Method for event classification and dispatching and building a unified event type system

Abstract

The application provides a method for event classification and dispatching and building a unified event type system. The method for event classification comprises: obtaining features of an event, the features of the event comprising event content; obtaining features of a predefined event type, the features of the predefined event type comprising a classification path of the predefined event type; inputting the features of the event and the features of the predefined event type into a trained neural network matching model to obtain at least one output matching result. The method according to the example embodiment can realize adaptive event matching classification while reducing the use of promotion costs.

Description

Technical Field

[0001] This application relates to the fields of e-government and information processing technology, and more specifically, to methods and apparatus for event classification, methods and apparatus for event assignment, methods and apparatus for constructing a unified event type system, methods and apparatus for associating event types for departments, as well as computing devices and computer-readable media. Background Technology

[0002] As the economy continues to develop, departmental complexity increases, and departmental management and services become more sophisticated, the workload of each department is also constantly increasing. Traditionally, the management and service process involves the public or organizations submitting requests to departments through hotlines or other channels. Relevant staff then compile these requests into incident records. These incidents are then manually categorized and assigned to departments based on their respective responsibilities. Each department then processes and provides feedback upon receiving the incidents.

[0003] With the deepening implementation and advancement of "Internet Plus Government Services," the establishment of social operation centers, and the implementation of the "Social Brain," numerous solutions for intelligent event classification and assignment have emerged. This can reduce manual operations and improve the efficiency of social management. However, these solutions still have various problems, such as a lack of generalizability.

[0004] Therefore, it is necessary to continue developing new event classification and / or assignment schemes to improve the efficiency of social management. Summary of the Invention

[0005] This application aims to provide a method for event classification and event dispatch, which improves event dispatch efficiency and reduces the cost of use and promotion.

[0006] The user features and advantages of this application will become apparent from the following detailed description, or may be learned in part from practice of this application.

[0007] According to one aspect of this application, a method for event classification is provided, comprising: acquiring features of an event, the features of the event including event content; acquiring features of a predefined event type, the features of the predefined event type including a classification path of the predefined event type; inputting the features of the event and the features of the predefined event type into a trained neural network matching model, and obtaining at least one output matching result.

[0008] According to another aspect of this application, a method for event dispatch is proposed, comprising: obtaining features of an event, the features of the event including event content; obtaining features of a department, the features of the department including a classification path of a predefined event type associated with the department; inputting the features of the event and the features of the department into a trained neural network matching model, and obtaining at least one output matching result.

[0009] According to another aspect of this application, a method for constructing a unified event type system is provided, comprising: parsing at least one existing event classification system, constructing classification paths from the root node to the leaf node, wherein the end of each classification path is a leaf node type; counting the classification paths for each leaf node type, and selecting one as the preferred path; splitting all preferred paths into child node-parent node mapping relationships, and aggregating the same mapping relationships; starting from each leaf node, recursively tracing back to the parent node according to the mapping relationship to obtain all newly created paths; and constructing a tree using the newly created paths to obtain the event type system.

[0010] According to another aspect of this application, a method for associating event types with departments is provided, comprising: obtaining a unified event type system; obtaining a mapping relationship between the department and the unified event type system based on the mapping relationship between the department and an existing event classification system; obtaining all leaf nodes corresponding to the department based on the mapping relationship between the department and the unified event type system, as the event types associated with the department, wherein the features of the event types are also used as the features of the department.

[0011] According to another aspect of this application, a computing device is provided, comprising: a processor;

[0012] A memory, on which computer programs are stored;

[0013] The aforementioned method is implemented when the processor executes the computer program.

[0014] According to another aspect of this application, a computer-readable medium is provided having a computer program stored thereon, which, when executed by a processor, implements the aforementioned method.

[0015] According to some embodiments of this application, event classification is achieved by matching events and event types using a matching model, and event dispatch is achieved by matching events and departments. In traditional classification and dispatch models, the input layer and cross-layer are event content and event-related features, and the classification result is fixed. The matching model according to the embodiments of this application is based on text semantics and calculates semantic relevance, without being limited by the number of classifications, thus achieving adaptive event matching and classification and end-to-end event dispatch.

[0016] It should be understood that the above general description and the following detailed description are merely exemplary and do not limit this application. Attached Figure Description

[0017] The above and other objectives, features and advantages of this application will become more apparent from a detailed description of exemplary embodiments thereof with reference to the accompanying drawings.

[0018] Figure 1A A schematic diagram of a neural network matching model for event classification according to an embodiment of this application is shown.

[0019] Figure 1B A schematic diagram of a neural network matching model for event classification according to another embodiment of this application is shown.

[0020] Figure 2A A schematic diagram of a neural network matching model for event dispatch according to an embodiment of this application is shown.

[0021] Figure 2B A schematic diagram of a neural network matching model for event dispatch according to another embodiment of this application is shown.

[0022] Figure 3 A method for event classification according to an example embodiment is shown.

[0023] Figure 4 A method for event dispatch according to an example embodiment is shown.

[0024] Figure 5 This illustrates a method for constructing a unified event type system based on an example embodiment.

[0025] Figure 6 This illustrates a method for constructing a department-related event type according to an example embodiment.

[0026] Figure 7 An apparatus for event classification according to an example embodiment of this application is shown.

[0027] Figure 8 An apparatus for event dispatch according to an example embodiment of this application is shown.

[0028] Figure 9 An apparatus for constructing a unified event type system according to an example embodiment of this application is shown.

[0029] Figure 10 An apparatus for a department-related event type is shown according to an example embodiment of this application.

[0030] Figure 11 A block diagram of a computing device according to an example embodiment of this application is shown. Detailed Implementation

[0031] Exemplary embodiments will now be described more fully with reference to the accompanying drawings. However, these exemplary embodiments can be implemented in many forms and should not be construed as limited to the embodiments set forth herein; rather, they are provided so that this application will be thorough and complete, and will fully convey the concept of the exemplary embodiments to those skilled in the art. The same reference numerals in the drawings denote the same or similar parts, and therefore repeated descriptions of them will be omitted.

[0032] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0033] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0034] Before describing the embodiments of this application, some terms involved in the embodiments of this application will be explained.

[0035] Event: In this article, "event" refers to the demands and suggestions made by the public or organizations to relevant institutions and units through various channels in social management. These are generally recorded in written form or converted into written form for record-keeping.

[0036] Event Classification: These requests and suggestions are classified according to certain rules so that they can be forwarded to the appropriate departments for processing.

[0037] Incident assignment: Based on the incident description and classification, the above requests and suggestions are forwarded to the corresponding management departments for processing.

[0038] End-to-end: This article refers to a one-step method that goes directly from event description to assignment to the corresponding department, without first classifying events by fixed event types and then assigning them to departments based on mapping relationships.

[0039] Departmental authority and responsibility: A department refers to a relevant department, institution or individual established in social management. Departmental authority and responsibility correspond to the scope of jurisdiction (time, space, etc.) and the types of events that these departments, institutions or individuals are responsible for.

[0040] BERT: BERT is a language representation model that stands for Bidirectional Encoder Representations from Transformers. Transformers are deep learning components that, unlike recurrent neural networks (RNNs) which depend on sequence, can process entire sequences in parallel. This allows for the analysis of larger datasets and accelerates model training. Transformers use attention mechanisms to gather contextual information about words and encode it as rich vectors representing that context, thus processing (rather than processing individually) words related to all other words in a sentence simultaneously. Transformer models learn how to derive the meaning of a given word from every other word in a sentence. BERT is essentially a bidirectional feature encoder for Transformers, representing the current input as feature representations (or encoding, or representation), i.e., word vector representations.

[0041] Existing event dispatch solutions typically require two steps. First, all received events are categorized according to a given social event type system. Then, the categorization results are mapped to the corresponding departments based on the mapping relationship between departmental responsibilities and events, thus completing event dispatch.

[0042] While this model can quickly provide a feasible solution for each region or city, it lacks generalizability. Once the classification model is trained and deployed using historical data, the number of event categories and the mapping between event types and departments are fixed. This makes it unable to support the needs of adding, deleting, or modifying event types as social management becomes increasingly refined, as well as the dynamic adjustment of departmental responsibilities. Each change requires retraining and redeploying the system.

[0043] Furthermore, the classification systems of each city or region are not entirely consistent, requiring the training of independent classification models for each city or region. Moreover, the limited historical data for each city or region, coupled with independent classifications, prevents data and model integration, hindering model performance and making it unable to support dynamic additions, deletions, and modifications to event types, quantities, and departmental relationships. Therefore, it is necessary to retrain the model for each different region or city.

[0044] In view of this, this application proposes a classification and assignment method based on a matching model. Classification is achieved by matching events and event types, and assignment is achieved by matching events and departments. The technical solution of this application supports dynamic expansion of event types and adjustment of departmental responsibilities. Based on the technical solution of this application, intelligent event classification and end-to-end assignment algorithms can be implemented.

[0045] Furthermore, this application proposes a method for constructing a unified event type system, which can achieve the integration and expansion of different event type systems. Simultaneously, departmental responsibilities are also based on this event type system. Further, a tag-based representation can be constructed, utilizing the relevance between event text and type tags to achieve decoupling between the number of event categories and departmental responsibilities. Thus, adaptive intelligent event classification and end-to-end assignment can be achieved. Based on the system construction of this application, the experience of advanced cities or regions can be continuously accumulated, providing effective reference for the e-government construction of other cities or regions, and also supporting the dynamic adjustment of event types and departmental responsibilities.

[0046] The technical solution and its advantages of this application will be described in detail below with reference to the embodiments.

[0047] Figure 1A A schematic diagram of a neural network matching model for event classification according to an embodiment of this application is shown. Figure 1B A schematic diagram of a neural network matching model for event classification according to another embodiment of this application is shown.

[0048] like Figure 1A As shown, according to an embodiment, the neural network matching model for event classification may include a sequentially coupled encoding layer 110, feature cross-multiplication layer 120, and output layer 130. The encoding layer 110 may include a vectorized encoder, such as LSTM, GRU, BERT, or a BERT-based encoder. The feature cross-multiplication layer 120 may include a vector cross-multiplication layer and a fully connected layer. The output layer 130 may be a softmax layer, representing the matching of events and classification paths.

[0049] See Figure 1A According to the embodiment, the encoding layer can take the event content and the classification path of the predefined event type as inputs, and output vectorized codes Vc and Vt. The classification path can be a text sequence, which, when input into an encoder such as BERT, will produce a vectorized representation. For example, the classification path "employment -> labor security -> wage arrears" may be a 768-dimensional value after input and quantization encoding.

[0050] The output vector undergoes concatenation, subtraction, and other operations at the vector crossover layer. For example, (V1, V2, V3) represents the concatenation of three vectors, such as (v c ,v t ,|v c -v t |) is v c v t and |v c -v t | Concatenate three vectors. Assuming each vector is 100-dimensional, the concatenated vector will be 300-dimensional. | vc -v t | indicates that the corresponding elements of two vectors are subtracted, and the result is still a 100-dimensional vector.

[0051] Figure 1B The model shown is the same as Figure 1A They are essentially the same, the difference being that the features of the predefined event type input also include auxiliary labels associated with the predefined event type. Furthermore, the event features input may also include keywords extracted from the event content.

[0052] See Figure 1B According to the embodiment, the encoding layer can take as input event content, content keywords, and predefined event type classification paths and auxiliary labels, and output vectorized codes Vc, Vk, and Vt. The output vectors undergo concatenation, subtraction, and other operations in the vector cross-operation layer. (v c ,v t ,|v c -v t |) is v c v t and |v c -v t |Concatenate three vectors, (v k ,v t ,|v k -v t |) is v k v t and |v k -v t | Concatenate three vectors. Assuming each vector is 100-dimensional, the concatenated vector will be 300-dimensional.

[0053] Figure 2A and 2B The output layer can characterize the matching of events and types.

[0054] Figure 2A A schematic diagram of a neural network matching model for event dispatch according to an embodiment of this application is shown. Figure 2B A schematic diagram of a neural network matching model for event dispatch according to another embodiment of this application is shown.

[0055] Figure 2A and 2B The model shown is the same as Figure 1A and 1B The models shown have a consistent overall framework, and while they share common input features, they also exhibit some differences.

[0056] See Figure 2A For event dispatching, the input on the left side of the model is... Figure 1AThe inputs are the same, but the right-hand inputs are departmental features, such as the classification path of predefined event types associated with the department.

[0057] See Figure 2B For event dispatching, the input on the left side of the model is... Figure 1B In addition to the category paths and auxiliary labels for predefined event types associated with departments, the right-hand input can also include the department's authority and responsibility attributes.

[0058] Figure 2A and 2B The output layer can represent the matching of events and departments.

[0059] The matching model in the example embodiment classifies events by matching events with categories and dispatches events by matching events with departments. Traditional classification and dispatch models only have event content and event-related features in the input and cross-layers, directly outputting a 0-n classification or dispatch (e.g., 0-99 for 100 categories). After training, this traditional classification model has a fixed classification result, only able to classify 100 categories; it cannot classify more or newly added categories. However, the matching model in this application embodiment is based on text semantics, so it calculates semantic relevance and is not limited by the number of categories.

[0060] The matching model in this embodiment calculates the top N most similar paths across all classification paths. For example, if there are 800 classification paths in the event type system, when an event occurs, it will be matched against all 800 paths, and the top N most similar paths will be returned. The classification paths correspond to the semantics of the event type hierarchy.

[0061] According to the example implementation, the matching model is not limited to any type, abandoning the traditional approach of a fixed number of categories plus mapping relationships. Therefore, it can support functions such as dynamically expanding event types and adjusting the scope of departmental responsibilities.

[0062] Figure 3 A method for event classification according to an example embodiment is shown.

[0063] See Figure 3 In S301, the characteristics of the event are obtained. The characteristics of the event may include the event content.

[0064] For example, an example event message might be: "An elderly person is singing near the airplane model in the children's park, causing serious noise pollution. Please handle this."

[0065] According to some embodiments, the characteristics of an event may also include keywords extracted from the event content.

[0066] For example, keywords extracted from the aforementioned event content using algorithms such as TextRank may include: disturbing the peace #1.0 ^ singing #0.942511 ^ model #0.940166 ^ noise #0.648647 ^ airplane #0.646424 ^ children's park #0.640388 ^ elderly #0.63548 ^ handling #0.369955.

[0067] In S303, retrieve the characteristics of the predefined event type. The characteristics of the predefined event type may include the classification path of the predefined event type.

[0068] According to some embodiments, the method also includes pre-constructing a unified event type system, as will be described in detail below with reference to the accompanying drawings. The predefined event types may come from the constructed unified event type system.

[0069] According to some embodiments, the features of a predefined event type also include auxiliary tags associated with the predefined event type. The auxiliary tags may include keywords characterizing the predefined event type. For example, the auxiliary tags may be keywords extracted from historical events, original descriptive keywords for event categories, or manually set keywords.

[0070] For example, algorithms like TextRank can be used to extract keywords from the content of each historical event, such as verbs and nouns. Then, based on the original category labels of the historical event content, it is mapped to a newly established unified event type system, and the extracted keywords are labeled under the corresponding event type leaf nodes as auxiliary labels. Ultimately, each event type can be associated with the top N (e.g., 5) most frequent labels.

[0071] For example, the following are some examples of characteristics of event types:

[0072] Category path 1: Employment -> Labor Protection -> Wage Arrears Issues;

[0073] Auxiliary tag 1: unpaid wages, arrears, wages, recovery.

[0074] Category path 2: Residential -> Noise pollution -> Daily noise;

[0075] Secondary label 2: noise, disturbance, disturbing the peace, noisy.

[0076] In S305, the features of the event and the features of the predefined event type are input into the trained neural network matching model to obtain at least one matching result.

[0077] According to some example embodiments, the neural network matching model can be the aforementioned reference. Figure 1A and 1BThe matching model described includes sequentially coupled encoding layers, feature cross layers, and output layers, but this application is not limited thereto.

[0078] According to some embodiments, based on the output matching results, the N most similar event type paths (N is an integer greater than 0) can be selected as the classification results.

[0079] According to the example implementation, matching and classification are achieved by matching events with predefined event types, and dynamic expansion of event types is supported.

[0080] According to the example embodiment, the matching model does not limit the type, but instead calculates the similarity after considering hierarchical, multi-granular representations of events and event types. Therefore, the classification method according to the example embodiment has generalization ability and can support the dynamic expansion of event types.

[0081] According to the example implementation, by using a predefined event type system plus auxiliary labels, the mapping relationship is automatically learned and adjusted, making the event classification results more accurate.

[0082] Figure 4 A method for event dispatch according to an example embodiment is shown.

[0083] See Figure 4 In S401, the characteristics of the event are obtained. The characteristics of the event may include the event content.

[0084] According to some embodiments, the characteristics of an event may also include event attributes to accommodate situations where the responsibilities of different departments may overlap, meaning that the same event type may be assigned to different departments under different circumstances. Event attributes may include event time and / or event location, etc.

[0085] For example, an example event message might be: "An elderly person is singing near the airplane model in the children's park, causing serious noise pollution. Please handle this."

[0086] For the content of this event, keywords extracted using algorithms such as TextRank can include: disturbing the peace #1.0, singing #0.942511, model #0.940166, noise #0.648647, airplane #0.646424, children's park #0.640388, elderly #0.63548, handling #0.369955.

[0087] The attributes of an event may include:

[0088] Location of the incident: District D1, City C1;

[0089] Time of the incident: 2:00 AM on December 11, 2021.

[0090] According to some embodiments, the attribute features of an event can be extracted from the event content or entered by the user or data entry person when the event is reported.

[0091] In S403, retrieve the department's characteristics. The department's characteristics may include the classification path of the associated predefined event types.

[0092] According to some embodiments, the characteristics of a department may also include auxiliary labels corresponding to associated predefined event types. According to other embodiments, the characteristics of a department may also include the department's authority and responsibility attributes. Authority and responsibility attributes may include attribute keywords related to jurisdiction, such as jurisdictional area (which may include spatiotemporal relationships).

[0093] For example, the following shows an example of a department's characteristics:

[0094] Category path 1: Employment -> Labor Protection -> Wage Arrears Issues;

[0095] Supporting tag 1: unpaid wages, arrears, wages, recovery;

[0096] Authority and responsibility attribute 1: C1 City, D1 District.

[0097] Category path 2: Residential -> Noise pollution -> Daily noise;

[0098] Secondary label 2: noise, disturbance, disturbing the peace;

[0099] Attribute 2: City C1, District D1, at night, near schools, residential areas, and workplaces.

[0100] Category path 3: Employment -> Labor Protection -> Wage Arrears Issues;

[0101] Supporting tag 3: unpaid wages, arrears, wages, recovery;

[0102] Attribute of rights and responsibilities 3: C1 City, D3 District.

[0103] Classification path 4: Residential -> Noise pollution -> Daily noise;

[0104] Supplementary tag 4: noise, disturbance, disturbing the peace;

[0105] Attribute 4: City C1, District D1, daily operations, all areas.

[0106] In S405, the features of the event and the features of the department are input into the trained neural network matching model to obtain at least one matching result.

[0107] According to some example embodiments, the neural network matching model can be the aforementioned reference. Figure 2A and 2BThe matching model described includes sequentially coupled encoding layers, feature cross layers, and output layers, but this application is not limited thereto.

[0108] According to some implementations, based on the matching results, the N most similar event type paths (N is an integer greater than 0) are selected as the dispatch results.

[0109] According to some embodiments, the method also includes pre-defining event types for departmental associations based on a unified predefined event type system, as detailed below.

[0110] According to the example implementation, end-to-end dispatch is achieved by matching events and departments, supporting the dynamic expansion of departments.

[0111] According to the example implementation, the mapping relationship method is abandoned. End-to-end dispatch is achieved by matching events and departments, which supports department adjustment and has generalization ability.

[0112] According to the example implementation, by using an event type system and auxiliary tags, the event content and departmental responsibilities are represented in a unified way, which realizes the automatic learning and adjustment of the mapping relationship, making the end-to-end event assignment results more accurate.

[0113] Figure 5 This illustrates a method for constructing a unified event type system based on an example embodiment.

[0114] The unified event type system constructed according to the method of this example embodiment can be used in the aforementioned classification and dispatch methods.

[0115] See Figure 5 In S501, at least one existing event classification system is parsed, and a classification path from the root node to the leaf node is constructed, with the end of each classification path being a leaf node type.

[0116] According to the implementation example, the event classification of all existing cities or regions is parsed, and a classification path is constructed from the upper-level classification to the lower-level leaf node classification, i.e., the path is "first-level classification -> second-level classification -> ... -> n-level classification", and each leaf node is the most granular event type.

[0117] In S503, the classification paths for each leaf node type are counted, and one is selected as the preferred path.

[0118] According to the embodiment, the classification path with the most occurrences for each original leaf node type is counted. For classification paths with the same highest frequency, the shortest path is retained as the preferred path for that type, and the path length is equivalent to the type level.

[0119] According to some embodiments, for a classification path that appears only once, if there are many historical events under the classification path, it is retained; otherwise, the classification path can be discarded.

[0120] For example, consider a simple classification system for the following specific cities:

[0121] System A: Incident --> Fire Hazard --> Fire Exit

[0122] System B: Fire Protection --> Fire Equipment Maintenance --> Fire Lanes and Components --> Public Facilities --> Fire Lanes

[0123] C-system: Incident --> Fire Hazard --> Fire Exit

[0124] D System: Public Safety --> Production Safety --> Fire Safety Supervision --> Fire Exits

[0125] The preferred path is: Event --> Fire Hazard --> Fire Exit.

[0126] In S505, all preferred paths are split into child-parent node mappings and the same mappings are aggregated.

[0127] According to the embodiment, all preferred paths are split into a mapping relationship of "child node -> parent node". For example, "first-level category -> second-level category -> third-level category" is split into "third-level category -> second-level category" and "second-level category -> first-level category", and all the same mapping relationship is aggregated.

[0128] During aggregation, synonym replacement normalization can be performed first, and then duplicates can be removed. For example, "public facilities" and "public utility facilities" can be replaced with "public facilities", and "noise" and "noise" can be replaced with "noise".

[0129] In S507, starting from each leaf node, the parent node is recursively traced back according to the mapping relationship to obtain all newly created paths.

[0130] According to the embodiment, starting from each leaf node, the parent node is recursively traced back according to the aforementioned mapping relationship until there is no parent node or a loop occurs, thereby obtaining all paths.

[0131] In S509, a tree is constructed using newly created paths to obtain the event type system.

[0132] According to some implementations, during tree construction, pruning and merging of identical nodes can be performed to remove too few or unreasonable nodes, ultimately resulting in a more standardized tree.

[0133] For example, multiple parent nodes may exist due to differences in the categorization systems of different cities. In such cases, a rule-based voting mechanism can be used. For instance, if the national standard exists, it should be used first. If it doesn't exist in the national standard, the most frequent classification path for each type should be counted. If duplicates still exist, they can be manually modified.

[0134] According to some embodiments, associated auxiliary labels can also be added to the leaf nodes of the event type system.

[0135] Auxiliary tags can be keywords extracted from historical events, original descriptions of event categories, or manually selected keywords. According to some implementations, algorithms such as TextRank can be used to extract keywords from each historical event, for example, extracting verbs and nouns as keywords.

[0136] Then, based on the original category labels of historical event content, they are mapped to the newly established unified event type system. The extracted keywords are then labeled as auxiliary labels under the corresponding event type leaf nodes. Ultimately, each event type can be associated with the top N (e.g., 5) most frequent labels. Simultaneously, the association between event content and the event type system is completed, thus enabling the preparation of training data.

[0137] According to the example implementation, the traditional fixed number of categories is abandoned, and a unified event type system is used to represent event content, thereby supporting adaptive event classification and end-to-end event dispatch.

[0138] According to the example implementation, by constructing a unified event type system, the integration and expansion of different event type systems can be achieved. Furthermore, a tag-based representation can be constructed, utilizing the relevance between event text and type tags to decouple the number of event categories and departmental responsibilities. This enables adaptive intelligent event classification and end-to-end dispatch. The system construction based on this application can continuously accumulate the experience of advanced cities or regions, providing effective reference for the e-government construction of other cities or regions, and can also support the dynamic adjustment of event types.

[0139] Figure 6 This illustrates a method for predefined event types associated with departments, based on an example embodiment.

[0140] The unified event type system constructed according to the method of this example embodiment can be used in the aforementioned dispatch method.

[0141] See Figure 6 In S601, based on the mapping relationship between departments and the existing event classification system, the mapping relationship between departments and the unified event type system is obtained.

[0142] In S603, based on the mapping relationship between departments and the unified event type system, all leaf nodes corresponding to each department are obtained, serving as the event types associated with that department. The features of these event types also serve as the features of the departments. For example, the features of a department may include the classification path of the event type and auxiliary labels. Simultaneously, the association between departments and the event type system is completed, thus enabling the preparation of training data.

[0143] For example, the event types associated with the State Grid Corporation of China as a department (i.e., all corresponding leaf nodes) may include power manhole covers, cable manhole covers, power facilities, power facility signs, power poles, etc., and their characteristics may include the classification paths and auxiliary tags corresponding to these event types.

[0144] According to some embodiments, authority and responsibility attributes can also be added to the characteristics of departments.

[0145] In practical applications, the responsibilities and powers of different departments may overlap. The same type of event may be assigned to different departments under different circumstances. According to some embodiments, corresponding to the event attributes (time, location, etc.), the department characteristics can be supplemented with responsibilities and powers attributes, which may include attribute keywords related to jurisdiction, such as jurisdictional area (which may include spatiotemporal relationships).

[0146] According to the example embodiment, departments rely on a unified event type system, ensuring consistency in adaptive event classification and end-to-end dispatch. Furthermore, tagged and attribute-based representations can be constructed, utilizing the relevance of event text and type tags / attributes for recall, thereby decoupling the number of event categories and departmental responsibilities. The solution according to the embodiments of this application can support dynamic adjustment of event types and departmental responsibilities.

[0147] According to the example implementation, by constructing a dynamically expandable unified event type system, dynamic event classification can be effectively supported, thereby enabling experience accumulation across cities / regions and significantly aiding in the cold start of new cities / regions. The example implementation abandons the traditional method of fixed number of categories plus mapping relationships, utilizing a unified event type system to represent event content and departmental responsibilities, thus supporting adaptive event classification and end-to-end event dispatch.

[0148] After constructing a unified event type system and department-related event types according to the method of the embodiments of this application, the aforementioned neural network matching model can be trained using historical event data. Then, the trained model can be used for event classification and assignment.

[0149] Figure 7 An apparatus for event classification according to an example embodiment of this application is shown.

[0150] like Figure 7As shown, the device 700 for event classification according to the embodiment may include an event feature acquisition module 701, an event type feature acquisition module 703, and a neural network module 705.

[0151] The event feature acquisition module 701 is used to acquire the features of an event, which may include the event content.

[0152] The event type feature acquisition module 703 is used to acquire features of predefined event types, including the classification path of the predefined event type.

[0153] The neural network module 705 is used to output at least one classification matching result based on the features of the input event and the features of a predefined event type.

[0154] The apparatus according to this embodiment can perform functions similar to the methods described above. Other functions are described above and will not be repeated here.

[0155] Figure 8 An apparatus for event dispatch according to an example embodiment of this application is shown.

[0156] like Figure 8 As shown, the device 800 for event dispatch according to the embodiment may include an event feature acquisition module 801, a department feature acquisition module 803, and a neural network module 805.

[0157] The event feature acquisition module 801 is used to acquire the features of an event, including the event content.

[0158] The department feature acquisition module 803 is used to acquire the features of the department, which include the classification path of the associated predefined event types.

[0159] The neural network module 805 is used to output at least one assignment matching result based on the characteristics of the input event and the characteristics of the department.

[0160] The apparatus according to this embodiment can perform functions similar to the methods described above. Other functions are described above and will not be repeated here.

[0161] Figure 9 An apparatus for constructing a unified event type system according to an example embodiment of this application is shown.

[0162] like Figure 9 As shown, the apparatus 900 for constructing a unified event type system according to the embodiment may include a parsing module 901, a path selection module 903, a mapping aggregation module 905, a path module 907, and a construction module 909.

[0163] The parsing module 901 is used to parse at least one existing event classification system and construct classification paths from the root node to the leaf node, with the end of each classification path being a leaf node type.

[0164] The path selection module 903 is used to count the classification paths of each leaf node type and select one as the preferred path.

[0165] The mapping aggregation module 905 is used to split all preferred paths into child-parent node mapping relationships and aggregate the same mapping relationships.

[0166] The path module 907 is used to start from each leaf node and recursively backtrack to the parent node according to the mapping relationship to obtain all newly created paths.

[0167] Module 909 is used to build a tree using newly created paths to obtain an event type system.

[0168] The apparatus according to this embodiment can perform functions similar to the methods described above. Other functions are described above and will not be repeated here.

[0169] Figure 10 An apparatus for a department-related event type is shown according to an example embodiment of this application.

[0170] like Figure 10 As shown, the apparatus 1000 for department-related event types according to an embodiment may include a mapping module 1001 and a feature module 1003.

[0171] The mapping module 1001 is used to obtain the mapping relationship between departments and the unified event type system based on the mapping relationship between departments and the existing event classification system.

[0172] The feature module 1003 is used to obtain all leaf nodes corresponding to a department based on the mapping relationship between the department and the unified event type system. These leaf nodes serve as the event types associated with the department, and the features of the event types also serve as the features of the department.

[0173] The apparatus according to this embodiment can perform functions similar to the methods described above. Other functions are described above and will not be repeated here.

[0174] Figure 11 A block diagram of a computing device according to an example embodiment of this application is shown.

[0175] like Figure 11 As shown, the cloud computing device 30 includes a processor 12 and a memory 14. The cloud computing device 30 may also include a bus 22, a network interface 16, and an I / O interface 18. The processor 12, memory 14, network interface 16, and I / O interface 18 can communicate with each other via the bus 22.

[0176] The processor 12 may include one or more general-purpose CPUs (Central Processing Units), microprocessors, or application-specific integrated circuits, etc., for executing relevant program instructions.

[0177] Memory 14 may include a machine-readable medium in the form of volatile memory, such as random access memory (RAM), read-only memory (ROM), and / or cache memory. Memory 14 is used to store one or more programs containing instructions, as well as data. Processor 12 can read the instructions stored in memory 14 to perform the methods described above according to embodiments of this application. Electronic device 12 may further include other removable / non-removable, volatile / non-volatile storage devices.

[0178] The cloud computing device 12 can also communicate with one or more networks through the network interface 16. The network interface 16 can be a wired network interface, a wireless network interface, or a virtual network interface.

[0179] The cloud computing device 12 can also communicate with one or more external devices (such as audio input devices, audio output devices, cameras, keyboards, mice, monitors, various sensors, etc.) through the input / output (I / O) interface 18.

[0180] Bus 22 may include an address bus, a data bus, a control bus, etc. Bus 22 provides a path for exchanging information between components.

[0181] It should be noted that, in specific implementations, the cloud computing device 30 may also include other components necessary for normal operation. Furthermore, those skilled in the art will understand that the device described above may only include the components necessary for implementing the embodiments of this specification, and not necessarily all the components shown in the figures.

[0182] This application also provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of the above-described method. The computer-readable storage medium may include, but is not limited to, any type of disk, including floppy disks, optical disks, DVDs, CD-ROMs, microdrives, as well as magneto-optical disks, ROMs, RAMs, EPROMs, EEPROMs, DRAMs, VRAMs, flash memory devices, magnetic cards or optical cards, nanosystems (including molecular memory ICs), network storage devices, cloud storage devices, or any type of medium or device suitable for storing instructions and / or data.

[0183] This application also provides a computer program product including a non-transitory computer-readable storage medium storing a computer program operable to cause a computer to perform some or all of the steps of any of the methods described in the above method embodiments.

[0184] Those skilled in the art will clearly understand that the technical solutions of this application can be implemented using software and / or hardware. In this specification, "unit" and "module" refer to software and / or hardware capable of independently performing or cooperating with other components to perform a specific function, where the hardware may be, for example, a field-programmable gate array (FPGA), integrated circuit, etc.

[0185] It should be noted that, for the sake of simplicity, the foregoing method embodiments are all described as a series of actions. However, those skilled in the art should understand that this application is not limited to the described order of actions, as some steps may be performed in other orders or simultaneously according to this application. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential to this application.

[0186] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0187] In the several embodiments provided in this application, it should be understood that the disclosed apparatus can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. For example, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the coupling or direct coupling or communication connection shown or discussed may be through some service interface; the indirect coupling or communication connection between devices or units may be electrical or other forms.

[0188] The units described as separate components may or may not be physically separate. 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 units can be selected to achieve the purpose of this embodiment according to actual needs.

[0189] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0190] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage device. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application.

[0191] In the above embodiments, the descriptions of each embodiment have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions in other embodiments.

[0192] The embodiments of this application have been described and explained in detail above. It should be clearly understood that this application describes how specific examples are formed and used, but this application is not limited to any details of these examples. Rather, based on the teachings of the disclosure of this application, these principles can be applied to many other embodiments.

[0193] Through the description of the example embodiments, those skilled in the art will readily understand that the technical solutions according to the embodiments of this application have at least one or more of the following advantages.

[0194] According to the example implementation, by constructing a unified event type system that can be dynamically expanded, dynamic event classification can be effectively supported, thereby enabling experience accumulation across cities / regions and providing significant assistance for the cold start of new cities / regions.

[0195] According to the example implementation, by using an event type system plus event keywords, the event content and departmental responsibilities are represented in a unified way, realizing the automatic learning and adjustment of the mapping relationship, and realizing an adaptive event classification and end-to-end event dispatch model.

[0196] According to the example implementation, the matching model does not limit the type, but instead calculates the similarity after hierarchical, multi-granular representations of events and types. Therefore, it can support functions such as dynamically expanding event types and adjusting departmental responsibilities.

[0197] According to the example implementation, the traditional method of fixed number of categories plus mapping relationship is abandoned. Instead, a unified event type system is used to represent event content and departmental responsibilities, thereby realizing an adaptive event classification and end-to-end event dispatch model.

[0198] According to the example implementation, a multi-granular and multi-level matching algorithm for event content and type and department is constructed by using event content, keywords, event attributes, and features such as departmental authority and responsibility event type path and authority and responsibility attributes. This algorithm supports the dynamic expansion of event types and fine-grained (temporal and spatial) adjustment of departmental authority and responsibility.

[0199] The following terms will help to better understand the technical solution of this application:

[0200] Clause 1. A method for event classification, characterized in that it includes:

[0201] Acquire the characteristics of the event, including the event content;

[0202] Obtain the features of a predefined event type, wherein the features of the predefined event type include the classification path of the predefined event type;

[0203] The features of the event and the features of the predefined event type are input into a trained neural network matching model to obtain at least one output matching result.

[0204] Clause 2. The method as described in Clause 1, characterized in that,

[0205] The features of the predefined event type also include auxiliary tags associated with the predefined event type, the auxiliary tags including keywords characterizing the predefined event type.

[0206] Clause 3, the method as described in Clause 1, wherein the features of the event further include keywords extracted from the content of the event.

[0207] Clause 4. The method described in Clause 1, characterized in that it further includes: pre-constructing a unified event type system, wherein the predefined event types are derived from the unified event type system.

[0208] Clause 5. The method as described in Clause 1, wherein the neural network matching model comprises a sequentially coupled encoding layer, a feature crossover layer, and an output layer.

[0209] Clause 6. A method for event dispatching, characterized in that it includes:

[0210] Acquire the characteristics of the event, including the event content;

[0211] Obtain the characteristics of the department, including the classification path of the predefined event types associated with the department;

[0212] The features of the event and the features of the department are input into a trained neural network matching model to obtain at least one output matching result.

[0213] Clause 7. The method as described in Clause 6, characterized in that,

[0214] The characteristics of the event also include the event attributes of the event and keywords extracted from the event content;

[0215] The characteristics of the department also include auxiliary tags corresponding to predefined event types associated with the department, and the department's authority and responsibility attributes. The auxiliary tags include keywords that characterize the predefined event types associated with the department.

[0216] Clause 8. The method as described in Clause 7, characterized in that it further comprises:

[0217] Pre-build a unified event type system;

[0218] Predefine event types for the departments based on the unified event type system.

[0219] Clause 9. The method as described in Clause 6, wherein the neural network matching model comprises a sequentially coupled encoding layer, a feature crossover layer, and an output layer.

[0220] Clause 10. A method for constructing a unified event type system, characterized by comprising:

[0221] Parse at least one existing event classification system and construct classification paths from the root node to the leaf node, with the end of each classification path being the leaf node type;

[0222] Analyze the classification paths for each leaf node type and select one as the preferred path;

[0223] All preferred paths are broken down into child-parent node mappings, and identical mappings are aggregated.

[0224] Starting from each leaf node, recursively backtrack the parent node according to the mapping relationship to obtain all newly created paths;

[0225] The newly created path is used to construct a tree, thus obtaining the event type system.

[0226] Clause 11. The method as described in Clause 10, characterized in that it further comprises:

[0227] Add associated auxiliary tags to the leaf nodes of the event type system, the auxiliary tags including keywords that characterize the event type.

[0228] Clause 12. A method for department-related event types, characterized in that it includes:

[0229] Obtain a unified event type system;

[0230] Based on the mapping relationship between departments and existing event classification systems, the mapping relationship between the departments and the unified event type system is obtained;

[0231] Based on the mapping relationship between the department and the unified event type system, all leaf nodes corresponding to the department are obtained, which serve as the event types associated with the department. The characteristics of the event types also serve as the characteristics of the department.

[0232] Clause 13. The method as described in Clause 12, characterized in that it further comprises:

[0233] Add authority and responsibility attributes to the characteristics of the department.

[0234] Clause 14. A computing device, characterized in that it comprises:

[0235] processor;

[0236] A memory, on which computer programs are stored;

[0237] When the processor executes the computer program, it implements the method as described in any one of Clauses 1-3.

[0238] Exemplary embodiments of this application have been specifically shown and described above. It should be understood that this application is not limited to the detailed structures, arrangements, or implementation methods described herein; rather, this application is intended to cover various modifications and equivalent arrangements contained within the spirit and scope of the appended provisions.

Claims

1. A method for event classification, characterized in that, include: Acquire the characteristics of the event, including the event content; Obtain the features of a predefined event type, wherein the features of the predefined event type include the classification path of the predefined event type; The features of the event and the features of the predefined event type are input into a trained neural network matching model to obtain at least one output matching result; The method further includes: pre-constructing a unified event type system, wherein the predefined event types are derived from the unified event type system; The pre-built unified event type system includes: Parse at least one existing event classification system and construct classification paths from the root node to the leaf node, with the end of each classification path being the leaf node type; Analyze the classification paths for each leaf node type and select one as the preferred path; All preferred paths are broken down into child-parent node mappings, and identical mappings are aggregated. Starting from each leaf node, recursively backtrack the parent node according to the mapping relationship to obtain all newly created paths; The newly created path is used to construct a tree, resulting in the unified event type system.

2. The method as described in claim 1, characterized in that, The features of the predefined event type also include auxiliary tags associated with the predefined event type, the auxiliary tags including keywords characterizing the predefined event type.

3. The method as described in claim 1, characterized in that, The characteristics of the event also include keywords extracted from the event content.

4. The method as described in claim 1, characterized in that, Also includes: Add associated auxiliary tags to the leaf nodes of the unified event type system, the auxiliary tags including keywords that characterize the event type.

5. A method for event dispatching, characterized in that, include: Acquire the characteristics of the event, including the event content; Obtain the characteristics of the department, including the classification path of the predefined event types associated with the department; The features of the event and the features of the department are input into a trained neural network matching model to obtain at least one output matching result; The method further includes: Pre-build a unified event type system; Predefine event types for the departmental associations based on the unified event type system; The pre-built unified event type system includes: Parse at least one existing event classification system and construct classification paths from the root node to the leaf node, with the end of each classification path being the leaf node type; Analyze the classification paths for each leaf node type and select one as the preferred path; All preferred paths are broken down into child-parent node mappings, and identical mappings are aggregated. Starting from each leaf node, recursively backtrack the parent node according to the mapping relationship to obtain all newly created paths; The newly created path is used to construct a tree, resulting in the unified event type system.

6. The method as described in claim 5, characterized in that, The characteristics of the event also include the event attributes of the event and keywords extracted from the event content; The characteristics of the department also include auxiliary tags corresponding to predefined event types associated with the department, and the department's authority and responsibility attributes. The auxiliary tags include keywords that characterize the predefined event types associated with the department.

7. The method as described in claim 5, characterized in that, Also includes: Add associated auxiliary tags to the leaf nodes of the unified event type system, the auxiliary tags including keywords that characterize the event type.

8. The method according to any one of claims 5-7, characterized in that, Also includes: Obtain a unified event type system; Based on the mapping relationship between departments and existing event classification systems, the mapping relationship between the departments and the unified event type system is obtained; Based on the mapping relationship between the department and the unified event type system, all leaf nodes corresponding to the department are obtained, which serve as the event types associated with the department, and the characteristics of the event types also serve as the characteristics of the department. Add authority and responsibility attributes to the characteristics of the department.

Images