Map information task processing method and device, and electronic equipment

Abstract

The application relates to the computer technical field, in particular to a map information task processing method and device and electronic equipment, the method comprises the following steps: obtaining target content; a target road state classification model is used for classifying the road state type of the target content, and a classification result is obtained; if the classification result indicates that the target content belongs to a target road state type, a road entity of the target content is extracted, and a first road entity in the target content is obtained; the first road entity and a second road entity related to a target map information task are matched, and an entity matching result is obtained; if the entity matching result indicates that the first road entity and the second road entity are the same entity, the processing priority of the target map information task is determined according to the target content; the method provided by the application can automatically determine the processing level of the target map information task through the target content in the Internet, and the processing efficiency of the target map information task is improved.

Description

Technical Field

[0001] This application relates to the field of computer technology, specifically to a method, apparatus, and electronic device for processing map information tasks. Background Technology

[0002] As an important source of intelligence for internet maps, the timeliness of processing geographic information and internet intelligence is a key performance indicator. In related technologies, the processing priority of intelligence tasks is usually determined manually by the personnel involved, which is not only inefficient and costly, but also subject to subjective judgment errors, resulting in low accuracy. Summary of the Invention

[0003] In view of this, embodiments of this application propose a method, apparatus, and electronic device for processing map intelligence tasks, in order to solve the problem of low efficiency and accuracy of manually determining the processing priority of intelligence tasks.

[0004] The embodiments of this application are implemented using the following technical solutions:

[0005] In a first aspect, embodiments of this application provide a method for processing map intelligence tasks, comprising: acquiring target content; classifying the target content into road state types using a target road state classification model to obtain a classification result; wherein the target road state classification model refers to a road state classification model used to identify target road state types, and the target road state type is a road state type involved in the target map intelligence task; if the classification result indicates that the target content belongs to the target road state type, extracting road entities from the target content to obtain a first road entity in the target content; performing entity matching between the first road entity and a second road entity involved in the target map intelligence task to obtain an entity matching result; if the entity matching result indicates that the first road entity and the second road entity are the same entity, determining the processing priority of the target map intelligence task based on the target content.

[0006] Secondly, embodiments of this application provide a processing apparatus for a map intelligence task, comprising: an acquisition module for acquiring target content; a classification module for classifying the target content into road state types using a target road state classification model to obtain a classification result; wherein the target road state classification model is a road state classification model used to identify target road state types, and the target road state type is a road state type involved in the target map intelligence task; an entity extraction module for extracting road entities from the target content if the classification result indicates that the target content belongs to the target road state type, to obtain a first road entity in the target content; a matching module for performing entity matching between the first road entity and a second road entity involved in the target map intelligence task to obtain an entity matching result; and an output module for determining the processing priority of the target map intelligence task based on the target content if the entity matching result indicates that the first road entity and the second road entity are the same entity.

[0007] In some embodiments, the processing apparatus for map intelligence tasks further includes an element extraction module, used to extract key event element information from the target content if the entity matching result indicates that the first road entity and the second road entity are the same entity; and an update module, used to update the event state change timeline information corresponding to the target map intelligence task based on the key event element information.

[0008] In some implementations, the update module is further configured to delete the target map intelligence task from the map intelligence task set if, based on the event state change timeline information corresponding to the target map intelligence task, the road state of the second road entity changes to the target road state.

[0009] In some implementations, the target map intelligence task is any one of a set of map intelligence tasks; the map intelligence task processing device further includes a synchronization module for synchronizing the target content and the updated processing priority of the target map intelligence task to the map task processing platform; a receiving module for receiving the authenticity verification result returned by the map task processing platform; and an information sending module for sending a status update prompt message to the map application based on the road status of the first road entity in the target content if the authenticity verification result indicates that the target content is authentic, so that the map application updates the road status of the first road entity in the electronic map.

[0010] In some implementations, the target content is published content whose popularity value exceeds a threshold; the output module is specifically used to, if the entity matching result indicates that the first road entity and the second road entity are the same entity, use the processing priority corresponding to the popularity level of the target content as the processing priority of the target map intelligence task according to the correspondence between popularity level and processing priority.

[0011] In some embodiments, the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing device further includes an intelligence acquisition module for acquiring map intelligence content; an intelligence classification module for classifying the map intelligence content into road state types using multiple road state classification models applicable to different road state types, obtaining multiple road state type classification results; a state confirmation module for determining the road state type to which the map intelligence content belongs based on the multiple road state type classification results; an intelligence entity extraction module for extracting road entities from the map intelligence content, obtaining the road entities in the map intelligence content; and a task creation module for creating a map intelligence task for the map intelligence content based on the road state type to which the map intelligence content belongs and the road entities in the map intelligence content, and adding the created map intelligence task to the map intelligence task set.

[0012] In some embodiments, the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing device further includes a list acquisition module for acquiring road public opinion list information; the road public opinion list information includes the road name of at least one road and the road status indication information of at least one road; a type confirmation module for determining the road status type of at least one road based on the road status indication information of at least one road in the road public opinion list information; a task creation module for creating a corresponding map intelligence task for at least one road based on the road status type and the road name of at least one road; wherein, the road name of one road is the road entity involved in the corresponding map intelligence task, and the road status type of one road is the road status type involved in the corresponding map intelligence task; and an adding module for adding the created map intelligence task to the map intelligence task set.

[0013] In some embodiments, the processing device for map intelligence tasks further includes a first training data acquisition module, used to acquire training data corresponding to the road state classification model; the training data includes multiple road event descriptions and labeled road state types corresponding to each road event description; a first prediction module, used to classify the road event descriptions by the road state classification model to obtain predicted road state types for each road event description; a first loss calculation module, used to calculate a first loss based on the predicted road state types and the labeled road state types corresponding to each road event description; and a first adjustment module, used to adjust the parameters of the road state classification model based on the first loss until a first training termination condition is met.

[0014] In some implementations, the matching module includes a first semantic extraction unit, used to extract semantic features from the first road entity using a first semantic feature extraction network in the dual-tower network to obtain a first semantic feature; a second semantic extraction unit, used to extract semantic features from the second road entity using a second semantic feature extraction network in the dual-tower network to obtain a second semantic feature; a similarity calculation unit, used to calculate the semantic similarity between the first semantic feature and the second semantic feature; and a matching output unit, used to determine the entity matching result based on the semantic similarity.

[0015] In some embodiments, the processing device for map intelligence tasks further includes a second training data acquisition module for acquiring entity matching training data, wherein the entity matching training data includes multiple road entity pairs and matching labels corresponding to each road entity pair; the matching label corresponding to a road entity pair is used to indicate whether a first sample road entity in the road entity pair and a second sample road entity therein are the same entity; a first feature extraction module is used to extract semantic features from the first sample road entity by a first semantic feature extraction network in the dual-tower network to obtain first sample semantic features; a second feature extraction module is used to extract semantic features from the second sample road entity by a second semantic feature extraction network in the dual-tower network to obtain second sample semantic features; a similarity prediction module is used to calculate the semantic similarity corresponding to each road entity pair based on the first sample semantic features and the second sample semantic features; a second loss calculation module is used to calculate a second loss based on the semantic similarity corresponding to each road entity pair and the matching labels corresponding to each road entity pair; and a second adjustment module is used to adjust the parameters of the dual-tower network based on the second loss until a second training termination condition is reached.

[0016] Thirdly, embodiments of this application provide an electronic device, including: a processor; and a memory storing computer instructions, which, when executed by the processor, implement the above-described method.

[0017] Fourthly, embodiments of this application provide a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the above-described method.

[0018] Fifthly, embodiments of this application provide a computer program product, including computer instructions that, when executed by a processor, implement the above-described method.

[0019] This application provides a method, apparatus, and electronic device for processing map intelligence tasks. The method uses a target road state classification model to identify the target road state type involved in the target map intelligence task. It classifies the target content into road state types, and when the classification result indicates that the target content belongs to a target road state type, it performs entity matching between a first road entity in the target content and a second road entity involved in the target map intelligence task. If the first road entity and the second road entity represent the same entity, it indicates that the road event described in the target content (i.e., a road-related event) and the road event involved in the target map intelligence task are the same event. Therefore, the processing priority of the target map intelligence task can be determined through relevant information of the target content (such as popularity), facilitating prioritized processing of the target map intelligence task. This automatic determination of the processing priority based on the target content enables dynamic adjustment of the target map intelligence task's processing priority, ensuring timely processing of the target map intelligence task subsequently.

[0020] These or other aspects of this application will become more apparent in the following description of the embodiments. Attached Figure Description

[0021] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0022] Figure 1 A schematic diagram of an application scenario involving an embodiment of this application is shown.

[0023] Figure 2 This illustration shows a first flowchart of a map information task processing method provided in an embodiment of this application.

[0024] Figure 3 A schematic diagram of the structure of an entity extraction model according to an embodiment of this application is shown.

[0025] Figure 4 This illustration shows a second flowchart of a map information task processing method provided in an embodiment of this application.

[0026] Figure 5 This illustration shows a third flowchart of a map information task processing method provided in an embodiment of this application.

[0027] Figure 6 The diagram illustrates the fourth flow of a map information task processing method provided in an embodiment of this application.

[0028] Figure 7 The diagram illustrates the fifth step of a map information task processing method provided in an embodiment of this application.

[0029] Figure 8 The training process of a road state classification model provided in one embodiment of this application is illustrated.

[0030] Figure 9 A schematic diagram of the structure of a road condition classification model according to an embodiment of this application is shown.

[0031] Figure 10 An embodiment of this application is shown. Figure 2 A flowchart of step S140.

[0032] Figure 11 A schematic diagram of the training process of a dual-tower model according to an embodiment of this application is shown.

[0033] Figure 12 A schematic diagram of the structure of a twin-tower model according to an embodiment of this application is shown.

[0034] Figure 13 This paper illustrates an application flowchart of a map information task processing method provided in an embodiment of this application.

[0035] Figure 14 An illustration of the effect involved in one embodiment of this application is provided.

[0036] Figure 15 A schematic diagram of a map information task processing apparatus according to an embodiment of this application is shown.

[0037] Figure 16 A schematic diagram of an electronic device according to an embodiment of this application is shown. Detailed Implementation

[0038] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain this application, and should not be construed as limiting this application.

[0039] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are merely some embodiments of the present application, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present application without creative effort are within the scope of protection of the present application.

[0040] In the following description, the terms "first" and "second" are used merely to distinguish similar objects and do not represent a specific ordering of objects. It is understood that "first" and "second" may be interchanged in a specific order or sequence where permitted, so that the embodiments of this application described herein can be implemented in an order other than that illustrated or described herein.

[0041] In this document, "multiple" refers to two or more. "And / or" describes the relationship between associated objects, indicating that three relationships can exist. For example, A and / or B can represent: A alone, A and B simultaneously, or B alone. The character " / " generally indicates that the preceding and following associated objects are in an "or" relationship. In the following description, references to "some embodiments or some embodiment methods" describe a subset of all possible embodiments. However, it is understood that "some embodiments" can be the same subset or different subsets of all possible embodiments and can be combined with each other without conflict.

[0042] To facilitate understanding of this application, some terms will be explained below.

[0043] Map intelligence refers to intelligence related to maps, such as the geographical location, topography, transportation conditions, and resource distribution of a specific area or target.

[0044] Entity extraction, also known as Named Entity Recognition (NER), is a key technology in Natural Language Processing (NLP). Its main task is to automatically identify entities with specific meanings from unstructured text data and classify these entities into predefined categories. These entities are usually nouns with explicit referential meanings in the text, such as names of people, places, organizations, times, dates, numbers, etc.

[0045] As an important source of intelligence for internet maps, the timeliness of processing geographic information and internet intelligence is a key performance indicator. In related technologies, the processing priority of intelligence is usually determined manually by the personnel involved, which is not only inefficient and costly, but also subject to subjective judgment errors, resulting in low accuracy.

[0046] To address the aforementioned problems, this application provides a method, apparatus, and electronic device for processing map information tasks.

[0047] Please see Figure 1 , Figure 1 A schematic diagram of an application scenario according to an embodiment of this application is provided, including a terminal 10, an information platform 20 and a server 30, wherein the terminal 10 and the server 30 are connected via a wired or wireless network, and the server 30 and the information platform 20 are connected via a wired or wireless network.

[0048] Server 30 deploys road state classification models corresponding to various road state types. Server 30 can generate map intelligence tasks and manage multiple map intelligence tasks. Any map intelligence task can serve as the target map request task in this application. After obtaining the target content from information platform 20, server 30 can determine the processing priority of the target map intelligence task according to the method provided in this application. Subsequently, the processing priority of the target map intelligence task determined by server 30 is synchronized to terminal 10 so that users can process the map intelligence tasks according to their respective processing priorities.

[0049] When executing the method provided in this application, after obtaining the target content from the information platform 20, the server 30 classifies the target content into road state types using a target road state classification model to obtain classification results. The target road state classification model refers to a road state classification model used to identify the target road state type, and the target road state type is the road state type involved in the target map intelligence task. If the classification result indicates that the target content belongs to the target road state type, road entities are extracted from the target content to obtain the first road entity in the target content. The first road entity is matched with the second road entity involved in the target map intelligence task to obtain entity matching results. If the entity matching results indicate that the first road entity and the second road entity are the same entity, the processing priority of the target map intelligence task is determined.

[0050] Terminal 10 can be a smartphone, tablet, laptop, desktop computer, smart speaker, smart TV, wearable device, smartwatch, virtual reality device, vehicle terminal, smart TV, etc., but is not limited to these.

[0051] Server 30 can be a standalone physical server, a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN, and big data and artificial intelligence platforms.

[0052] In some embodiments, the method of this application can also be executed by terminal 10, which can deploy road state classification models corresponding to various road state types. Then, the processing priority of the target map intelligence task is determined by the method of this application.

[0053] The present application will now be described in detail with reference to the embodiments.

[0054] Please see Figure 2 , Figure 2 A first flowchart illustrating the map intelligence task processing method provided in this application embodiment is given. The map intelligence task processing method includes steps S110-S150:

[0055] S110, Obtain the target content.

[0056] The target content refers to the content published on the information platform; the information platform refers to the platform on which content can be published, such as social media platforms, news forums, information search platforms, etc.

[0057] In some implementations, considering the large amount of content published on the information platform, in order to obtain effective content from the massive amount of published content, the target content can be determined as any shared content in the information platform's popular content list. The target content can be news, public account articles, blogs, notes, discussion topics, etc.; whereby, an information platform's popular content list displays the most popular content on the information platform.

[0058] In some implementations, the target content can also be published content whose popularity value exceeds a threshold. The popularity value of the published content can be calculated in different ways; for example, the popularity value of the published content can be determined based on one or more of the following: number of searches, number of views, number of comments, number of reposts, number of likes, and number of favorites within a specified time period.

[0059] In some embodiments, both the popularity and timeliness of the published content can be considered, and published content with timeliness exceeding a timeliness threshold and popularity exceeding a popularity threshold can be selected as target content. The timeliness of published content can be determined based on its publication time; the later the publication time, the higher the timeliness, and the earlier the publication time, the lower the timeliness.

[0060] In some embodiments, target content can be obtained in real time from multiple information platforms. The obtained target content can be multiple items, and each item of target content can be processed according to the method of this application.

[0061] S120. The target content is classified into road condition types using the target road condition classification model to obtain the classification results. The target road condition classification model is a road condition classification model used to identify the target road condition type. The target road condition type is the road condition type involved in the target map intelligence task.

[0062] In this context, "target map intelligence task" refers to a map intelligence task that needs to be matched with target content. When there are multiple map intelligence tasks, any one of them can be used as the target map intelligence task in this application.

[0063] Map intelligence tasks refer to tasks created for map intelligence content that needs to be processed. These tasks can include verifying the authenticity of map intelligence content, or classifying the public opinion level of the map intelligence content. A higher public opinion level indicates a higher priority for processing. Map intelligence tasks can also include detecting the state evolution of traffic and road elements involved in the map intelligence content.

[0064] Map information content refers to information related to traffic and road elements on a map. Traffic and road elements include urban roads, bridges, highways, and other roads that provide passage. In addition to information about traffic and road elements, map information content also indicates the road conditions of those elements. Therefore, map information content can also be considered as information indicating the road conditions of traffic and road elements.

[0065] The road condition of traffic road elements refers to the state of traffic road elements related to traffic flow. Road conditions include road opening, road closure, road traffic restrictions (which can be further subdivided into odd-even license plate restrictions, tidal flow lane restrictions, sports event restrictions, construction restrictions, rain and snow weather restrictions, traffic accident restrictions, road maintenance restrictions, etc.), changes in road traffic direction, changes in road traffic lights, and whether parking is allowed or prohibited on the road.

[0066] In this application, a road entity involved in a map intelligence task refers to the name of the traffic road element involved in the map intelligence content corresponding to the map intelligence task. A road state type involved in a map intelligence task refers to the road state type to which the road state of the traffic road element in the map intelligence content corresponding to the map intelligence task belongs. Road state types can be set according to actual needs; for example, road state types may include road closure, road restriction, traffic light change, traffic direction change, and whether parking is allowed on the road, etc. Among these, the road closure category involves road opening and road closure; further, the road closure category can be subdivided into the opening and closing of new roads, and the repeated opening and closing of old roads; the road restriction category refers to restrictions on road traffic, such as odd-even license plate restrictions, tidal flow lane restrictions, sports event restrictions, construction restrictions, etc.

[0067] In this application, a road state classification model is pre-deployed to identify each road state type, with one road state classification model used to identify one road state type. The target road state types involved in the target map intelligence task can be manually labeled based on the map intelligence content involved in the target map intelligence task, or they can be pre-determined by classifying the road state types of the map intelligence content involved in the target map intelligence task using road state classification models corresponding to multiple road state types.

[0068] A road condition classification model corresponding to a road condition type is a classification model constructed using one or more neural networks, such as recurrent neural networks or fully connected neural networks. In some embodiments, considering that the target content is primarily text-based, the road condition classification model can be a classification model for processing text. In some embodiments, the road condition classification model can also be a classification model for processing multimodal content (e.g., text + image).

[0069] One road state classification model for a given road state type can perform either binary classification or multi-class classification of the input content. Taking road state classification model M for road state type A as an example, if binary classification is used, model M identifies whether the input content belongs to road state type A or not. If multi-class classification (involving at least three categories), model M identifies which of the at least three categories involved in road state type A the input content belongs to. The at least three categories involved in road state type A include multiple categories belonging to road state type A and one category representing what does not belong to road state type A.

[0070] Therefore, the classification result output by a road condition classification model for target content can indicate whether the target content belongs to the road condition type corresponding to the road condition classification model.

[0071] In some embodiments, considering the overall length of the target content (e.g., news, blog posts, etc.), the target road condition classification model would have a long processing time for classifying the road condition type, leading to excessive processing pressure. Therefore, key content can be extracted from the target content. Key content can be key text and key images. Key text content includes at least one of the target content's title or summary, while key images include video covers. This key content is then input into the target road condition classification model for road condition type classification, yielding the classification result. Of course, if the target content is short, such as a trending topic, the entire target content can be input into the target road condition classification model for road condition type classification.

[0072] In some embodiments, if the classification result output by the target road state classification model for target content indicates that the target content does not belong to the road state type corresponding to the target road state classification model, it means that there is no connection between the target content and the target map intelligence task, and it cannot be used to determine the processing priority of the target map intelligence task. If there are other map intelligence tasks that have not been matched with the target content, they can continue to be matched with the target content in a similar manner according to the next map intelligence task; if the target content has no connection with all map intelligence tasks, the target content can be deleted, or the target content can be labeled as invalid.

[0073] S130. If the classification result indicates that the target content belongs to the target road state type, extract the road entity from the target content to obtain the first road entity in the target content.

[0074] Road entity extraction refers to the automatic identification of road entities from unstructured text data; where road entities refer to road names.

[0075] In this application, for ease of distinction, the road entities extracted from the target content are referred to as the first road entities. The road entities involved in the target map intelligence task are referred to as the second road entities. It is understood that if the classification result output by the target road state classification model for the target content indicates that the target content belongs to the target road state type, then the probability of correlation between the target content and the map intelligence content of the target map intelligence task is relatively high.

[0076] Entity extraction models (such as recurrent neural networks and hidden Markov models) can be used to extract road entities from target content. These models can extract road entities from the text content of the target content, obtaining the first road entity. For example, the key text content of the target content (at least one of the title and abstract), or the entire text content of the target content, can be input into the entity extraction model for road entity extraction.

[0077] In some embodiments, the entity extraction model can be a BERT model, a Transformer model, or a neural network model with other structures. See also Figure 3 , Figure 3 An exemplary structural diagram of an entity extraction model is provided.

[0078] Figure 3 The entity extraction model shown is a BERT+Bi-LSTM (Bidirectional Long Short-Term Memory)+CRF (Conditional Random Field) model, which includes an input layer, an embedding layer, a BERT layer, a Bi-LSTM layer, a CRF layer, and an output layer. The function of each layer is as follows:

[0079] Input layer: Text content used to input target content; a start identifier, namely the [CLS] identifier (classification identifier), can be added before the start position of the text content.

[0080] In some implementations, the input text length requirement of the entity extraction model can also be considered. For example, if the entity extraction model requires the input text length to be K, i.e., max_len = K, when the text length (number of words) of the text content exceeds K, the first K words of the text content are extracted as the input content of the input layer. If the text length of the text content is less than K, a whitespace marker is added to the end of the text content so that the number of words of the text content after adding the whitespace marker is K. The whitespace marker is, for example, [PAD].

[0081] The embedding layer is used to perform line embedding processing on the fixed-length text content input from the input layer, obtaining a vector representation of each word in the text content, such as... Figure 3 In the case of {x0, x1, ..., xN}.

[0082] The BERT layer is used to extract features from the vector representation of each word in the text content, thus obtaining the features of each word in the text content.

[0083] The Bi-LSTM layer is used to capture bidirectional dependencies in the text content by combining the outputs of the forward and backward LSTM networks. This allows for further feature extraction of each word predicted by the BERT layer, resulting in the hidden state representation of each word in the text content.

[0084] The CRF layer is used to establish a road entity label transition matrix based on the hidden state representation of each word in the text content output by the Bi-LSTM layer. By learning the transition probabilities between road entity labels, the road entity label of each word in the text content is determined. The road entity label includes labels that indicate that it is a road entity and labels that indicate that it is not a road entity.

[0085] The output layer is used to output the text that identifies the corresponding road recognition tag as a road entity.

[0086] S140. Perform entity matching between the first road entity and the second road entity involved in the target map intelligence mission to obtain the entity matching result.

[0087] Understandably, if the classification result indicates that the target content belongs to the target road state type, it means that the target content and the road state type involved in the target map intelligence task are the same road state type (e.g., road openness and closure). However, the target content and the target map intelligence task may involve different roads. For example, the target content involves the openness and closure of road A, while the target map intelligence task involves the openness and closure of road B. Therefore, if the classification result indicates that the target content belongs to the target road state type, it is still necessary to perform entity matching between the first road entity of the target content and the second road entity involved in the target map intelligence task to determine whether the target content and the target map intelligence task involve the same road and the same road state type.

[0088] The entity matching result is used to indicate whether the first road entity and the second road entity represent the same entity.

[0089] In some implementations, the first road entity and the second road entity can be considered to represent the same entity if the edit distance between them is less than a distance threshold. Alternatively, the semantic feature similarity between them can be calculated if it is greater than a similarity threshold.

[0090] In other embodiments, a road entity set can be pre-constructed, comprising multiple road entity subsets. Different road entity subsets represent different road entities, and multiple road entities within a single road entity subset represent the same road entity. That is, multiple road entities within a single road entity subset can be multiple names for the same road. Therefore, within the road entity set, it is determined whether a first road entity and a second road entity reside in the same road entity subset. If they do, it indicates that the first road entity and the second road entity represent the same entity; otherwise, it indicates that they are not the same entity.

[0091] S150. If the entity matching result indicates that the first road entity and the second road entity are the same entity, determine the processing priority of the target map intelligence task based on the target content.

[0092] The road state type of the target content is the same as that of the target road state type involved in the target map intelligence task, and the first road entity in the target content is also the same as the second road entity involved in the target map intelligence task. This indicates that the road event described in the target content (i.e., the road-related event) is the same event as the road event described in the map intelligence content of the target map intelligence task. Therefore, relevant information of the target content (such as popularity, discussion level, real-time status, etc.) can be used to determine the processing priority of the target map intelligence task.

[0093] In some implementations, the target content is published content whose popularity value exceeds a threshold; step S150 may specifically include: if the entity matching result indicates that the first road entity and the second road entity are the same entity, according to the correspondence between popularity level and processing priority, the processing priority corresponding to the popularity level of the target content is used as the processing priority of the target map intelligence task.

[0094] The popularity level can be determined based on the popularity value of the target content; specifically, the popularity value range to which the popularity value of the target content belongs is determined, and the popularity level corresponding to the popularity value range is determined as the popularity level of the target content.

[0095] In other implementations, if the target content is any published content in the popular content list of the information platform, the popularity level of the target content can also be determined according to the ranking position of the target content in the popular content list; for example, the ranking position of the target content in the popular content list can be determined as the popularity level of the target content.

[0096] In some implementations, determining the processing priority of target map intelligence tasks can be achieved by adding priority tags to the target map intelligence tasks; in subsequent processing, target map intelligence tasks with priority tags are processed first.

[0097] In other implementations, determining the processing priority of a target map intelligence task can also involve determining the sorting position of the target map intelligence task in the queue of map intelligence tasks to be processed. For example, if the entity matching result indicates that the first road entity and the second road entity are the same entity, the current position of the target map intelligence task in the task sequence is moved forward by K positions, where K is a positive integer greater than zero, representing an increase in its processing priority. The value of K can be determined based on the popularity level of the target content.

[0098] If the target map intelligence task is to verify the authenticity of map intelligence content, after increasing the task processing priority of the target map intelligence task, subsequent verification of the target map intelligence task can be prioritized. If the target map intelligence task is to classify the public opinion level of map intelligence content, after increasing the task processing priority of the target map intelligence task, subsequent public opinion level classification of the target map intelligence task can be prioritized. If the target map intelligence task is also a task to detect the state evolution of traffic and road elements involved in the map intelligence content, subsequent updates to the real-time road status of the second road entities involved in the target map intelligence task can be prioritized.

[0099] The method provided in this application classifies the target content into road state types using a target road state classification model corresponding to the target road state types involved in the target map intelligence task. When the classification result indicates that the target content belongs to the target road state type, entity matching is performed between the first road entity in the target content and the second road entity involved in the target map intelligence task. If the first road entity and the second road entity represent the same entity, it indicates that the road event described in the target content (i.e., the road-related event) is the same event involved in the target map intelligence task. Therefore, the processing priority of the target map intelligence task can be determined through relevant information of the target content (such as popularity), facilitating the priority processing of the target map intelligence task. This automatic determination of the processing priority of the map intelligence content of the target map intelligence task based on the target content enables dynamic adjustment of the processing priority of the target map intelligence task, allowing for timely processing of target map intelligence tasks with higher popularity.

[0100] In some implementations, please refer to Figure 4 , Figure 4 A second flowchart of the map information task processing method provided in this application embodiment is given. The map information task processing method further includes steps S210-S220:

[0101] S210. If the entity matching result indicates that the first road entity and the second road entity are the same entity, extract key event element information from the target content.

[0102] Among them, key event element information can be the key event elements of the road event described by the target content, such as the time, location, road condition, and participants of the event.

[0103] In some implementations, key event elements in the target content can be extracted by using pre-defined regular expression matching for each key event element, or by using a key event element extraction model (such as a large language model or other deep learning models) to extract key event element information from the target content.

[0104] S220. Update the timeline information of event status changes corresponding to the target map intelligence task based on key event element information.

[0105] In this embodiment, an event state change timeline is maintained for the second road entity involved in the target map intelligence task. The event state change timeline presents the relevant states of the second road entity at different times in chronological order.

[0106] For example, the target map intelligence task is "the opening of Bridge A", and the target content is "sections a and b of Bridge A were successfully joined on XX day, and it is expected that the entire line will be open to traffic this month". Through the target content, the key event element information that can be extracted includes the time information "XX day", the road name information "sections a and b of Bridge A", and the status change information "joined". Therefore, the event status change timeline information corresponding to the target map intelligence task can be updated. For example, the event status change timeline can be "Bridge A - Sections a and b joined on XX day". If other target content is matched later, such as "Bridge A was successfully opened to traffic on YY day", the event status change timeline can be updated to "Bridge A - Sections a and b joined on XX day - Open to traffic on YY day".

[0107] In the above implementation, the event state change timeline information corresponding to the target map intelligence task is updated by using key event element information in the target content, thereby realizing real-time updates of the road state of the second road entity involved in the target map intelligence task. This makes it easier for users to intuitively understand the state evolution process of the second road entity involved in the target map intelligence task.

[0108] In some implementations, please refer to [the relevant documentation]. Figure 4 The target map intelligence task is any one of the map intelligence task sets; after step S220, the processing method for the map intelligence task further includes step S230:

[0109] S230. If, based on the event state change timeline information corresponding to the target map intelligence task, it is determined that the road state of the second road entity has changed to the target road state, the target map intelligence task is deleted from the map intelligence task set.

[0110] The target road state refers to the road state defined by the task removal conditions set for the target map intelligence task. Specifically, if the road state of the second road entity involved in the target map intelligence task evolves to the target road state, the task removal conditions set for the target map intelligence task are met. In other words, if the road state of the second road entity involved in the target map intelligence task evolves to the target road state, it indicates that there is no need to continue monitoring the road state of the second road entity; therefore, the target map intelligence task is removed from the map intelligence task set.

[0111] In the above implementation, when it is determined that the road state of the second road entity has changed to the target road state, the target map intelligence task is deleted from the map intelligence task set, thereby realizing the dynamic updating of the map intelligence task set.

[0112] In some implementations, please refer to Figure 5 , Figure 5 A third flowchart of the map information task processing method provided in this application embodiment is given. After step S150, the map information task processing method further includes steps S310-S330:

[0113] S310. Synchronize the processing priority of the updated target content and target map intelligence tasks to the map task processing platform.

[0114] Among them, the map task processing platform refers to the platform on which map information tasks are processed manually.

[0115] Understandably, the map task processing platform will display map intelligence tasks with higher processing priority based on the processing priority of the updated target map intelligence tasks. In this way, staff can focus on the target map intelligence tasks with higher processing priority and thus process them first.

[0116] In the map task processing platform, staff can manually verify the authenticity of the target content and the map intelligence content of the target map intelligence task by combining the map intelligence content of the target map intelligence task with the target content.

[0117] Furthermore, in some embodiments, within the map task processing platform, staff can, after manually verifying the authenticity of the target content and the map intelligence content of the target map intelligence task, also determine whether to raise the public opinion level of the map intelligence content involved in the target map intelligence task based on the target content.

[0118] S320: Receive the authenticity verification result returned by the map task processing platform.

[0119] The authenticity verification result is used to indicate whether the target content is authentic. In some implementations, the authenticity of the target content can be determined manually.

[0120] In other implementations, a pre-defined information source whitelist can be used. The whitelist contains reliable information sources (such as official websites and authoritative platforms). The authenticity of the target content is then determined based on its information source. In this case, the information source information of the target content also needs to be synchronized to the map task processing platform. Specifically, if the information source of the target content is on the whitelist, the target content is considered authentic; if the information source is not on the whitelist, the authenticity of the target content is considered questionable, and further manual verification of the target content's authenticity can be conducted.

[0121] S330. If the authenticity verification result indicates that the target content is authentic, a status update prompt message is sent to the map application according to the road status of the first road entity in the target content, so that the map application updates the road status of the first road entity in the electronic map.

[0122] Because the target content is highly time-sensitive, if the authenticity verification result indicates that the target content is authentic, it means that the road status of the first road entity in the target content is the latest road status in the actual environment. Therefore, based on the road status of the first road entity in the target content, a status update notification is sent to the map application to update the road status of the first road entity in the electronic map. This causes the map application to update the road status of the first road entity in the electronic map to match the road status in the target content, ensuring that the updated road status of the first road entity in the electronic map is consistent with the actual road status. This avoids inconsistencies between the road status of a road entity in the electronic map and its actual road status, which could lead to unusable routes recommended by the electronic map or incorrect navigation provided by the electronic map.

[0123] In some implementations, please refer to Figure 6 , Figure 6 The fourth flowchart of the map intelligence task processing method provided in this application embodiment is shown, where the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing method further includes steps S410-S450:

[0124] S410, Obtain map information.

[0125] In some implementations, map information content can be obtained by searching multiple platforms using keyword retrieval. The keywords can be determined based on pre-defined targets of interest. For example, if the pre-defined target of interest is Bridge A, then "Bridge A" can be used as the keyword for retrieval.

[0126] In other implementations, map information can also be uploaded manually.

[0127] S420. By using multiple road condition classification models applicable to different road condition types, the map information content is classified into road condition types to obtain multiple road condition type classification results.

[0128] Specifically, by using multiple road condition classification models applicable to different road condition types, the road condition type of each map intelligence content is classified to obtain the road condition type classification result of that map intelligence content.

[0129] S430. Based on the classification results of multiple road status types, determine the road status type to which the map information content belongs.

[0130] S440. Extract road entities from the map information content to obtain the road entities in the map information content.

[0131] Step S440 is similar to step S130 in the previous embodiment, and will not be described again here.

[0132] S450. Based on the road status type and road entities in the map intelligence content, create a map intelligence task for the map intelligence content and add the created map intelligence task to the map intelligence task set.

[0133] The road status type to which the map intelligence content belongs is taken as the road status type involved in the map intelligence task created for the map intelligence content, and the road entities in the map intelligence content are taken as the road entities involved in the map intelligence task created for the map intelligence content.

[0134] In the above implementation, the road entities and their respective road status types in each map intelligence content are automatically extracted as the road entities and road status types involved in the map intelligence task for creating map intelligence content. Subsequently, it is not necessary to repeatedly extract road entities and classify road status types for the map intelligence content involved in the map intelligence task.

[0135] In some implementations, please refer to Figure 7 , Figure 7The fifth flowchart of the map intelligence task processing method provided in this application embodiment is shown, where the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing method further includes steps S510-S540:

[0136] S510. Obtain road public opinion list information; the road public opinion list information includes the road name of at least one road and the road status indication information of at least one road.

[0137] Road condition information describes the road conditions that require attention for that road.

[0138] In some implementations, the road-related public opinion list information can be created manually by staff, or it can be provided by parties with road-related concerns.

[0139] S520. Based on the road status indication information of at least one road in the road public opinion list, determine the road status type of at least one road.

[0140] Understandably, based on the road status type, multiple road states belonging to that road status type can be identified; for example, if the road status type is "open / closed", then "open" and "closed" can be identified as two road states. The road status indication information is used to describe the road status that needs to be monitored for that road. Therefore, if the road status indicated by the road status indication information belongs to one of the multiple road states included in a road status type, then that road status type is identified as the road status type to which the road belongs.

[0141] S530. Based on the road status type of at least one road and the road name of at least one road, create a corresponding map intelligence task for at least one road; wherein, the road name of one road is the road entity involved in the corresponding map intelligence task, and the road status type of one road is the road status type involved in the corresponding map intelligence task.

[0142] S540. Add the created map intelligence task to the map intelligence task set.

[0143] In the above implementation, by setting up a road public opinion list and creating a map intelligence task based on the road status type of at least one road and the road name of at least one road in the road public opinion list information, attention to the specified road target is achieved.

[0144] In some implementations, to ensure the accuracy of road condition classification models for various road condition types, it is necessary to pre-train the road condition classification models for each type. Please refer to [link / reference]. Figure 8 , Figure 8 The training process of the road state classification model provided in the embodiments of this application is illustrated. For each road state classification model, training is performed according to the following steps S610-S640:

[0145] S610. Obtain the training data corresponding to the road state classification model; the training data includes multiple road event descriptions and the labeled road state types corresponding to each road event description.

[0146] The road event description refers to text that describes an event with roads as the main subject. Road event descriptions can include collected road-related news, discussion topics, published notes, etc., without specific limitations. The road status type corresponding to a road event description refers to the road status type labeled according to the road status presented in the road event description. In specific embodiments, this can be determined manually.

[0147] S620. The road condition classification model is used to classify the road condition type of the event description content of each road, and the predicted road condition type of each road event description content is obtained.

[0148] The predicted road state type of a road event description refers to the road state type predicted by the road state classification model based on the road event description.

[0149] In some implementations, if the road state classification model is a binary classification model, the predicted road state type output by the road state classification model for each road event description is used to indicate whether the road state described in the road event description belongs to the road state type corresponding to the road state classification model.

[0150] Specifically, the road state classification model classifies the road state type of a road event description and obtains the first probability that the road event description belongs to the road state type corresponding to the road state classification model. If the first probability is greater than the first probability threshold, the road event description is considered to belong to the road state type corresponding to the road state classification model; otherwise, the road event description is considered not to belong to the road state type corresponding to the road state classification model.

[0151] In addition, the road condition classification model can also be a multi-classification model. The specific process of road condition classification is similar to that described above, and will not be repeated here.

[0152] S630. Calculate the first loss based on the predicted road state type of each road event description and the labeled road state type corresponding to each road event description.

[0153] The first loss can be calculated using a loss calculation function, such as the cross-entropy loss function or the mean squared error loss function.

[0154] S640. Based on the first loss, adjust the parameters of the road state classification model until the first training termination condition is met.

[0155] The first training termination condition can be either the first loss being less than the first loss threshold, or the number of training iterations reaching the first training iteration threshold.

[0156] In some embodiments, the road state classification model can be a BERT model, a Transformer model, or a neural network model with other structures. See also Figure 9 , Figure 9 The diagram shows a schematic of the structure of a road state classification model according to an embodiment of this application. The road state classification model includes an input layer, an embedding layer, a feature extraction layer, a feature output layer, a classification layer, and a classification output layer.

[0157] The input layer is used to input road event descriptions; in some implementations, a category identifier, namely the [CLS] identifier, can be added before the beginning of the road event description.

[0158] In some implementations, the input text length requirement of the road condition classification model can also be considered. For example, if the road condition classification model requires the input text length to be K, i.e., max_len = K, when the text length (number of words) of the text content exceeds K, the first K words of the text content are extracted as the input content of the input layer. If the text length of the text content is less than K, a whitespace marker is added to the end of the text content so that the number of words of the text content after adding the whitespace marker is K. The whitespace marker is, for example, [PAD].

[0159] The embedding layer is used to embed the fixed-length text content input from the input layer, obtaining a vector representation of each word in the road event description content, such as... Figure 9 In the case of {x0, x1, ..., xN}.

[0160] The feature extraction layer is used to extract features from the vector representation of each word obtained from the embedding layer. In some implementations, the feature extraction layer can be... Figure 9 The feature extraction layer of the pre-trained BERT (Bidirectional Encoder Representations from Transformers) model.

[0161] The feature output layer is used to output the features or feature codes extracted by the feature extraction layer for each word in the description of road events. The output CLS at the first position is obtained by weighted summation of all token vectors through a self-attention mechanism, which can be used as input to the classification layer.

[0162] The classification layer is used to perform classification prediction based on the feature encoding output by the feature output layer, obtaining the predicted probability that the road event description belongs to the road state type corresponding to the road state classification model. The classification layer can be... Figure 4 The linear classification layer in the model includes a random deactivation layer (Dropout layer) and a fully connected layer. Activation functions can be deployed in the classification layer.

[0163] The classification output layer is used to determine whether the road event description belongs to the road state type corresponding to the road state classification model based on the predicted probability of the road event description and a pre-set probability threshold, and outputs text indicating whether the road event description belongs to the road state type corresponding to the road state classification model; for example... Figure 9 In this model, the road state classification model corresponds to the road state type "road open or closed". Therefore, the classification output layer will eventually output the text indicating whether the road event description belongs to the road state type "road open or closed". For example, if the preset probability threshold is 0.5, the predicted probability of the road event description is greater than 0.5, and the road event description is considered to belong to the road state type corresponding to the road state classification model. Otherwise, the road event description is considered not to belong to the road state type corresponding to the road state classification model.

[0164] In some implementations, entity matching can also be performed using a dual-tower model; please refer to [link / reference]. Figure 10 , Figure 10 The embodiments provided in this application are given Figure 2 The flowchart of step S140 is shown below. Step S140 may include steps S710-S740:

[0165] S710. The first semantic feature extraction network in the dual-tower network extracts semantic features from the first road entity to obtain the first semantic feature.

[0166] The dual-tower model, also known as the dual-tower neural network model or dual-tower architecture, is a common model structure in deep learning. It consists of two independent neural networks that typically have similar structures but do not share parameters. Each network is responsible for processing a specific type of input data and mapping it to a high-dimensional vector space. In the dual-tower model, the output vectors of the two towers can be used to calculate similarity or matching scores.

[0167] S720. The second semantic feature extraction network in the dual-tower network extracts semantic features from the second road entity to obtain the second semantic features.

[0168] S730. Calculate the semantic similarity between the first semantic feature and the second semantic feature.

[0169] There are various ways to calculate semantic similarity, such as cosine similarity, Euclidean distance, Manhattan distance, etc.

[0170] S740. Determine the entity matching result based on semantic similarity.

[0171] Specifically, if the semantic similarity is greater than the similarity threshold, the entity matching result is determined to indicate that the first road entity and the second road entity match; otherwise, the entity matching result is determined to indicate that the first road entity and the second road entity do not match.

[0172] In some implementations, please refer to Figure 11 , Figure 11 A schematic diagram of the training process of the dual-tower model provided in the embodiments of this application is given. The training process of the dual-tower model may include steps S810-S860:

[0173] S810. Obtain entity matching training data. The entity matching training data includes multiple road entity pairs and matching labels corresponding to each road entity pair. The matching label corresponding to a road entity pair is used to indicate whether the first sample road entity in the road entity pair and the second sample road entity therein are the same entity.

[0174] Among them, the road entity pairs include positive sample entity pairs consisting of a first sample road entity and a second sample road entity belonging to the same entity, and negative sample entity pairs consisting of a first sample road entity and a second sample road entity belonging to different entities.

[0175] S820. The first semantic feature extraction network in the dual-tower network extracts semantic features from the first sample road entity to obtain the semantic features of the first sample.

[0176] S830. The second semantic feature extraction network in the dual-tower network extracts semantic features from the second sample road entity to obtain the semantic features of the second sample.

[0177] S840. Calculate the semantic similarity of each road entity pair based on the semantic features of the first sample and the semantic features of the second sample.

[0178] There are various ways to calculate semantic similarity, such as cosine similarity, Euclidean distance, Manhattan distance, etc.

[0179] S850. Calculate the second loss based on the semantic similarity of each road entity pair and the matching label of each road entity pair.

[0180] The second loss can be calculated using a loss calculation function, such as the cross-entropy loss function or the mean squared error loss function.

[0181] S860. Based on the second loss, adjust the parameters of the dual-tower network until the second training termination condition is met.

[0182] The second training termination condition can be either the second loss being less than the second loss threshold, or the number of training iterations reaching the second training iteration threshold.

[0183] In some implementations, please refer to Figure 12 , Figure 12 A schematic diagram of the structure of the dual-tower model provided in the embodiments of this application is given. The dual-tower model includes a first semantic feature extraction network, a second semantic feature extraction network, and a matching layer. Each semantic feature extraction network includes a feature extraction layer, a pooling layer, and an output layer.

[0184] by Figure 12 Taking the first semantic feature extraction network as an example, the feature extraction layer of the first semantic feature extraction network is used to extract the semantic features of the first road entity in the road entity pair. Then, the pooling layer performs pooling processing on the semantic features extracted by the feature extraction layer. Finally, the output layer outputs the pooled first semantic features.

[0185] Then, the cosine similarity between the first semantic feature output by the first semantic feature extraction network and the second semantic feature output by the second semantic feature extraction network is calculated through the matching layer, and the entity matching result is determined based on the cosine similarity result.

[0186] In some implementations, please refer to Figure 13 , Figure 13 An application flowchart of the map information task processing method provided in the embodiments of this application is given.

[0187] The process involves acquiring map intelligence content and target content. Multiple road state classification models applicable to different road condition types are used to classify the acquired map intelligence content, thereby determining the road state type to which the map intelligence content belongs. A deduplication algorithm is then used to remove duplicate map intelligence content under the same road state type. Next, the second road entity is extracted from the map intelligence content, and map intelligence tasks are created for the map intelligence content based on its road state type and the second road entity within it. These created map intelligence tasks are then added to a map intelligence task set.

[0188] For each target map intelligence task in the map intelligence task set, the target content is classified by the road state classification model corresponding to the target road state type involved in the target map intelligence task. If the classification result indicates that the target content belongs to the target road state type, the road entity is extracted from the target content to obtain the first road entity in the target content.

[0189] Finally, the first road entity is matched with the second road entity involved in the target map intelligence task to obtain the entity matching result. If the entity matching result indicates that the first road entity and the second road entity are the same entity, a priority processing tag is added to the target map intelligence task; otherwise, no priority processing tag is added to the target map intelligence task.

[0190] All map intelligence tasks in the map intelligence task set, along with successfully matched target content, are pushed to the map task processing platform, which then prioritizes map intelligence tasks with priority processing tags.

[0191] Please see Figure 14 , Figure 14 The embodiments provided in this application are given. Figure 13 The provided application flowchart is shown in the image.

[0192] After collecting map intelligence, determine the map intelligence task based on the map intelligence content; such as... Figure 14 The map information collected indicates that "AAA Bridge is expected to open to traffic on March 2nd". This allows us to identify the target map information task for "AAA Bridge opening to traffic". If the target content matches the road entity "AAA Bridge" and the road status type is "open / closed", a priority processing tag is added to the target map information task "AAA Bridge opening to traffic", thereby increasing the priority level of the target map information task.

[0193] Compared to existing large-scale public opinion service systems and manual judgment methods, large-scale public opinion service systems suffer from functional redundancy and high maintenance costs; while manual judgment methods are costly, lack standardized practices, and are inefficient. The map intelligence task processing method proposed in this application, based on a multimodal text model (including the classification model, entity recognition model, and dual-tower model in the aforementioned embodiments), accurately matches the target content with the map intelligence task, solving the problem of how to identify high-profile and urgent map intelligence tasks in real time. Not only is the system small in scale and has a short development cycle, but it can also achieve the recognition effect of large-scale public opinion service systems in actual business practice.

[0194] In some implementations, please refer to Figure 15 , Figure 15A schematic diagram of a map information task processing apparatus 900 provided in an embodiment of this application is given. The map information task processing apparatus 900 includes:

[0195] Module 910 is used to acquire target content.

[0196] The classification module 920 is used to classify the target content into road state types using the target road state classification model to obtain the classification results. The target road state classification model is a road state classification model used to identify the target road state type. The target road state type is the road state type involved in the target map intelligence task.

[0197] The entity extraction module 930 is used to extract road entities from the target content if the classification result indicates that the target content belongs to the target road state type, and obtain the first road entity in the target content.

[0198] The matching module 940 is used to perform entity matching between the first road entity and the second road entity involved in the target map intelligence task, and obtain the entity matching result.

[0199] The output module 950 is used to determine the processing priority of the target map intelligence task based on the target content if the entity matching result indicates that the first road entity and the second road entity are the same entity.

[0200] In some embodiments, the map intelligence task processing device 900 further includes an element extraction module, used to extract key event element information from the target content if the entity matching result indicates that the first road entity and the second road entity are the same entity; and an update module, used to update the event status change timeline information corresponding to the target map intelligence task based on the key event element information.

[0201] In some implementations, the update module is also used to delete the target map intelligence task from the map intelligence task set if it is determined, based on the event state change timeline information corresponding to the target map intelligence task, that the road state of the second road entity has changed to the target road state.

[0202] In some implementations, the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing device 900 further includes a synchronization module for synchronizing the target content and the updated processing priority of the target map intelligence task to the map task processing platform; a receiving module for receiving the authenticity verification result returned by the map task processing platform; and an information sending module for sending a status update prompt message to the map application based on the road status of the first road entity in the target content if the authenticity verification result indicates that the target content is authentic, so that the map application updates the road status of the first road entity in the electronic map.

[0203] In some implementations, the target content is published content whose popularity value exceeds a threshold; the output module is specifically used to, if the entity matching result indicates that the first road entity and the second road entity are the same entity, take the processing priority corresponding to the popularity level of the target content as the processing priority of the target map intelligence task according to the correspondence between popularity level and processing priority.

[0204] In some implementations, the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing device 900 further includes an intelligence acquisition module for acquiring map intelligence content; an intelligence classification module for classifying the map intelligence content into road state types using multiple road state classification models applicable to different road state types, obtaining multiple road state type classification results; a state confirmation module for determining the road state type to which the map intelligence content belongs based on the multiple road state type classification results; an intelligence entity extraction module for extracting road entities from the map intelligence content, obtaining the road entities in the map intelligence content; and a task creation module for creating map intelligence tasks for the map intelligence content based on the road state type to which the map intelligence content belongs and the road entities in the map intelligence content, and adding the created map intelligence tasks to the map intelligence task set.

[0205] In some implementations, the target map intelligence task is any one of the map intelligence task set; the map intelligence task processing device 900 further includes a list acquisition module for acquiring road public opinion list information; the road public opinion list information includes the road name of at least one road and the road status indication information of at least one road; a type confirmation module for determining the road status type of at least one road based on the road status indication information of at least one road in the road public opinion list information; a task creation module for creating a corresponding map intelligence task for at least one road based on the road status type and the road name of at least one road; wherein, the road name of a road is the road entity involved in the corresponding map intelligence task, and the road status type of a road is the road status type involved in the corresponding map intelligence task; and an adding module for adding the created map intelligence task to the map intelligence task set.

[0206] In some embodiments, the map information task processing device 900 further includes a first training data acquisition module, used to acquire training data corresponding to the road state classification model; the training data includes multiple road event description contents and labeled road state types corresponding to each road event description content; a first prediction module, used to classify the road event description contents into road state types by the road state classification model to obtain the predicted road state types of each road event description content; a first loss calculation module, used to calculate a first loss based on the predicted road state types of each road event description content and the labeled road state types corresponding to each road event description content; and a first adjustment module, used to adjust the parameters of the road state classification model based on the first loss until a first training termination condition is met.

[0207] In some implementations, the matching module 940 includes a first semantic extraction unit, used to extract semantic features from a first semantic feature extraction network in the dual-tower network to obtain a first semantic feature; a second semantic extraction unit, used to extract semantic features from a second semantic feature extraction network in the dual-tower network to obtain a second semantic feature; a similarity calculation unit, used to calculate the semantic similarity between the first semantic feature and the second semantic feature; and a matching output unit, used to determine the entity matching result based on the semantic similarity.

[0208] In some embodiments, the map intelligence task processing device 900 further includes a second training data acquisition module for acquiring entity matching training data, which includes multiple road entity pairs and matching labels corresponding to each road entity pair; the matching label corresponding to a road entity pair is used to indicate whether a first sample road entity in the road entity pair and a second sample road entity therein are the same entity; a first feature extraction module is used to extract semantic features from the first sample road entity using a first semantic feature extraction network in the dual-tower network to obtain first sample semantic features; a second feature extraction module is used to extract semantic features from the second sample road entity using a second semantic feature extraction network in the dual-tower network to obtain second sample semantic features; a similarity prediction module is used to calculate the semantic similarity of each road entity pair based on the first sample semantic features and the second sample semantic features; a second loss calculation module is used to calculate a second loss based on the semantic similarity of each road entity pair and the matching labels corresponding to each road entity pair; and a second adjustment module is used to adjust the parameters of the dual-tower network based on the second loss until a second training termination condition is met.

[0209] Figure 16A schematic diagram of a computer system suitable for implementing the electronic device of the embodiments of this application is shown. The electronic device can be the terminal described above, used to implement the map information task processing method provided in this application. It should be noted that... Figure 16 The computer system 1300 of the electronic device shown is merely an example and should not impose any limitation on the functionality and scope of use of the embodiments of this application.

[0210] like Figure 16 As shown, the computer system 1300 includes a Central Processing Unit (CPU) 1301, which can perform various appropriate actions and processes, such as executing the methods described in the above embodiments, based on programs stored in Read-Only Memory (ROM) 1302 or programs loaded from storage portion 1308 into Random Access Memory (RAM) 1303. The RAM 1303 also stores various programs and data required for system operation. The CPU 1301, ROM 1302, and RAM 1303 are interconnected via a bus 1304. An Input / Output (I / O) interface 1305 is also connected to the bus 1304.

[0211] The following components are connected to I / O interface 1305: an input section 1306 including a keyboard, mouse, microphone, etc.; an output section 1307 including a cathode ray tube (CRT), liquid crystal display (LCD), and speakers, etc.; a storage section 1308 including a hard disk, etc.; and a communication section 1309 including a network interface card such as a LAN (Local Area Network) card and a modem, etc. The communication section 1309 performs communication processing via a network such as the Internet. A drive 1310 is also connected to I / O interface 1305 as needed. Removable media 1311, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., are installed on drive 1310 as needed so that computer instructions read from them can be loaded into storage section 1308 as needed.

[0212] In particular, according to embodiments of this application, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, embodiments of this application include a computer program product comprising computer instructions. When these computer instructions are executed by the central processing unit (CPU) 1301, various functions defined in the system of this application are performed.

[0213] This application also provides a computer-readable storage medium storing computer instructions that, when executed by a processor, implement the methods described in any of the above method embodiments.

[0214] It should be noted that the computer-readable storage medium shown in the embodiments of this application can be a computer-readable signal medium, a computer-readable storage medium, or any combination of the two. Computer-readable storage media can be, for example, but not limited to, electrical, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatuses, or devices, or any combination thereof. More specific examples of computer-readable storage media may include, but are not limited to: electrical connections having one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM), flash memory, optical fiber, portable compact disc read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof. In this application, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this application, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such transmitted data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. The computer-readable signal medium can also be any computer-readable storage medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable storage medium can be transmitted using any suitable medium, including but not limited to wireless, wired, etc., or any suitable combination thereof.

[0215] In the embodiments of this application, the terms "module" or "unit" refer to computer instructions or a portion of computer instructions that have a predetermined function and work together with other related parts to achieve a predetermined goal. These instructions can be implemented, wholly or partially, using software, hardware (e.g., processing circuitry or memory), or a combination thereof. Similarly, a processor (or multiple processors or memory) can be used to implement one or more modules or units. Furthermore, each module or unit can be part of an overall module or unit that functions as a whole.

[0216] The above are merely preferred embodiments of this application and are not intended to limit this application in any way. Although this application has disclosed preferred embodiments as above, it is not intended to limit this application. Any person skilled in the art can make some modifications or alterations to the above-disclosed technical content to create equivalent embodiments without departing from the scope of the technical solution of this application. Any simple modifications, equivalent changes and alterations made to the above embodiments based on the technical essence of this application without departing from the scope of the technical solution of this application shall still fall within the scope of the technical solution of this application.

Claims

1. A method for processing map information tasks, characterized in that, include: Obtain the target content; The target content is classified into road condition types using the target road condition classification model to obtain the classification results. The target road condition classification model refers to a road condition classification model used to identify the target road condition type, where the target road condition type is the road condition type involved in the target map intelligence task. If the classification result indicates that the target content belongs to the target road state type, road entity extraction is performed on the target content to obtain the first road entity in the target content; The first road entity is matched with the second road entity involved in the target map intelligence task to obtain the entity matching result. If the entity matching result indicates that the first road entity and the second road entity are the same entity, the processing priority of the target map intelligence task is determined according to the target content.

2. The method according to claim 1, characterized in that, After performing entity matching between the first road entity and the second road entity involved in the target map intelligence task to obtain the entity matching result, the method further includes: If the entity matching result indicates that the first road entity and the second road entity are the same entity, extract key event element information from the target content; Based on the key event element information, update the event status change timeline information corresponding to the target map intelligence task.

3. The method according to claim 2, characterized in that, The target map intelligence task is any one of the map intelligence task set; After updating the event status change timeline information corresponding to the target map intelligence task based on the key event element information, the method further includes: If, based on the event state change timeline information corresponding to the target map intelligence task, it is determined that the road state of the second road entity has changed to the target road state, the target map intelligence task is deleted from the map intelligence task set.

4. The method according to claim 1, characterized in that, If the entity matching result indicates that the first road entity and the second road entity are the same entity, after determining the processing priority of the target map intelligence task based on the target content, the method further includes: The updated processing priority of the target content and the target map intelligence task will be synchronized to the map task processing platform. Receive the authenticity verification result returned by the map task processing platform; If the authenticity verification result indicates that the target content is authentic, a status update prompt message is sent to the map application based on the road status of the first road entity in the target content, so that the map application updates the road status of the first road entity in the electronic map.

5. The method according to claim 1, characterized in that, The target content refers to published content whose popularity value exceeds a threshold; If the entity matching result indicates that the first road entity and the second road entity are the same entity, the processing priority of the target map intelligence task is determined based on the target content, including: If the entity matching result indicates that the first road entity and the second road entity are the same entity, the processing priority corresponding to the heat level of the target content is taken as the processing priority of the target map intelligence task according to the correspondence between heat level and processing priority.

6. The method according to any one of claims 1 to 5, characterized in that, The target map intelligence task is any one of the map intelligence task set; the method further includes: Obtain map information; By using multiple road condition classification models applicable to different road condition types, the map information content is classified into road condition types, resulting in multiple road condition type classification results. Based on the classification results of the multiple road status types, determine the road status type to which the map information content belongs; Road entities are extracted from the map intelligence content to obtain the road entities in the map intelligence content; Based on the road status type to which the map intelligence content belongs and the road entities in the map intelligence content, a map intelligence task is created for the map intelligence content, and the created map intelligence task is added to the map intelligence task set.

7. The method according to any one of claims 1 to 5, characterized in that, The target map intelligence task is any one of the map intelligence task set; The method further includes: Obtain road public opinion list information; the road public opinion list information includes the name of at least one road and the road status indication information of at least one road; Based on the road status indication information of at least one road in the road public opinion list information, determine the road status type of at least one of the roads; Based on the road status type of at least one of the roads and the road name of at least one of the roads, a corresponding map intelligence task is created for at least one of the roads; wherein, the road name of one of the roads is the road entity involved in the corresponding map intelligence task, and the road status type of one of the roads is the road status type involved in the corresponding map intelligence task; Add the created map intelligence task to the map intelligence task set.

8. The method according to any one of claims 1 to 5, characterized in that, Before the target content is classified into road state types using the target road state classification model to obtain the classification result, the method further includes: The road condition classification models are trained according to the following process: Obtain the training data corresponding to the road state classification model; the training data includes multiple road event descriptions and the labeled road state types corresponding to each road event description; The road state classification model is used to classify the road state type of each road event description to obtain the predicted road state type of each road event description. The first loss is calculated based on the predicted road state type of each road event description and the labeled road state type corresponding to each road event description. Based on the first loss, adjust the parameters of the road state classification model until the first training termination condition is met.

9. The method according to any one of claims 1 to 5, characterized in that, The step of matching the first road entity with the second road entity involved in the target map intelligence task to obtain the entity matching result includes: The first semantic feature is extracted from the first road entity by the first semantic feature extraction network in the dual-tower network to obtain the first semantic feature; The second semantic feature is extracted from the second road entity by the second semantic feature extraction network in the dual-tower network to obtain the second semantic feature; Calculate the semantic similarity between the first semantic feature and the second semantic feature; The entity matching result is determined based on the semantic similarity.

10. The method according to claim 9, characterized in that, Before performing entity matching between the first road entity and the second road entity involved in the target map intelligence task to obtain the entity matching result, the method further includes: Acquire entity matching training data, which includes multiple road entity pairs and matching labels corresponding to each road entity pair; the matching label corresponding to a road entity pair is used to indicate whether the first sample road entity in the road entity pair and the second sample road entity therein are the same entity. The semantic features of the first sample road entity are extracted by the first semantic feature extraction network in the dual-tower network to obtain the semantic features of the first sample. The second semantic feature extraction network in the dual-tower network extracts semantic features from the second sample road entity to obtain the second sample semantic features. Based on the semantic features of the first sample and the semantic features of the second sample, the semantic similarity of each road entity pair is calculated. The second loss is calculated based on the semantic similarity of each road entity pair and the matching label of each road entity pair. Based on the second loss, adjust the parameters of the dual-tower network until the second training termination condition is met.

11. A processing device for map information tasks, characterized in that, include: The acquisition module is used to acquire the target content; The classification module is used to classify the target content into road state types using a target road state classification model to obtain classification results; the target road state classification model refers to a road state classification model used to identify the target road state type, and the target road state type is the road state type involved in the target map intelligence task. An entity extraction module is used to extract road entities from the target content if the classification result indicates that the target content belongs to the target road state type, thereby obtaining the first road entity in the target content. The matching module is used to perform entity matching between the first road entity and the second road entity involved in the target map intelligence task, and obtain the entity matching result; The output module is used to determine the processing priority of the target map intelligence task based on the target content if the entity matching result indicates that the first road entity and the second road entity are the same entity.

12. An electronic device, characterized in that, include: processor; A memory storing computer instructions that, when executed by the processor, implement the method as described in any one of claims 1-10.

13. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that, when executed by a processor, implement the method as described in any one of claims 1-10.

14. A computer program product comprising computer instructions, characterized in that, When executed by a processor, the computer instructions implement the method described in any one of claims 1-10.

Images