Basic-level governance duty intelligent matching method, system and device and storage medium

Abstract

The invention discloses a grassroots governance job intelligent matching method, system and device and a storage medium, which are one-to-one corresponding schemes, and in the scheme, a training data set is constructed based on a post knowledge base and a contradictory dispute mediation database, and a job matching model is obtained through training, on one hand, the matching relationship between existing contradictory dispute events and corresponding posts is learned, and on the other hand, the matching relationship between the post knowledge base and the contradictory dispute mediation database is learned; on the other hand, different data are mapped to the same semantic space for matching, so that the matching efficiency and accuracy of events and processing posts can be greatly improved; generally speaking, the invention provides an automatic scheme from the level of data processing based on an artificial intelligence model (namely a job matching model), so that the burden and the application cost of primary-level workers can be reduced, the mediation efficiency is improved, the service satisfaction of residents is improved, and the social governance level of the primary-level is improved.

Description

Technical Field

[0001] The present invention relates to the technical fields of deep learning and natural language processing, and particularly relates to an intelligent matching method, system, device and storage medium for grass-roots governance positions and tasks. Background Art

[0002] With the improvement of productivity and the progress of technology, the social division of labor has been further refined, and the depth and breadth of economic and social management have increased significantly. To match the development of society, institutional functions have become more complex, and the need for cross-level and departmental collaboration has increased substantially. County-level grass-roots social governance will also face digital, platform-based, and intelligent transformation. The problem of work task division in the process of grass-roots governance platformization essentially requires an understanding of specific work matters, and at the same time, it is necessary to master the specific responsibilities of all positions, and reasonably allocate positions and corresponding specific tasks according to the work matters. If it is completely realized by manual labor, not only the processing efficiency is low, but also the allocated personnel are required to have a lot of professional knowledge, and the implementation difficulty is very high.

[0003] With the development of artificial intelligence and deep learning, it has become possible to achieve this function through advanced language models. Most of the existing position matching methods focus on the person-position matching in enterprise management and talent recruitment. For example, the Chinese patent application for invention "Method and System for Employee Position Matching and Deployment Based on Artificial Intelligence" with the publication number CN119494522A, the Chinese patent application for invention "A Person-Position Matching Recommendation Method Based on BERT and Latent Semantic Algorithm Model" with the authorization announcement number CN119377490B, the Chinese patent application for invention "Deep Learning Intelligent Person-Position Matching Method and System Based on Large Language Model and Multiple Prompts" with the publication number CN119358985A, the Chinese patent application for invention "Management Method and System Based on Person-Position Intelligent Matching Algorithm Model" with the publication number CN119204834A, the Chinese patent application for invention "An Agent Method for Person-Position Matching Based on Large Model Development" with the publication number CN118657501A, etc. However, there is currently no systematic research on the matching scheme of data required for grass-roots social contradiction and dispute mediation and related positions and tasks. Summary of the Invention

[0004] The purpose of the present invention is to provide an intelligent matching method, system, device and storage medium for grass-roots governance positions and tasks, introduce natural language processing technology to process position data and contradiction and dispute mediation data, reduce manual processing steps, improve the mediation efficiency of contradiction and dispute events, and improve the level of social governance.

[0005] The purpose of the present invention is achieved by the following technical solutions: An intelligent matching method for grass-roots governance positions and tasks includes: Constructing a position knowledge base by using descriptions of each position and its responsibilities; Collect the existing disposal archives of conflict and dispute events, sort out the written descriptions therein, and construct a conflict and dispute mediation database; Combine the post knowledge base and the conflict and dispute mediation database to construct a training data set; Use the training data set to train a selected pre-trained model to obtain an event-post matching model; Input the description of the conflict and dispute event to be processed into the event-post matching model, and the event-post matching model performs matching in the post knowledge base to determine the corresponding disposal post.

[0006] A grass-roots governance event-post intelligent matching system for implementing the foregoing method, comprising: A post knowledge base construction unit for constructing a post knowledge base by using the descriptions of each post and its responsibilities; A conflict and dispute mediation database construction unit for collecting the existing disposal archives of conflict and dispute events, sorting out the written descriptions therein, and constructing a conflict and dispute mediation database; A training data set construction unit for combining the post knowledge base and the conflict and dispute mediation database to construct a training data set; A model training unit for using the training data set to train a selected pre-trained model to obtain an event-post matching model; An intelligent matching unit for inputting the description of the conflict and dispute event to be processed into the event-post matching model, and the event-post matching model performs matching in the post knowledge base to determine the corresponding disposal post.

[0007] A processing device, comprising: one or more processors; a memory for storing one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the foregoing method.

[0008] A readable storage medium storing a computer program, which implements the foregoing method when the computer program is executed by a processor.

[0009] It can be seen from the technical solution provided by the present invention above that by constructing a training data set based on the post knowledge base and the conflict and dispute mediation database and training to obtain an event-post matching model, on the one hand, learning the matching relationship between existing conflict and dispute events and corresponding posts, and on the other hand, mapping different data to the same semantic space for matching, can greatly improve the matching efficiency and accuracy of events and processing posts; Generally speaking, the present invention provides an automated solution from the data processing level based on an artificial intelligence model (i.e., the event-post matching model), which can reduce the burden and operation cost of grass-roots staff, improve the mediation efficiency, improve the satisfaction of residents' services, and improve the level of grass-roots social governance. Description of the Drawings

[0010] To more clearly illustrate the technical solutions of the embodiments of the present invention, the following will briefly introduce the accompanying drawings required for the description of the embodiments. Obviously, the accompanying drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, without creative efforts, other accompanying drawings can be obtained based on these drawings.

[0011] Figure 1 It is a flowchart of an intelligent matching method for grass-roots governance positions provided by an embodiment of the present invention; Figure 2 It is a flowchart of position matching provided by an embodiment of the present invention; Figure 3 It is a flowchart of training and updating of a position matching model provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of an intelligent matching system for grass-roots governance positions provided by an embodiment of the present invention; Figure 5 It is a schematic diagram of a processing device provided by an embodiment of the present invention. Specific Embodiments

[0012] Next, in combination with the accompanying drawings in the embodiments of the present invention, the technical solutions in the embodiments of the present invention will be clearly and completely described. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts fall within the protection scope of the present invention.

[0013] First, the following explanations will be given to the terms that may be used in this article: The description of terms such as "comprising", "including", "containing", "having" or other similar semantics should be interpreted as non-exclusive inclusion. For example: including a certain technical feature element (such as raw materials, components, ingredients, carriers, dosage forms, materials, dimensions, parts, components, mechanisms, devices, steps, processes, methods, reaction conditions, processing conditions, parameters, algorithms, signals, data, products or articles, etc.) should be interpreted as not only including the clearly listed certain technical feature element, but also including other well-known technical feature elements in the art that are not clearly listed.

[0014] The term "consisting of" means excluding any technical feature elements not explicitly listed. If this term is used in a claim, it will make the claim a closed type, so that it does not include technical feature elements other than those explicitly listed, except for the related conventional impurities. If this term only appears in a sub-clause of a claim, then it only limits the elements explicitly listed in that sub-clause, and the elements recorded in other sub-clauses are not excluded from the overall claim.

[0015] The following will describe in detail a method, system, device, and storage medium for intelligent matching of grass-roots governance positions provided by the present invention. The content not described in detail in the embodiments of the present invention belongs to the prior art well-known to those skilled in the art. In the embodiments of the present invention, those not specified in specific conditions are carried out according to the conventional conditions in the art or the conditions recommended by the manufacturer. The reagents or instruments not specified in the production manufacturer in the embodiments of the present invention are all conventional products that can be obtained through commercial purchase.

[0016] Embodiment 1 The embodiment of the present invention provides a method for intelligent matching of grass-roots governance positions, as Figure 1 shown, mainly including the following steps: Step 1: Construct a position knowledge base by using the descriptions of each position and its responsibilities.

[0017] In the embodiment of the present invention, taking the area served by the grass-roots governance platform as the boundary, the positions and their responsibilities involved in all contradiction and dispute mediation within the area are sorted out to construct a position knowledge base, denoted as , where each represents a piece of position data, and a piece of position data includes the description of the position and its responsibilities, i = 1, 2,..., m, and m is the length of the position knowledge base.

[0018] Step 2: Collect the existing case files for handling contradiction and dispute events, sort out the text descriptions therein, and construct a contradiction and dispute mediation database.

[0019] Step 3: Combine the position knowledge base and the contradiction and dispute mediation database to construct a training data set.

[0020] In the embodiments of the present invention, the format of a single piece of training data in the training dataset is: <query, positive sample, negative sample, association score>; where the query is the text description to be queried, and the job and responsibility descriptions in the job knowledge base, or the event descriptions in the contradiction and dispute mediation database, or a custom text describing the contradiction and dispute event are used as the query; the positive sample is a piece of text related to the query, and the job and responsibility descriptions in the job knowledge base or the event descriptions in the contradiction and dispute mediation database are used as the positive sample; the negative sample is text unrelated to the query, and the job and responsibility descriptions in the job knowledge base or randomly generated unrelated text are used as the negative sample; there is one or more negative samples; the association score is the score between the query and the positive sample.

[0021] Preferably, a single piece of training data is constructed in the following manner: Taking the event description in the contradiction and dispute mediation database as the query, taking the job and responsibility description matching the query as the positive sample, taking the job and responsibility description not matching the query as the negative sample, and the association score is the score between the query and the positive sample; Or, taking a custom text describing the contradiction and dispute event as the query, taking the corresponding event description in the contradiction and dispute mediation database as the positive sample, taking randomly generated unrelated text as the negative sample, and the association score is the score between the query and the positive sample; Or, taking the job and responsibility description in the job knowledge base as the query, taking the other job and responsibility descriptions associated with the query as the positive sample, taking the job and responsibility descriptions excluding the query and the positive sample as the negative sample, and the association score is the score between the query and the positive sample.

[0022] Step 4: Use the training dataset to train the selected pre-trained model to obtain a job-position matching model.

[0023] In the embodiments of the present invention, the positive sample and negative sample contrast learning method is adopted; the sum of the number of positive samples and negative samples corresponding to each query is denoted as N, where there is one positive sample and N - 1 negative samples; in the training process, the method of dynamically mining hard negative samples is adopted, that is, only some hard negative samples are used for all data, and after several training steps, the top N - 1 hardest-to-distinguish negative samples are mined to replace the N - 1 negative samples corresponding to the query and used in the subsequent training.

[0024] The contrast learning loss used in training is expressed as: ; Where is the contrast learning loss, B is the training data batch size, Denote the similarity between the query of the $i$-th training data in a batch and the $j$-th sample, where $j = 1$ is the positive sample and $j = 2, 3, \ldots, N$ are negative samples; Denote the association score between the query of the $i$-th training data and the positive sample, which is a hyperparameter during training.

[0025] Step 5: Input the description of the contradiction and dispute event to be processed into the event-post matching model. The event-post matching model performs matching in the job knowledge base to determine the corresponding handling position.

[0026] In the embodiment of the present invention, the event-post matching model extracts feature vectors from the description of the contradiction and dispute event to be processed , and extracts corresponding feature vectors from the description of each position and its responsibilities in the job knowledge base, denoted as , where $m$ is the length of the job knowledge base; calculate the similarity between the feature vector and the feature vector , select the descriptions of positions and their responsibilities with similarity higher than the threshold, and determine the corresponding handling positions.

[0027] Preferably, it further includes: regularly collecting the handling files of contradiction and dispute events, updating the contradiction and dispute mediation database, combining the updated contradiction and dispute mediation database to update the training data set, and updating the event-post matching model.

[0028] In order to more clearly show the technical solution provided by the present invention and the technical effects produced, the method provided by the embodiment of the present invention will be described in detail below with specific embodiments.

[0029] 1. Construct a grass-roots social contradiction and dispute mediation job knowledge base (abbreviated as job knowledge base).

[0030] In the embodiment of the present invention, taking the area served by the grass-roots governance platform as the boundary (generally at the county level), sort out all the positions and their responsibilities in the area that may be involved in grass-roots social contradiction and dispute mediation, and construct a job knowledge base based on the atomic functions of the power and responsibility list, the description of job responsibilities and performance behaviors, denoted as , and each represents a piece of job data. By sorting out the types of grass-roots institutions, form the types of administrative institutions under each administrative region at all levels, including county (district) level, town (sub-district) level, village (community), etc. Then sort out the positions set by each type of administrative institution below, and further divide the corresponding job responsibilities of each position according to the types of events it is responsible for handling. Form a full-scale job knowledge base from top to bottom of county (district), town (sub-district), and village (community), with clear institutional job structures and clear job responsibilities. Support the addition, deletion, modification, and query maintenance of the job knowledge base, support multi-condition joint query of data, and dynamic update.

[0031] In the embodiments of the present invention, in the job and responsibility description, the job description includes the job name and the name of the affiliated institution, and the responsibility description is a text description of the job responsibilities, which is processed in the following standard way: (1) Job name: filled in according to the government's job catalog, for example, the agricultural economy position; (2) Institution name: filled in according to the division of the government's job catalog, for example, the Agricultural and Rural Office of the Town (Sub-district); (3) Job responsibilities: filled in according to the government's job responsibility settings, for example, mediating contradictions and disputes related to the development of rural collective economy, providing economic value of crop cultures, etc., and promoting reconciliation between the two parties. As shown in Table 1, examples of some entries in the job knowledge base are provided.

[0032] Table 1: Examples of Some Entries in the Job Knowledge Base Serial Number Institution Name Position Name Job Responsibilities 1 Village and Town Construction Comprehensive Service Center of the Town (Sub-district) Traffic Post Mediate relevant contradictions and disputes such as traffic inconvenience caused by road construction and construction in the village and town, and impassability caused by villagers blocking the road for various reasons, and promote reconciliation between the two parties. 2 Agriculture and Rural Affairs Office of the Town (Sub-district) Water Conservancy Post Mediate relevant contradictions and disputes such as flood control and drought relief, irrigation and drainage, and promote reconciliation between the two parties. 3 Agriculture and Rural Affairs Office of the Town (Sub-district) Agricultural Economics Post Mediate relevant contradictions and disputes such as the development of rural collective economy and the economic value of crop breeding, and promote reconciliation between the two parties.

[0033] II. Construct a contradiction and dispute mediation database.

[0034] In the embodiments of the present invention, from the existing archives of social contradiction and dispute incident handling, the text descriptions of the contradiction and dispute incidents are sorted out, and combined with the actual situation and the grass-roots social contradiction and dispute mediation job knowledge base, a reasonable job is matched for the incident to form a contradiction and dispute mediation database, denoted as , each item e represents a contradiction and dispute mediation data, including the incident description, incident type, and associated institution-job name; it is processed in the following standard way: (1) Incident description: the specific incident description reported by grid workers or the public, etc. Example: "On a certain date in a certain month and year, a certain person in a certain village in a certain town reported that there was construction waste piled up at the door, affecting travel." (2) Incident type: the incident type of the specific incident reported by grid workers or the public, etc. Example: "Contradiction and dispute related to environmental protection." (3) Associated institution-job name: the job name with a unique identifier. Example: "Environmental Protection Position of the Emergency Management and Ecological Environment Protection Office of the Town (Sub-district)". As shown in Table 2, examples of some data in the contradiction and dispute mediation database are provided.

[0035] Table 2: Examples of Some Data in the Contradiction and Dispute Mediation Database <{ Serial Number Event Description (Longer Text) Event Type Associated Post 1 Associated Post 2 Associated Post 3 Associated Post 4 Associated Post 5 1 XXX from XX Village, XX Town reported that due to environmental renovation, a parking lot was built in front of his own house, and the trees and vegetable gardens he planted were cleared, and now he requests compensation. Mediate relevant contradictions and disputes related to environmental protection Beautiful Post of the Agriculture and Rural Affairs Office of the Town (Sub-district) 2 XXX from XX Village, XX Town reported that construction waste was piled up in front of his door, affecting travel. Mediate relevant contradictions and disputes related to environmental protection Environmental Protection Post of the Emergency Management and Ecological Environment Protection Office of the Town (Sub-district) Urban Management Squadron Post of the Comprehensive Administrative Law Enforcement Brigade of the Town (Sub-district) Environmental Protection Post of the Ecological Environment Branch of the County (District) 3 On a certain date in a certain year and month, some people reported that the XX City construction site west of XX Community in XX Town did not comply with the legal construction time and caused noise pollution to the residents. Mediate relevant contradictions and disputes related to environmental protection Environmental Protection Post of the Emergency Management and Ecological Environment Protection Office of the Town (Sub-district) Planning Post of the Village and Town Construction Comprehensive Service Center of the Town (Sub-district) Urban Management Squadron Post of the Comprehensive Administrative Law Enforcement Brigade of the Town (Sub-district) Environmental Protection Post of the Ecological Environment Branch of the County (District) Real Estate Management Post of the Housing and Urban-Rural Development Bureau of the County (District) 4 On a certain date in a certain year and month, some people reported that the XXXX barbecue restaurant on XX Road Street in XX District had noise pollution, and a conflict occurred between the two parties. Mediate relevant contradictions and disputes related to environmental protection Environmental Protection Post of the Emergency Management and Ecological Environment Protection Office of the Town (Sub-district) Urban Management Squadron Post of the Comprehensive Administrative Law Enforcement Brigade of the Town (Sub-district) Environmental Protection Post of the Ecological Environment Branch of the County (District)

[0036] III. Construct a training data set.

[0037] In the embodiments of the present invention, in order to enable the pre-trained semantic feature extraction model to correctly represent the semantic space required for the job-incident matching task, the pre-trained model needs to be fine-tuned, so a dedicated fine-tuning data set D needs to be constructed according to the task requirements.

[0038] In the embodiments of the present invention, the format of a single piece of training data in the training dataset is: <query, positive sample, negative sample, correlation score>, where the query is a piece of text, the positive sample is a piece of text related to the query, the negative sample is text unrelated to the query and the number can be one or more, and the correlation score is the correlation score between the query and the positive sample, generally taking values between 0 and 1. Based on the above analysis, the following three methods are adopted in the present invention to construct the required dataset: (1) Use the event description text in the contradiction and dispute mediation database as the query, the matching post and responsibility description text as the positive sample, and the non-matching post and responsibility description text as the negative sample, and set the correlation score to 1.

[0039] (2) Use "a piece of text describing a contradiction and dispute event" or other sentences that can express this meaning as the query, the event description text in the contradiction and dispute mediation database as the positive sample, and some randomly generated irrelevant text as the negative sample, and set the correlation score to 1.

[0040] (3) Manually annotate the correlation scores, that is, assign reasonable correlation scores to all pairs of post and responsibility description texts. For example, for "Party Building Work Office of Town (Sub-district) - Women's Federation Post: Mediate contradictions and disputes related to the protection of women's rights and interests, two-cancer screening and assistance, etc. in family relations, and promote reconciliation between the two parties." and "County (District) Women's Federation - Comprehensive Post: Mediate contradictions and disputes related to women's rights and interests, etc. in emotional marriage and family, and promote reconciliation between the two parties.", a correlation score of 0.7 can be assigned, while for "County (District) Ethnic and Religious Affairs Bureau - Comprehensive Post: Mediate contradictions and disputes related to ethnic and religious affairs, etc., and promote reconciliation between the two parties.", a correlation score of 0 can be assigned. Then use the post and responsibility description text as the query, the post and responsibility description text related to it as the positive sample, and the unrelated ones as the negative sample, and set the correlation score to the annotated correlation score; in addition, text enhancement can also be performed on the query, such as using methods like deleting parts unrelated to the responsibilities in the description, swapping the order if there are multiple responsibilities, and replacing with synonyms, and use the enhanced text as the positive sample, and set the correlation score to 1.

[0041] Based on this, after annotating the correlation scores of every two post and responsibility descriptions, for a single post and responsibility description, when it is used as the query, if the correlation score between other post and responsibility descriptions and this query is greater than 0, it is a positive sample, and if it is equal to 0, it is a negative sample.

[0042] IV. Train the pre-trained model.

[0043] In the embodiments of the present invention, the pre-trained model is a representation of a pre-trained semantic feature extraction model, which can map the post knowledge base to a reasonable semantic space and use the cosine distance in the same semantic space to match suitable handling posts for events. Exemplarily, the pre-trained model can select semantic feature extraction models such as stella-mrl-large-zh-v3.5-1792d and BGE-large-zh-v1.5; stella-mrl-large-zh-v3.5-1792d is a Matryoshka embedding model, stella is the model name, mrl refers to the model using Matryoshka Representation Learning (Russian doll representation learning), large indicates that the model belongs to the large parameter version, zh is the language identifier (i.e., the model is optimized for Chinese text), v3.5 is the version number, and 1792d indicates that the default vector dimension of the model is 1792 dimensions; BGE-large-zh-v1.5 is a high-performance semantic representation model, BGE is the model name, and v1.5 is the version number.

[0044] For an ideal semantic space represented by a semantic feature extraction model, it should satisfy the following characteristics: (1) The feature vector of the text description of the contradiction and dispute event should be close to the feature vector corresponding to the matching handling post and its responsibilities description, and far from the feature vector corresponding to the non-matching post and its responsibilities description.

[0045] (2) The feature vector corresponding to the text "a text describing a contradiction and dispute event" should be close to the feature vector of the contradiction and dispute event description (data in the contradiction and dispute mediation database), and far from the feature vectors of texts unrelated to the contradiction and dispute, random garbled codes, etc.

[0046] (3) There should also be a reasonable distribution in the semantic space between different post and responsibilities descriptions. For example, the feature vector of "Women's Federation Post in the Party Building Work Office of the Town (Sub-district): Mediate contradictions and disputes related to the protection of women's rights and interests, two-cancer screening and assistance, etc. in family relations, and promote reconciliation between the two parties." should be close to the feature vector of "Comprehensive Post in the Women's Federation of the County (District): Mediate contradictions and disputes related to women's rights and interests, etc. in emotional marriage and family, and promote reconciliation between the two parties.", while far from the feature vector of "Comprehensive Post in the Ethnic and Religious Affairs Bureau of the County (District): Mediate contradictions and disputes related to ethnic and religions, etc., and promote reconciliation between the two parties."

[0047] Therefore, the present invention fine-tunes and trains the pre-trained model (such as: BGE-large-zh-v1.5) based on the above training dataset D.

[0048] During the training process, a contrastive learning paradigm with a single positive sample and multiple negative samples is adopted. For all data entries in the training data, the sum of the number of positive and negative samples should be consistent with N (for example, N = 8, that is, one positive sample and seven negative samples, and the value of N depends on the computing power resources and training settings. Generally, N - 1 is not greater than the minimum number of negative samples in the dataset). According to the construction method of the dataset, the number of negative samples is not constant. Therefore, in the training process of the present invention, a method of dynamically mining hard negative samples is adopted, that is, only a part of the hard negative samples (N - 1) are used for all data, and after a certain number of training steps, the top N - 1 hard-to-distinguish negative samples are mined according to the current model to replace the corresponding N - 1 negative samples for the query and used in the subsequent training. This not only ensures the full utilization of negative samples but also enhances the normality of the semantic space after fine-tuning. The training uses soft-infoNCE-Loss (smooth information noise contrast loss), and the calculation formula is as follows: ; In the formula, B represents the batch size of the training data during training, N is the sum of the number of positive and negative samples in a single piece of training data, represents the similarity between the query of the i-th piece of training data in a batch and the j-th sample. Generally, j = 1 is the positive sample, and j = 2, 3,..., N are negative samples, represents the association score between the query of the i-th piece of training data and the positive sample, is a hyperparameter during training, generally taking values from 10 to 50.

[0049] Finally, a job-position matching model is obtained through training 。

[0050] V. Job-position intelligent matching process.

[0051] In the embodiment of the present invention, the job-position matching model is used to calculate the feature vectors of all job positions and responsibility texts , and then the job-position matching model is used to calculate the feature vector of the input contradiction and dispute event description . is calculated with one by one for similarity. Those with a similarity greater than the threshold T are the corresponding matching disposal job positions. Thus, one or more matching disposal job positions can be selected. T generally takes 0.5. As Figure 2 shown, it is the flowchart of job-position intelligent matching.

[0052] Based on the above solution, the matching result can be transmitted to the grass-roots governance platform, and the grass-roots governance platform automatically recommends a disposal plan and forms a mediation team.

[0053] VI. Job-position matching model training and update process.

[0054] In the embodiments of the present invention, in order to improve the intelligent matching effect of events and positions, feedback data can be collected in the application to update the position knowledge base, the contradiction and dispute mediation database, the associated scores, etc., and then dynamically update the event-position matching model, as Figure 3 shown, which shows the training and updating process of the event-position matching model.

[0055] Through the description of the above embodiments, those skilled in the art can clearly understand that the above embodiments can be implemented by software, or can be implemented by means of software plus a necessary general hardware platform. Based on such an understanding, the technical solutions of the above embodiments can be embodied in the form of a software product, and the software product can be stored in a non-volatile storage medium (which can be a CD-ROM, a USB flash drive, a mobile hard disk, etc.), including several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0056] Embodiment 2 The present invention also provides a grass-roots governance event-position intelligent matching system, which is mainly used to implement the method provided in the foregoing embodiments, as Figure 4 shown, the system mainly includes: A position knowledge base construction unit, which is used to construct a position knowledge base by using the descriptions of each position and its responsibilities; A contradiction and dispute mediation database construction unit, which is used to collect the disposal files of existing contradiction and dispute events, sort out the text descriptions therein, and construct a contradiction and dispute mediation database; A training data set construction unit, which is used to construct a training data set by combining the position knowledge base and the contradiction and dispute mediation database; A model training unit, which is used to train a selected pre-trained model by using the training data set to obtain an event-position matching model; An intelligent matching unit, which is used to input the description of the contradiction and dispute event to be processed into the event-position matching model, and the event-position matching model performs matching in the position knowledge base to determine the corresponding disposal position.

[0057] Considering that the main processing details involved in the above system have been introduced in detail in the previous embodiments, they will not be elaborated here.

[0058] Those skilled in the art can clearly understand that for the convenience and simplicity of description, only the above division of each functional module is used as an example for illustration. In practical applications, the above functions can be allocated to different functional modules according to needs, that is, the internal structure of the system is divided into different functional modules to complete all or part of the functions described above.

[0059] Embodiment 3 The present invention also provides a processing device, such as Figure 5 as shown, which mainly includes: one or more processors; a memory for storing one or more programs; wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method provided in the foregoing embodiments.

[0060] Furthermore, the processing device further includes at least one input device and at least one output device; in the processing device, the processor, the memory, the input device, and the output device are connected through a bus.

[0061] In the embodiments of the present invention, the specific types of the memory, the input device, and the output device are not limited; for example: The input device can be a touch screen, an image acquisition device, a physical button, or a mouse, etc.; The output device can be a display terminal; The memory can be a Random Access Memory (RAM), or a non-volatile memory, such as a disk memory.

[0062] Embodiment 4 The present invention also provides a readable storage medium storing a computer program, which implements the method provided in the foregoing embodiments when the computer program is executed by a processor.

[0063] In the embodiments of the present invention, as a computer-readable storage medium, the readable storage medium can be disposed in the foregoing processing device, for example, as the memory in the processing device. In addition, the readable storage medium can also be various media that can store program codes, such as a USB flash drive, a mobile hard disk, a Read-Only Memory (ROM), a magnetic disk, or an optical disc.

[0064] As described above, the above are only the preferred specific embodiments of the present invention, but the protection scope of the present invention is not limited thereto. Any changes or substitutions that can be easily thought of by those skilled in the art within the technical scope disclosed by the present invention should be covered by the protection scope of the present invention. Therefore, the protection scope of the present invention should be subject to the protection scope of the claims. The information disclosed in the background art part of this article is only intended to deepen the understanding of the overall background art of the present invention, and should not be regarded as an admission or any form of implication that this information constitutes the prior art known to those skilled in the art.

Claims

1. An intelligent matching method for grass-roots governance positions, characterized in that, Including: Construct a job knowledge base by using the descriptions of each position and its responsibilities; Collect the existing disposal files of conflict and dispute events, sort out the text descriptions therein, and construct a conflict and dispute mediation database; Combine the job knowledge base and the conflict and dispute mediation database to construct a training data set; Use the training data set to train a selected pre-trained model to obtain a job-event matching model; Input the description of the conflict and dispute event to be processed into the job-event matching model, and the job-event matching model performs matching in the job knowledge base to determine the corresponding disposal position.

2. The intelligent matching method for grass-roots governance positions according to claim 1, characterized in that The constructing of the job knowledge base by using the descriptions of each position and its responsibilities includes: Taking the area served by the grass-roots governance platform as the boundary, sort out all the positions and their responsibilities related to the mediation of contradictions and disputes within the area, and construct a position knowledge base, denoted as , where each represents a piece of position data. A piece of position data includes the position and its responsibility description, i = 1, 2, …, m, where m is the length of the position knowledge base.

3. A method for intelligent matching of grass-roots governance positions according to claim 1, characterized in that, The format of a single piece of training data in the training data set is: <query, positive sample, negative sample, correlation score>; Among them, the query is the text description to be queried, and the descriptions of positions and responsibilities in the job knowledge base, or the event descriptions in the conflict and dispute mediation database, or a custom text describing the conflict and dispute event are used as the query; The positive sample is a text related to the query, and the descriptions of positions and responsibilities in the job knowledge base or the event descriptions in the conflict and dispute mediation database are used as the positive sample; The negative sample is text unrelated to the query, and the descriptions of positions and responsibilities in the job knowledge base or randomly generated unrelated text are used as the negative sample; there are one or more negative samples; The correlation score is the score between the query and the positive sample.

4. The intelligent matching method for grass-roots governance positions according to claim 3, wherein A single piece of training data is constructed in the following way: Use the event description in the conflict and dispute mediation database as the query, use the descriptions of positions and responsibilities that match the query as the positive sample, use the descriptions of positions and responsibilities that do not match the query as the negative sample, and the correlation score is the score between the query and the positive sample; Or, use a custom text describing the conflict and dispute event as the query, use the corresponding event description in the conflict and dispute mediation database as the positive sample, use randomly generated unrelated text as the negative sample, and the correlation score is the score between the query and the positive sample; Or, use the descriptions of positions and responsibilities in the job knowledge base as the query, use the descriptions of other positions and responsibilities associated with the query as the positive sample, use the descriptions of positions and responsibilities except the query and the positive sample as the negative sample, and the correlation score is the score between the query and the positive sample.

5. A grass-roots governance position intelligent matching method according to claim 3 or 4, characterized in that The training of the selected pre-trained model by using the training data set to obtain a job-event matching model includes: Adopt the way of contrastive learning between positive samples and negative samples; record the sum of the number of positive samples and negative samples as N, where there is one positive sample and N - 1 negative samples; Adopt the way of dynamically mining hard negative samples during the training process, that is, only use some hard negative samples for all data, and after several training steps, mine the top N - 1 hardest-to-distinguish negative samples to replace the N - 1 negative samples corresponding to the query and use them in the subsequent training; The contrastive learning loss used in the training is expressed as: ； Among them, is the contrastive learning loss, B is the batch size of training data, represents the similarity between the query of the i-th training data in a batch and the j-th sample, where j = 1 is the positive sample and j = 2, 3, …, N are negative samples; represents the association score between the query of the i-th training data and the positive sample, is a hyperparameter during training.

6. The intelligent matching method for grass-roots governance positions according to claim 1, characterized in that The job-event matching model performs matching in the job knowledge base to determine the corresponding disposal position includes: The described job-event matching model extracts feature vectors from the descriptions of the contradiction and dispute events to be processed , and respectively extracts the corresponding feature vectors from each job and its responsibility description in the job knowledge base, denoted as , where m is the length of the job knowledge base; Calculate the eigenvectors separately and the eigenvectors for their similarity. Select the job positions and duty descriptions with similarity higher than the threshold, and determine the corresponding handling positions.

7. A method for intelligent matching of grass-roots governance positions according to claim 1, characterized in that, Also including: Regularly collect the disposal archives of conflict and dispute events, update the conflict and dispute mediation database, and combine the updated conflict and dispute mediation database to update the training dataset, and update the event-position matching model.

8. An intelligent matching system for grass-roots governance positions, characterized in that, For implementing the method according to any one of claims 1 to 7, comprising: A position knowledge base construction unit for constructing a position knowledge base by using the descriptions of each position and its responsibilities; A conflict and dispute mediation database construction unit for collecting the existing disposal archives of conflict and dispute events, sorting out the text descriptions therein, and constructing a conflict and dispute mediation database; A training dataset construction unit for constructing a training dataset by combining the position knowledge base and the conflict and dispute mediation database; A model training unit for training a selected pre-trained model by using the training dataset to obtain an event-position matching model; An intelligent matching unit for inputting the description of the conflict and dispute event to be processed into the event-position matching model, and the event-position matching model performs matching in the position knowledge base to determine the corresponding disposal position.

9. A processing device, characterized in that, Comprising: One or more processors; A memory for storing one or more programs; Wherein, when the one or more programs are executed by the one or more processors, the one or more processors implement the method according to any one of claims 1 to 7.

10. A readable storage medium stores a computer program, characterized in that, When the computer program is executed by the processor, it implements the method according to any one of claims 1 to 7.

Images