Method, device, electronic device and storage medium for intelligent identification of water accumulation points

Abstract

The embodiment of the present invention discloses a method, device, electronic device and storage medium for intelligent identification of waterlogging points. The method includes: using crawler technology to obtain text information related to rainstorm waterlogging from social media, and performing text sentence segmentation and keyword reverse screening preprocessing operations to obtain target text information; classifying the target text information, and labeling the classified text information with categories; constructing a fine-grained geographic location automatic extraction algorithm, and obtaining target fine-grained waterlogging point location information from waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm; determining the longitude and latitude information corresponding to the target fine-grained waterlogging point location information, and encoding and visualizing the target fine-grained waterlogging point location based on the longitude and latitude information. By adopting this solution, fine-grained waterlogging point location information can be mined from social media text information efficiently and at low cost, greatly improving the utilization value of social media big data.

Description

Technical Field

[0001] The embodiments of the present invention relate to the technical field of disaster prevention and mitigation, and in particular to a method, device, electronic device and storage medium for intelligently identifying waterlogging points. Background Art

[0002] In recent years, extreme rainfall events have frequently occurred worldwide, causing significant casualties and economic losses. Compared to rural areas, floods inflict more severe damage in densely populated and densely built-up areas. Urban waterlogging is characterized by spatial dispersion and dynamic changes, making it extremely difficult to collect complete and timely location information.

[0003] Traditional technologies for capturing real-time flooding locations include manual exploration and ground-based sensors, which guarantee accurate measurements but have limited range. Remote sensing, radar satellites, and other technologies can collect real-time information, but data quality is easily affected by cloud cover and revisit cycles. Low-altitude drone photography can generate high-resolution data, but large-scale collection of real-time data on urban flooding locations is still limited by multiple factors, including airspace control, rainy weather operations, and power shortages. The emergence of social media has made it easier to quickly obtain information about flooding. However, the proportion of geotags embedded in social media is limited, and the locations in user profiles often reflect the registered location, which is often inaccurate or outdated. Users of social media platforms are privacy-conscious, and most users are reluctant to share their locations, resulting in very limited directly accessible geotagged data. The relatively rich geographic descriptions in social media texts offer new avenues for mining location data, but the mining and application of such information is still in its infancy.

[0004] Therefore, how to obtain the location information of waterlogging points from social media data is a technical problem that needs to be solved urgently by those skilled in the art. Summary of the Invention

[0005] The embodiments of the present invention provide a method, device, electronic device and storage medium for intelligently identifying waterlogging points, so as to efficiently and cost-effectively mine fine-grained waterlogging point location information from social media text information, greatly improving the utilization value of social media big data and helping to promote the construction of smart cities.

[0006] In a first aspect, an embodiment of the present invention provides a method for intelligently identifying water accumulation points, comprising:

[0007] Using crawler technology to obtain text information related to rainstorm waterlogging from social media, and preprocessing the text information to obtain target text information; the preprocessing includes text sentence segmentation and keyword reverse screening;

[0008] Classify the target text information and label the classified text information; the classified text information includes waterlogging information and loss information;

[0009] Constructing a fine-grained geographic location automatic extraction algorithm, and obtaining target fine-grained waterlogging point location information from waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm;

[0010] Determine the longitude and latitude information corresponding to the target fine-grained water accumulation point location information, and encode and visualize the target fine-grained water accumulation point location based on the longitude and latitude information.

[0011] In a second aspect, an embodiment of the present invention further provides a device for intelligently identifying water accumulation points, comprising:

[0012] A text information acquisition and preprocessing module is used to use crawler technology to obtain text information related to rainstorm waterlogging from social media and preprocess the text information to obtain target text information; the preprocessing includes text sentence segmentation and keyword reverse screening;

[0013] The text information classification and labeling module is used to classify the target text information and label the classified text information into categories; wherein the classified text information includes water accumulation information and loss information;

[0014] A fine-grained waterlogging point location information extraction module is used to construct a fine-grained geographic location automatic extraction algorithm and obtain target fine-grained waterlogging point location information from waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm;

[0015] The fine-grained waterlogging point location information visualization module is used to determine the longitude and latitude information corresponding to the target fine-grained waterlogging point location information, and encode and visualize the target fine-grained waterlogging point location based on the longitude and latitude information.

[0016] In a third aspect, an embodiment of the present invention further provides an electronic device, the electronic device comprising:

[0017] one or more processors;

[0018] a storage device for storing one or more programs;

[0019] When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent water accumulation point identification method described in any embodiment of the present invention.

[0020] In a fourth aspect, an embodiment of the present invention further provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method for intelligently identifying water accumulation points described in any embodiment of the present invention.

[0021] In a fifth aspect, an embodiment of the present invention further provides a computer program product, comprising a computer program, which, when executed by a processor, implements the method for intelligently identifying water accumulation points as described in any embodiment of the present invention.

[0022] The embodiments of the present invention provide a method, device, electronic device, and storage medium for intelligent identification of waterlogging points. The method uses crawler technology to obtain text information related to rainstorm waterlogging from social media and preprocesses it to obtain target text information. The target text information is classified into waterlogging information and loss information and labeled. A fine-grained geographic location automatic extraction algorithm is constructed, and target fine-grained waterlogging point location information is obtained from the waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm. The latitude and longitude information corresponding to the target fine-grained waterlogging point location information is determined, and the target fine-grained waterlogging point location is encoded and visualized based on the latitude and longitude information. The technical solution of the embodiments of the present invention integrates rapid data collection, processing, and classification, fine-grained waterlogging point location mining, location encoding, and visualization. Fine-grained waterlogging point location information is mined from social media text information efficiently and at low cost, greatly improving the utilization value of social media big data and contributing to the construction of smart cities. BRIEF DESCRIPTION OF THE DRAWINGS

[0023] Other features, objects, and advantages of the present invention will become more apparent upon reading the detailed description of non-limiting embodiments made with reference to the following drawings. The drawings are for the purpose of illustrating preferred embodiments only and are not to be considered as limiting the present invention. Like reference characters are used throughout the drawings to denote like parts. In the drawings:

[0024] Figure 1 This is a flow chart of a method for intelligently identifying water accumulation points provided by the first embodiment of the present invention;

[0025] Figure 2 This is a flow chart of a method for intelligently identifying water accumulation points provided in the second embodiment of the present invention;

[0026] Figure 3 This is a flowchart of a method for obtaining keywords based on crawler technology provided in the second embodiment of the present invention;

[0027] Figure 4 A schematic diagram of a text binary classification model architecture provided in the second embodiment of the present invention;

[0028] Figure 5A schematic diagram of a multi-label text classification model architecture provided in Example 2 of the present invention;

[0029] Figure 6 This is a flow chart of a method for intelligently identifying waterlogging points provided in the third embodiment of the present invention;

[0030] Figure 7 A schematic diagram of a named entity recognition model architecture provided in Example 3 of the present invention;

[0031] Figure 8 This is a schematic diagram of the architecture of a method for extracting fine-grained water accumulation points provided by the third embodiment of the present invention;

[0032] Figure 9 A schematic diagram of fine-grained water accumulation point output logic provided in the third embodiment of the present invention;

[0033] Figure 10 This is a schematic structural diagram of a device for intelligently identifying water accumulation points provided by a fourth embodiment of the present invention;

[0034] Figure 11 This is a structural diagram of an electronic device provided in Example 5 of the present invention. DETAILED DESCRIPTION

[0035] The present invention will be further described in detail below with reference to the accompanying drawings and examples. It will be understood that the specific embodiments described herein are intended only to illustrate the present invention and are not intended to limit the present invention. It should also be noted that, for ease of description, the accompanying drawings only illustrate portions relevant to the present invention, not all structures.

[0036] Before discussing the exemplary embodiments in more detail, it should be mentioned that some exemplary embodiments are described as processes or methods depicted as flow charts. Although the flow charts describe the various operations (or steps) as sequential processes, many of the operations (or steps) therein can be implemented in parallel, concurrently, or simultaneously. In addition, the order of the various operations can be rearranged. The process can be terminated when its operation is completed, but can also have additional steps not included in the accompanying drawings. The process can correspond to a method, function, procedure, subroutine, subprogram, etc.

[0037] The acquisition, storage, use, and processing of data in the technical solution of this application are in compliance with the relevant provisions of national laws and regulations. It should be noted that in the embodiments of this application, certain software, components, or models, etc., which are already available in the industry, may be mentioned. These should be considered as exemplary and are only intended to illustrate the feasibility of implementing the technical solution of this application. It does not mean that the applicant has already or necessarily used such a solution.

[0038] It should be noted that the terms "first," "second," and the like in the description and claims of the present invention and the accompanying drawings are used to distinguish similar objects and are not necessarily used to describe a specific order or precedence. It should be understood that the terms used in this manner are interchangeable where appropriate, such that the embodiments of the present invention described herein can be practiced in an order other than that illustrated or described herein.

[0039] Example 1

[0040] Figure 1 This is a flow chart of a method for intelligently identifying waterlogging points, provided in Example 1 of the present invention. This embodiment is applicable to intelligently identifying waterlogging points from social media. The method of this embodiment can be performed by an intelligent waterlogging point identification device, which can be implemented using hardware and / or software. The device can be configured in a server for intelligent waterlogging point identification. The method specifically includes the following steps:

[0041] S110 , using crawler technology to obtain text information related to rainstorm waterlogging from social media, and preprocessing the text information to obtain target text information.

[0042] Crawler technology, also known as web crawlers, spiders, or robots, refers to a technology that automatically retrieves information from the internet. Crawler technology uses programs to automatically access network resources, capture useful data, and store it on local or remote servers. Crawler technology can automatically acquire large amounts of data, greatly improving the efficiency and accuracy of data acquisition while also providing more possibilities for data analysis and mining.

[0043] Social media can refer to virtual communities and online platforms used for creating, sharing, and exchanging opinions, ideas, and experiences. Social media allows users to have greater choice and editing capabilities, allowing them to form a kind of reading and listening community. Social media can be presented in a variety of formats, including text, images, music, and video. Preprocessing can refer to data cleaning and filtering of useless information from text; this filtering can include removing information that does not include flooding points associated with heavy rain.

[0044] The embodiment of the present invention uses crawler technology to obtain text information related to rainstorm waterlogging from social media texts, and performs data cleaning and useless information filtering on the text information to obtain target text information containing rainstorm waterlogging waterlogging points.

[0045] S120: Classify the target text information and label the classified text information by category.

[0046] The geographical location mentioned in the target text information does not always refer to the location of the waterlogging point, but may also include geographical location information of secondary disasters caused by heavy rain and geographical location information containing opinions on the occurrence of heavy rain disasters.

[0047] In this embodiment of the present invention, target text information is classified and the classified text information is labeled. The classified text information includes waterlogging information, loss information, and other information. The waterlogging information may refer to information describing the location of waterlogging points, the loss information may refer to information describing secondary disaster losses caused by heavy rain and waterlogging, and the other information is used to describe opinions on the disasters caused by heavy rain and waterlogging. The waterlogging information, loss information, and other information may all contain location information related to heavy rain and waterlogging.

[0048] S130: Construct a fine-grained geographic location automatic extraction algorithm, and obtain target fine-grained waterlogging point location information from waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm.

[0049] The fine-grained automatic geographic location extraction algorithm may refer to a technology for automatically identifying and extracting geographic location information from text data. The fine-grained automatic geographic location extraction algorithm is capable of identifying specific location names, such as streets, communities, roads, and buildings, rather than just rough location information at the country, continent, city, or district level. In an embodiment of the present invention, the fine-grained automatic geographic location extraction algorithm is used to obtain more fine-grained geographic location information of flooding points caused by rainstorms from waterlogging information and loss information, making the acquired geographic location information of flooding points more accurate and detailed.

[0050] S140: Determine the longitude and latitude information corresponding to the target fine-grained water accumulation point location information, and encode and visualize the target fine-grained water accumulation point location based on the longitude and latitude information.

[0051] Encoding the target fine-grained waterlogging point location may refer to the process of performing coordinate conversion on the longitude and latitude information corresponding to the target fine-grained waterlogging point location information. In an embodiment of the present invention, after determining the target fine-grained waterlogging point location information, the corresponding longitude and latitude information is determined, and the longitude and latitude information is converted into World Geodetic System coordinates using a coordinate conversion tool to complete the encoding of the target fine-grained waterlogging point location information. After determining the World Geodetic System coordinates of the target fine-grained waterlogging point location information, the target fine-grained waterlogging point location is marked on a map and visually displayed.

[0052] The embodiment of the present invention provides an intelligent waterlogging point identification method. The method uses crawler technology to obtain text information related to rainstorm waterlogging from social media, pre-processes the text information to obtain target text information, classifies the target text information, and labels the classified text information with categories. The classified text information includes waterlogging information and loss information. A fine-grained geographic location automatic extraction algorithm is constructed, and target fine-grained waterlogging point location information is obtained from the waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm. The latitude and longitude information corresponding to the target fine-grained waterlogging point location information is determined, and the target fine-grained waterlogging point location is encoded and visualized based on the latitude and longitude information. The technical solution of the embodiment of the present invention integrates rapid data collection, processing and classification, fine-grained waterlogging point location mining, location encoding and visualization, and efficiently and cost-effectively mines fine-grained waterlogging point location information from social media text information, greatly improving the utilization value of social media big data and helping to promote the construction of smart cities.

[0053] Example 2

[0054] Figure 2 This is a flow chart of a method for intelligently identifying water points provided by the second embodiment of the present invention. The embodiment of the present invention further optimizes the above embodiment on the basis of the above embodiment, and the embodiment of the present invention can be combined with various optional solutions in one or more of the above embodiments. Figure 2 As shown, the intelligent water accumulation point identification method provided in the embodiment of the present invention may include the following steps:

[0055] S210: Acquire text information related to rainstorm and waterlogging from social media using crawler technology, and pre-process the text information to obtain target text information.

[0056] Crawling technology was used to collect text information on heavy rain from social media platforms and perform data preprocessing. This technology was used to gather text information on social media platforms related to heavy rain and urban flooding. Both personal and official accounts often post information related to urban flooding on social media platforms. Using web crawling technology based on simulated logins, text information was rapidly collected based on specified social media types, time ranges, and keywords. User addresses, posting times, and related information were automatically stored in a database.

[0057] As an optional but non-limiting implementation, the crawler technology is used to obtain text information related to rainstorm waterlogging from social media, and the text information is pre-processed to obtain target text information, including but not limited to steps A1-A2:

[0058] Step A1: Using web crawler technology based on simulated login, obtain text information related to rainstorm waterlogging points based on preset social media types, time ranges, and keywords.

[0059] Among them, the Octopus collector is used to automatically collect text information related to flooding points caused by heavy rain. The collection ideas can be found in Figure 3 . By simulating the user's browsing path, the internal collection logic is extracted to complete data collection. For page-turning loops and single-page microblog loops, the corresponding path language needs to be set to ensure the accuracy of the crawling results. In the step of editing and optimizing the crawling fields, delete the fields that do not need to be crawled; in the main crawling fields, the text information needs to set the path language separately, because the text information contains long texts and needs to be accurately positioned. The number of likes, comments, and reposts also needs to be formatted, and regular expressions are used to obtain the corresponding numbers. By modifying the keywords and time separately, you can obtain batches of URLs and realize the rapid collection of text information.

[0060] For example, consider the heavy rain in City A. The time range is set to January 1, 2018, to December 31, 2022. The keywords are 'waterlogging in City A,' 'flooding in City A,' 'inundation in City A,' 'waterlogging in City A,' 'flood in City A,' and 'flood in City A.' The social media type is 'original.' Using the aforementioned crawler mechanism, we retrieved 7,243 pieces of original text information.

[0061] Step A2: performing data cleaning and useless information filtering on the text information, eliminating information irrelevant to the rainstorm waterlogging waterlogging points, so as to obtain the target text information.

[0062] Among them, preprocessing includes data cleaning and useless information filtering, data cleaning includes sentence-level cleaning and symbol-level cleaning, and sentence-level cleaning includes keyword reverse screening, text segmentation, waterlogging-related keyword screening, and text similarity deduplication processing.

[0063] Keyword reverse screening can refer to removing locations that have the same name as a waterlogged location but do not correspond to the location. For example, if the location of point B1 in area A is needed and the text information includes the location of point B2, but point B2 belongs to a street in area C, point B2 will be removed; however, the names of B1 and B2 are the same.

[0064] Text segmentation can be performed by dividing text into sentences based on punctuation. First, it can eliminate a large number of sentences that do not contain information related to rainstorms and flooding, improving data quality. Second, it helps determine the category of sentences in text classification tasks, improving classification accuracy. Third, it can eliminate interference from place names in areas not affected by flooding.

[0065] The present invention uses the MinHash-LSH algorithm for text similarity deduplication, removing text with a similarity greater than 70%. Symbol-level cleaning can be achieved by using regular expressions (the 're' module) to remove special symbols such as @, Uniform Resource Locator (URL), 'folded', 'full text', and 'web link'.

[0066] The purpose of useless information filtering is to remove useless information, which can be converted into a text binary classification task. The embodiment of the present invention uses the TextCNN model to perform text binary classification to filter out useful information; "useful information" refers to text information that actually reports on flooding points during heavy rain, and vice versa is "useless information". The TextCNN model consists of an input layer, a convolutional layer, a pooling layer, a fully connected layer, and an output layer. The network architecture of TextCNN can be found in Figure 4 . Before entering the input layer, Word2Vec is first used to convert the collected text information into a language word vector that the computer can understand; among them, Word2Vec is a model for learning word vectors, which can convert words into real number vectors. These vectors can capture the semantic and grammatical relationships between words. Due to the small amount of sample data, the negative sampling skip-gram model (word2vec / skip-gram with negative sampling) is selected. The convolution layer uses three sizes of convolution kernels (3, 4 and 5) to convolve the input vector to achieve feature mapping. In the pooling layer, the maximum pooling function is used to capture the most prominent features. The fully connected layer is responsible for connecting the word vectors. Finally, the Softmax activation function is used to generate the corresponding class output. Among them, the skip-gram model is a language model widely used in the field of natural language processing, especially in the Word2Vec algorithm. It trains the model by predicting the occurrence of a word in the context to capture the semantic information of the word.

[0067] This embodiment of the present invention uses the preprocessing of 7,243 original text messages as an example. These 7,243 original microblogs were reversely screened, yielding a total of 4,222 original microblogs. The text was segmented using the "." keyword, yielding a total of 17,939 sentences. After filtering for flood-related keywords, 4,486 sentences were obtained. After deduplication, a total of 3,453 target text messages were obtained. After symbol-level cleaning, a total of 3,452 target text messages were obtained. Before filtering out useless information, the training dataset must be annotated, hyperparameters adjusted, and the TextCNN model trained. 600 useful and useless messages were annotated in a 1:1 ratio. Due to the limited training dataset, a five-fold cross-validation approach was used to evaluate the model training performance. The average accuracy on the test set was 89.67%, the average precision was 86.69%, the average recall was 93.88%, and the average F1 score was 90.03%. A model with a 100% recall rate was selected to predict the text message type, yielding 1,782 useful messages and filtering out 1,670 useless messages.

[0068] In the embodiment of the present invention, crawler technology is used to obtain a large amount of text information related to flooding points caused by heavy rain from social media, thereby improving the efficiency of obtaining text information and reducing the cost of manual acquisition. The text information is cleaned and the TextCNN model is used to filter out useless information, which can quickly filter out useless information and help improve the efficiency of system recognition.

[0069] S220: Classify the target text information and label the classified text information.

[0070] After data collection and preprocessing, textual information reporting flooding locations can be obtained. However, the geographic locations mentioned in this textual information do not always refer to flooding locations. This depends primarily on the specific subject matter of the text, such as flooding caused by heavy rain, information on secondary disaster losses resulting from heavy rain, and opinions expressed regarding the disaster. To accurately locate flooding locations, multi-label text classification is required to classify useful information into flooding, losses, and other information. Manual annotation is performed according to the category definitions, and the results are used to train the model.

[0071] As an optional but non-limiting implementation, the target text information is classified and the classified text information is labeled as a category, including but not limited to steps B1-B2:

[0072] Step B1: Analyze the target text information to determine the subject content described by the target text information.

[0073] Step B2: Classify the target text information into waterlogging information, loss information, and other information according to the subject content, and label the classified text information.

[0074] The loss information is used to describe the loss information of secondary disasters caused by heavy rain and waterlogging, and the other information is used to describe the views on the disasters brought about by heavy rain and waterlogging.

[0075] The target text information is classified into waterlogging information, loss information, and other information, and the classified text information is annotated. The annotated text information is used to train a text classification model, which adopts the ALBERT+BiGRU+Attention+FC model.

[0076] Among them, see Figure 5 In the model, ALBERT is responsible for converting text into vectors. BiGRU (BidirectionalGRU) + Attention is a feature extraction layer used to extract the global semantic features of the vector matrix output by the text vector representation layer. The forward and backward GRU layers are superimposed to form BiGRU, and the output is determined by the collaboration of the two layers of GRU units. Therefore, the bidirectional gated recurrent unit (BiGRU) can extract deep features containing contextual information. The attention mechanism (Attention) can focus on details according to the target, perceive important semantic features in the text context, and distinguish the importance by assigning different weights to each hidden layer node, so that the model pays attention to the key words and sentence information of the text. The fully connected layer (Fully Connected, FC) can map the results to the corresponding label category. By compressing the output of the FC layer with a sigmoid function, the probability of the target text information belonging to each type can be obtained.

[0077] Optionally, the 600 manually annotated data items were divided into training, validation, and test sets in an 8:1:1 ratio. The annotated samples were fed into the model for training. The word embedding dimension of the ALBERT layer was 312, the hidden dimension of the BiGRU layer was 128, and the dimension of the attention layer was 32. The experimental model was built using the Keras deep learning framework on Windows 10 and Python version 3.6. The Adam optimizer was used to minimize the final objective function during training. At Epoch 30, the test set Acc was 70.00% and the HL was 0.14. This means that 70% of the target text in the test set was fully predicted to the correct category, and the difference between the predicted label and the true label was 0.14. This model was used to perform multi-label text classification on 1782 useful information items, assigning the target text items appropriate labels. The "water accumulation information|loss information" category was the most common, with 527 items, while the "loss information|other information" category was the least common, with 14 items. Thirty-two items were classified as "blank." This indicates that the probability that the target text information belongs to any of the three categories is less than 0.5. Based on the principles of multi-label text classification, the target text information categories "flooding information | other information" and "loss information | other information" were merged into the "flooding information" and "loss information" categories, respectively. After processing, there were 954 target text information items tagged with "flooding information" and 802 target text information items tagged with "loss information." A total of 1,229 flooding and loss information items were obtained, but only 5.04% of the target text information contained inherent fine-grained geotags. The fine-grained geolocations for the remaining 1,167 flooding and loss information items remain to be discovered.

[0078] S230: Construct a fine-grained geographic location automatic extraction algorithm, and obtain target fine-grained waterlogging point location information from waterlogging information and loss information based on the fine-grained geographic location automatic extraction algorithm.

[0079] S240: Determine the longitude and latitude information corresponding to the target fine-grained water accumulation point location information, and encode and visualize the target fine-grained water accumulation point location based on the longitude and latitude information.

[0080] After determining the target fine-grained waterlogging point location information, the standardized name and longitude and latitude information corresponding to each target fine-grained waterlogging point location are automatically retrieved. The obtained longitude and latitude information is encrypted based on the Martian coordinate system. The coordinate conversion toolkit is then used to convert all coordinate systems into the World Geodetic System coordinates to complete the geocoding. The encoded waterlogging points are then visualized.

[0081] An embodiment of the present invention provides an intelligent waterlogging identification method. It uses crawler technology to acquire large amounts of text information related to waterlogging caused by heavy rain from social media, improving the efficiency of text information acquisition and reducing manual acquisition costs. The text information is cleaned and useless information is filtered using a TextCNN model, which can quickly filter out useless information and help improve the efficiency of system recognition. The model is used to classify and annotate target text information, significantly reducing the manual and time costs of model training. The location of the target fine-grained waterlogging point is encoded and visualized, facilitating timely processing of the waterlogging point and promoting the construction of smart cities.

[0082] Example 3

[0083] Figure 6 This is a flow chart of a method for intelligently identifying water points provided by the third embodiment of the present invention. The embodiment of the present invention further optimizes the above embodiment on the basis of the above embodiment, and the embodiment of the present invention can be combined with various optional solutions in one or more of the above embodiments. Figure 6 As shown, the intelligent water accumulation point identification method provided in the embodiment of the present invention may include the following steps:

[0084] S610: Acquire text information related to rainstorm and waterlogging from social media using crawler technology, and pre-process the text information to obtain target text information.

[0085] S620: Classify the target text information and label the classified text information.

[0086] S630: Construct a language rule-based geographic location extraction algorithm, and use the language rule-based geographic location extraction algorithm to process the waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information.

[0087] Among them, the geographic location extraction algorithm based on language rules can refer to establishing a corresponding semantic location concept hierarchical model by analyzing the geographic description features in natural language, and defining location extraction rules accordingly, including place name extraction rules and spatial relationship extraction rules. The extraction of location information is achieved by identifying place names and spatial relationships. In the embodiment of the present invention, a geographic location extraction algorithm based on language rules is used to process waterlogging information, loss information and geographic location identification results to obtain the first fine-grained waterlogging point location information. Among them, the geographic location identification result is obtained by using a general named entity recognition model to perform named entity recognition of the geographic location of the waterlogging point on the target text information.

[0088] Prior art typically uses a Universal Chinese Named Entity Recognition (UCNER) model trained on a Chinese general corpus to extract flooded locations from waterlogging and damage information. However, the place names identified by the UCNER model are incomplete, imprecise, and noisy, making them inadequate for direct visualization of flooded areas. To address these issues, embodiments of the present invention propose a Water Location Refinement (GLR) algorithm to improve UCNER's recognition results and obtain fine-grained flooded areas.

[0089] Among them, see Figure 7 The network architecture of the UCNER model is BERT+BiLSTM+CRF. The model input is a sentence and the output is the label corresponding to each word in the sentence. BERT represents a pre-trained language representation model. In the BERT layer, the input text is converted into a semantically rich vector. Each vector contains three kinds of information, namely word embedding (E T ), the sentence embedding of the sentence where the word is located (E S ) and position embedding (E P ). Subsequently, the embedding layer is input into the Transformer encoder model. During the training process, the model's attention mechanism adjusts its attention to individual words in the context through weight changes; therefore, word vectors are generated through deep bidirectional encoding. BiLSTM is a bidirectional long short-term memory network. The BiLSTM layer extracts contextual and semantic features from these vectors to achieve semantic encoding. This layer consists of a forward LSTM and a backward LSTM, and can obtain contextual information by capturing forward and backward information in the sentence sequence. Conditional Random Field (CRF) is a commonly used sequence labeling model. The CRF layer assigns the label with the highest probability to each character in the sentence and determines the best category after obtaining the latent vector output by the BiLSTM layer.

[0090] The embodiment of the present invention selects the People's Daily corpus (PFR) to train the NER model. PFR is a unique named entity recognition dataset, which is annotated using the BIO system. In PFR, there are three different types of named entities, namely people (PER), places (LOC), and organizations (ORG). However, for location tags, PFR only provides the "LOC" tag, and does not provide more fine-grained location tags such as "street", "region" or "building". In addition, most of the locations in PFR are broad areas, such as continents, countries, cities and regions. The training effect of PFR on the named entity recognition model is shown in Table 1. The PFR-NER model predicts the largest number of LOC tags, and the recall rate can reach 96.05%, indicating that the model can recognize the vast majority of place names in the text, and the recognition effect meets the requirements of the present invention.

[0091] Table 1 PFR training effect on named entity recognition model

[0092]

[0093] The evaluation criteria for geographic named entity recognition (GNE) methods are based on precision (P), recall (R), and the F1 value, which serve as evaluation metrics for recognition models. These metrics are also fundamental indicators for evaluating models in fields such as information search, knowledge classification, and entity recognition. Precision (P) represents the accuracy of the predicted GNE results. Recall (R) represents the probability that a given GNE in the original data is accurately identified as a GNE name.

[0094] Alternatively, 1,167 pieces of flooding and loss information were fed into the trained PFR-NER model to extract the place names contained in each target text message. Of these, 999 target text messages contained place names, accounting for 85.60% of the target text messages. However, most of the recognized place names were not granular enough.

[0095] To solve the problems of UCNER being unable to recognize special positioning words and complete complex fine-grained water accumulation points, and outputting descriptions of the same place with different degrees of accuracy as multiple addresses at the same time, see Figure 8 ,This paper proposes a geographic location refinement (GLR) algorithm, which includes two sub-algorithms, namely the rule-based ,geolocation extraction algorithm (RGLE) and the category label ,based geolocation filtering algorithm (CGLF).

[0096] As an optional but non-limiting implementation, the language rule-based geographic location extraction algorithm is used to process the waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information, including but not limited to steps C1-C6:

[0097] Step C1: Segment the waterlogging information and loss information, and divide different geographical locations into different segments.

[0098] Step C2: De-noising the geographic location recognition result to obtain a de-noised geographic location recognition result.

[0099] Step C3: Detect the segment to determine whether the segment contains a target recognition result; the target recognition result is a denoised geographic location recognition result.

[0100] Step C4: If the segment contains a target recognition result, the target recognition result is used as the third fine-grained water accumulation point location information.

[0101] Step C5: If the segment contains at least two target recognition results, determine whether the at least two target recognition results belong to the same geographical location based on the distance between the two adjacent target recognition results, and combine the at least two target recognition results belonging to the same geographical location to obtain the fourth fine-grained water accumulation point location information.

[0102] Step C6: Determine whether the waterlogging information and loss information contain special locating words, and determine whether the special locating words are connected to the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information. If they are connected, combine the special locating words with the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information to obtain the first fine-grained waterlogging point location information.

[0103] The rule-based location extraction algorithm takes as input the flooding and loss information and their corresponding UCNER location recognition results, and outputs the first-level fine-grained flooding point location information. The specific steps include the following three steps:

[0104] In the first step, the input data is preprocessed, specifically including segmenting the waterlogging information and loss information and denoising the geographic location recognition results. Segmenting the waterlogging information and loss information can simplify the difficulty of distinguishing complex geographic locations. In Chinese expressions, different geographic locations are usually separated by punctuation marks such as ','; ','. Therefore, each target text information is divided into different segments according to the above punctuation marks. In this way, for sentences that mention multiple locations at the same time, different geographic locations can be correctly divided into different segments. Next, in order to eliminate the interference of repeated place names in the geographic location recognition results, the geographic location recognition results corresponding to each waterlogging information and loss information are cleaned of noise vocabulary, that is, the geographic location recognition results are denoised. Through the operations in the initialization stage, each segment of waterlogging information and loss information corresponds to the denoised geographic location recognition result, that is, the target recognition result.

[0105] The second step is to include the target recognition result (ner s ) to extract the geographic location from the fragment. Check each fragment in turn to see if it contains the target recognition result. If it does, process it in two different cases. If there is only one target recognition result corresponding to the fragment, the corresponding target recognition result is output as the third fine-grained water point location information. If the target recognition result contains m ner i , that is, ner s =[ner1,ner2,…,ner m ], the program will further determine whether multiple target recognition results belong to the same geographical location. The judgment standard is adjacent ner i The distance between characters is based on the adjacent ner i The character distance between them determines whether multiple target recognition results belong to the same geographical location. Adjacent target recognition results belonging to the same geographical location are combined to obtain the fourth fine-grained water accumulation point location information.

[0106] In the third step, the algorithm checks whether any special locating terms are missing from the flooding and loss information. Special locating terms not identified by UCNER are stored in a list. These include, but are not limited to, intersections, road sections, junctions, crossroads, junctions, directions, ramps, bridge tunnels, bridge bottoms, and entrances and exits. Chinese expressions typically place special locating terms at the end of the geographic location. Using regular expressions, the algorithm checks whether the third- or fourth-granularity flooding point location information is connected to the special locating terms. If so, they are combined to obtain the first-granularity flooding point location information.

[0107] As an optional but non-limiting implementation, if the segment contains at least two target recognition results, determining whether the at least two target recognition results belong to the same geographic location based on the distance between the two adjacent target recognition results, and combining the at least two target recognition results belonging to the same geographic location to obtain fourth fine-grained waterlogging point location information includes but is not limited to steps D1-D2:

[0108] Step D1: Determine whether the first distance between the first target recognition result and the second target recognition result is less than or equal to a preset distance threshold, and combine the first target recognition result and the second target recognition result when the first distance is less than or equal to the preset distance threshold, and output the combined fine-grained geographic location.

[0109] Step D2: Determine whether the second distance between the combined fine-grained geographic location and the third target recognition result is less than or equal to a preset distance threshold. If the second distance is greater than the preset distance threshold, the combined fine-grained geographic location and the third target recognition result are used as the fourth fine-grained water accumulation point location information.

[0110] Among them, the character distance between the target recognition results is used to determine whether multiple target recognition results belong to the same geographical location. When the first distance between the first target recognition result and the second target recognition result is less than or equal to the preset distance threshold, the two adjacent target recognition results in the same segment and the characters between them will be merged into a fine-grained geographical location and enter the next round of judgment cycle as a new target recognition result; judge whether the second distance between the new target recognition result and the third target recognition result is less than or equal to the preset distance threshold. If the character distance between the new target recognition result and the third target recognition result is greater than the preset distance threshold, the content of the target recognition result remains unchanged. When ner m Enter the loop and judge the end, and the process ends. In order to distinguish the target recognition result before processing from the target recognition result before processing, the target recognition result after processing can be assigned a value.

[0111] The embodiment of the present invention can combine place names belonging to the same geographical location into a complex fine-grained geographical location through a geographical location extraction algorithm based on language rules, thereby ensuring the accuracy of the complex fine-grained geographical location output result.

[0112] S640. Construct a geographic location filtering algorithm based on category labels, and use the geographic location filtering algorithm based on category labels to prioritize the first fine-grained waterlogging point location information to obtain the second fine-grained waterlogging point location information, and use the second fine-grained waterlogging point location information as the target fine-grained waterlogging point location information.

[0113] The category label-based geolocation filtering algorithm is a method for optimizing geolocation data. It combines geolocation information with category labels to filter and optimize data. The linguistic rule-based geolocation extraction algorithm can complete the locations extracted by UCNER into waterlogged areas, while the category label-based geolocation filtering algorithm is used to determine which waterlogged areas should be output.

[0114] Optionally, for target text information without category labels, it is necessary to extract fine-grained geographic location from the target text information; for target text information with category labels, the corresponding category labels can be directly matched.

[0115] As an optional but non-limiting implementation, the method of using a category label-based geographic location filtering algorithm to prioritize the first fine-grained waterlogging point location information to obtain the second fine-grained waterlogging point location information includes but is not limited to steps E1-E3:

[0116] Step E1: Construct a priority list representing the fine-grained level of water accumulation point locations.

[0117] Step E2: Use geographic location category labels to label the first fine-grained waterlogging point location information, and assign the labeled first fine-grained waterlogging point location information to a corresponding priority list.

[0118] Step E3: Retrieve the first fine-grained waterlogging point location information in the priority list according to priority, and select the first fine-grained waterlogging point location information that exists at the highest priority position from the first priority, second priority, and third priority as the second fine-grained waterlogging point location information.

[0119] Among them, the geographic location category labels include roads, intersections, points of interest, communities, streets, districts and cities; roads, intersections and points of interest are the first priority, communities are the second priority, streets are the third priority, districts are the third priority, and cities are the fifth priority.

[0120] Among them, see Figure 9 , the embodiment of the present invention uses 7 geographic location category labels to represent the accuracy level of waterlogging points, and the geographic place name database and Baidu map location search API are used to provide category labels. Among them, the geographic place name database can only provide location category labels for communities, streets, districts and cities, and the Baidu map location search API can provide category labels at the road, intersection and point of interest (POI) levels. In the present invention, district-level or city-level locations have a lower priority because they are too rough and cannot make effective contributions to risk warning or emergency management. Therefore, these low-priority waterlogging locations will be ignored. Although communities and subdistricts represent regions, they can still provide valuable information for citizens in a small range. Category labels such as roads, intersections and POIs can provide more refined geographic locations for waterlogging points, so they have the highest priority.

[0121] The category label-based geolocation filtering algorithm takes as input the first-level fine-grained waterlogging location information for each target text and outputs the normalized second-level fine-grained waterlogging location information and its corresponding geolocation category label. This algorithm is divided into three main tasks: data preparation, priority matching, and sequential output, which can be expressed as follows:

[0122] In the data preparation task, it is necessary to create a new list to save the results and the specific category labels of the geographic locations. First, create five lists, namely the first priority list, the second priority list, the third priority list, the fourth priority list, and the fifth priority list, to store the output results of different priorities. Secondly, search the geographic location of the waterlogging points in the geographic place name database and the PlaceAPI in turn, and obtain their corresponding geographic information (the second fine-grained waterlogging point location information, geographic location category labels, and longitude and latitude). The results matched by the geographic place name database are saved in the priority list, and the results matched by the PlaceAPI are uniformly saved in a folder, where each geographic location corresponds to a Json format file, and the file is named after the geographic location.

[0123] During the priority matching task, the program prioritizes the search within the priority list, storing the second-fine-grained waterlogging location information and its location category label corresponding to category tags with priority levels 2, 3, 4, and 5 in the corresponding priority list. For locations not found in the priority list, the program automatically searches the Json files in the folder. If information is found, if the location's category label is "road" or "intersection," the corresponding second-fine-grained waterlogging location information and location category label are saved. If it is not, the corresponding second-fine-grained waterlogging location information is saved with the category "POI." This is because the folder does not specify a location as a "POI," but rather a more specific category, such as a toll booth or city park. Finally, the information stored with the three location category labels is stored in the first priority list. At this point, all five priority lists contain locations with the corresponding category labels.

[0124] In the sequential output task, each target text message is used as a judgment unit. The priority lists are searched in descending order to see if they are empty. If the current priority list is not empty, all the information in the current priority list is output. If it is empty, the next priority list is checked. If the third priority list is empty, the output result for the target text message is empty, indicating that the target text message does not contain fine-grained geographic location.

[0125] Using the optional GLR algorithm, fine-grained geolocations were identified in 425 of the 999 target text messages containing place names, accounting for 42.54% of all target text messages containing place names. A random sample of 100 target text messages contained 156 fine-grained geolocations. The PFR-NER model correctly identified 108 fine-grained geolocations, with an accuracy rate of 69.23%. Adding the GLR algorithm to the dataset resulted in 30 additional correct geolocations, increasing the accuracy rate to 88.46%, a 27.78% improvement.

[0126] S650: Determine the longitude and latitude information corresponding to the target fine-grained water accumulation point location information, and encode and visualize the target fine-grained water accumulation point location based on the longitude and latitude information.

[0127] The embodiment of the present invention provides an intelligent waterlogging point identification method. By constructing a linguistic rule-based geographic location extraction algorithm and using the linguistic rule-based geographic location extraction algorithm to process waterlogging information, loss information, and geographic location identification results, a first fine-grained waterlogging point location information is obtained. The geographic location identification result is obtained by using a general named entity recognition model to perform waterlogging point geographic location named entity recognition on target text information. A category label-based geographic location filtering algorithm is constructed and used to prioritize the first fine-grained waterlogging point location information to obtain a second fine-grained waterlogging point location information. The second fine-grained waterlogging point location information is used as the target fine-grained waterlogging point location information. The technical solution of the embodiment of the present invention can accurately locate the geographic location of the identified waterlogging point to the road, intersection, or POI, and efficiently and cost-effectively mine fine-grained waterlogging points from social media text, greatly improving the utilization value of social media big data. The full-process technology from data collection to fine-grained waterlogging point visualization is low-cost and fast, which helps promote the construction of smart cities.

[0128] Example 4

[0129] Figure 10 This is a schematic diagram of the structure of a device for intelligently identifying water accumulation points provided by the fourth embodiment of the present invention. The technical solution of this embodiment is applicable to the case of intelligent identification of water accumulation points. The device can be implemented by software and / or hardware and is generally integrated into any electronic device with network communication function, including but not limited to: servers, computers, personal digital assistants and other devices. Figure 10As shown, the intelligent water accumulation point identification device provided in this embodiment may include: a text information acquisition and preprocessing module 1010, a text information classification and labeling module 1020, a fine-grained water accumulation point location information extraction module 1030 and a fine-grained water accumulation point location information visualization module 1040; wherein,

[0130] The text information acquisition and preprocessing module 1010 is used to acquire text information related to rainstorm waterlogging from social media using crawler technology and preprocess the text information to obtain target text information; the preprocessing includes text sentence segmentation and keyword reverse screening;

[0131] The text information classification and labeling module 1020 is used to classify the target text information and label the classified text information by category; wherein the classified text information includes waterlogging information and loss information;

[0132] The fine-grained waterlogging point location information extraction module 1030 is used to construct a fine-grained geographic location automatic extraction algorithm and obtain target fine-grained waterlogging point location information from the waterlogging information and loss information according to the fine-grained geographic location automatic extraction algorithm;

[0133] The fine-grained waterlogging point location information visualization module 1040 is configured to determine the longitude and latitude information corresponding to the target fine-grained waterlogging point location information, and to encode and visualize the target fine-grained waterlogging point location based on the longitude and latitude information.

[0134] Based on the above embodiment, optionally, the text information acquisition and preprocessing module is specifically used to:

[0135] Using web crawler technology based on simulated login, text information related to flooding points caused by heavy rain is obtained based on preset social media types, time ranges, and keywords;

[0136] Performing data cleaning and useless information filtering on the text information, eliminating information irrelevant to the rainstorm waterlogging point, so as to obtain the target text information;

[0137] Among them, data cleaning includes sentence-level cleaning and symbol-level cleaning. Sentence-level cleaning includes keyword reverse screening, text sentence segmentation, waterlogging-related keyword screening, and text similarity deduplication processing.

[0138] Based on the above embodiment, optionally, the text information classification and annotation module is specifically used to:

[0139] Analyze the target text information to determine the subject content described by the target text information;

[0140] Classifying the target text information into waterlogging information, loss information, and other information according to the subject content, and labeling the classified text information by category;

[0141] The loss information is used to describe the loss information of secondary disasters caused by heavy rain and waterlogging, and the other information is used to describe the views on the disasters brought about by heavy rain and waterlogging.

[0142] Based on the above embodiment, optionally, the fine-grained water accumulation point location information extraction module is specifically used to:

[0143] A geographic location extraction algorithm construction unit is configured to construct a language rule-based geographic location extraction algorithm, and use the language rule-based geographic location extraction algorithm to process the waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information; wherein the geographic location identification results are obtained by performing named entity recognition of the geographic location of the waterlogging point on the target text information using a general named entity recognition model;

[0144] A geographic location filtering algorithm construction unit is used to construct a geographic location filtering algorithm based on category labels, and use the geographic location filtering algorithm based on category labels to prioritize the first fine-grained waterlogging point location information to obtain the second fine-grained waterlogging point location information, and use the second fine-grained waterlogging point location information as the target fine-grained waterlogging point location information.

[0145] Based on the above embodiment, optionally, the geographic location extraction algorithm construction unit is specifically configured to:

[0146] Segment the waterlogging and loss information, dividing different geographical locations into different segments;

[0147] De-noising the geographic location recognition result to obtain a denoised geographic location recognition result;

[0148] Detecting the segment to determine whether the segment contains a target recognition result; the target recognition result is a denoised geographic location recognition result;

[0149] If the segment contains a target recognition result, the target recognition result is used as the third fine-grained water accumulation point location information;

[0150] If the segment contains at least two target recognition results, determining whether the at least two target recognition results belong to the same geographical location based on the distance between the two adjacent target recognition results, and combining the at least two target recognition results belonging to the same geographical location to obtain fourth fine-grained water accumulation point location information;

[0151] Determine whether the waterlogging information and loss information contain special locating words, and determine whether the special locating words are connected to the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information. If connected, combine the special locating words with the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information to obtain the first fine-grained waterlogging point location information.

[0152] Based on the above embodiment, optionally, the geographic location extraction algorithm construction unit is further configured to:

[0153] determining whether a first distance between the first target recognition result and the second target recognition result is less than or equal to a preset distance threshold, and combining the first target recognition result and the second target recognition result if the first distance is less than or equal to the preset distance threshold, and outputting a combined fine-grained geographic location;

[0154] Determine whether the second distance between the combined fine-grained geographic location and the third target recognition result is less than or equal to the preset distance threshold. If the second distance is greater than the preset distance threshold, the combined fine-grained geographic location and the third target recognition result are used as the fourth fine-grained water accumulation point location information.

[0155] Based on the above embodiment, optionally, the geographic location filtering algorithm construction unit is specifically configured to:

[0156] Construct a priority list representing a fine-grained level of water accumulation location;

[0157] Using geographic location category labels to mark the first fine-grained waterlogging point location information, and assigning the marked first fine-grained waterlogging point location information to a corresponding priority list;

[0158] Retrieving the first fine-grained waterlogging point location information in the priority list according to the priority, and selecting the first fine-grained waterlogging point location information that exists at the highest priority position from the first priority, the second priority, and the third priority as the second fine-grained waterlogging point location information;

[0159] Among them, the geographic location category labels include roads, intersections, points of interest, communities, streets, districts and cities; roads, intersections and points of interest are the first priority, communities are the second priority, streets are the third priority, districts are the third priority, and cities are the fifth priority.

[0160] The intelligent water accumulation point identification device provided in the embodiment of the present invention can execute the intelligent water accumulation point identification method provided in any embodiment of the present invention mentioned above, and has the corresponding functions and beneficial effects of executing the intelligent water accumulation point identification method. For detailed process, please refer to the relevant operations of the intelligent water accumulation point identification method in the above embodiment.

[0161] Example 5

[0162] Figure 11 1 is a structural diagram of an electronic device provided in Example 5 of the present invention. The electronic device 10 is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device may also represent various forms of mobile devices, such as personal digital processing, cellular phones, smart phones, wearable devices (such as helmets, glasses, watches, etc.) and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely examples and are not intended to limit the implementation of the present invention described and / or required herein.

[0163] like Figure 11 As shown, the electronic device 10 includes at least one processor 11 and a memory, such as a read-only memory (ROM) 12, a random access memory (RAM) 13, etc., which is communicatively connected to the at least one processor 11. The memory stores a computer program that can be executed by the at least one processor. The processor 11 can perform various appropriate actions and processes according to the computer program stored in the read-only memory (ROM) 12 or the computer program loaded from the storage unit 18 into the random access memory (RAM) 13. Various programs and data required for the operation of the electronic device 10 can also be stored in the RAM 13. The processor 11, ROM 12, and RAM 13 are connected to each other via a bus 14. An input / output (I / O) interface 15 is also connected to the bus 14.

[0164] Multiple components in the electronic device 10 are connected to the I / O interface 15, including an input unit 16, such as a keyboard, a mouse, etc.; an output unit 17, such as various types of displays, speakers, etc.; a storage unit 18, such as a magnetic disk, an optical disk, etc.; and a communication unit 19, such as a network card, a modem, a wireless communication transceiver, etc. The communication unit 19 allows the electronic device 10 to exchange information / data with other devices via a computer network such as the Internet and / or various telecommunication networks.

[0165] Processor 11 can be any general-purpose and / or specialized processing component with processing and computing capabilities. Some examples of processor 11 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various specialized artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any other suitable processor, controller, microcontroller, etc. Processor 11 executes the various methods and processes described above, such as the method for intelligently identifying waterlogging points.

[0166] In some embodiments, the method for intelligently identifying water accumulation points may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 18. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 10 via ROM 12 and / or communication unit 19. When the computer program is loaded into RAM 13 and executed by processor 11, one or more steps of the method for intelligently identifying water accumulation points described above may be performed. Alternatively, in other embodiments, processor 11 may be configured to execute the method for intelligently identifying water accumulation points in any other appropriate manner (e.g., by means of firmware).

[0167] Various embodiments of the systems and techniques described herein can be implemented in digital electronic circuit systems, integrated circuit systems, field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), system-on-chip systems (SOCs), programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments can include being implemented in one or more computer programs that are executable and / or interpreted on a programmable system that includes at least one programmable processor, which can be a special purpose or general purpose programmable processor that can receive data and instructions from a storage system, at least one input device, and at least one output device, and transmit data and instructions to the storage system, the at least one input device, and the at least one output device.

[0168] Computer programs for implementing the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when the computer program is executed by the processor, the functions / operations specified in the flowcharts and / or block diagrams are implemented. The computer program may be executed entirely on the machine, partially on the machine, as a stand-alone software package, partially on the machine and partially on a remote machine, or entirely on a remote machine or server.

[0169] In the context of the present invention, computer-readable storage media can be tangible media that can contain or store a computer program for use with an instruction execution system, device or equipment or used in combination with an instruction execution system, device or equipment. Computer-readable storage media can include but are not limited to electronic, magnetic, optical, electromagnetic, infrared or semiconductor systems, devices or equipment, or any suitable combination of the foregoing. Alternatively, computer-readable storage media can be machine-readable signal media. More specific examples of machine-readable storage media can include electrical connections based on one or more lines, portable computer disks, hard disks, random access memories (RAM), read-only memories (ROM), erasable programmable read-only memories (EPROM or flash memory), optical fibers, portable compact disk read-only memories (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination of the foregoing.

[0170] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user can provide input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, the feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including acoustic input, voice input, or tactile input).

[0171] The systems and techniques described herein can be implemented in a computing system that includes back-end components (e.g., as a data server), or a computing system that includes middleware components (e.g., an application server), or a computing system that includes front-end components (e.g., a user computer with a graphical user interface or web browser through which a user can interact with implementations of the systems and techniques described herein), or a computing system that includes any combination of such back-end components, middleware components, or front-end components. The components of the system can be interconnected by any form or medium of digital data communication (e.g., a communication network). Examples of communication networks include: a local area network (LAN), a wide area network (WAN), a blockchain network, and the Internet.

[0172] A computing system may include clients and servers. The clients and servers are typically remote from each other and typically interact via a communication network. This client-server relationship arises through computer programs running on the respective computers, creating a client-server relationship. The server may be a cloud server, also known as a cloud computing server or cloud host. This server is a hosting product within the cloud computing service ecosystem that addresses the management difficulties and limited scalability of traditional physical hosting and VPS services.

[0173] Example 6

[0174] An embodiment of the present invention also provides a computer program product, including a computer program, which, when executed by a processor, implements the intelligent water accumulation point identification method provided in any embodiment of the present application.

[0175] The computer program product may be implemented by writing computer program code for performing the operations of the present invention in one or more programming languages, or a combination thereof, including object-oriented programming languages ​​such as Java, Smalltalk, C++, and conventional procedural programming languages ​​such as "C" or similar programming languages. The program code may be executed entirely on the user's computer, partially on the user's computer, as a stand-alone software package, partially on the user's computer and partially on a remote computer, or entirely on a remote computer or server. In cases involving a remote computer, the remote computer may be connected to the user's computer via any type of network, including a local area network (LAN) or a wide area network (WAN), or may be connected to an external computer (e.g., via the Internet using an Internet service provider).

[0176] It should be understood that the various forms of the processes shown above can be used to reorder, add, or delete steps. For example, the steps described in the present invention can be performed in parallel, sequentially, or in a different order, as long as the desired results of the technical solution of the present invention can be achieved. This is not limited herein.

[0177] The above specific embodiments do not limit the scope of protection of the present invention. Those skilled in the art will appreciate that various modifications, combinations, sub-combinations, and substitutions may be made based on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of the present invention are intended to be included within the scope of protection of the present invention.

Claims

1. A method for intelligently identifying water accumulation points, characterized in that: The method comprises: Using crawler technology to obtain text information related to rainstorm waterlogging from social media, and preprocessing the text information to obtain target text information; the preprocessing includes text sentence segmentation and keyword reverse screening; Classify the target text information and label the classified text information; the classified text information includes waterlogging information and loss information; A language rule-based geographic location extraction algorithm is constructed, and the language rule-based geographic location extraction algorithm is used to process the waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information; wherein the geographic location identification results are obtained by performing named entity recognition of the geographic location of the waterlogging point on the target text information using a general named entity recognition model; a category label-based geographic location filtering algorithm is constructed, and the category label-based geographic location filtering algorithm is used to prioritize the first fine-grained waterlogging point location information to obtain second fine-grained waterlogging point location information, and the second fine-grained waterlogging point location information is used as the target fine-grained waterlogging point location information; Determining the longitude and latitude information corresponding to the target fine-grained waterlogging point location information, and encoding and visualizing the target fine-grained waterlogging point location based on the longitude and latitude information; The language rule-based geographic location extraction algorithm is used to process the waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information, including: Segment the waterlogging and loss information, dividing different geographical locations into different segments; De-noising the geographic location recognition result to obtain a denoised geographic location recognition result; Detecting the segment to determine whether the segment contains a target recognition result; the target recognition result is a denoised geographic location recognition result; If the segment contains a target recognition result, the target recognition result is used as the third fine-grained water accumulation point location information; If the segment contains at least two target recognition results, determining whether the at least two target recognition results belong to the same geographical location based on the distance between the two adjacent target recognition results, and combining the at least two target recognition results belonging to the same geographical location to obtain fourth fine-grained water accumulation point location information; Determine whether the waterlogging information and loss information contain special locating words, and determine whether the special locating words are connected to the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information. If connected, combine the special locating words with the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information to obtain the first fine-grained waterlogging point location information.

2. The method according to claim 1, characterized in that The crawler technology is used to obtain text information related to rainstorm waterlogging from social media, and the text information is pre-processed to obtain target text information, including: Using web crawler technology based on simulated login, text information related to rainstorm waterlogging is obtained based on preset social media types, time ranges, and keywords; Performing data cleaning and useless information filtering on the text information, eliminating information irrelevant to the rainstorm waterlogging point, so as to obtain the target text information; Among them, data cleaning includes sentence-level cleaning and symbol-level cleaning. Sentence-level cleaning includes keyword reverse screening, text sentence segmentation, waterlogging-related keyword screening, and text similarity deduplication processing.

3. The method according to claim 1, characterized in that The process of classifying the target text information and labeling the classified text information by category includes: Analyze the target text information to determine the subject content described by the target text information; Classifying the target text information into waterlogging information, loss information, and other information according to the subject content, and labeling the classified text information by category; The loss information is used to describe the loss information of secondary disasters caused by heavy rain and waterlogging, and the other information is used to describe the views on the disasters brought about by heavy rain and waterlogging.

4. The method according to claim 1, wherein If the segment contains at least two target recognition results, determining whether the at least two target recognition results belong to the same geographic location based on the distance between the two adjacent target recognition results, and combining the at least two target recognition results belonging to the same geographic location to obtain fourth fine-grained waterlogging point location information, including: determining whether a first distance between the first target recognition result and the second target recognition result is less than or equal to a preset distance threshold, and combining the first target recognition result and the second target recognition result if the first distance is less than or equal to the preset distance threshold, and outputting a combined fine-grained geographic location; Determine whether the second distance between the combined fine-grained geographic location and the third target recognition result is less than or equal to the preset distance threshold. If the second distance is greater than the preset distance threshold, the combined fine-grained geographic location and the third target recognition result are used as the fourth fine-grained water accumulation point location information.

5. The method according to claim 1, characterized in that The method of using a category label-based geographic location filtering algorithm to prioritize the first fine-grained waterlogging point location information to obtain the second fine-grained waterlogging point location information includes: Construct a priority list representing a fine-grained level of water accumulation location; Using geographic location category labels to mark the first fine-grained waterlogging point location information, and assigning the marked first fine-grained waterlogging point location information to a corresponding priority list; Retrieving the first fine-grained waterlogging point location information in the priority list according to the priority, and selecting the first fine-grained waterlogging point location information that exists at the highest priority position from the first priority, the second priority, and the third priority as the second fine-grained waterlogging point location information; Among them, the geographic location category labels include roads, intersections, points of interest, communities, streets, districts and cities; roads, intersections and points of interest are the first priority, communities are the second priority, streets are the third priority, districts are the third priority, and cities are the fifth priority.

6. An intelligent water point identification device, characterized in that: The device comprises: A text information acquisition and preprocessing module is used to use crawler technology to obtain text information related to rainstorm waterlogging from social media and preprocess the text information to obtain target text information; the preprocessing includes text sentence segmentation and keyword reverse screening; The text information classification and labeling module is used to classify the target text information and label the classified text information into categories; wherein the classified text information includes water accumulation information and loss information; A fine-grained waterlogging point location information extraction module is used to construct a language rule-based geographic location extraction algorithm, and use the language rule-based geographic location extraction algorithm to process waterlogging information, loss information, and geographic location identification results to obtain first fine-grained waterlogging point location information; wherein the geographic location identification result is obtained by performing waterlogging point geographic location named entity recognition on target text information using a general named entity recognition model; a fine-grained waterlogging point location information second extraction module is used to construct a category label-based geographic location filtering algorithm, and use the category label-based geographic location filtering algorithm to prioritize the first fine-grained waterlogging point location information to obtain second fine-grained waterlogging point location information, and use the second fine-grained waterlogging point location information as the target fine-grained waterlogging point location information; A fine-grained waterlogging point location information visualization module is used to determine the longitude and latitude information corresponding to the target fine-grained waterlogging point location information, and encode and visualize the target fine-grained waterlogging point location based on the longitude and latitude information; The fine-grained waterlogging point location information extraction module is specifically used to: Segment the waterlogging and loss information, dividing different geographical locations into different segments; De-noising the geographic location recognition result to obtain a denoised geographic location recognition result; Detecting the segment to determine whether the segment contains a target recognition result; the target recognition result is a denoised geographic location recognition result; If the segment contains a target recognition result, the target recognition result is used as the third fine-grained water accumulation point location information; If the segment contains at least two target recognition results, determining whether the at least two target recognition results belong to the same geographical location based on the distance between the two adjacent target recognition results, and combining the at least two target recognition results belonging to the same geographical location to obtain fourth fine-grained water accumulation point location information; Determine whether the waterlogging information and loss information contain special locating words, and determine whether the special locating words are connected to the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information. If connected, combine the special locating words with the third fine-grained waterlogging point location information or the fourth fine-grained waterlogging point location information to obtain the first fine-grained waterlogging point location information.

7. An electronic device, characterized in that: include: one or more processors; a storage device for storing one or more programs; When the one or more programs are executed by the one or more processors, the one or more processors implement the intelligent water accumulation point identification method described in any one of claims 1-5.

8. A storage medium containing computer-executable instructions, characterized in that: When executed by a computer processor, the computer executable instructions are used to execute the intelligent water accumulation point identification method as described in any one of claims 1 to 5.

Images