment words under the public concern attribute is divided;

[0022]

[0023] In the formula, Attributes for public attention The i-th public opinion sentiment word under Sentiment words The emotional polarity;

[0024] (2) Based on the attributes of public concern The sentiment polarity of public opinion sentiment words under the above conditions is used to calculate the public attention attribute Emotional urgency :

[0025] .

[0026] Furthermore, the topic popularity is calculated based on the frequency of public opinion expression. The specific calculation formula is:

[0027]

[0028] In the formula, For attributes The topic popularity, For the public to The frequency of expressing opinions, The number of attributes that the public is concerned about.

[0029] Furthermore, the public’s preference opinions, emotional urgency and topic popularity are integrated to obtain the public opinion weight of the attribute that the public is concerned about. The specific calculation formula is:

[0030]

[0031] In the formula, For attributes The weight of public opinion, For attributes The public preference measure, For attributes The public's emotional urgency, For attributes The popularity of public topics.

[0032] Furthermore, the fusion model suggested attribute set and the attribute set of public concern is specifically to take the union of the model suggested attribute set S and the attribute set of public concern A to obtain the standard attribute set Z;

[0033] If an attribute in the standard attribute set Z does not belong to the attribute set S, the model recommendation weight of the attribute is set to 0; if an attribute in the standard attribute set Z does not belong to the attribute set A, the public opinion weight of the attribute is set to 0; then the weights of each attribute in the standard attribute set Z are calculated according to the following formula:

[0034]

[0035] In the formula, is the number of attributes in the standard attribute set Z, For attributes The model suggests weights, For attributes The weight of public opinion, and They are large language models and public attribute A and B are weighting factor sets, and the values ​​of the weighting factors are determined by experts through consultation.

[0036] Furthermore, the current public events are divided into different development stages, and the corresponding response decisions are made based on the current state and the real-time public opinion generation stage; and at different development stages, the large language model and the public's attributes are adjusted in real time. The weighting factor , .

[0037] A system for determining attributes and weights of public event response decisions, comprising:

[0038] The model suggestion attribute generation module is used to: input the current public event status, public opinion, and decision-making experience of historical public events into the large language model, learn from the decision-making experience of historical public events, and generate a list of attributes for the current public event response decision, which is recorded as the model suggestion attribute set;

[0039] The model suggestion weight generation module is used to: input the model suggestion attribute set, the current state of public events and public opinions into the large language model, and generate the weight of each attribute in the model suggestion attribute set, which is recorded as the model suggestion weight;

[0040] The public concern attribute extraction module is used to: obtain the public opinions on the current public events and extract the attribute set of public concern;

[0041] The public opinion weight determination module is used to: for each attribute that the public is concerned about: on the one hand, obtain the public's preference measurement and emotional urgency through preference opinion evaluation and emotional opinion analysis; on the other hand, calculate the topic popularity based on the frequency of public opinion expression; then combine the public's preference measurement, emotional urgency and topic popularity to obtain the public opinion weight of the attribute that the public is concerned about;

[0042] The standard attribute and weight fusion module is used to: fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for current public event response decisions; at the same time, comprehensively consider the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for current public event response decisions.

[0043] Compared with the prior art, the present invention has the following technical effects:

[0044] 1. By integrating the analytical capabilities of large language models and public opinions, attribute sets and weights are generated intelligently, so that when major public emergencies occur, decision-making attribute sets and their weights can be quickly determined, making emergency response more rapid and efficient.

[0045] 2. Strengthen public participation in the decision-making process. By integrating public opinions on social media, the decision-making process for public events has changed from expert decisions to public participation, making decision-making closer to public opinion and enhancing social acceptance of decisions.

[0046] 3. By learning historical experience through large language models and introducing real-time public opinions through social media, experts can flexibly adjust decision-making priorities based on real-time information during the process of formulating decision attributes, thereby improving the adaptability of decision-making results.

[0047] 4. Promoted scientific decision-making. By quantitatively analyzing the weights of large model recommendations and public opinion weights, the determination of decision attributes and weights no longer depends solely on the subjective opinions of experts, providing data support for decision-making. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 It is the framework of the method and system described in the embodiments of the present application;

[0049] Figure 2 This is the process of determining the large language model decision attribute set in the embodiment of the present application;

[0050] Figure 3 It is the large language model decision attribute weight mining process of the embodiment of the present application;

[0051] Figure 4 This is the public discussion topic mining process of the embodiment of the present application;

[0052] Figure 5 This is the public opinion preference evaluation process of the embodiment of the present application. DETAILED DESCRIPTION

[0053] The following is a detailed description of an embodiment of the present invention. This embodiment is based on the technical solution of the present invention, and provides a detailed implementation method and a specific operation process to further explain the technical solution of the present invention.

[0054] Example 1

[0055] This embodiment provides a method for determining the attributes and weights of public event response decisions, using a generatively trained large language model (LLM) to determine a set of decision-making attributes when a major public emergency occurs based on historical case experience. The large language model can generate a comprehensive list of key decision attributes by learning historical event data, current event status, and public opinion information in social media. These attributes cover all aspects of the event, such as the scope of impact, urgency, resource requirements, etc. In addition, the present invention introduces public opinion into the dynamic adjustment process of attribute weights, so that public event response decisions can be closer to the actual situation and better reflect the public's immediate needs and concerns.

[0056] The attributes and weight determination method of the public event response decision provided in this embodiment are as follows: Figure 1 The framework diagram shown includes the following basic steps:

[0057] Step 1: The large language model determines the attribute set, such as Figure 2As shown in the figure. The current state of public events, public opinions, and decision-making experience of historical public events are input into the large language model. By learning the decision-making experience of historical public events, the large language model is guided to generate a list of attributes for the current public event response decision based on the scenario introduction and answer specification requirements of the decision-making scenario, which is recorded as the model recommended attribute set. This step not only ensures the integrity of the attribute set, but also predicts the attributes that may appear in new events through the model's predictive ability.

[0058] The large language model determines the basic attribute set of new event decision-making based on the decision-making case experience of historical public events. The input and output of the model are carried out in a question-and-answer format.

[0059] First, clarify the question setting criteria and templates. Due to the model framework structure design, the large language model requires clear questions to enable the model to better understand the requirements and quickly obtain accurate answers, which is very important for improving decision-making efficiency and ensuring smooth human-computer interaction. To this end, the present invention proposes question setting criteria and templates for decision-making scenarios, as shown in Table 1 and Table 2, respectively.

[0060]

[0061]

[0062] Secondly, import the decision-making scenario and clarify the decision-making role. The excellent context perception ability of the large language model gives it the ability to set roles. Before participating in decision-making problems, it clarifies the decision-making scenario and role, guides it to refer to historical documents, and specifies the answer format and requirements, which can effectively ensure the quality of subsequent output content. To this end, the scenario introduction and answer specification requirements of the decision-making scenario are proposed, as shown in Table 3.

[0063]

[0064] In summary, we can obtain the attribute set recognized by the large language model.

[0065] Step 2: Perform weight mining based on the large language model, such as Figure 3 The model-recommended attribute set, the current state of public events, and public opinion are input into the large language model to generate the weight of each attribute in the model-recommended attribute set, which is recorded as the model-recommended weight.

[0066] LLM in the analysis and decision-making stage Based on the state statistics of the output, the recommended attribute experience weights and explanation information are defined as ,Right now:

[0067]

[0068] in, Import and standardize requirements for decision-making scenarios, For specific questions, For attributes The large model recommends weights, The number of attributes suggested for the model.

[0069] Step 3: Extract public attention attributes, such as Figure 4 As shown in the figure, obtain the public opinions on the current public event and extract the attribute set of public concern.

[0070] The public opinion information aggregation process obtains the public's real-time opinion expressions at the current stage, crawls the public's real-time opinion preferences at the current stage on social media, preprocesses the text data, performs word segmentation, removes stop words, constructs word vectors, constructs sentence vectors, etc., converts the text data into a data form that can be processed by machine learning, and finally obtains the attribute set of public concern through LDA topic clustering.

[0071] Step 4: For each attribute of public concern: on the one hand, obtain the public's preference measure and emotional urgency through preference opinion evaluation and emotional opinion analysis respectively; on the other hand, calculate the topic popularity based on the frequency of public opinion expression; then, combine the public's preference measure, emotional urgency and topic popularity to obtain the public opinion weight of the attribute of public concern.

[0072] Step 4.1, preference opinion evaluation. The degree words of the public opinions under the same public concern attribute are used as the preference opinion expression of its probabilistic language terminology. The frequency of the degree words of the same level is used as the probability of preference of the probabilistic language term set. The public preference value under each decision attribute is calculated with the help of the comparison operator of the probabilistic language term set. .like Figure 5 shown.

[0073] First, establish a six-level granularity language term set LTSs and a corresponding probabilistic language term set PLTS:

[0074]

[0075] in, For the Level-granular language terms, arrive Corresponding to: {super, most, exceptionally, more, slightly, not a bit} respectively.

[0076]

[0077] in, It is a language term and its corresponding probability The combination of .

[0078] Then, a preference dictionary (Preference NLP, PreNLP) of the language term set LTSs is established. With the help of the established preference dictionary, the public's preference opinions are standardized, and then the decision-making stage is statistically analyzed. The probability of expressing the public's preference opinions at each level of granularity language terminology under each decision attribute .

[0079] Finally, aggregate the probability of expressing the preference opinions of each public concern attribute at each level of granularity language terminology , get the public attention attribute The preference measure is defined as ,Right now:

[0080] ,in

[0081] in, Decision attribute The total frequency of preference opinions expressed, is the number of attributes that the public cares about, is the probability of expressing the preference opinion of the k-th level granularity language term; is the weight vector of the language term set, which is determined according to the preference dictionary PreNLP, so it is not an arithmetic progression.

[0082] Step 4.2, sentiment analysis. The preference opinions in step 4.1 only express the public's opinion tendency, and cannot tell the public's emotional urgency for the attribute opinion. This embodiment integrates the emotional urgency based on the preference opinions and combines the decision attribute topic popularity to obtain the final public opinion weight.

[0083] The method based on the sentiment dictionary method in the present invention analyzes the sentiment urgency of public opinion content under decision-making attributes, and realizes the division and measurement of sentiment polarity at different granularities according to the sentiment polarity of sentiment words in the sentiment dictionary. According to the characteristics of social media data, the BosonNLP Chinese sentiment dictionary is selected as the training word bag.

[0084] Firstly, based on the sentiment polarity of each sentiment word provided by the sentiment dictionary, the sentiment polarity of public opinion sentiment words under the public concern attribute is divided;

[0085]

[0086] In the formula, Attributes for public attention The i-th public opinion sentiment word under Sentiment words The emotional polarity;

[0087] Then, according to the public attention attribute The sentiment polarity of public opinion sentiment words under the above conditions is used to calculate the public attention attribute Emotional urgency :

[0088] .

[0089] Step 4.3: Mining the popularity of public topics. Based on the frequency of public opinions expressed under each attribute, calculate the popularity of public topics. The topic popularity is defined as ,Right now

[0090]

[0091] In the formula, For attributes The topic popularity, For the public to The frequency of expressing opinions, The number of attributes that the public is concerned about.

[0092] 4.4, Comprehensive calculation of public opinion weight. Based on the popularity of decision-making attribute topics, the public opinion preference and public opinion emotional urgency are integrated to obtain the public opinion weight. .Right now

[0093] .

[0094] Step 5: Fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for the current public event response decision-making; at the same time, combine the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for the current public event response decision-making.

[0095] The fusion model recommended attribute set and the attribute set of public concern are specifically performed by taking the union of the model recommended attribute set S and the attribute set A of public concern to obtain the standard attribute set Z.

[0096] If an attribute in the standard attribute set Z does not belong to the attribute set S, the model recommendation weight of the attribute is set to 0; if an attribute in the standard attribute set Z does not belong to the attribute set A, the public opinion weight of the attribute is set to 0; then the weights of each attribute in the standard attribute set Z are calculated according to the following formula:

[0097]

[0098] In the formula, is the number of attributes in the standard attribute set Z, For attributes The model suggests weights, For attributes The weight of public opinion, and They are large language models and public attribute A and B are weighting factor sets, and the values ​​of the weighting factors are determined by experts through consultation.

[0099] This step 5 obtains the comprehensive attribute weights by weighted fusion of the large model suggestion weights and public opinion weights, ensuring the effective integration and weighing of multi-source information in the decision-making process.

[0100] In a more preferred embodiment, the current public event can also be divided into different development stages, and the corresponding response decision can be made based on the current stage status and the real-time public opinion generation stage; and at different development stages, the large language model and the public's attribute can be adjusted in real time. The weighting factor , , and obtain the most appropriate public event response decisions at each stage of development.

[0101] Example 2

[0102] This embodiment provides a system for determining attributes and weights of public event response decisions, including:

[0103] The model suggestion attribute generation module is used to: input the current public event status, public opinion, and decision-making experience of historical public events into the large language model, learn from the decision-making experience of historical public events, and generate a list of attributes for the current public event response decision, which is recorded as the model suggestion attribute set;

[0104] The model suggestion weight generation module is used to: input the model suggestion attribute set, the current state of public events and public opinions into the large language model, and generate the weight of each attribute in the model suggestion attribute set, which is recorded as the model suggestion weight;

[0105] The public concern attribute extraction module is used to: obtain the public opinions on the current public events and extract the attribute set of public concern;

[0106] The public opinion weight determination module is used to: for each attribute that the public is concerned about: on the one hand, obtain the public's preference measurement and emotional urgency through preference opinion evaluation and emotional opinion analysis; on the other hand, calculate the topic popularity based on the frequency of public opinion expression; then combine the public's preference measurement, emotional urgency and topic popularity to obtain the public opinion weight of the attribute that the public is concerned about;

[0107] The standard attribute and weight fusion module is used to: fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for current public event response decisions; at the same time, comprehensively consider the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for current public event response decisions.

[0108] The specific working principles and implementation methods of the modules of the attribute and weight determination system for public event response decision-making described in this embodiment are the same as those in Example 1 and will not be repeated in this embodiment.

[0109] The above embodiments are preferred embodiments of the present application. Ordinary technicians in this field can also make various changes or improvements on this basis. Without departing from the overall concept of the present application, these changes or improvements should fall within the scope of protection required by the present application.

[0110] References:

[0111] [1] Chen Xiaohong, Xu Xuanhua, Zeng Jianghong. Multi-attribute large group decision making method based on entropy weight[J]. Systems Engineering and Electronics, 2007(07): 1086-1089.

[0112] [2] Guo Kaihong, Li Wenli. Multi-attribute group decision-making method and its extension under unknown weight information[J]. Chinese Journal of Management Science, 2011, 19(05): 94-103.

[0113] [3] Xu Xuanhua, Liu Ying, Chen Xiaohong. Large group risk emergency decision-making method based on group pressure[J]. Journal of Systems Engineering, 2022, 37(4): 460-476.

[0114] [4] Xu X, Yin X, Chen X. A large-group emergency risk decision method based on data mining of public attribute preferences[J]. Knowledge-BasedSystems, 2019, 163: 495-509.

[0115] [5] Cao J, Xu X H, Yin X P, et al. A risky large group emergencydecision-making method based on topic sentiment analysis[J]. Expert Systemswith Applications, 2022, 195: 116527。

Claims

1. A method for determining attributes and weights of public event response decisions, characterized in that: include: Step 1: Input the current state of public events, public opinions, and decision-making experience of historical public events into the large language model. By learning the decision-making experience of historical public events, a list of attributes for the current public event response decision is generated, which is recorded as the model recommendation attribute set. Step 2: Input the model suggested attribute set, the current state of the public event, and public opinion into the large language model to generate the weight of each attribute in the model suggested attribute set, which is recorded as the model suggested weight; Step 3, obtaining public opinions on current public events and extracting attribute sets of public concern; Step 4: For each attribute that the public cares about: on the one hand, through preference opinion evaluation and sentiment opinion analysis, the public's preference measurement and sentiment urgency are obtained respectively; On the other hand, based on the frequency of public opinion expression, the topic popularity is calculated; then the public preference measurement, emotional urgency and topic popularity are combined to obtain the public opinion weight of the public's attention attribute; The public's preference opinions are evaluated on the attributes of public concern to obtain the public's preference measurement, including: (1) Establishing a six-level granularity language term set LTSs and its preference dictionary and probabilistic language term set. The language term set includes six-level granularity language terms: super, most, extraordinarily, more, slightly, and not at all; the preference dictionary is used to store the correspondence between the public's preference opinion expressions and the terms in the language term set; (2) Use the preference dictionary to normalize the preference opinion expression of the public's concerned attributes into language terms of various levels of granularity, and then calculate the probability of each language term of each level of granularity of the public's concerned attributes to obtain the preference opinion expression of the probabilistic language term set of the public's opinion; (3) Aggregate the probabilistic language term sets of each public concern attribute to express the preference opinions and obtain the public concern attribute Preference measure : ； In the formula, For properties of public concern The total frequency of preference opinions expressed, The number of attributes that are of public concern; is the intermediate variable, , is the probability of the k-th level granularity language term, is the weight vector of the k-th level granularity language term in the language term set LTSs; The public's emotional urgency is obtained by analyzing the attributes of public concern, including: (1) Based on the sentiment polarity of each sentiment word provided by the sentiment dictionary, the sentiment polarity of public opinion sentiment words under the public concern attribute is divided; ； In the formula, Attributes for public attention The i-th public opinion sentiment word under Sentiment words The emotional polarity; (2) Based on the attributes of public concern The sentiment polarity of public opinion sentiment words under the above conditions is used to calculate the public attention attribute Emotional urgency : ； Step 5: Fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for the current public event response decision-making; at the same time, combine the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for the current public event response decision-making.

2. The method for determining attributes and weights of public event response decisions according to claim 1, characterized in that: When using a large language model to generate an attribute list, the model input and output are conducted in a question-answering manner: first, the question setting criteria and question input template are clarified; Secondly, public event response decisions are imported into the big language model as scenarios, and the decision-making role of the big language model is clarified to guide the big language model to refer to the decision-making experience of historical public events and output a list of attributes in accordance with the specified answer specifications.

3. The method for determining attributes and weights of public event response decisions according to claim 1, characterized in that: When extracting the attribute set of public concern, we first segment the obtained public opinion text by combining the question-specific dictionary and the stop dictionary, then build word vectors and sentence vectors based on word2vec technology, and finally obtain the attribute set of public concern by LDA topic clustering.

4. The method for determining attributes and weights of public event response decisions according to claim 1, characterized in that: The topic popularity is calculated based on the frequency of public opinion expression. The specific calculation formula is: ； In the formula, For attributes The topic popularity, For the public to The frequency of expressing opinions, The number of attributes that the public is concerned about.

5. The method for determining attributes and weights of public event response decisions according to claim 1, characterized in that: Combining the public's preference measurement, emotional urgency and topic popularity, we can get the public opinion weight of the attribute that the public is concerned about. The specific calculation formula is: ； In the formula, For attributes The weight of public opinion, For attributes The public preference measure, For attributes The public's emotional urgency, For attributes The popularity of public topics.

6. The method for determining attributes and weights of public event response decisions according to claim 1, characterized in that: The fusion model suggested attribute set and the attribute set of public concern is specifically to take the union of the model suggested attribute set S and the attribute set of public concern A to obtain the standard attribute set Z; If an attribute in the standard attribute set Z does not belong to the attribute set S, the model recommendation weight of the attribute is set to 0; if an attribute in the standard attribute set Z does not belong to the attribute set A, the public opinion weight of the attribute is set to 0; then the weights of each attribute in the standard attribute set Z are calculated according to the following formula: ； In the formula, is the number of attributes in the standard attribute set Z, For attributes The model suggests weights, For attributes The weight of public opinion, and They are large language models and public attribute A and B are weighting factor sets, and the values ​​of the weighting factors are determined by experts through consultation.

7. The method for determining attributes and weights of public event response decisions according to claim 6, characterized in that: According to the different development stages of current public events, corresponding response decisions are made based on the current state and real-time public opinion generation stage; and at different development stages, the large language model and public opinion on attributes are adjusted in real time. The weighting factor , .

8. A system for determining attributes and weights of public event response decisions, characterized in that: include: The model suggestion attribute generation module is used to: input the current public event status, public opinion, and decision-making experience of historical public events into the large language model, learn from the decision-making experience of historical public events, and generate a list of attributes for the current public event response decision, which is recorded as the model suggestion attribute set; The model suggestion weight generation module is used to: input the model suggestion attribute set, the current state of public events and public opinions into the large language model, and generate the weight of each attribute in the model suggestion attribute set, which is recorded as the model suggestion weight; The public concern attribute extraction module is used to: obtain the public opinions on the current public events and extract the attribute set of public concern; The public opinion weight determination module is used to: for each attribute that the public is concerned about: on the one hand, obtain the public's preference measurement and emotional urgency respectively through preference opinion evaluation and emotional opinion analysis; On the other hand, based on the frequency of public opinion expression, the topic popularity is calculated; then the public preference measurement, emotional urgency and topic popularity are combined to obtain the public opinion weight of the public's attention attribute; The public's preference opinions are evaluated on the attributes of public concern to obtain the public's preference measurement, including: (1) Establishing a six-level granularity language term set LTSs and its preference dictionary and probabilistic language term set. The language term set includes six-level granularity language terms: super, most, extraordinarily, more, slightly, and not at all; the preference dictionary is used to store the correspondence between the public's preference opinion expressions and the terms in the language term set; (2) Use the preference dictionary to normalize the preference opinion expression of the public's concerned attributes into language terms of various levels of granularity, and then calculate the probability of each language term of each level of granularity of the public's concerned attributes to obtain the preference opinion expression of the probabilistic language term set of the public's opinion; (3) Aggregate the probabilistic language term sets of each public concern attribute to express the preference opinions and obtain the public concern attribute Preference measure : ； In the formula, For properties of public concern The total frequency of preference opinions expressed, The number of attributes that are of public concern; is the intermediate variable, , is the probability of the k-th level granularity language term, is the weight vector of the k-th level granularity language term in the language term set LTSs; The public's emotional urgency is obtained by analyzing the attributes of public concern, including: (1) Based on the sentiment polarity of each sentiment word provided by the sentiment dictionary, the sentiment polarity of public opinion sentiment words under the public concern attribute is divided; ； In the formula, Attributes for public attention The i-th public opinion sentiment word under Sentiment words The emotional polarity; (2) Based on the attributes of public concern The sentiment polarity of public opinion sentiment words under the above conditions is used to calculate the public attention attribute Emotional urgency : ； The standard attribute and weight fusion module is used to: fuse the model recommended attribute set and the attribute set of public concern to obtain the standard attribute set for current public event response decisions; at the same time, comprehensively consider the model recommended weights and public opinion weights of each attribute in the attribute set to obtain the attribute weight set for current public event response decisions.

Images