Multi-mode linkage emergency response system for modular ambulance

Abstract

The invention relates to a multi-mode linkage modular ambulance emergency response system, and relates to the field of intelligent control technology, comprising: evaluating the urgency of each segmented data block to obtain an evaluation result; sorting the data blocks according to the evaluation result, and extracting information related to the rescue type and the location of the incident from the sorted data blocks; using named entity recognition technology to identify key information such as the location name and address, and combining geographic information data to perform geographic positioning and coordinate conversion on the location of the incident; sorting the data blocks according to the urgency, and sending the key information first; according to the rescue type, an emergency response control module selects a corresponding rescue mode from a predefined rescue mode library, and each rescue mode has a corresponding medical equipment configuration; and according to the location of the incident, a corresponding path is calculated for the modular ambulance.

Description

Technical Field

[0001] The present invention relates to the field of intelligent control technology, and in particular to a multi-mode linkage cabin ambulance emergency response system. Background Art

[0002] When receiving emergency rescue requests, some traditional modular ambulance systems can only process text data in a specific format, and have limited processing capabilities for non-text requests (such as voice, images, etc.). This results in the system being unable to quickly and accurately obtain key information under non-standard rescue requests, thus delaying rescue time.

[0003] Secondly, when processing rescue request text data, some existing systems lack in-depth analysis and utilization of semantic similarities between data blocks. The information in a rescue request usually includes multiple aspects such as the location of the incident, the condition of the injured, and the type of rescue required. These aspects may appear in different ways in the text and have complex relationships with each other. Due to the lack of effective semantic analysis methods, the system may find it difficult to accurately identify and segment this information, which in turn affects the subsequent information extraction and urgency assessment. Summary of the invention

[0004] The technical problem to be solved by the present invention is to overcome the deficiencies of the prior art and provide a multi-mode linkage modular ambulance emergency response system that can achieve rapid and accurate positioning to improve the efficiency and quality of emergency rescue.

[0005] In order to solve the above technical problems, the basic concept of the technical solution adopted by the present invention is:

[0006] In the first aspect, a multi-mode linkage cabin ambulance emergency response system includes:

[0007] A square cabin ambulance, which has a medical cabin inside;

[0008] The emergency response control module receives emergency rescue requests from different channels; for non-text requests, converts them into corresponding text data, and pre-processes the text data to obtain pre-processed data; performs semantic analysis on the pre-processed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured person's condition, and instructions for the type of rescue required; by calculating the semantic similarity between the data blocks, the pre-processed data is divided into relatively independent data blocks; the urgency of each divided data block is evaluated to obtain an evaluation result; according to the evaluation result, the data blocks are sorted, and information related to the rescue type and the location of the incident is extracted from the sorted data blocks; named entity recognition technology is used to identify key information such as the location name and address, and combined with geographic information data, the location of the incident is geographically located and coordinates are converted; the data blocks are sorted according to the urgency, and key information is sent first; according to the rescue type, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library, and each rescue mode has a corresponding medical equipment configuration; according to the location of the incident, a corresponding path is calculated for the cabin ambulance;

[0009] The multi-mode linkage module receives instructions from the emergency response control module, which include the selected rescue mode and the planned path. According to the instructions, the multi-mode linkage module controls the medical equipment in the modular ambulance to perform linkage operations so that the ambulance driver can navigate to the scene of the incident according to the planned path.

[0010] Furthermore, semantic analysis is performed on the preprocessed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured person's condition, and instructions for the type of rescue required, including:

[0011] Receive pre-processed text data, use natural language processing technology to segment the text, remove stop words, and tag parts of speech; extract keywords, phrases or sentences as features;

[0012] Use named entity recognition technology to identify place names, building names, and street name entities in text;

[0013] Classify the identified place name entities into incident location data blocks through rule matching;

[0014] Search for descriptions related to the injured person's condition through keyword matching, use sentiment analysis technology to determine the sentiment expressed in the text, build description rules for the injured person's condition, and classify the matched text into the injured person's condition data block;

[0015] Through keyword matching, we search for words related to the rescue type, use context information to determine the type of rescue needs mentioned in the text, build a rescue type classifier, and classify the identified relevant information into the required rescue type data block.

[0016] Furthermore, by calculating the semantic similarity between data blocks, the preprocessed data is divided into relatively independent data blocks, including:

[0017] For each sentence in the preprocessed data, tokenize it into individual words;

[0018] Convert each word into a vector and store the vector representation of each sentence in a matrix, where each row corresponds to the vector of a sentence;

[0019] Create an empty similarity matrix whose size is equal to the square of the number of sentences to store the similarity scores between sentences, traverse each sentence and use it as the source sentence. For the source sentence, traverse all other sentences as the target sentences, calculate the similarity score between the source sentence and the target sentence, and fill the calculated similarity score into the corresponding position of the similarity matrix to obtain the final similarity matrix;

[0020] The number of groups into which the sentences are to be clustered is determined by analyzing the structure of the final similarity matrix, and K sentences are randomly selected as the initial clustering centers. For each sentence, its similarity with all clustering centers is calculated, and the sentence is assigned to the group where the corresponding clustering center is located; for each group, the average value of all sentence vectors in the group is calculated, and the average value is set as the new clustering center. The data points are assigned to the clustering centers and the clustering centers are updated repeatedly until the preset number of iterations is reached to obtain the clustering results, wherein each group in the clustering results represents a data block containing semantically similar sentences, and a sentence list of each group is output as the segmented data block.

[0021] Furthermore, for each sentence, its similarity with all cluster centers is calculated, including:

[0022] For each sentence, Calculate its similarity with all cluster centers;

[0023] in, Indicates sentence vectors; Indicates Cluster center vectors; is the index, representing each dimension in the vector; This means that each vector has elements; Indicates The first sentence in the vector Component, that is, the sentence vector is The value of the dimension; Indicates The first cluster center vector Quantity; Indicates sentence vectors and The similarity between the cluster center vectors.

[0024] Furthermore, an urgency evaluation is performed on each segmented data block to obtain an evaluation result, including:

[0025] Set a variable total_score to 0 to store the sum of the urgency scores; create an empty hash table keyword_weights to store urgent keywords and their corresponding initial weights; create an empty hash table document_frequency to record the number of documents in which each urgent keyword appears in multiple data blocks.

[0026] Define a global variable total_documents to represent the total number of data blocks in the data set; traverse the list of urgent keywords, take each keyword as the key and its corresponding weight as the value, and store them in the hash table keyword_weights;

[0027] Traverse the entire data set and perform word segmentation on each data block; for each word after word segmentation, if it is an urgent keyword, add 1 to the corresponding count in document_frequency; after the traversal is completed, calculate the inverse document frequency of each urgent keyword, update the keyword_weights hash table, and multiply the weight of each keyword by its inverse document frequency;

[0028] Perform word segmentation on the current data block, split it into individual words or phrases, and traverse the results after word segmentation. For each word, specifically include:

[0029] Look up the word in the hash table keyword_weights. If the word is found and is an urgent keyword, obtain its corresponding weight; calculate the word frequency in the current data block; multiply the word frequency in the current data block by its weight in keyword_weights to obtain the word's contribution to the urgency of the current data block, and add it to total_score; after completing the matching of all keywords, word frequency calculation and weight accumulation, total_score is the urgency score of the data block.

[0030] Furthermore, according to the location of the incident, a corresponding path is calculated for the square cabin ambulance, including:

[0031] Determine the starting point and the end point. The starting point is the current location of the square cabin ambulance, marked as S; the end point is the hospital or emergency treatment center, marked as D;

[0032] The actual road network is abstracted into a graph model, where intersections or key locations are used as graph vertices, road segments are used as edges, and each edge has a weight value, which represents the travel time or distance through the corresponding road segment;

[0033] Define a heuristic function for estimating the cost from any vertex to the end point;

[0034] Create lists, including open lists and closed lists. The open list is used to store vertices to be explored and initially contains the starting point S. The closed list is used to store vertices that have been explored and is initially empty.

[0035] For each vertex, calculate the actual cost from the starting point to the vertex according to the heuristic function;

[0036] Calculate the heuristic cost of each vertex according to the heuristic function, that is, the estimated cost from the vertex to the end point;

[0037] Add the actual cost to the heuristic cost to get the total cost of each vertex, and use the total cost as the basis for selecting the next vertex to explore; select the vertex with the smallest total cost from the open list As the current exploration vertex, if the current exploration vertex is the end point D, then the path has been found and the loop is exited. If the currently explored vertex If it is not the end point, remove it from the open list and add it to the closed list; traverse the current exploration vertex All neighbor vertices n of ; for each neighbor vertex n, if it is already in the closed list, it is skipped; otherwise, the calculation starts from the current exploration vertex The actual cost to the neighbor vertex n is calculated, and the actual cost and total cost of the neighbor vertex n are updated. If the neighbor vertex n is not in the open list, it is added to the open list and the iteration is continued until the end point D is found or the open list is empty. If the end point D is found, backtrack along the predecessor vertex of each vertex from the end point D until returning to the starting point S to form a path from the starting point S to the end point D.

[0038] Furthermore, the calculation formula of the heuristic function is:

[0039] ;

[0040] in, From the current vertex Set out to the next vertex along the path; is the current vertex , the distance between the coordinates (x1, y1) and the end point D, the coordinates (x2, y2); From the current vertex To the next vertex The actual length of the road between is the current vertex To the next vertex The straight-line distance between is the slope angle of the terrain; is the adjustment factor.

[0041] In the second aspect, a multi-mode linkage shelter ambulance emergency response method is applied to the system, and the method includes:

[0042] Receive emergency rescue requests from different channels; for non-text requests, convert them into corresponding text data, and pre-process the text data to obtain pre-processed data;

[0043] Perform semantic analysis on the preprocessed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured's condition, and instructions on the type of rescue required; divide the preprocessed data into relatively independent data blocks by calculating the semantic similarity between the data blocks;

[0044] Evaluate the urgency of each segmented data block to obtain the evaluation result; sort the data blocks according to the evaluation result, and extract information related to the rescue type and the location of the incident from the sorted data blocks; use named entity recognition technology to identify the key information of the location name and address, and combine geographic information data to geolocate and convert the coordinates of the location of the incident; sort the data blocks according to their urgency and send the key information first;

[0045] According to the rescue type, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library. Each rescue mode has a corresponding medical equipment configuration; according to the location of the incident, a corresponding path is calculated for the modular ambulance;

[0046] Receive instructions from the emergency response control module, which include the selected rescue mode and the planned path; according to the instructions, the multi-mode linkage module controls the medical equipment in the modular ambulance to perform linkage operations, so that the ambulance driver can navigate to the scene of the incident according to the planned path.

[0047] After adopting the above technical solution, the present invention has the following beneficial effects compared with the prior art:

[0048] It can receive and process emergency rescue requests from different channels, including non-text requests such as voice, images, etc. By converting them into text data and pre-processing them, it effectively expands the information receiving range of the system and improves the accuracy and efficiency of information processing.

[0049] Through semantic analysis technology, the system can intelligently identify different data blocks describing the location of the incident, the condition of the injured, and the type of rescue required, and calculate the semantic similarity between data blocks to achieve accurate segmentation of pre-processed data. This helps to subsequently assess the urgency and extract information for each data block, further improving the level of intelligent data processing.

[0050] Combining named entity recognition technology and geographic information data, the system can accurately identify location names, key address information, and quickly perform geographic positioning and coordinate conversion. This greatly improves the positioning accuracy of the incident site and provides strong support for rescue route planning and navigation.

[0051] According to the rescue type, the corresponding rescue mode is selected from the predefined rescue mode library, and each rescue mode is ensured to have corresponding medical equipment configuration. This flexible selection mechanism enables the system to provide the most suitable rescue plan according to different situations, improving the pertinence and effectiveness of rescue.

[0052] The multi-mode linkage module can receive instructions from the emergency response control module and control the medical equipment in the modular ambulance to perform linkage operations according to the instructions. This not only simplifies the rescue process, but also improves the automation and coordination of the rescue process, which helps to improve the overall rescue efficiency. BRIEF DESCRIPTION OF THE DRAWINGS

[0053] The drawings described herein are used to provide a further understanding of the present application and constitute a part of the present application. The illustrative embodiments of the present application and their descriptions are used to explain the present application and do not constitute an improper limitation on the present application. Some specific embodiments of the present application will be described in detail in an illustrative and non-restrictive manner with reference to the drawings. The same reference numerals in the drawings indicate the same or similar components or parts. It should be understood by those skilled in the art that these drawings are not necessarily drawn to scale. In the drawings:

[0054] Figure 1 It is a schematic diagram of the multi-mode linkage cabin ambulance emergency response system of the present invention.

[0055] Figure 2 It is a schematic flow chart of the emergency response method of the multi-mode linkage modular ambulance of the present invention. DETAILED DESCRIPTION

[0056] In order to enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below in conjunction with the drawings in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in the field without creative work should fall within the scope of protection of the present application.

[0057] The following embodiment of the present application takes the multi-mode linkage modular ambulance emergency response system as an example to explain the solution of the present application in detail, but this embodiment cannot limit the scope of protection of the present application.

[0058] like Figure 1 As shown, the present invention provides a multi-mode linkage cabin ambulance emergency response system, including:

[0059] A square cabin ambulance, which has a medical cabin inside;

[0060] The emergency response control module receives emergency rescue requests from different channels; for non-text requests, converts them into corresponding text data, and pre-processes the text data to obtain pre-processed data; performs semantic analysis on the pre-processed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured person's condition, and instructions for the type of rescue required; by calculating the semantic similarity between the data blocks, the pre-processed data is divided into relatively independent data blocks; the urgency of each divided data block is evaluated to obtain an evaluation result; according to the evaluation result, the data blocks are sorted, and information related to the rescue type and the location of the incident is extracted from the sorted data blocks; named entity recognition technology is used to identify key information such as the location name and address, and combined with geographic information data, the location of the incident is geographically located and coordinates are converted; the data blocks are sorted according to the urgency, and key information is sent first; according to the rescue type, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library, and each rescue mode has a corresponding medical equipment configuration; according to the location of the incident, a corresponding path is calculated for the cabin ambulance;

[0061] The multi-mode linkage module receives instructions from the emergency response control module, which include the selected rescue mode and the planned path. According to the instructions, the multi-mode linkage module controls the medical equipment in the modular ambulance to perform linkage operations so that the ambulance driver can navigate to the scene of the incident according to the planned path.

[0062] In the embodiment of the present invention, emergency rescue requests from different channels can be received and processed, including non-text requests such as voice, image, etc. By converting them into text data and preprocessing them, the information receiving range of the system is effectively expanded, and the accuracy and efficiency of information processing are improved. Through semantic analysis technology, the system can intelligently identify different data blocks describing the location of the incident, the condition of the injured, and the type of rescue required, and calculate the semantic similarity between the data blocks to achieve accurate segmentation of the preprocessed data. This helps to subsequently evaluate the urgency and extract information for each data block, further improving the intelligent level of data processing. Combined with named entity recognition technology and geographic information data, the system can accurately identify the location name, address key information, and quickly perform geographic positioning and coordinate conversion. This greatly improves the positioning accuracy of the location of the incident and provides strong support for the planning and navigation of the rescue path. According to the rescue type, the corresponding rescue mode is selected from the predefined rescue mode library, and it is ensured that each rescue mode has a corresponding medical equipment configuration. This flexible selection mechanism enables the system to provide the most suitable rescue plan according to different situations, improving the pertinence and effectiveness of rescue. The multi-mode linkage module can receive instructions from the emergency response control module and control the medical equipment in the modular ambulance to perform linkage operations according to the instructions. This not only simplifies the rescue process, but also improves the automation and coordination of the rescue process, which helps to improve the overall rescue efficiency.

[0063] In another preferred embodiment of the present invention, a medical cabin is provided inside the modular ambulance. The medical cabin is equipped with medical equipment and medicines to provide emergency medical assistance to patients on site or during transportation. The design of the medical cabin takes into account the efficiency of space utilization and the convenience of medical operations, ensuring that medical staff can quickly and effectively perform treatment work in a limited space.

[0064] The emergency response control module is the core part of the emergency response system of the modular ambulance, responsible for receiving and processing emergency rescue requests from different channels. The following is the specific implementation process of the module processing non-text requests and converting them into text data:

[0065] The emergency response control module first receives emergency rescue requests from different channels, which may include phone calls, text messages, social media, dedicated rescue applications, etc. The requests can be in the form of voice, image, video, or other non-text formats. For non-text requests, such as voice or image, the emergency response control module will use corresponding technologies to convert them into text data. For example, for voice requests, speech recognition technology can be used to convert them into text; for image requests, image recognition and character recognition technologies need to be combined to extract key information and convert it into text. The converted text data will be preprocessed to improve the accuracy of subsequent analysis. The preprocessing steps may include removing noise, standardizing the text (such as unifying case, removing punctuation marks, etc.), word segmentation, removing stop words (such as common words like "of", "is", etc.), and stemming or lemmatization. The preprocessed text data will be output as preprocessed data for more in-depth analysis and processing in subsequent steps. These data will contain clearer and more standardized information, facilitating operations such as semantic analysis, information extraction, and emergency level assessment by the system.

[0066] In a preferred embodiment of the present invention, semantic analysis is performed on the preprocessed data to identify different data blocks. The data blocks are descriptions of the incident location, descriptions of the injured person's condition, and instructions on the type of rescue required, including:

[0067] Receive the preprocessed text data, perform word segmentation, stop word removal, and part-of-speech tagging on the text using natural language processing techniques; extract keywords, phrases, or sentences as features, specifically including:

[0068] The emergency response control module receives the preprocessed text data through a predetermined data interface or message queue; ensures that the received data format is correct and performs necessary data cleaning and verification to remove possible transmission errors or data corruption; uses word segmentation tools (such as jieba segmentation, HanLP, etc.) to perform word segmentation on Chinese text, splitting the continuous text into independent words; filters out stop words in the word segmentation results according to the stop word list (containing commonly used but not very helpful words for semantic analysis, such as "of", "is", etc.); uses a part-of-speech tagging tool to assign a part-of-speech tag (such as noun, verb, adjective, etc.) to each word; extracts keywords in the text based on methods such as word frequency statistics, TF-IDF (term frequency-inverse document frequency), uses phrase extraction algorithms (such as rule-based methods, n-gram models, etc.) to extract meaningful phrases from the text, and selects key sentences as features according to the importance or information content of the sentences, which can be achieved through methods such as sentence-level TF-IDF, TextRank, etc.

[0069] Named entity recognition technology is used to identify place names, building names, and street name entities in the text, including: using a pre-trained named entity recognition (NER) model, such as the deep learning-based BiLSTM-CRF model or the Transformer model (such as BERT+CRF), to perform entity recognition on the text; the NER model will identify entities such as place names, building names, street names, etc. in the text, and mark them out, and based on the marking results, extract these entity names and their corresponding categories.

[0070] The identified place name entities are classified as incident location data blocks through rule matching, specifically including: creating a list of keywords that may indicate the location of the incident, such as "site", "incident", "place of occurrence", "location", "area", etc.; if the place name entity appears next to these keywords or in the context of the keywords (for example, in the same sentence or phrase), it is considered as a potential incident location. Define specific scene keywords for different types of emergency situations, such as "fire", "car accident", "collapse", "explosion", etc. When these specific scene keywords appear at the same time as the place name entity, the possibility of the place name entity as the incident location is increased. Define some exclusionary keywords, such as "non-incident location", "nearby", "surrounding", "except", etc. If the place name entity is associated with these exclusionary keywords, the possibility of it being the incident location is reduced or excluded. Utilize structured information in the text, such as the pattern of "[type of event] occurred at [place name]". Extract information that conforms to this structure through regular expressions or pattern matching, where the [place name] part is identified as the incident location. Analyze the text context around the place name entity to find clues that may indicate that the place is the incident location. For example, if the text mentions "rescue teams have rushed to [place name]", then [place name] is likely to be the location of the incident. If available, combine external data sources (such as map APIs) to verify the geographic location information of the place name entity. Enhance the recognition accuracy of the incident location based on the proximity of the geographic location or other relevant features (such as traffic flow, historical event data, etc.).

[0071] Search for descriptions related to the injured person's condition through keyword matching, use sentiment analysis technology to determine the sentiment tendency expressed in the text, construct description rules for the injured person's condition, and classify the matched text into the injured person's condition data block, specifically including: defining a list of keywords related to the injured person's condition, such as "injured", "bleeding", "coma", etc.; using keyword matching technology to search for descriptions related to the injured person's condition in the text; using sentiment analysis technology (such as dictionary-based methods, deep learning models, etc.) to analyze the sentiment tendency of the matched text to determine whether the injured person's condition is positive, negative or neutral; construct description rules for the injured person's condition, and combine keyword matching and sentiment analysis results to classify the text that meets the rules as the injured person's condition data block;

[0072] Through keyword matching, we search for words related to rescue types, use context information to determine the type of rescue needs mentioned in the text, build a rescue type classifier, and classify the identified relevant information into required rescue type data blocks, including: creating an exhaustive list of keywords that are associated with various possible rescue types. For example:

[0073] Fire related: fire, fire intensity, burning, fireworks, fire alarm, etc.

[0074] Car accident related: car accident, car crash, traffic accident, vehicle collision, rollover, etc.

[0075] Heart disease related: heart disease, myocardial infarction, angina pectoris, cardiac arrest, cardiopulmonary resuscitation, etc.

[0076] This list should be continuously expanded and updated according to actual needs to cover more potential rescue scenarios.

[0077] Next, keyword matching techniques (such as regular expression matching, exact matching, or fuzzy matching) are used to search the received text data for items that match the words in the above keyword list. This can help quickly locate text fragments that may contain rescue type information. Once the words that match the keywords are found, the context information needs to be further used to determine the specific type of rescue needs. This includes:

[0078] Analyze the words before and after the matching words to understand the details of the incident. For example, words such as "serious" and "many injured" that may appear before and after "car accident" can provide more clues about the scale of the accident.

[0079] Check sentence structures, such as parallel relationships, causal relationships, etc., to infer the logical relationship and urgency of the event. For example, the sentence structure "the patient has lost consciousness due to a heart attack" clearly points out the urgency and type of rescue. In order to automatically classify the rescue type more accurately, a classifier based on supervised learning methods can be built. The specific steps include:

[0080] Collect and annotate a batch of text data with known rescue types as a training set. This data should cover various rescue scenarios, and each text should have a clear rescue type label. Select Naive Bayes to learn a classification model based on the training data. In this process, feature engineering techniques can be used to extract features that are more helpful for classification, such as n-gram features, TF-IDF features, etc. Use cross-validation and other techniques to evaluate the performance of the model, and make necessary adjustments and optimizations based on the evaluation results. Finally, when new text data is received, the trained classifier can be used to automatically predict the rescue type to which it belongs. The prediction results will be classified as the required rescue type data block.

[0081] In an embodiment of the present invention, the text is finely processed by natural language processing technology, including word segmentation, removal of stop words, part-of-speech tagging, etc., so that keywords, phrases or sentences can be more accurately extracted as features. This helps to more accurately identify different data blocks in subsequent steps and reduce the possibility of misjudgment or omission of information. Named entity recognition technology is used to specifically identify entities such as place names, building names, street names, etc. in the text, so that the location of the incident can be quickly located. This technology is crucial for emergency rescue because it can significantly shorten the time for the rescue team to arrive at the scene, thereby improving rescue efficiency. Through keyword matching and sentiment analysis technology, the system can automatically find descriptions related to the condition of the injured and judge the emotional tendency expressed in the text. This not only helps rescuers to fully understand the specific conditions of the injured, but also provides them with additional information about the emotional state of the injured, so that they can be more careful and thoughtful during the rescue process. Through keyword matching and context information analysis, the system can accurately identify the type of rescue demand mentioned in the text. This helps the rescue team to make the most appropriate rescue plan selection according to the actual situation and ensure the pertinence and effectiveness of the rescue operation. The identified relevant information is classified into data blocks of the location of the incident, the condition of the injured, and the type of rescue required, so that rescue decision-makers can grasp the key information at a glance. This structured information display method greatly improves decision-making efficiency and reduces decision-making errors caused by confusing or missing information.

[0082] In a preferred embodiment of the present invention, the preprocessed data is divided into relatively independent data blocks by calculating the semantic similarity between the data blocks, including:

[0083] For each sentence in the preprocessed data, tokenize it into individual words; convert each word into a vector and store the vector representation of each sentence in a matrix, where each row corresponds to a sentence vector, including:

[0084] For each sentence in the preprocessed data, use a tokenization tool (such as jieba for Chinese, NLTK for English, etc., depending on the language being processed) to split the sentence into individual words; when tokenizing, you can choose whether to remove stop words (such as commonly used words like "的", "是", "在" in Chinese that contribute less to semantics); after tokenization, each sentence will be converted into a list of words; select a pre-trained word vector model (such as Word2Vec, GloVe, FastText, or BERT, etc.), which can convert each word into a vector of a fixed dimension; for each word obtained after tokenization, convert it into a vector by querying the word vector model; for each sentence, combine the vectors of all its words in some form (such as averaging, summing, or encoding through an RNN model, etc.) to obtain a vector representing the semantics of the entire sentence; store the vector representation of each sentence in a matrix, where each row corresponds to the vector of a sentence.

[0085] Create an empty similarity matrix with a size equal to the square of the number of sentences, which is used to store the similarity scores between sentences. Traverse each sentence and use it as the source sentence. For the source sentence, traverse all other sentences as the target sentences. Calculate the similarity score between the source sentence and the target sentences, and fill the calculated similarity score into the corresponding position in the similarity matrix to obtain the final similarity matrix. Specifically, it includes:

[0086] Create an empty similarity matrix with a size equal to the square of the number of sentences (i.e., N×N, where N is the total number of sentences). This matrix will be used to store the similarity scores between sentences. Traverse each sentence and use it as the source sentence. For each source sentence, traverse all other sentences as the target sentences. For each pair of source and target sentences, use a certain similarity metric method (such as cosine similarity, Euclidean distance, etc.) to calculate their similarity score. Fill the calculated similarity score into the corresponding position in the similarity matrix. Specifically, if the source sentence is the i-th row of the matrix and the target sentence is the j-th column of the matrix, then fill the similarity score into the position of the i-th row and j-th column of the matrix. Repeat the above process until the similarity scores between all sentence pairs are calculated and filled into the matrix, thus obtaining the final similarity matrix.

[0087] By analyzing the structure of the final similarity matrix, the number of groups into which the sentences are to be clustered is determined. K sentences are randomly selected as the initial cluster centers. For each sentence, its similarity with all cluster centers is calculated, and the sentence is assigned to the group where the corresponding cluster center is located. For each group, the average value of all sentence vectors in the group is calculated, and the average value is set as the new cluster center. The data points are assigned to the cluster centers and the cluster centers are updated repeatedly until the preset number of iterations is reached to obtain the clustering results. In the clustering results, each group represents a data block, which contains semantically similar sentences. The sentence list of each group is output as the segmented data block, which specifically includes:

[0088] To analyze the structure of the similarity matrix, the optimal number of clustering groups K can be determined by observing the heat map of the matrix or using specific algorithms (such as the Elbow Method, Silhouette Coefficient Analysis, etc.); the selected K value should be able to reflect the natural grouping in the data, that is, to make the similarity within the group high and the similarity between groups low.

[0089] Randomly select K sentences from the dataset and use their vector representations as the initial cluster centers, ensuring that these initial cluster centers are relatively evenly distributed in the data space to avoid the clustering results being affected by the initial selection. For each sentence in the dataset, calculate its similarity with all K cluster centers and assign each sentence to the group with the cluster center that is most similar to it.

[0090] For each cluster group, calculate the average value (or median, depending on the specific situation) of all sentence vectors in the group, set the calculated average value as the new cluster center of the group, repeat the above steps, that is, reallocate the sentences to the nearest cluster center, and update the position of the cluster center. After each iteration, the quality of clustering can be evaluated, for example, using the silhouette coefficient or other clustering evaluation indicators. When the preset number of iterations is reached, or the quality of the clustering results is no longer significantly improved, stop the iteration. After clustering is completed, each cluster group represents a data block containing semantically similar sentences. Output the sentence list of each cluster group as the segmented data block.

[0091] In an embodiment of the present invention, by dividing the data into relatively independent data blocks, the data blocks can be processed in parallel, thereby improving the speed and efficiency of data processing. The clustered data blocks are more similar in semantics, which makes it easier for data analysts to understand and interpret the data. Through clustering, similar data can be clustered together, so that storage and computing resources can be allocated more efficiently. For example, for data blocks of certain specific topics, more resources can be allocated for in-depth analysis. In machine learning tasks, using clustered data as input can improve the training efficiency and accuracy of the model. Because similar data points are classified together, the model can more easily learn the patterns and laws in the data. In search and recommendation systems, clustering can more accurately locate user queries or points of interest to specific data blocks, thereby improving the accuracy of search results and the relevance of recommended content. In the clustering process, isolated points or noise data that are significantly different from most data may be identified and excluded, thereby improving the quality and purity of the data.

[0092] In a preferred embodiment of the present invention, for each sentence, its similarity with all cluster centers is calculated, including:

[0093] For each sentence, Calculate its similarity with all cluster centers;

[0094] in, Indicates sentence vectors; Indicates Cluster center vectors; is the index, representing each dimension in the vector; This means that each vector has elements; Indicates The first sentence in the vector Component, that is, the sentence vector is The value of the dimension; Indicates The first cluster center vector Quantity; Indicates sentence vectors and The similarity between the cluster center vectors.

[0095] In an embodiment of the present invention, cosine similarity is an indicator that measures the degree of similarity between two vectors in direction. In text processing, it is particularly suitable for capturing the semantic similarity between sentences, even if the sentences differ in length or word order. By calculating the similarity between sentences and cluster centers in this way, sentences can be more accurately assigned to cluster groups that are most similar to their semantics, thereby improving the accuracy of clustering. Cosine similarity focuses on the direction of the vector rather than the length, so the influence of individual noise words or outliers that may appear in the sentence will be relatively weakened. This helps to reduce the interference of these unfavorable factors on the results during the clustering process, making the clustering results more robust and reliable. By pre-calculating the sentence vector and the cluster center vector, and repeatedly using these vectors for similarity calculation during the clustering process, it is possible to avoid recalculating the vector representation of the sentence at each iteration. This can significantly reduce the amount of calculation and accelerate the execution speed of the clustering algorithm, especially when processing large-scale data sets. The method does not rely on a specific sentence vector representation method and can be flexibly combined with various existing or customized word vector models and sentence encoding technologies. This provides users with a wider range of choices, allowing them to choose the most appropriate vector representation method based on specific tasks and data characteristics to further improve clustering results. Since cosine similarity is a measure based on vector direction, each group in the clustering result has a relatively clear semantic directionality. This makes it easier for users to understand the semantic theme or concept represented by each group when analyzing the clustering results, improving the interpretability and practicality of the clustering results.

[0096] In a preferred embodiment of the present invention, an urgency evaluation is performed on each segmented data block to obtain an evaluation result, including:

[0097] Set a variable total_score to 0 to store the sum of the urgency scores; create an empty hash table keyword_weights to store urgent keywords and their corresponding initial weights; create an empty hash table document_frequency to record the number of documents in which each urgent keyword appears in multiple data blocks.

[0098] Define a global variable total_documents to represent the total number of data blocks in the data set; traverse the list of urgent keywords, take each keyword as the key and its corresponding weight as the value, and store them in the hash table keyword_weights;

[0099] Traverse the entire data set and perform word segmentation on each data block; for each word after word segmentation, if it is an urgent keyword, add 1 to the corresponding count in document_frequency; after the traversal is completed, calculate the inverse document frequency of each urgent keyword, update the keyword_weights hash table, and multiply the weight of each keyword by its inverse document frequency;

[0100] Perform word segmentation on the current data block, split it into individual words or phrases, and traverse the results after word segmentation. For each word, specifically include:

[0101] Look up the word in the hash table keyword_weights. If the word is found and is an urgent keyword, obtain its corresponding weight; calculate the word frequency in the current data block; multiply the word frequency in the current data block by its weight in keyword_weights to obtain the word's contribution to the urgency of the current data block, and add it to total_score; after completing the matching of all keywords, word frequency calculation and weight accumulation, total_score is the urgency score of the data block.

[0102] In an embodiment of the present invention, by evaluating the urgency of data blocks, data blocks can be prioritized. In scenarios such as emergency response, customer service, or news reporting, this helps to quickly identify and handle the most urgent situations, thereby improving work efficiency and response speed. After understanding the urgency of data blocks, organizations can allocate resources more effectively. For example, in news reports, more resources can be allocated to data blocks reporting emergency events to ensure timely and accurate information delivery. When faced with a large amount of data, it may be difficult for decision makers to quickly determine which information is most critical. Through urgency assessment, decision makers can quickly focus on the most important and urgent information, thereby making faster and more accurate decisions. When monitoring social media, news reports, or other data sources to identify potential risks, urgency assessment can help detect situations that may cause crises early, so as to intervene and manage them in time. In a content recommendation system, based on the interests and preferences of users, combined with the urgency of data blocks, more personalized and timely content recommendations can be provided to users. In an era of information overload, users often find it difficult to filter out truly important content from massive amounts of information. Through urgency assessment, information platforms can provide users with filtered high-quality content, thereby improving user experience and satisfaction. Urgency assessment can be used as part of a machine learning or artificial intelligence system, enabling the system to process and analyze data more intelligently, improving automation and work efficiency.

[0103] According to the evaluation results, the data blocks are sorted, and information related to the rescue type and the location of the incident is extracted from the sorted data blocks; named entity recognition technology is used to identify key information such as the location name and address, and combined with geographic information data, the location of the incident is geographically located and the coordinates are converted; the data blocks are sorted according to the urgency, and key information is sent first; according to the rescue type, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library. Each rescue mode has a corresponding medical equipment configuration, including:

[0104] A list of data blocks that have been evaluated for urgency, each data block contains text content and its corresponding urgency score; the data blocks are sorted in descending order according to the urgency score, ensuring that the most urgent data block is at the top of the list, and the sorted list of data blocks.

[0105] For each data block, natural language processing (NLP) techniques, such as keyword matching, regular expressions, or more advanced entity recognition methods, are used to identify the rescue type (such as fire, earthquake, medical emergency, etc.) and incident location information (such as location name, address, etc.) in the text, and the identified rescue type and incident location information are stored as structured data for subsequent processing.

[0106] Use named entity recognition (NER) technology to specifically identify key information of place names and addresses (such as cities, streets, house numbers, etc.), and combine with geographic information data (such as geocoding services, map APIs, etc.) to convert the identified place names and addresses into specific geographic coordinates (such as longitude and latitude). For fuzzy location information that cannot be directly converted, you can try to use geospatial search or user manual confirmation to assist in positioning; sort the data blocks according to their urgency, determine the priority of information sending, and use appropriate communication protocols (such as HTTP, MQTT, etc.) to send the most urgent data blocks and their related rescue types and incident location information (including geographic coordinates) to the emergency response system or relevant personnel first; the emergency response control module selects a matching rescue mode from the predefined rescue mode library based on the rescue type. These rescue modes may include standard operating procedures, resource allocation plans, and equipment configurations for different emergency situations. According to the selected rescue mode, the corresponding emergency response process is automatically or semi-automatically triggered, including notifying relevant personnel, allocating resources, and preparing necessary medical equipment configurations.

[0107] In a preferred embodiment of the present invention, a corresponding path is calculated for the modular ambulance according to the location of the incident, including:

[0108] Determine the starting point and the end point. The starting point is the current location of the modular ambulance, marked as S; the end point is the hospital or emergency treatment center, marked as D; abstract the actual road network into a graph model, in which intersections or key locations are the vertices of the graph, and road sections are the edges. Each edge has a weight value, representing the travel time or distance through the corresponding road section; define a heuristic function for estimating the cost from any vertex to the end point; create a list, including an open list and a closed list. The open list is used to store vertices to be explored and initially contains the starting point S; the closed list is used to store vertices that have been explored and is initially empty; for each vertex, calculate the actual cost from the starting point to the vertex according to the heuristic function; calculate the heuristic cost of each vertex according to the heuristic function, that is, the estimated cost from the vertex to the end point;

[0109] Add the actual cost to the heuristic cost to get the total cost of each vertex, and use the total cost as the basis for selecting the next vertex to explore; select the vertex with the smallest total cost from the open list As the current exploration vertex, if the current exploration vertex is the end point D, then the path has been found and the loop is exited. If the currently explored vertex If it is not the end point, remove it from the open list and add it to the closed list; traverse the current exploration vertex All neighbor vertices n of ; for each neighbor vertex n, if it is already in the closed list, it is skipped; otherwise, the calculation starts from the current exploration vertex The actual cost to the neighbor vertex n is calculated, and the actual cost and total cost of the neighbor vertex n are updated. If the neighbor vertex n is not in the open list, it is added to the open list and the iteration is continued until the end point D is found or the open list is empty. If the end point D is found, backtrack along the predecessor vertex of each vertex from the end point D until returning to the starting point S to form a path from the starting point S to the end point D.

[0110] In an embodiment of the present invention, by planning the optimal path for the modular ambulance, the driving time to the scene of the incident can be significantly reduced. In emergency rescue scenarios, time saving is directly related to the improvement of rescue efficiency, which helps to save more lives and reduce property losses. Path planning not only takes into account the distance factor, but can also be dynamically adjusted according to actual conditions (such as road conditions, traffic congestion, etc.). This means that rescue resources can be allocated more reasonably to avoid wasting time and resources on congested or inefficient routes. Through the application of graph models and heuristic functions, scientific and quantitative basis can be provided for decision makers. This helps to make fast and accurate decisions in complex and changing emergency situations. Automated path planning reduces the need for manual intervention and improves the intelligence and automation level of the emergency response system. This can not only reduce the burden on operators, but also maintain the efficient operation of the system under insufficient manpower or high-pressure environments. Through accurate path planning and real-time road condition updates, the risk of modular ambulances encountering obstacles during driving can be reduced, thereby improving the reliability and success rate of rescue missions. A rescue system that can respond to emergencies quickly and accurately will enhance the public's trust and satisfaction with rescue services.

[0111] In a preferred embodiment of the present invention, the calculation formula of the heuristic function is:

[0112] ;

[0113] in, From the current vertex Set out to the next vertex along the path; is the current vertex , the distance between the coordinates (x1, y1) and the end point D, the coordinates (x2, y2); From the current vertex To the next vertex The actual length of the road between is the current vertex To the next vertex The straight-line distance between is the slope angle of the terrain; is the adjustment factor.

[0114] In an embodiment of the present invention, the heuristic function comprehensively considers multiple factors, including the straight-line distance from the current vertex to the end point, the actual length of the road, the terrain slope, etc. This enables the function to more accurately estimate the cost from any vertex to the end point, thereby improving the accuracy of path planning. By introducing the terrain slope angle and adjustment coefficient, the function can be flexibly adjusted according to different terrain conditions. For example, in areas with large slopes, the function will appropriately increase the estimated cost to reflect the difficulty of the vehicle driving in these areas. This adaptability makes path planning more in line with actual conditions. The design of the heuristic function is to reduce the search space and guide the algorithm to quickly find the optimal path. By reasonably estimating the cost from the current vertex to the end point, the function can help the algorithm avoid unnecessary exploration during the search process, thereby improving the efficiency of path planning. Since the function takes into account multiple factors, including road length and terrain slope, it can show better robustness when facing complex and changeable road networks. Even when some road information is missing or changed, the function can provide relatively reliable cost estimates to ensure the stability and reliability of path planning. The design of the heuristic function has certain flexibility and scalability. By adjusting the parameters in the formula or introducing other relevant factors, the performance of the function can be further optimized to adapt to path planning tasks under different scenarios and requirements.

[0115] A multi-mode linkage shelter ambulance emergency response method is applied to the system, and the method includes:

[0116] Receive emergency rescue requests from different channels; for non-text requests, convert them into corresponding text data, pre-process the text data to obtain pre-processed data; perform semantic analysis on the pre-processed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured person's condition, and instructions for the type of rescue required; by calculating the semantic similarity between the data blocks, divide the pre-processed data into relatively independent data blocks; evaluate the urgency of each segmented data block to obtain an evaluation result; according to the evaluation result, sort the data blocks, and extract information related to the rescue type and the location of the incident from the sorted data blocks; use named entity recognition technology to identify key information such as the location name and address, and combine geographic information data to geolocate and convert the coordinates of the location of the incident; sort the data blocks according to their urgency and send key information first; according to the rescue type, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library, and each rescue mode has a corresponding medical equipment configuration; according to the location of the incident, calculate a corresponding path for the cabin ambulance;

[0117] Receive instructions from the emergency response control module, which include the selected rescue mode and the planned path; according to the instructions, the multi-mode linkage module controls the medical equipment in the modular ambulance to perform linkage operations, so that the ambulance driver can navigate to the scene of the incident according to the planned path.

[0118] The embodiment of the present invention further provides a computing device, comprising: a processor, a memory storing a computer program, wherein when the computer program is executed by the processor, the method described above is executed. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0119] The embodiment of the present invention also provides a computer-readable storage medium storing instructions, which, when executed on a computer, enable the computer to execute the method described above. All implementations in the above method embodiment are applicable to this embodiment and can achieve the same technical effect.

[0120] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit it. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or replace some or all of the technical features therein with equivalents. However, these modifications or replacements do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. The multi-mode linkage shelter ambulance emergency response system is characterized by: include: A square cabin ambulance, which has a medical cabin inside; Emergency response control module, receiving emergency rescue requests from different channels; For a request in a non-text form, convert it into corresponding text data, and pre-process the text data to obtain pre-processed data; Perform semantic analysis on the preprocessed data to identify different data blocks, which are information describing the location of the incident, descriptions of the injured's condition, and instructions on the type of rescue required; divide the preprocessed data into relatively independent data blocks by calculating the semantic similarity between the data blocks; Evaluate the urgency of each segmented data block to obtain an evaluation result; sort the data blocks according to the evaluation result, and extract information related to the rescue type and the location of the incident from the sorted data blocks; Use named entity recognition technology to identify key information such as location names and addresses, and combine geographic information data to perform geographic positioning and coordinate conversion of the incident location; sort data blocks according to their urgency, and send key information first; according to the type of rescue, the emergency response control module selects the corresponding rescue mode from the predefined rescue mode library, and each rescue mode has a corresponding medical equipment configuration; according to the location of the incident, calculate a corresponding path for the modular ambulance; The multi-mode linkage module receives instructions from the emergency response control module, including the selected rescue mode and the planned path; According to the instructions, the multi-mode linkage module controls the medical equipment in the modular ambulance to perform linkage operations, so that the ambulance driver can navigate to the scene of the incident according to the planned route; According to the location of the incident, a corresponding path is calculated for the square cabin ambulance, including: Determine the starting point and the end point. The starting point is the current location of the square cabin ambulance, marked as S; the end point is the hospital or emergency treatment center, marked as D; The actual road network is abstracted into a graph model, where intersections or key locations are used as graph vertices, road segments are used as edges, and each edge has a weight value, which represents the travel time or distance through the corresponding road segment; Define a heuristic function for estimating the cost from any vertex to the end point; Create lists, including open lists and closed lists. The open list is used to store vertices to be explored and initially contains the starting point S. The closed list is used to store vertices that have been explored and is initially empty. For each vertex, calculate the actual cost from the starting point to the vertex according to the heuristic function; Calculate the heuristic cost of each vertex according to the heuristic function, that is, the estimated cost from the vertex to the end point; Add the actual cost to the heuristic cost to get the total cost of each vertex, and use the total cost as the basis for selecting the next vertex to explore; select the vertex with the smallest total cost from the open list As the current exploration vertex, if the current exploration vertex is the end point D, then the path has been found and the loop is exited. If the currently explored vertex If it is not the end point, remove it from the open list and add it to the closed list; traverse the current exploration vertex All neighbor vertices n of ; for each neighbor vertex n, if it is already in the closed list, it is skipped; otherwise, the calculation starts from the current exploration vertex The actual cost to the neighbor vertex n is updated, and the actual cost and total cost value of the neighbor vertex n are updated. If the neighbor vertex n is not in the open list, it is added to the open list and iterated until the end point D is found or the open list is empty. If the end point D is found, backtrack along the predecessor vertex of each vertex from the end point D until returning to the starting point S, so as to form a path from the starting point S to the end point D; The calculation formula of the heuristic function is: ; in, From the current vertex Set out to the next vertex along the path; is the current vertex , the distance between the coordinates (x1, y1) and the end point D, the coordinates (x2, y2); From the current vertex To the next vertex The actual length of the road between is the current vertex To the next vertex The straight-line distance between is the slope angle of the terrain; is the adjustment factor.

2. The multi-mode linkage modular ambulance emergency response system according to claim 1, characterized in that: Semantic analysis is performed on the preprocessed data to identify different data blocks, which are information describing the location of the incident, the condition of the injured, and the type of rescue required, including: Receive pre-processed text data, use natural language processing technology to segment the text, remove stop words, and tag parts of speech; extract keywords, phrases or sentences as features; Use named entity recognition technology to identify place names, building names, and street name entities in text; Classify the identified place name entities into incident location data blocks through rule matching; Search for descriptions related to the injured person's condition through keyword matching, use sentiment analysis technology to determine the sentiment expressed in the text, build description rules for the injured person's condition, and classify the matched text into the injured person's condition data block; Through keyword matching, we search for words related to the rescue type, use context information to determine the type of rescue needs mentioned in the text, build a rescue type classifier, and classify the identified relevant information into the required rescue type data block.

3. The multi-mode linkage modular ambulance emergency response system according to claim 2, characterized in that: By calculating the semantic similarity between data blocks, the preprocessed data is divided into relatively independent data blocks, including: For each sentence in the preprocessed data, tokenize it into individual words; Convert each word into a vector and store the vector representation of each sentence in a matrix, where each row corresponds to the vector of a sentence; Create an empty similarity matrix whose size is equal to the square of the number of sentences to store the similarity scores between sentences, traverse each sentence and use it as the source sentence. For the source sentence, traverse all other sentences as the target sentences, calculate the similarity score between the source sentence and the target sentence, and fill the calculated similarity score into the corresponding position of the similarity matrix to obtain the final similarity matrix; The number of groups into which the sentences are to be clustered is determined by analyzing the structure of the final similarity matrix, and K sentences are randomly selected as the initial clustering centers. For each sentence, its similarity with all clustering centers is calculated, and the sentence is assigned to the group where the corresponding clustering center is located; for each group, the average value of all sentence vectors in the group is calculated, and the average value is set as the new clustering center. The data points are assigned to the clustering centers and the clustering centers are updated repeatedly until the preset number of iterations is reached to obtain the clustering results, wherein each group in the clustering results represents a data block containing semantically similar sentences, and a sentence list of each group is output as the segmented data block.

4. The multi-mode linkage modular ambulance emergency response system according to claim 3 is characterized in that: For each sentence, calculate its similarity with all cluster centers, including: For each sentence, Calculate its similarity with all cluster centers; in, Indicates sentence vectors; Indicates Cluster center vectors; is the index, representing each dimension in the vector; This means that each vector has elements; Indicates The first sentence in the vector Component, that is, the sentence vector is The value of the dimension; Indicates The first cluster center vector Quantity; Indicates sentence vectors and The similarity between the cluster center vectors.

5. The multi-mode linkage modular ambulance emergency response system according to claim 4 is characterized in that: An urgency assessment is performed on each segmented data block to obtain an assessment result, including: Set a variable total_score to 0 to store the sum of the urgency scores; create an empty hash table keyword_weights to store the urgent keywords and their corresponding initial weights; create an empty hash table document_frequency to record the number of documents in which each urgent keyword appears in multiple data blocks; Define a global variable total_documents to represent the total number of data blocks in the data set; traverse the list of urgent keywords, take each keyword as the key and its corresponding weight as the value, and store them in the hash table keyword_weights; Traverse the entire data set and perform word segmentation on each data block; for each word after word segmentation, if it is an urgent keyword, add 1 to the corresponding count in document_frequency; after the traversal is completed, calculate the inverse document frequency of each urgent keyword, update the keyword_weights hash table, and multiply the weight of each keyword by its inverse document frequency; Perform word segmentation on the current data block, split it into individual words or phrases, and traverse the results after word segmentation. For each word, specifically include: Look up the word in the hash table keyword_weights. If the word is found and is an urgent keyword, obtain its corresponding weight; calculate the word frequency in the current data block; multiply the word frequency in the current data block by its weight in keyword_weights to obtain the word's contribution to the urgency of the current data block, and add it to total_score; after completing the matching of all keywords, word frequency calculation and weight accumulation, total_score is the urgency score of the data block.

Images