A method and device for intelligent recommendation of complex events in grassroots social grid governance

Abstract

This invention discloses an intelligent recommendation method and device for complex event mechanisms in grassroots social grid governance. The method includes: connecting to a third-party information platform to obtain pending events in grassroots social grid governance via interface or direct database connection; extracting a first feature from the pending events; using a simple / complex classification model, classifying the pending events into simple or complex events based on the first feature; if the pending event is a complex event, recommending a corresponding processing mechanism for it through vectorized retrieval and machine learning methods, obtaining a first probability and a second probability of belonging to each processing mechanism, ranking the first and second probabilities, and recommending the processing mechanism with the highest probability. This enables rapid response and effective handling of complex events. The application of intelligent technology can further improve the accuracy and efficiency of governance, injecting new vitality into grassroots social governance.

Description

Technical Field

[0001] This invention belongs to the field of digital social governance, and in particular relates to an intelligent recommendation method and device for complex event mechanisms in grassroots social grid governance. Background Technology

[0002] At present, there is no authoritative and scientific method for handling complex events in grassroots social grid governance. Mediation is mainly carried out through means such as conflict source governance and multi-party consultation. These methods rely more on the level of economic development and expert experience, resulting in problems such as unclear responsibilities between horizontal organizations and uncoordinated vertical relationships. There is a lack of scientific and standardized guiding norms for grassroots social grid governance.

[0003] It is widely acknowledged that deep learning models have achieved good results in the field of natural language processing. In the complex events of grassroots social grid governance, the mainstream intelligent recommendation technology currently mainly uses natural language models such as BERT and Transformer to encode and classify text for output.

[0004] In practice, it has been found that current intelligent mechanism recommendation technologies, especially in the case of complex mechanism relationships and event elements, rely heavily on deep learning models for classification and prediction, which makes it difficult to achieve both accuracy and efficiency. Summary of the Invention

[0005] To address the problems existing in the prior art, the purpose of this application is to provide an intelligent recommendation method and apparatus for complex event mechanisms in grassroots social grid governance.

[0006] According to a first aspect of the embodiments of this application, an intelligent recommendation method for complex event mechanisms in grassroots social grid governance is provided, comprising: Connect to a third-party information platform to obtain pending events for grassroots social grid governance through interfaces or direct database connections, and extract the first feature from the pending events; Using a simple / complex classification model, based on the first feature, the events to be processed are divided into simple events or complex events; If the event to be processed is a complex event, then a corresponding processing mechanism is recommended for it through vectorized retrieval and machine learning methods. The first probability and the second probability of it belonging to each processing mechanism are obtained respectively. The first probability and the second probability are ranked, so as to recommend the processing mechanism with the highest probability.

[0007] Furthermore, a first feature is extracted from the event to be processed, specifically: If the event to be processed is structured data, then the first feature is extracted directly; If the event to be processed is unstructured data, the first feature is extracted through manual annotation and feature extraction.

[0008] Furthermore, a vectorized retrieval method is used to recommend appropriate processing mechanisms for the event to be processed, and the first probability of belonging to each processing mechanism is obtained, including: The text content of the event to be processed is vectorized and encoded. The encoded events are retrieved from the Elasticsearch database that stores event cases to obtain the most similar events. Based on the processing mechanism corresponding to each similar event, the first probability of the event to be processed belonging to each processing mechanism is set.

[0009] Furthermore, machine learning methods are used to recommend appropriate processing mechanisms for the events to be processed, and a second probability of belonging to each processing mechanism is obtained, including: Perform semantic reshaping on the events to be processed; Using a large language model, keywords are extracted and quantified from the semantically reshaped events to obtain event severity quantification values. Extract the second feature of the event to be processed, which includes the number of people involved, the quantitative value of the amount involved, the quantitative value of the severity of the event, the event type, and the event theme; Based on the second feature, the event to be processed is classified using a pre-trained multi-classification model to obtain the second probability of it belonging to each processing mechanism.

[0010] Furthermore, using a large language model, keywords are extracted and quantified from the semantically reshaped events to obtain event severity quantification values, specifically: Using a large language model, keywords are extracted from the semantically reshaped events to be processed; Based on the semantic ambiguity of the keywords and the degree of influence of the corresponding behaviors, the keywords are quantified to obtain the event severity quantification value.

[0011] Furthermore, the quantified value of the target amount is obtained in the following manner: The maximum value among the extracted amounts from the events to be processed is taken as the target amount; The target amount is mapped to the corresponding quantified value according to the predetermined segmentation range, thus obtaining the target amount quantified value.

[0012] According to a second aspect of the embodiments of this application, an intelligent recommendation device for complex event mechanisms in grassroots social grid governance is provided, comprising: The feature extraction module is used to connect to a third-party information platform to obtain pending events of grassroots social grid governance through interface or direct database connection, and extract the first feature from the pending events. The simplified / traditional classification module is used to classify the event to be processed into a simple event or a complex event based on the first feature using a simplified / traditional classification model. The mechanism recommendation module is used to recommend corresponding processing mechanisms for complex events by using vectorized retrieval and machine learning methods. It obtains the first probability and the second probability of each processing mechanism, sorts the first probability and the second probability, and recommends the processing mechanism with the highest probability.

[0013] According to a third aspect of the embodiments of this application, a computer program product is provided, including a computer program / instruction that, when executed by a processor, implements the method as described in the first aspect.

[0014] According to a fourth aspect of the embodiments of this application, an electronic device is provided, comprising: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors perform the method as described in the first aspect.

[0015] According to a fourth aspect of the embodiments of this application, a computer-readable storage medium is provided that stores computer instructions thereon, which, when executed by a processor, implement the steps of the method as described in the first aspect.

[0016] The technical solutions provided by the embodiments of this application may include the following beneficial effects: As described in the above embodiments, this application provides a method and apparatus for intelligent recommendation of mechanisms for complex events in grassroots social grid governance within the field of social governance. This application utilizes event data reported by grassroots staff or from third-party software platforms, fully combining expert experience and machine learning models to classify events into simple and complex types. For complex events, it then uses vectorized retrieval and machine learning methods for intelligent mechanism recommendation. Deep neural network models and big oracle models are used to fully analyze the inherent semantics, elements, and other factors of complex events, enabling prediction of processing mechanisms and achieving rapid response and effective handling of complex events. The application of intelligent technology further enhances the accuracy and efficiency of governance, injecting new vitality into grassroots social governance.

[0017] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0018] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0019] Figure 1 This is a flowchart illustrating an intelligent recommendation method for complex event mechanisms in grassroots social grid governance, based on an exemplary embodiment.

[0020] Figure 2 This is a flowchart illustrating step S3 according to an exemplary embodiment.

[0021] Figure 3 This is a block diagram of an intelligent recommendation device for complex event mechanisms in grassroots social grid governance, illustrated according to an exemplary embodiment.

[0022] Figure 4 This is a schematic diagram of an electronic device according to an exemplary embodiment. Detailed Implementation

[0023] Exemplary embodiments will now be described in detail, examples of which are illustrated in the accompanying drawings. When the following description relates to the drawings, unless otherwise indicated, the same numbers in different drawings represent the same or similar elements. The embodiments described in the following exemplary embodiments do not represent all embodiments consistent with this application.

[0024] The terminology used in this application is for the purpose of describing particular embodiments only and is not intended to be limiting of the application. The singular forms “a,” “the,” and “the” used in this application and the appended claims are also intended to include the plural forms unless the context clearly indicates otherwise. It should also be understood that the term “and / or” as used herein refers to and includes any or all possible combinations of one or more of the associated listed items.

[0025] It should be understood that although the terms first, second, third, etc., may be used in this application to describe various information, such information should not be limited to these terms. These terms are only used to distinguish information of the same type from one another. For example, without departing from the scope of this application, first information may also be referred to as second information, and similarly, second information may also be referred to as first information. Depending on the context, the word "if" as used herein may be interpreted as "when," "when," or "in response to determination."

[0026] Figure 1 This is a flowchart illustrating an intelligent recommendation method for complex event mechanisms in grassroots social grid governance, based on an exemplary embodiment, such as... Figure 1 As shown, the method may include the following steps: S1: Connect to a third-party information platform to obtain pending events for grassroots social grid governance through interfaces or direct database connections, and extract the first feature from the pending events; S2: Using a simple / complex classification model, based on the first feature, the events to be processed are divided into simple events or complex events; S3: If the event to be processed is a complex event, then a corresponding processing mechanism is recommended for it through vectorized retrieval and machine learning methods. The first probability and the second probability of it belonging to each processing mechanism are obtained respectively. The first probability and the second probability are sorted to recommend the processing mechanism with the highest probability.

[0027] As described above, this application relates to a method and apparatus for intelligent recommendation of mechanisms for complex events in grassroots social grid governance within the field of social governance. This application utilizes event data reported by grassroots staff or from third-party software platforms, fully combining expert experience and machine learning models to categorize events into simple and complex types. For complex events, it then employs vectorized retrieval and machine learning methods for intelligent mechanism recommendation. Deep neural network models and big oracle models are used to fully analyze the inherent semantics, elements, and other factors of complex events, enabling the prediction of processing mechanisms and facilitating rapid response and effective handling of complex events. Furthermore, the application of intelligent technologies further enhances the accuracy and efficiency of governance, injecting new vitality into grassroots social governance.

[0028] In a specific implementation of S1, the system connects to a third-party information platform to obtain pending events for grassroots social grid governance through an interface or direct database connection, and extracts a first feature from the pending events. Specifically, in the field of social governance, different regions have different information system platforms, whose data structures are largely similar, including basic user information, case details, and reporting time. These can be reported by staff or by users themselves. The reported events are then manually annotated using expert experience and structured through methods such as element extraction to extract a first feature. This first feature includes: the number of people involved, the amount involved, the details of the individuals involved, escalating behavior, potential violence, and the degree of urgency. In one embodiment, it can be set as follows: Number of people involved: at least one person. If the same incident involves multiple people, there is a possibility of a collective incident. Target Amount: The minimum target amount is 0 yuan. Generally, the larger the target amount, the higher the severity of the event. Information on those involved: There are two scenarios, namely, those who died and those who did not. Escalating Behaviors: Based on expert experience, a weighted list of escalating behaviors is generated by assigning weights to sensitive words (such as debts, strikes, etc.) contained in the event content.

[0029] Potential violence: Based on expert experience, potential violence (such as verbal altercations, fractures, traffic accidents, etc.) is divided into 9 levels and a form is constructed.

[0030] Urgency level: Based on expert experience, events are divided into four levels: major and urgent, major but not urgent, urgent but not major, and not major but urgent.

[0031] That is, based on six indicators representing the characteristics of events, the unstructured data of the grassroots social grid governance events are processed through manual annotation and element extraction technology to form structured data. The manual annotation includes assessments of potential violence, escalating behaviors, urgency, etc.

[0032] It should be noted that the acquired events may also be data that has been structured according to the above requirements, in which case no further structuring is required.

[0033] In a specific implementation of S2, a simple / complex classification model is adopted, and based on the first feature, the event to be processed is divided into simple events or complex events. Specifically, the structured data is binary-classified using a pre-trained machine learning model (logistic regression model) to classify it into simple events or complex events, as shown in Table 1 below: Table 1 Examples of Simple and Complex Events I live in XX Street, XX City. I need to apply for a temporary residence permit recently. A police officer from XX Community contacted me and asked me to go to the community to apply for the permit. I have some questions about this. Where should I go to apply for a temporary residence permit? Is there a fee? Please tell me the specific procedure. Simple events The XX Cement Products Factory near XX Village, XX Street, XX City is currently producing stone powder. When the wind is strong, all the dust is blown into the villagers' houses. The environmental protection department has previously ordered it to rectify the situation, but it has not yet made any environmental protection announcements and is still producing stone powder. Please have the relevant department verify and rectify the situation. Complex events If it is a simple event, no further processing will be performed; if it is a complex event, the processing in step S3 will continue.

[0034] In a specific implementation of S3, if the event to be processed is a complex event, a corresponding processing mechanism is recommended for it using vectorized retrieval and machine learning methods. The first probability and second probability of belonging to each processing mechanism are obtained, and the first and second probabilities are ranked to recommend the processing mechanism with the highest probability. Figure 2 As shown.

[0035] Specifically, this application involves three processing mechanisms: the first mechanism, the second mechanism, and the third mechanism. The first mechanism is operated by or led by the township level; the second mechanism is operated by or led by the village / community level; and the third mechanism is operated by or led by the district / county level. Specifically, the first mechanism addresses complex events at the township level, which require collaboration among multiple departments and are handled through a multi-faceted collaborative approach. The second mechanism addresses complex events at the village / community level, involving collective interests such as road and bridge construction, requiring consultation among multiple parties, with village / community cadres and residents negotiating to resolve the issues. The third mechanism addresses complex events at the district / county level, typically characterized by suddenness, urgency, or significant harm, by convening relevant officials and experts for discussion and assessment to arrive at a decision.

[0036] Specifically, by using a vectorized retrieval method to recommend appropriate processing mechanisms for the event to be processed, and obtaining the first probability of it belonging to each processing mechanism, the following can be included: S11: Vectorize the text content of the event to be processed; This application employs the M3E model series for text vectorization encoding. Since the events involved in this application are typically short text data, the basic Roberta model is used for encoding. Its input length is approximately 512 tokens, which is sufficient for most event text encoding tasks; if it exceeds this length, it is truncated. Its output is a 768-dimensional vector, meaning that the output vector length is fixed regardless of the input text length. In other words, an event is represented using a 768-dimensional vector. S12: Retrieve the encoded events in the Elasticsearch database that stores event cases, and obtain the most similar events; In practical applications, as the knowledge base grows, challenges arise in scaling the knowledge base vectors and ensuring efficient search capabilities. To address this issue, Elasticsearch software is introduced to enable CRUD operations on the knowledge base, as well as vectorized or keyword-based retrieval.

[0037] A 768-dimensional vector is stored in an Elasticsearch database, and cosine similarity is used as a metric to compare the similarity between events. Cosine similarity assesses the similarity between two non-zero vectors by measuring the cosine of the angle between them. Its value ranges from -1 to 1, where 1 indicates that they are in exactly the same direction, 0 indicates that they are orthogonal, and -1 indicates that they are in completely opposite directions. In text retrieval, cosine similarity is often used to assess the similarity between text vectors.

[0038] In Elasticsearch, the following steps are used to achieve vectorized retrieval of cosine values: S121: Create an index in Elasticsearch and define a field of type dense_vector to store the vector representation of the event, and store the encoded text in the dense_vector field; S123: Use Elasticsearch's vector retrieval feature to query the event with the most similar vector. In practice, this can be done by calculating the cosine similarity between the query vector and each event vector in the index, and then sorting the results based on the similarity score. S124: Results returned. Elasticsearch will return a list of events sorted by similarity score. These events are most similar to the query vector and contain the corresponding similarity scores. The top K events will be returned.

[0039] S13: Based on the processing mechanism corresponding to each similar event, set the first probability that the event to be processed belongs to each processing mechanism; Specifically, based on the processing mechanism corresponding to each similar event, the first probability of the event to be processed belonging to each processing mechanism is set. For example, if five similar events are retrieved, three of which are processed using the third mechanism and two of which are processed using the second mechanism, then the first probability of the third mechanism is set to 60%, the first probability of the second mechanism is set to 40%, and the first probability of the first mechanism is set to 0.

[0040] In practical implementation, a first probability can be set by combining the similarity between each similar event and the event to be processed. For example, if five similar events are retrieved, and events 1, 2, and 4 (with similarities of 95%, 90%, and 80% respectively) are processed using the third mechanism, while events 3 and 5 are processed using the second mechanism (with similarities of 85% and 60% respectively), then the first probability of the third mechanism is (95%+90%+80%) / 3*60%=54%, the first probability of the second mechanism is (85%+60%) / 2*40%=29%, and the first probability of the first mechanism is 0. It should be noted that this is only an example of setting the first probability based on similarity; other combinations can also be used in practical implementation.

[0041] While vectorized retrieval methods offer a simple and convenient way to obtain probabilities, they can sometimes be inaccurate and have limited recognition capabilities, such as when the similar events retrieved have low similarity to the event to be processed. Therefore, to achieve more accurate case classification, this invention employs a combination of Large Language Model (LLM) and machine learning for cases that cannot be directly matched using simple retrieval techniques, thus successfully classifying events.

[0042] Specifically, by using machine learning methods to recommend appropriate processing mechanisms for the event to be processed, and obtaining the second probability of it belonging to each processing mechanism, the following can be included: S21. Perform semantic reshaping on the event to be processed; Specifically, this step is part of event preprocessing. Events often have common problems such as colloquialisms and typos. Large language models can perform semantic reshaping to address these issues, thereby improving event quality.

[0043] Specifically, for semantically ambiguous event texts, feature extraction alone cannot effectively obtain key information, thus potentially affecting classification results. To address this type of text, this invention further employs the semantic reshaping function of large-scale modeling techniques to reprocess the text. In the first stage, a large language model is used for text rewriting to refine the case description while preserving its semantic content. This method ensures structural and semantic accuracy, thereby enhancing the semantic interpretation capabilities of traditional classification models. Large language models perform well in sentence restructuring within legal text analysis. As a generative model trained on a broad text corpus, large language models exhibit strong generalization capabilities in tasks related to text rewriting. Rewriting case texts using large language models not only preserves the original meaning but also improves clarity and coherence. Utilizing the semantic understanding of large language models, relevant keywords are placed in appropriate positions within the case text, enhancing readability. In this rewriting stage, large language models identify key semantic elements and seamlessly integrate them into the revised text, enriching the narrative. For subsequent classification tasks, accurate classification requires in-depth semantic understanding. Traditional machine learning methods heavily rely on feature engineering, which may not provide the semantic analysis depth required for accurate classification. Therefore, this application turns to neural network models for case classification, integrating the text rewriting and semantic interpretation functions of large language models. This greatly enhances the ability of neural network models to process and classify complex legal texts, thereby ensuring data readiness for machine learning tasks and improving model efficiency.

[0044] S22. Using a large language model, extract keywords from the semantically reshaped event to be processed and quantify them to obtain the event severity quantification value; For example, the first mechanism is used to handle complex events involving multiple departments; the second mechanism is used to handle complex events involving the common interests of residents in the same village or community, such as bridge construction or community reform; and the third mechanism is mainly used to handle urgent and major emergencies, such as landslides and urban flooding. Starting with the differences between the three mechanisms, and combining expert experience and large-scale models, we can achieve feature extraction and manual labeling of complex events, constructing a supervised machine learning dataset where each complex event is labeled using one of the three mechanisms.

[0045] This invention enables comprehensive analysis of word frequencies in the training set. Several feature groups are established to refine the keyword set, and representative keywords are selected to populate these sets based on statistical results and expert insights. Compared to traditional methods, large language models (MLMs) demonstrate superior capabilities in Named Entity Recognition (NER), thus they are used as feature extraction tools to extract key features. MLMs possess a deep understanding of the overall semantic logic of sentences, keywords, and text. They also effectively handle ambiguity in case texts; even without explicit keyword indication in the sentence, they can discern the semantics of keywords. During keyword extraction, cases where keywords are explicitly present in the text are classified as semantically precise. Conversely, cases where the text implies the meaning of keywords without directly mentioning them are considered semantically ambiguous. After keyword extraction, the feature groups constructed in the above steps are used to evaluate the cases, assigning scores to cases of varying complexity.

[0046] Based on the advice of legal experts and the importance of certain keywords, for sensitive keywords such as "house demolition," "law," and "traffic accident," regardless of whether they are semantically precise or ambiguous, as long as the large language model successfully extracts them, the corresponding feature will receive 1 point. Specifically, for "law," 1 point is awarded as long as any word in the keyword group is present. For "domestic violence" and "threat," a score of 1 is assigned if they are semantically precise, and 0.5 points if they are semantically ambiguous. For feature groups 1 and 2, based on the degree of adverse impact caused by their respective behaviors, 0.5 and 0.2 points are assigned when the semantics are precise, and half the points are assigned when the semantics are ambiguous, thus obtaining a quantitative value for the severity of the event.

[0047] S23: Extract the second feature of the event to be processed, the second feature including the number of people involved, the quantitative value of the amount involved, the quantitative value of the event severity, the event type, and the event theme; Specifically, the monetary value described in the event description is also important for determining the complexity of the case. Therefore, a "quantity" feature group is introduced. The maximum values ​​of various quantities in the text are extracted to calculate features, and the quantities are divided into six intervals using a box segmentation method, set in one embodiment as: [0, 1000), [1000, 10000), [10000, 50000), [50000, 100000), and [100000, +∞). Each interval corresponds to a different score range from 0 to 5. The event topic is obtained by filtering keywords extracted from the event content, and the event type is a manually labeled multi-level event tag.

[0048] S23. Based on the second feature, the event to be processed is classified using a pre-trained multi-classification model to obtain the second probability of it belonging to each processing mechanism; Specifically, the traditional machine learning model logistic regression algorithm is adopted. In this embodiment, it is a three-class classification model, which is trained using the second feature of the event and the annotation processing mechanism as labels.

[0049] After obtaining the first probability and the second probability of the event to be processed belonging to each mechanism, the first probability and the second probability are sorted, and the mechanism with the highest probability value is determined as the recommended mechanism for the event to be processed.

[0050] Corresponding to the aforementioned embodiments of the intelligent recommendation method for complex event mechanisms in grassroots social grid governance, this application also provides embodiments of an intelligent recommendation device for complex event mechanisms in grassroots social grid governance.

[0051] Figure 3 This is a block diagram of an intelligent recommendation device for complex event mechanisms in grassroots social grid governance, illustrated according to an exemplary embodiment. (Refer to...) Figure 3 The device may include: Feature extraction module 21 is used to connect to a third-party information platform to obtain pending events of grassroots social grid governance through interface or direct database connection, and extract the first feature from the pending events; The simplified / traditional classification module 22 is used to classify the event to be processed into a simple event or a complex event based on the first feature using a simplified / traditional classification model. The mechanism recommendation module 23 is used to recommend corresponding processing mechanisms for the event to be processed if the event to be processed is a complex event, by means of vectorized retrieval and machine learning, to obtain the first probability and the second probability of the event belonging to each processing mechanism, to sort the first probability and the second probability, and to recommend the processing mechanism with the highest probability.

[0052] Regarding the apparatus in the above embodiments, the specific manner in which each module performs its operation has been described in detail in the embodiments related to the method, and will not be elaborated upon here.

[0053] For the device embodiments, since they basically correspond to the method embodiments, the relevant parts can be referred to in the description of the method embodiments. The device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this application according to actual needs. Those skilled in the art can understand and implement this without creative effort.

[0054] Accordingly, this application also provides a computer program product, including a computer program / instruction, which, when executed by a processor, implements the intelligent recommendation method for complex event mechanisms in grassroots social grid governance as described above.

[0055] Accordingly, this application also provides an electronic device, comprising: one or more processors; a memory for storing one or more programs; and, when the one or more programs are executed by the one or more processors, causing the one or more processors to implement the intelligent recommendation method for complex event mechanisms in grassroots social grid governance as described above. Figure 4 The diagram shown is a hardware structure diagram of any device with data processing capabilities, which is the intelligent recommendation device for complex event mechanisms in grassroots social grid governance provided by an embodiment of the present invention. (Except for...) Figure 4 In addition to the processor, memory, and network interface shown, any data processing device in the embodiment may also include other hardware depending on the actual function of the data processing device, which will not be described in detail here.

[0056] Accordingly, this application also provides a computer-readable storage medium storing computer instructions, which, when executed by a processor, implement the intelligent recommendation method for complex event mechanisms in grassroots social grid governance as described above. The computer-readable storage medium can be an internal storage unit of any data-processing device as described in any of the foregoing embodiments, such as a hard disk or memory. The computer-readable storage medium can also be an external storage device, such as a plug-in hard disk, smart media card (SMC), SD card, flash card, etc., equipped on the device. Furthermore, the computer-readable storage medium can include both internal storage units of any data-processing device and external storage devices. The computer-readable storage medium is used to store the computer program and other programs and data required by the data-processing device, and can also be used to temporarily store data that has been output or will be output.

[0057] Other embodiments of this application will readily occur to those skilled in the art upon consideration of the specification and practice of the disclosure herein. This application is intended to cover any variations, uses, or adaptations of this application that follow the general principles of this application and include common knowledge or customary techniques in the art not disclosed herein.

[0058] It should be understood that various modifications and changes can be made to this application without departing from the scope described above and shown in the accompanying drawings.

Claims

1. An intelligent recommendation method for complex event mechanisms in grassroots social grid governance, characterized in that, include: Connect to a third-party information platform to obtain pending events for grassroots social grid governance through interfaces or direct database connections, and extract the first feature from the pending events; Using a simple / complex classification model, based on the first feature, the events to be processed are divided into simple events or complex events; If the event to be processed is a complex event, then a corresponding processing mechanism is recommended for it through vectorized retrieval and machine learning methods. The first probability and the second probability of it belonging to each processing mechanism are obtained respectively. The first probability and the second probability are sorted to recommend the processing mechanism with the highest probability. Specifically, a vector-based retrieval method is used to recommend appropriate processing mechanisms for the event to be processed, and the first probability of belonging to each processing mechanism is obtained, including: The text content of the event to be processed is vectorized and encoded. The encoded events are retrieved from the Elasticsearch database that stores event cases to obtain the most similar events. Based on the processing mechanism corresponding to each similar event, set the first probability that the event to be processed belongs to each processing mechanism; Specifically, machine learning methods are used to recommend appropriate processing mechanisms for the events to be processed, and the second probability of each event belonging to a particular processing mechanism is obtained, including: Perform semantic reshaping on the events to be processed; Using a large language model, keywords are extracted and quantified from the semantically reshaped events to obtain event severity quantification values. Extract the second feature of the event to be processed, which includes the number of people involved, the quantitative value of the amount involved, the quantitative value of the severity of the event, the event type, and the event theme; Based on the second feature, the event to be processed is classified using a pre-trained multi-classification model to obtain the second probability of it belonging to each processing mechanism.

2. The method according to claim 1, characterized in that, The first feature is extracted from the event to be processed, specifically as follows: If the event to be processed is structured data, then the first feature is extracted directly; If the event to be processed is unstructured data, the first feature is extracted through manual annotation and feature extraction.

3. The method according to claim 1, characterized in that, Using a large language model, keywords are extracted and quantified from the semantically restructured events to obtain event severity quantification values, specifically: Using a large language model, keywords are extracted from the semantically reshaped events to be processed; Based on the semantic ambiguity of the keywords and the degree of influence of the corresponding behaviors, the keywords are quantified to obtain the event severity quantification value.

4. The method according to claim 1, characterized in that, The quantified value of the target amount is obtained in the following way: The maximum value among the extracted amounts from the events to be processed is taken as the target amount; The target amount is mapped to the corresponding quantified value according to the predetermined segmentation range, thus obtaining the target amount quantified value.

5. An intelligent recommendation device for complex event mechanisms in grassroots social grid governance, characterized in that, To implement the method of claim 1, the method comprises: The feature extraction module is used to connect to a third-party information platform to obtain pending events of grassroots social grid governance through interface or direct database connection, and extract the first feature from the pending events. The simplified / traditional classification module is used to classify the event to be processed into a simple event or a complex event based on the first feature using a simplified / traditional classification model. The mechanism recommendation module is used to recommend corresponding processing mechanisms for complex events by using vectorized retrieval and machine learning methods. It obtains the first probability and the second probability of each processing mechanism, sorts the first probability and the second probability, and recommends the processing mechanism with the highest probability.

6. A computer program product comprising a computer program / instructions, characterized in that, When the computer program / instruction is executed by the processor, it implements the method as described in any one of claims 1-4.

7. An electronic device, characterized in that, include: One or more processors; Memory, used to store one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-4.

8. A computer-readable storage medium storing computer instructions thereon, characterized in that, When executed by the processor, this instruction implements the steps of the method as described in any one of claims 1-4.

Images