A method and system for determining event attributes

By combining structured and narrative information, using deep learning models to determine event attributes, the problem of customer service's limited ability to handle security events is solved, the accuracy and efficiency of event attribute determination is improved, and security risks are reduced.

CN111858725BInactive Publication Date: 2024-11-12BEIJING DIDI INFINITY TECH & DEV CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202010365398.4
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2020-04-30
Publication Date
2024-11-12
Estimated Expiration
Not applicable · inactive patent

AI Technical Summary

Technical Problem

Customer service has limited ability when handling security incidents, and subjective factors affect the judgment results, resulting in misjudgment, and it is difficult for the existing technology to effectively determine the event attributes.

Method used

By obtaining the work ticket information generated by event description, combining structured information and narrative information, using the text transformation model and extraction model to extract features, and determining event attributes through feature fusion and classification models, providing event attribute determination methods and systems.

Benefits of technology

It improves the accuracy of event attribute determination, reduces customer service work pressure, reduces safety hazards, and prevents vicious car accidents.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN111858725B_ABST
    Figure CN111858725B_ABST
Patent Text Reader

Abstract

The embodiment of the present application discloses a method, system, device and storage medium for determining event attributes. The method comprises: obtaining work order information generated according to the event description of the target event, the work order information at least comprising structured information and narrative information; obtaining a first feature according to the narrative information; obtaining a second feature according to the structured information; fusing the first feature and the second feature to obtain a fused feature; and obtaining the event attribute of the target event based on the fused feature. The present application can improve the accuracy of event attribute determination.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present application relates to the field of data processing technology, and in particular to a method and system for determining event attributes. Background Art

[0002] With the development of shared transportation services, people are increasingly accustomed to handling safety issues through customer service, but the ability of customer service to handle safety incidents is limited. In addition, subjective factors have a greater impact on the judgment results of customer service, which may cause misjudgment of safety incidents. Therefore, it is necessary to provide a method and system for determining event attributes. Summary of the invention

[0003] The first aspect of the present application provides an event attribute determination method. The event attribute detection and determination method comprises: obtaining work order information generated according to the event description of the target event, the work order information at least comprising structured information and narrative information; obtaining a first feature according to the narrative information; obtaining a second feature according to the structured information; fusing the first feature and the second feature to obtain a fused feature; and obtaining the event attribute of the target event based on the fused feature.

[0004] The second aspect of the present application provides an event attribute determination system, including: a work order acquisition module, used to obtain work order information generated according to the event description of the target event, the work order information at least including structured information and narrative information; a first feature extraction module, used to obtain a first feature based on the narrative information; a second feature extraction module, used to obtain a second feature based on the structured information; a feature fusion module, used to fuse the first feature and the second feature to obtain a fused feature; and a determination module, used to obtain the event attribute of the target event based on the fused feature.

[0005] In some embodiments, the work order information is generated by customer service based on the event description and / or the result of processing the event.

[0006] In some embodiments, the structured information in the work order information includes at least one of the following information: the order number, license plate number, telephone number, whether the police have been called, whether the police have filed a case, whether the user has requested processing, and the urgency of the processing; the narrative information in the work order information includes at least one of the following information: the user's description of the incident, the police's handling results description, and the customer service's handling results description.

[0007] In some embodiments, the first feature extraction module processes the narrative information using a text conversion model to obtain the first feature.

[0008] In some embodiments, the text conversion model includes at least one of the following deep learning models: Fasttext, HAN, Text CNN, Transformer, LR, and XG Boost.

[0009] In some embodiments, a training module is also included, which is used to: obtain a sample set, the sample set includes multiple work order information, use the narrative information in the work order information as input, and use the event attributes corresponding to the work order information given by the customer service as identifiers to train the text conversion model.

[0010] In some embodiments, the second feature extraction module generates the second feature by processing structured information through an extraction model.

[0011] In some embodiments, the extraction model uses rule-based and / or AC automaton methods to perform feature extraction.

[0012] In some embodiments, the feature fusion module directly concatenates the first feature and the second feature or processes the first feature and the second feature through a set algorithm to generate a combined feature to obtain a fused feature.

[0013] In some embodiments, the determination module uses a classification model to process the fusion features to obtain a classification result of the event attribute; the classification result of the event attribute includes at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, and whether the event is repeatable.

[0014] In some embodiments, the classification model includes at least one of the following deep learning models: XG Boost, GBDT, Adaboost, random forest.

[0015] In some embodiments, a training module is also included, which is used to: obtain a sample set, the sample set includes multiple work order information, use the fusion features corresponding to the work order information as input, and use the event attributes corresponding to the work order information given by the customer service as identifiers to train the classification model.

[0016] The third aspect of the present application provides an event attribute determination device, comprising at least one processor and at least one memory; the at least one memory is used to store computer instructions; the at least one processor is used to execute at least part of the computer instructions to implement the operations described above.

[0017] A fourth aspect of the present application provides a computer-readable storage medium, which stores computer instructions. When the computer instructions are executed by a processor, the operations described above are implemented. BRIEF DESCRIPTION OF THE DRAWINGS

[0018] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. Among them:

[0019] Figure 1 is a schematic diagram of an application scenario of an event attribute determination system according to some embodiments of the present application;

[0020] Figure 2 is a schematic block diagram of an event attribute determination system according to some embodiments of the present application;

[0021] Figure 3 is an exemplary flow chart of a method for determining event attributes according to some embodiments of the present application; and

[0022] Figure 4 These are examples of common structured information and narrative information in some embodiments of the present application. DETAILED DESCRIPTION

[0023] In order to more clearly illustrate the technical solutions of the embodiments of the present application, the following is a brief introduction to the drawings required for the description of the embodiments. Obviously, the drawings described below are only some examples or embodiments of the present application. For ordinary technicians in this field, the present application can also be applied to other similar scenarios based on these drawings without creative work. Unless it is obvious from the language environment or otherwise explained, the same reference numerals in the figures represent the same structure or operation.

[0024] As shown in this application and claims, unless the context clearly indicates an exception, the words "a", "an", "an" and / or "the" do not refer to the singular and may also include the plural. Generally speaking, the terms "include" and "comprise" only indicate the inclusion of the steps and elements that have been clearly identified, and these steps and elements do not constitute an exclusive list. The method or device may also include other steps or elements.

[0025] Although the present application makes various references to certain modules in the system according to embodiments of the present application, any number of different modules may be used and run on the vehicle client and / or server. The modules are illustrative only, and different aspects of the system and method may use different modules.

[0026] Flowcharts are used in the present application to illustrate the operations performed by the system according to the embodiments of the present application. It should be understood that the preceding or following operations are not necessarily performed accurately in order. On the contrary, various steps may be processed in reverse order or simultaneously. At the same time, other operations may also be added to these processes, or a certain step or several steps of operations may be removed from these processes.

[0027] In addition, although the system and method disclosed in this application are mainly described with respect to transportation services, it should be understood that this is only an exemplary embodiment. The system and method of the present invention may be applicable to any other on-demand service, such as housekeeping services, take-out services, etc. In some embodiments, the system and method of the present application can be applied to different transportation systems, including land, sea, aerospace, etc. or any combination of the above examples. The means of transportation used in the transportation system may include taxis, private cars, ride-sharing, buses, trains, motor vehicles, high-speed railways, subways, ships, airplanes, spacecraft, hot air balloons, unmanned vehicles, etc. or combinations thereof. The transportation system may also include any transportation system suitable for management and / or distribution, for example, a system for transmitting and / or receiving express delivery. The application scenarios of the system or method of the present application may also include web pages, browser plug-ins, client terminals, customized systems, internal analysis systems, artificial intelligence robots, etc. or any combination thereof.

[0028] It can be understood that in some actual application scenarios, customer service needs to handle a large number of events that occur during on-demand services every day and quickly give processing results. As an example only, in shared transportation services, users rely on customer service to handle security incidents, but the actual ability of customer service to handle security incidents is limited. For example, a customer service may assess an incident with potential safety hazards as a normal incident. In some embodiments, the relevant information of security incidents handled by customer service can be processed by computer algorithms to find out whether there are any missed security hazards in the security incidents handled by customer service, and intervene in them to prevent the occurrence of vicious ride accidents. In some embodiments, customer service processes security incidents based on the event description provided by the user to obtain work order information. The relevant event description and the processing results of the customer service can be included in the work order information corresponding to the event. By processing the work order information by computer algorithms, potential safety hazards can be detected again, thereby effectively reducing the work pressure of customer service.

[0029] Figure 1 It is a schematic diagram of an application scenario of an event attribute determination system according to some embodiments of the present application.

[0030] like Figure 1 As shown, the exemplary event attribute determination system 100 may include a server 110 , a network 120 , a user terminal 130 , and a storage module 140 .

[0031] In some embodiments, the server 110 can be used to determine event attributes. The server 110 can be a single server or a server group. A server group can be centralized, such as a data center. A server group can also be distributed, such as a distributed system. The server 110 can be local or remote. In some embodiments, the server 110 can include a control processor 112 for executing instructions (program code) of the server 110. For example, the control processor 112 can execute instructions of an event detection program, and then analyze and process the work order information through a certain algorithm to determine the detection result.

[0032] In some embodiments, the terminal 130 includes a service requester terminal and / or a customer service terminal. As an example only, the service requester may be an individual who initiates a vehicle request. In some embodiments, the service requester may send an event description of a target event that occurs during the transportation service to a service end or a customer service terminal through its terminal. In some embodiments, the customer service terminal may receive an event description of a target event sent by the service requester terminal, and process the target event to generate work order information and / or processing results. In some embodiments, the work order information may include structured information, narrative information, and / or event attributes initially determined by the customer based on the event description. In some embodiments, the customer service terminal may return the processing result to the service requester terminal, or send the work order information to the server 110 for further processing. The terminal 130 includes, but is not limited to, one or a combination of mobile phones 130-1, tablet computers 130-2, laptop computers 130-3, etc. The server 110 may access the work order information stored in or stored in the storage module 140, or transmit the detection result to the user terminal 130 through the network 120.

[0033] In some embodiments, the storage module 140 may refer to a device with a storage function. The storage module 140 is mainly used to store the event description sent from the user terminal 130 and various data generated during the operation of the server 110. The storage module 140 may be local or remote. The connection or communication between the system database and other modules of the system may also be wired or wireless. The network 120 may provide a channel for information exchange. The network 120 may be a single network or a combination of multiple networks. The network 120 may include but is not limited to one or a combination of a local area network, a wide area network, a public network, a private network, a wireless local area network, a virtual network, a metropolitan area network, a public switch telephone network, etc. The network 120 may include multiple network access points, such as wired or wireless access points, base stations (such as 120-1, 120-2) or network exchange points, through which the data source is connected to the network 120 and information is sent through the network 120.

[0034] It should be understood by those skilled in the art that when an element of the event attribute determination system 100 is executed, the element may be executed by an electrical signal and / or an electromagnetic signal. For example, when the user terminal 130 processes a task, such as when a user makes an event description, the user terminal 130 may operate a logic circuit in its processor to process such a task. When the user terminal 130 issues an instruction to the server 110, the processor of the user terminal 130 may generate an electrical signal encoding the instruction. Then, the processor of the user terminal 130 may send the electrical signal to an output port. If the user terminal 130 communicates with the server 110 via a wired network, the output port may be physically connected to a cable, which further transmits the electrical signal to the input port of the server 110. If the user terminal 130 communicates with the server 110 via a wireless network, the output port of the user terminal 130 may be one or more antennas, which convert the electrical signal into an electromagnetic signal. Similarly, the storage module 140 may process a task by the operation of a logic circuit in its processor, and receive instructions and / or information from the server 110 via an electrical signal or an electromagnetic signal. In an electronic device, such as a user terminal 130, a storage module 140, and / or a server 110, when a processor processes an instruction, issues an instruction, and / or performs an action, the instruction and / or the action is performed via an electrical signal. For example, when the server 110 retrieves data from the storage module 140, it may send an electrical signal to a reading device of the storage medium, which may read structured information or narrative information in the storage medium. The structured information or narrative information may be transmitted to the processor via a bus of the electronic device in the form of an electrical signal. Here, an electrical signal may refer to one electrical signal, a series of electrical signals, and / or at least two discrete electrical signals.

[0035] Figure 2 It is a schematic block diagram of an event attribute determination system according to some embodiments of the present application.

[0036] like Figure 2 As shown, in some embodiments, the event attribute determination system 200 may include: a work order acquisition module 210, used to acquire work order information generated according to the event description of the target event, the work order information including at least structured information and narrative information; a first feature extraction module 220, used to obtain a first feature based on the narrative information; a second feature extraction module 230, used to obtain a second feature based on the structured information; a feature fusion module 240, used to fuse the first feature and the second feature to obtain a fused feature; and a determination module 250, used to obtain the event attribute of the target event based on the fused feature.

[0037] In some embodiments, the work order acquisition module 210 may be used to perform step 301. In some embodiments, the work order information is generated by the customer service based on the event description and the result of the event processing.

[0038] In some embodiments, the structured information in the work order information may include at least one of the following information: order number, license plate number, telephone number, whether to call the police, whether the police have filed a case, whether the user has requested processing, and the urgency of the processing; the narrative information in the work order information may include at least one of the following information: user description of the event, description of the police processing result, and description of the customer service processing result. For more relevant information, see step 301.

[0039] In some embodiments, the first feature extraction module 220 can be used to perform step 303. In some embodiments, the first feature extraction module can use a text conversion model to process the narrative information to obtain the first feature. In some embodiments, the text conversion model can include at least one of the following deep learning models: Fast text, HAN, Text CNN, Transformer, LR and XG Boost. In some embodiments, the method for training the text conversion model may include: obtaining a sample set, the sample set including multiple work order information, using the narrative information in the work order information as input, and using the event attribute corresponding to the work order information given by the customer service as an identifier to train the text conversion model. For more relevant information, see step 303.

[0040] In some embodiments, the second feature extraction module 230 can be used to perform step 305. In some embodiments, the second feature extraction module can generate the second feature by processing the structured information through an extraction model. In some embodiments, the extraction model can use a rule-based and / or AC automaton method to extract features. For more information, see step 305.

[0041] In some embodiments, the feature fusion module 240 may be used to perform step 307. In some embodiments, the feature fusion module may directly concatenate the first feature and the second feature or generate a combined feature by processing the first feature and the second feature through a set algorithm to obtain a fused feature. For more information, see step 307.

[0042] In some embodiments, the determination module 250 can be used to perform step 309. In some embodiments, the determination module can use a classification model to process the fusion feature to obtain a classification result of the event attribute. In some embodiments, the classification result of the event attribute obtained by the determination module may include at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, and whether the event is repeatable. In some embodiments, the classification model may include at least one of the following deep learning models: XG Boost, GBDT, Adaboost, and random forest. In some embodiments, the method for training the classification model may include: obtaining a sample set, the sample set including multiple work order information, using the fusion feature corresponding to the work order information as input, and using the event attribute corresponding to the work order information given by the customer service as an identifier to train the classification model. For more relevant information, see step 309.

[0043] It should be understood that Figure 2 The system and its modules shown can be implemented in various ways. For example, in some embodiments, the system and its modules can be implemented by hardware, software or a combination of software and hardware. Among them, the hardware part can be implemented using dedicated logic; the software part can be stored in a memory and executed by an appropriate instruction execution system, such as a microprocessor or a dedicated design hardware. Those skilled in the art will understand that the above-mentioned method and system can be implemented using computer executable instructions and / or included in a processor control code, such as a carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. Such code is provided on the carrier medium such as a disk, CD or DVD-ROM, a programmable memory such as a read-only memory (firmware), or a data carrier such as an optical or electronic signal carrier. The system and its modules of the present application can not only be implemented by hardware circuits such as ultra-large-scale integrated circuits or gate arrays, semiconductors such as logic chips, transistors, etc., or programmable hardware devices such as field programmable gate arrays, programmable logic devices, etc., but can also be implemented by software such as executed by various types of processors, and can also be implemented by a combination of the above-mentioned hardware circuits and software (for example, firmware).

[0044] It should be noted that the above description of the system and its modules is only for convenience of description and does not limit the present application to the scope of the embodiments. It is understandable that, after understanding the principle of the system, those skilled in the art may arbitrarily combine the modules or form a subsystem to connect with other modules without deviating from the principle. For example, in some embodiments, Figure 2The work order acquisition module 210, the first feature extraction module 220, etc. disclosed in the disclosure may be different modules in a system, or one module may realize the functions of two or more of the above modules. For another example, each module may share a storage module 140, or each module may have its own storage module 140. Such variations are all within the protection scope of this application.

[0045] Figure 3 It is an exemplary flow chart shown according to some embodiments of the present application, used to illustrate the implementation steps of the event attribute determination method.

[0046] In some other embodiments of the present application, a method for determining event attributes is provided. The method 300 may include the following steps:

[0047] Step 301 : Acquire work order information generated according to the event description of the target event, wherein the work order information may include at least structured information and narrative information. In some embodiments, this step may be performed by the work order acquisition module 210 in the system 200 .

[0048] In some embodiments, the event description may include a description of the event provided by the user. In some embodiments, the event description provided by the user includes, but is not limited to, video, voice, text, etc. The customer service processes the event description to obtain work order information. In some embodiments, the customer service processes the event description manually or semi-automatically, and semi-automation may refer to the use of various software and hardware to help the customer service process information.

[0049] In some embodiments, work order information may include structured information and narrative information. Structured information may refer to information that can be decomposed into multiple interrelated components after analysis, and each component has a clear hierarchical structure. The use and maintenance of structured information can be managed through a database, and there are certain operating specifications. For example, structured information can be represented and stored using a relational database, and is expressed as two-dimensional data. Data is in units of lines, and a line of data represents the information of an entity, and the attributes of each line of data are the same. The storage and arrangement of data are regular, which facilitates subsequent query and modification operations. Structured information can be applied to, for example, enterprise ERP, financial systems, medical HIS databases, education cards, government administrative approvals, and other core databases. The storage solution for structured information may include high-speed storage application requirements, data backup requirements, data sharing requirements, and data disaster recovery requirements.

[0050] In some embodiments, structured information may include order numbers, license plate numbers, telephone numbers, and information that can be expressed as "yes or no" related to the event, such as whether to call 110, whether the police have filed a case, etc. It may also refer to information with hierarchical levels, such as whether the user requests processing and the urgency of the processing. In some embodiments, structured information may not conform to the form of a relational database or other data table, but contains relevant tags to separate semantic elements and to stratify records and fields. Therefore, structured information is also called a self-describing structure. Structured information belongs to the same type of entity but can have different attributes, and even if they are combined together, the order of these attributes is not important. Common structured information includes XML and JSON. Structured information can be applied to mail systems, WEB clusters, teaching resource libraries, data mining systems, archive systems, etc. The storage solution for structured information can include basic storage needs such as data storage, data backup, data sharing, and data archiving.

[0051] In some embodiments, narrative information may refer to unstructured data with irregular or incomplete data structure, no predefined data model, and inconvenient to be represented by a two-dimensional logical table of a database. Narrative information may include office documents, texts, pictures, various reports, images, and audio / video information of all formats, etc. The formats of narrative information are very diverse, and the standards are also diverse. Moreover, technically, unstructured information is more difficult to standardize and understand than structured information. Therefore, storage, retrieval, publication, and utilization require more intelligent information technology, such as mass storage, intelligent retrieval, knowledge mining, content protection, value-added development and utilization of information, etc. The storage solution for narrative information may include data storage, data backup, and data sharing, etc.

[0052] In some embodiments, narrative information may refer to the user's description of the incident recorded by the customer service, the description of the police's handling result, and the description of the customer service's handling result. For example, the description of the user's problem is "the car was scratched", the customer service handling is "arranging a specialist and giving feedback to the traffic police", and the result is "user approval". Narrative information can be obtained by recording, saving videos, saving call records, etc.

[0053] In some embodiments, the work order information may be generated by the customer service based on the event description and the result of the event processing. In some embodiments, the work order information may refer to the information obtained after the user reflects the event to the customer service and the event is initially processed by the customer service. The work order information is converted into structured information and / or narrative information by the customer service in a manual or semi-automatic manner. For example, the process of the user's conversation with the customer service is recorded, and the recording is converted into structured information and / or narrative information.

[0054] In some embodiments, the structured information in the work order information includes at least one of the following information: the order number, license plate number, telephone number, whether the police were called, whether the police filed a case, whether the user requested processing and the urgency of the processing, etc.; the narrative information in the work order information includes at least one of the following information: the user's description of the incident, the police's handling results description, and the customer service's handling results description, etc.

[0055] like Figure 4 As shown, the customer service presents the work order information as structured information and / or narrative information based on the event description. The narrative information obtained includes: 1. Description of the user's problem: The user reported that he picked up a passenger, and the passenger hit the door on the bicycle when driving, causing the door to be scratched. At present, he has asked the passenger for 100 yuan and is worried that the passenger will complain about him. 2. Customer service processing: We are sorry to bring you a bad experience. We attach great importance to the problem you reported. We will strengthen the control of passengers in the future. We will immediately feedback your problem to the relevant department. The relevant customer service specialist will contact you within 2 hours to handle it. Please keep the phone open. 3. Result: User approval. The structured information obtained may include whether it is recommended that the user call 110 ("No"), the police handling result ("Unknown"), the user complaint ("Request for processing"), and whether the vehicle was maliciously damaged ("Unknown") and other information.

[0056] Step 303 , obtaining a first feature according to the narrative information. In some embodiments, this step may be performed by the first feature extraction module 220 in the system 200 .

[0057] In some embodiments, obtaining the first feature according to the narrative information includes: using a text conversion model to process the narrative information to obtain the first feature. The text conversion model includes at least one of the following deep learning models: Fast text, HAN, Text CNN, Transformer, LR, and XG Boost.

[0058] Fast text is a fast text classifier. Text classification refers to the classification of documents into one or more categories, which can be evaluation scores, degree of danger, degree of urgency, harassment information, etc. In order to build a classifier, label data is required. Label data refers to data and the category (i.e., identifier or label) corresponding to this data. For example, after being processed by customer service, the customer service manually gives each work order a label as to whether it is dangerous. Among them, the data with category labels is historical data for a long time. A sequence of words (a piece of text or a sentence) is input into the Fast text model, and the model outputs the probability that this word sequence belongs to different categories. The words and phrases in the sequence form a feature vector, which is mapped to the hidden layer through linear transformation, and the hidden layer is mapped to the label. Fast text classifies narrative information in this way.

[0059] HAN (Hierarchy Attention Network) is a model that uses the attention mechanism to represent hierarchical words and sentences in sentences and documents. The model corresponds to the hierarchical structure of documents: words constitute sentences, and sentences constitute documents, so the model also constructs text vector expressions in these two parts. Secondly, different words and sentences have different amounts of information and cannot be treated uniformly, so the Attention mechanism is introduced. In addition to improving the accuracy of the model, the introduction of the Attention mechanism can also analyze and visualize the importance of words and sentences, enhancing interpretability. HAN can include a word sequence encoder, a word-level attention layer, a sentence sequence encoder, a sentence-level attention layer, and the like.

[0060] Transformer can process all words or symbols in a sequence in parallel while incorporating context with more distant words using a self-attention mechanism. Self-attention refers to an attention mechanism that associates different positions of a single sequence. This is done by processing all words in parallel and letting each word pay attention to other words in the sentence in multiple processing steps.

[0061] LR (logistic regressive) or XG Boost (extreme gradient boosting) can assign specific weight values ​​to each word or character in the text of the training sample according to the corresponding classification. For example, in the process of training XG Boost using the training sample, XG Boost can give the weight values ​​of the words or characters in the sample to indicate the importance of each word or character to the model training.

[0062] In some embodiments, the method for training the text conversion model includes: obtaining a sample set, the sample set includes multiple work order information, taking the narrative information in the work order information as input, taking the event attribute corresponding to the work order information given by the customer service as an identifier, and training the text conversion model. Among them, the event attribute given by the customer service may include at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, whether the event is repeatable, whether the event is urgent, and whether the event can be warned. In Fast text, the sample set is labeled narrative information. The sample set can be narrative information with a very large classification category and a large enough data set to avoid overfitting. The narrative information is input into the model, and the model can output the probability that the narrative information belongs to different categories. The model can also output the label category with the maximum probability corresponding to the narrative information. In HAN, the sample set is labeled narrative information. The processor can convert the labeled narrative information into a word vector sequence by preprocessing it, and input the word vector sequence into the model. Preprocessing refers to preliminary processing of the labeled narrative information. In some embodiments, preprocessing can include unifying lowercase, removing garbled characters, removing abbreviations and numbers, removing stop words, etc. Stop words refer to some frequently appearing words that do not have much effect on classification. Stop words that appear in the text can be removed according to the stop word list. During the training process, the model trains the weights of words in sentences, trains the weights of sentences in documents, and outputs the vector representation of the text. In some embodiments, a softmax classifier can also be used to classify the entire text. The softmax classifier can output the probabilities to which the corresponding information belongs. The model can output the label category with the maximum probability corresponding to the narrative information. In Transformer, the sample set is labeled narrative information. Input narrative information into the model. Output the probabilities of narrative information corresponding to different categories. Output the label category with the maximum probability corresponding to the narrative information. In logistic regression (LR) or XG Boost, the sample set is labeled narrative information. Input narrative information into the model, assign a specific weight value to each character or word of the training sample in the sample set, and output the label category with the maximum weight corresponding to the narrative information.

[0063] Step 305 , obtaining a second feature according to the structured information. In some embodiments, this step may be performed by the second feature extraction module 230 in the system 200 .

[0064] In some embodiments, obtaining the second feature according to the structured information includes: generating the second feature by processing the structured information through an extraction model. For example, "whether to recommend the user to call 110". If it is "no", the extraction method is: "none" → feature_110_0 → the corresponding dimension of feature_110_0 is 1; if it is "yes", the extraction method is: "yes" → feature_110_1 → the corresponding dimension of feature_110_1 is 1. In some embodiments, the extraction model can use rule-based and / or AC automaton methods for feature extraction. AC automaton is a string search algorithm used to match substrings in a finite set of "dictionaries" in a string of input strings. A dictionary refers to a collection of elements, some of which may include words and sentences in the work order information. For example, "user processing" corresponds to feature_120. If it is "processing required", the extraction method is: "yes" → feature_120_1 → the corresponding dimension in feature_120_1 is 1; if it is "processing not required", the extraction method is "no" → feature_120_0 → the corresponding dimension in feature_120_0 is 1. The AC automaton can match all strings at the same time.

[0065] Step 307 , fusing the first feature and the second feature to obtain a fused feature. In some embodiments, this step may be performed by the feature fusion module 240 in the system 200 .

[0066] In some embodiments, the method of fusing the first feature and the second feature to obtain a fused feature includes: directly splicing or processing the first feature and the second feature by a set algorithm to generate a combined feature. In some embodiments, the method of generating a combined feature includes but is not limited to: feature combination, feature dimensionality reduction, and feature crossover. Feature combination refers to obtaining a new feature by some linear superposition or nonlinear superposition of features. Common feature combination methods may include the Cartesian product method. For example, A={0,1}, B={0,1}, then A×B={(0,0),(0,1),(1,0),(1,1)}. The feature combination method may also adopt the decision tree + LR method. In some embodiments, the gradient boosting tree + logistic regression (GBDT+LR) method may be adopted. Gradient boosting tree + logistic regression (GBDT+LR) is a method of automatic feature extraction. GBDT is a gradient boosted decision tree. First, a decision tree is constructed. First, a decision tree is constructed on the residual of the existing model and the actual sample output. It is continuously iterated. Each iteration will produce a classification feature with a large gain. Therefore, the number of leaf nodes in the decision tree constructed by GBDT is the size of the feature space obtained, and the feature is used as the input of the LR model. The decision tree is represented in the form of a node tree. Each node makes a binary decision based on the characteristics of the data, and each leaf node of the tree contains a prediction result.

[0067] The method of feature combination can also be implemented based on the principle of principal component analysis (PCA), singular value decomposition (SVD) feature dimension reduction, and feature crossover based on factorization machine (FM). PCA refers to the synthesis of multiple indicators into a few independent comprehensive indicators (i.e., principal components), each of which can reflect most of the information of the original variable, and the information contained is not repeated. The basic principle of principal component analysis (PCA) is to project the sample data in a matrix into a new space. For a matrix, diagonalizing it, that is, the process of generating characteristic roots and eigenvectors, is also the process of projecting it on a standard orthogonal basis. The characteristic direction reflects the distribution state of information, and the eigenvalue defines the length of the direction. For example, the larger the eigenvalue, the longer the length of the characteristic direction, and the more information the original data in this direction has. Singular value decomposition (SVD) is an important matrix decomposition method. SVD can be applied to numerical data types, and a smaller data set can be used to represent the original data set. Factorization machine (FM) can solve the problem of feature combination when data is sparse.

[0068] Step 309 : Based on the fusion feature, obtain the event attribute of the target event. In some embodiments, this step can be performed by the determination module 250 in the system 200 .

[0069] In some embodiments, obtaining the classification result of the event attribute based on the fusion feature includes: using a classification model to process the fusion feature to obtain the classification result of the event attribute. In some embodiments, the classification result of the event attribute includes at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, and whether the event is repeatable. In some embodiments, the event attribute classification may not be limited to the above four types, and may also include whether the event is urgent, whether the event can be warned, etc. In some embodiments, the output result of the classification result may be a binary classification result, that is, only two results of safe or unsafe are output, or a multi-classification result, that is, the probability of the output event being a safe event. For example, the binary classification result is specifically represented as a safe output of 1 and an unsafe output of 0. For another example, the multi-classification result is specifically represented as a probability output of 0.8 for a safe event, a probability of 0.1 for an unsafe event, and a probability of 0.1 for a harassment event. The classification model includes at least one of the following deep learning models: XG Boost, GBDT, Adaboost, and random forest.

[0070] Adaboost is a boosting tree. Different weak classifiers are trained for the same training set, and a strong classifier is constructed through Adaboost. Boosting tree refers to the promotion of weak learning algorithms to strong learning algorithms. Classifier refers to the function or model that can map the data records in the database to one of the given categories, so that it can be applied to data prediction. Classifier is a general term for the method of classifying samples in data mining, which at least includes decision trees, logistic regression, naive Bayes, neural networks and other algorithms. There are two kinds of weights in the Adaboost algorithm, one is the weight of the data, and the other is the weight of the weak classifier. Among them, the weight of the data is mainly used for the weak classifier to find the decision point with the smallest classification error. After finding it, the weight of the weak classifier is calculated using this minimum classification error. The larger the classifier weight, the greater the weight of the weak classifier in the final decision.

[0071] Random forests (RF) refers to a classifier that uses multiple trees to train and predict samples. The category of its output is determined by the mode of the category output by individual trees.

[0072] In some embodiments, the method for training the classification model includes: obtaining a sample set, the sample set including multiple work order information, using the fusion features corresponding to the work order information as input, and using the event attributes corresponding to the work order information given by the customer service as an identifier to train the classification model.

[0073] During the Adaboost model training process, weak classifiers are trained first, and the trained weak classifiers will participate in the next iteration. In the Nth iteration, there are a total of N weak classifiers, of which N-1 weak classifiers are trained previously, and their various parameters will no longer change. This time, the Nth classifier is trained. The relationship between the weak classifiers is that the Nth weak classifier is more likely to classify the data that the first N-1 weak classifiers did not classify correctly, and the final classification output depends on the comprehensive effect of these N classifiers. Adaboost generally uses a single-layer decision tree as its weak classifier. It should be noted that even if the sample set has multi-dimensional features, a single-layer decision tree can only choose one dimension to make a decision. In the neural network, the classified data is converted into a vector form and input into the neural network for classification.

[0074] It should be noted that the above description of the relevant process is only for example and illustration, and does not limit the scope of application of the present application. For those skilled in the art, various modifications and changes can be made to the process under the guidance of the present application. However, these modifications and changes are still within the scope of the present application. For example, decision tree + logistic regression can be used in addition to LBDT + LR (such as RF + LR, XG Boost + LR, etc.).

[0075] In other embodiments of the present application, an event attribute determination device is provided, comprising at least one processor and at least one memory; the at least one memory is used to store computer instructions; the at least one processor is used to execute at least part of the computer instructions to implement the operations described above.

[0076] In some further embodiments of the present application, a computer-readable storage medium for event attribute determination is provided, wherein the storage medium stores computer instructions, and when the computer instructions are executed by a processor, the operations described above are implemented.

[0077] It should be noted that the above description is only for the convenience of description and does not limit the present application to the scope of the embodiments. It is understandable that, after understanding the principle of the present application, those skilled in the art can make various modifications and changes in form and details to the implementation of the above process without departing from this principle. However, these changes and modifications do not depart from the scope of the present application.

[0078] The beneficial effects that may be brought about by the embodiments of the present application include but are not limited to: (1) the work order information contains both narrative information and structured information. To a certain extent, the narrative information can also reflect the customer service's processing results of the event. The event attributes are obtained by combining the processing results of the two types of information, thereby improving the accuracy of determining the event attributes; (2) the narrative information and structured information are processed separately and input into the model, without the need to perform relevant conversion on the two types of information, thereby improving processing efficiency.

[0079] It should be noted that different embodiments may produce different beneficial effects. In different embodiments, the beneficial effects that may be produced may be any one or a combination of the above, or any other possible beneficial effects.

[0080] The above content describes the present application and / or some other examples. According to the above content, the present application can also make different variations. The subject matter disclosed in the present application can be realized in different forms and examples, and the present application can be applied to a large number of applications. All applications, modifications and changes required for protection in the following claims belong to the scope of the present application.

[0081] At the same time, the present application uses specific words to describe the embodiments of the present application. For example, "one embodiment", "an embodiment", and / or "some embodiments" refer to a certain feature, structure or characteristic related to at least one embodiment of the present application. Therefore, it should be emphasized and noted that "one embodiment", "an embodiment", "an alternative embodiment", "another embodiment" or "another embodiment" mentioned twice or more in different positions in this specification do not necessarily refer to the same embodiment. In addition, some features, structures or characteristics in one or more embodiments of the present application can be appropriately combined.

[0082] Those skilled in the art will appreciate that the contents disclosed in this application may be subject to many variations and improvements. For example, the various system components described above are all implemented by hardware devices, but may also be implemented only by software solutions. For example: installing the system on an existing server. In addition, the provision of location information disclosed herein may be implemented by a firmware, a combination of firmware / software, a combination of firmware / hardware, or a combination of hardware / firmware / software.

[0083] All or part of the software may sometimes be communicated over a network, such as the Internet or other communications network. Such communications enable software to be loaded from one computer device or processor to another. For example, from a management server or host computer of a radiation therapy system to a hardware platform of a computer environment, or other computer environment implementing the system, or a system of similar functionality associated with providing information required to determine the parameters of the wheelchair target structure. Therefore, another medium capable of transmitting software elements may also be used as a physical connection between local devices, such as light waves, radio waves, electromagnetic waves, etc., transmitted through cables, optical cables or air. Physical media such as cables, wireless connections or optical cables and the like used to carry carriers may also be considered as media that carry software. As used herein, unless limited to tangible "storage" media, other terms referring to computer or machine "readable media" refer to media that participate in the process of executing any instructions by the processor.

[0084] The computer program coding required for the operation of each part of the present application can be written in any one or more programming languages, including object-oriented programming languages ​​such as Java, Scala, Smalltalk, Eiffel, JADE, Emerald, C++, C#, VB.NET, Python, etc., conventional procedural programming languages ​​such as C language, Visual Basic, Fortran 2003, Perl, COBOL 2002, PHP, ABAP, dynamic programming languages ​​such as Python, Ruby and Groovy, or other programming languages, etc. The program coding can be run completely on the user's computer, or run on the user's computer as an independent software package, or run partly on the user's computer and partly on the remote computer, or run completely on the remote computer or server. In the latter case, the remote computer can be connected to the user's computer through any network form, for example, a local area network (LAN) or a wide area network (WAN), or connected to an external computer (for example, via the Internet), or in a cloud computing environment, or used as a service such as software as a service (SaaS).

[0085] In addition, unless explicitly stated in the claims, the order of the processing elements and sequences described in this application, the use of alphanumeric characters, or the use of other names are not intended to limit the order of the processes and methods of this application. Although the above disclosure discusses some invention embodiments that are currently considered useful through various examples, it should be understood that such details are only for illustrative purposes, and the attached claims are not limited to the disclosed embodiments. On the contrary, the claims are intended to cover all modifications and equivalent combinations that are consistent with the essence and scope of the embodiments of this application. For example, although the system components described above can be implemented by hardware devices, they can also be implemented only by software solutions, such as installing the described system on an existing server or mobile device.

[0086] Similarly, it should be noted that in order to simplify the description of the disclosure of this application and thus help understand one or more embodiments of the invention, in the above description of the embodiments of this application, multiple features are sometimes combined into one embodiment, figure or description thereof. However, this disclosure method does not mean that the features required by the object of this application are more than the features mentioned in the claims. In fact, the features of the embodiments are less than all the features of the single embodiment disclosed above.

[0087] In some embodiments, numbers describing attributes and quantities are used. It should be understood that such numbers used to describe the embodiments are modified by the modifiers "about", "approximately" or "substantially" in some examples. Unless otherwise specified, "about", "approximately" or "substantially" indicate that the numbers are allowed to vary by ±20%. Accordingly, in some embodiments, the numerical parameters used in the specification and claims are approximate values, which may change according to the required features of individual embodiments. In some embodiments, the numerical parameters should take into account the specified significant digits and adopt the general method of retaining digits. Although the numerical domains and parameters used to confirm the breadth of their range in some embodiments of the present application are approximate values, in specific embodiments, the setting of such numerical values ​​is as accurate as possible within the feasible range.

[0088] Each patent, patent application, patent application disclosure, and other materials, such as articles, books, instructions, publications, documents, objects, etc., cited in this application are hereby incorporated by reference in their entirety. Except for application history documents that are inconsistent with or conflicting with the content of this application, documents that limit the broadest scope of the claims of this application (currently or later attached to this application) are also excluded. It should be noted that if the descriptions, definitions, and / or use of terms in the attached materials of this application are inconsistent or conflicting with the content described in this application, the descriptions, definitions, and / or use of terms in this application shall prevail.

[0089] Finally, it should be understood that the embodiments described in this application are only used to illustrate the principles of the embodiments of the present application. Other variations may also fall within the scope of the present application. Therefore, as an example and not a limitation, the alternative configurations of the embodiments of the present application may be considered to be consistent with the teachings of the present application. Accordingly, the embodiments of the present application are not limited to the embodiments explicitly introduced and described in the present application.

Claims

1. A method for determining event attributes, the method being implemented by at least one processor, the method comprising: Acquire work order information generated according to the event description of the target event, the work order information at least including structured information and narrative information, the work order information is generated by the customer service based on the event description and / or the result of the event processing, and the structured information in the work order information includes at least one of the following information: an order number, a license plate number, a telephone number, whether to call the police, whether the police have filed a case, whether the user requests processing, and the urgency of the processing required; The narrative information in the work order information includes at least one of the following information: a description of the event by the user, a description of the police handling result, and a description of the customer service handling result; A first feature is obtained according to the narrative information, wherein the narrative information refers to unstructured data with irregular or incomplete data structure, no predefined data model, and not convenient to be represented by a two-dimensional logical table of a database, including: using a text conversion model to process the narrative information to obtain the first feature; the first feature reflects a first judgment result of an event attribute of the target event, wherein the method of training the text conversion model includes: Acquire a sample set, wherein the sample set includes multiple work order information; Taking the narrative information in the work order information as input and the event attribute corresponding to the work order information given by the customer service as an identifier, training the text conversion model; Obtaining a second feature according to the structured information, including: generating the second feature by processing the structured information through an extraction model; fusing the first feature and the second feature to obtain a fused feature, including: directly concatenating or processing the first feature and the second feature by a set algorithm to generate a combined feature, wherein the processing includes feature combination, feature dimensionality reduction, and feature crossover; Based on the fusion feature, the event attributes of the target event are obtained, including: using a classification model to process the fusion feature to obtain a classification result of the event attribute, the classification result of the event attribute includes at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, whether the event is repeatable, whether the event is urgent, and whether the event can be warned, and the output of the classification result is a binary classification result or a multi-classification result.

2. The method according to claim 1, characterized in that The text conversion model includes at least one of the following deep learning models: Fast text, HAN, Text CNN, Transformer, LR and XG Boost.

3. The method according to claim 1, characterized in that The extraction model uses a rule-based and / or AC automaton method to perform feature extraction.

4. The method according to claim 1, characterized in that: The classification model includes at least one of the following deep learning models: XG Boost, GBDT, Adaboost, and random forest.

5. The method according to claim 1, characterized in that The method of training the classification model includes: Acquire a sample set, wherein the sample set includes multiple work order information; The classification model is trained by taking the fusion features corresponding to the work order information as input and the event attributes corresponding to the work order information provided by the customer service as identifiers.

6. An event attribute determination system, characterized in that: include: A work order acquisition module is used to acquire work order information generated according to the event description of the target event, wherein the work order information includes at least structured information and narrative information, and the work order information is generated by the customer service based on the event description and / or the result of the event processing. The structured information in the work order information includes at least one of the following information: an order number, a license plate number, a telephone number, whether a police report is made, whether a case is filed by the police, whether a user requests processing, and the urgency of the processing required; The narrative information in the work order information includes at least one of the following information: a description of the event by the user, a description of the police handling result, and a description of the customer service handling result; A first feature extraction module is used to obtain a first feature according to the narrative information, wherein the narrative information refers to unstructured data with irregular or incomplete data structure, no predefined data model, and inconvenient to be represented by a two-dimensional logical table of a database, including: using a text conversion model to process the narrative information to obtain the first feature; the first feature reflects a first judgment result of an event attribute of the target event, wherein a method for training the text conversion model includes: Acquire a sample set, wherein the sample set includes multiple work order information; Taking the narrative information in the work order information as input and the event attribute corresponding to the work order information given by the customer service as an identifier, training the text conversion model; A second feature extraction module is used to obtain a second feature according to the structured information, which includes: generating the second feature by processing the structured information through an extraction model; a feature fusion module, used for fusing the first feature and the second feature to obtain a fused feature, including: directly concatenating or processing the first feature and the second feature by a set algorithm to generate a combined feature, wherein the processing includes feature combination, feature dimensionality reduction and feature crossover; and A determination module is used to obtain the event attributes of the target event based on the fusion features, including: using a classification model to process the fusion features to obtain a classification result of the event attributes, the classification result of the event attributes including at least one of the following: whether the event is safe, whether the event is accurate, whether the event is traceable, whether the event is repeatable, whether the event is urgent, and whether the event can be warned, and the output of the classification result is a binary classification result or a multi-classification result.

7. An event attribute determination device, characterized in that: The apparatus comprises at least one processor and at least one memory; The at least one memory is used to store computer instructions; The at least one processor is configured to execute at least part of the computer instructions to implement the method according to any one of claims 1 to 5.

8. A computer-readable storage medium for determining event attributes, characterized in that: The storage medium stores computer instructions, and when the computer instructions are executed by a processor, the method according to any one of claims 1 to 5 is implemented.