Voice interaction-based medical navigation service method and system

Abstract

The application provides an old-age adaptation medical navigation service method and system based on voice interaction, relates to the technical field of intelligent medical treatment, and comprises the following steps: performing semantic analysis and repair on the voice of medical needs of an old user to generate a structured medical demand description; then, mapping a medical knowledge graph, combining cognitive load evaluation and space-time accessibility matching, and screening a candidate medical service scheme; finally, according to multi-dimensional path optimization rules, a final navigation scheme is generated in the internal space topology of a medical institution. The application can effectively understand and complete the fragmented and jumping medical demands of the old people, and provide personalized in-hospital navigation services with low cognitive load and high accessibility.

Description

Technical Field

[0001] This invention relates to the field of smart healthcare technology, and in particular to a method and system for age-friendly medical navigation services based on voice interaction. Background Technology

[0002] This invention relates to the field of smart healthcare and age-friendly services, specifically to an age-friendly medical navigation service method based on voice interaction.

[0003] In existing smart healthcare service systems, guidance for elderly users' medical needs and in-hospital navigation services typically rely on standardized information input interfaces or simplified voice command recognition. The conventional approach is for the system to match the user's voice or text input with a pre-set medical keyword database, directly mapping it to the corresponding department or service. Subsequently, based on the physical location information of the hospital departments, the system plans the shortest spatial distance or least time-consuming route for the user, presenting this route to the user in the form of a two-dimensional map or simple arrow indicators. These systems aim to improve the efficiency of healthcare services and reduce the time users spend finding their way in complex medical environments.

[0004] Elderly users' expressions are often fragmented, discontinuous, and semantically disjointed. Their medical needs are scattered across multiple discourse segments and are frequently accompanied by vague descriptions of their health status. Existing systems' rigid keyword matching mechanisms struggle to effectively parse this complex and incomplete semantic information, easily leading to errors in intent recognition or omission of key needs, thus failing to form accurate and complete descriptions of medical needs. The generated navigation paths typically only consider optimal physical distance, neglecting the physiological and cognitive burden on elderly users. For example, paths may include excessively long continuous walking distances, steep inclines, or a lack of necessary rest stops, placing additional physical strain and anxiety on elderly users. Furthermore, standard route planning fails to consider the complexity of the medical needs themselves (such as cross-disciplinary consultations) in conjunction with the cognitive difficulty of the route, resulting in users being unable to complete subsequent services smoothly even after reaching the designated location due to the complexity of the process. Summary of the Invention

[0005] This invention provides a method and system for age-friendly medical navigation services based on voice interaction, which can solve the problems in the prior art.

[0006] A first aspect of the present invention provides a method for providing age-friendly medical navigation services based on voice interaction, comprising: Semantic analysis is performed on medical demand voice information to extract medical intent type and health status description features. A temporal association graph is constructed based on the fragmentation features and semantic jump features of elderly users’ speech. By identifying breakpoint nodes and isolated nodes in the temporal association graph, missing dimension reasoning and semantic link repair are performed on the medical intent type and health status description features to generate a structured medical demand description. The structured medical demand description is mapped using a medical knowledge graph. The associated department set and cross-diagnosis and treatment path are identified and their complexity is quantified to obtain the path complexity index value. Based on the path complexity index value, service paths are constructed and cognitive load is assessed. A set of service paths with assessment values ​​lower than the preset cognitive load boundary is selected and matched with the service time period information in the medical resource database for spatiotemporal accessibility to obtain a set of candidate medical service solutions. Based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, which includes slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. The slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule are adaptively weighted and assigned. The path is solved for the service nodes in the candidate medical service scheme set by combining the internal spatial topology of the medical institution, and a navigation scheme is generated.

[0007] Semantic analysis of medical demand voice information is performed to extract features related to medical intent type and health status description, including: The medical demand voice information is subjected to acoustic feature extraction and speech recognition conversion to obtain a textualized expression sequence. Dialect feature recognition is performed on the textual expression sequence, and dialect words in the textual expression sequence are converted into standard medical terminology by constructing a dialect word mapping relationship for elderly users, thus obtaining a standardized text sequence; Syntactic structure analysis and semantic role labeling are performed on the standardized text sequence to identify symptom description components, time description components, and medical visit purpose description components in the standardized text sequence. Based on the medical symptom ontology knowledge base, the non-standard symptom expressions of the symptom description components are identified and mapped to standard symptom classifications. The non-standard expressions in the symptom description components are converted into standard symptom classification identifiers, and the time description components are converted into time urgency identifiers. Semantic association analysis is performed between the standard symptom classification identifier and the description component of the purpose of medical visit to extract the medical intent type; The standard symptom classification identifier is combined with the time urgency identifier to extract the health status description features.

[0008] Based on the fragmented and semantic jump features of elderly users' speech, a temporal correlation graph is constructed. By identifying breakpoints and isolated nodes in the temporal correlation graph, missing dimension inference and semantic link repair are performed on the medical intent type and the health status description features to generate a structured medical need description, including: The medical intent type and the health status description features are timestamped to obtain a time-series labeled information sequence; the semantic jump position is identified by calculating the semantic association strength between adjacent information segments, and the time-series labeled information sequence is divided into multiple discourse segment units based on the semantic jump position; The temporal association graph is constructed using the utterance fragment units as nodes and the temporal relationships marked by timestamps as edges; In the temporal correlation graph, the isolated node is identified by identifying the node with an edge degree of zero, and the breakpoint node is identified by identifying the node where the temporal relationship between the preceding and following nodes is interrupted. Semantic content analysis is performed on the isolated nodes and the breakpoint nodes to identify the missing information dimensions in the medical intent type and the health status description features; Based on medical domain knowledge reasoning rules, the missing information dimension is filled in to obtain reasoning completion information; the reasoning completion information is semantically associated with the isolated node and the breakpoint node, and a new semantic link connection is established in the temporal association graph to complete the semantic link repair. By integrating all node information in the repaired temporal correlation graph, the structured medical demand description is generated.

[0009] The structured medical needs description is mapped using a medical knowledge graph to identify related department sets and cross-diagnosis paths, and the path complexity index is obtained by performing complexity quantification, including: The health status description features are extracted from the structured medical needs description. After extracting symptom entities from the health status description features, semantic matching is performed with disease entity nodes in the medical knowledge graph to identify a set of candidate disease entities associated with the health status description features. The department nodes associated with each disease entity node in the candidate disease entity set are extracted along the department affiliation edges in the medical knowledge graph. The extracted department nodes are deduplicated to form the associated department set. For each department node in the associated department set, query the medical knowledge graph for the diagnosis and treatment relationship path between the department node and the disease entity node in the candidate disease entity set; Identify department nodes that are simultaneously associated with multiple disease entity nodes as cross-department nodes, extract all diagnosis and treatment relationship paths from the cross-department nodes to different disease entity nodes, and summarize all diagnosis and treatment relationship paths to form the cross-diagnosis and treatment path; The path complexity index value is obtained by weighting the path length, the number of intermediate nodes, and the number of disease entities involved in each treatment relationship path in the cross-treatment path.

[0010] Based on the path complexity index, service paths are constructed and cognitive load is assessed. A set of service paths with assessment values ​​below a preset cognitive load boundary is selected and matched with service time period information in the medical resource database for spatiotemporal accessibility. This yields a candidate medical service solution set, including: Based on the path complexity index value, the department nodes in the associated department set are prioritized and the corresponding department location information and treatment process information are extracted, and department location node sequence and treatment process node sequence are constructed respectively to form a service path; For each path in the service path, the spatial transfer distance and the number of spatial transfers are calculated based on the sequence of department location nodes, and the number of process steps and the number of process decision branches are calculated based on the sequence of medical process nodes. The spatial transfer distance, the number of spatial transfers, the number of process steps, and the number of process decision branches are weighted and summed to obtain the cognitive load assessment value. The service path set is formed by filtering service paths whose cognitive load assessment values ​​are lower than the preset cognitive load boundary. Extract the service time information of the corresponding department for each service path from the medical resource database; Obtain the current geographical location and expected consultation time of elderly users, calculate the arrival time to the corresponding department of each service path in the service path set based on the current geographical location, determine whether the time difference between the arrival time and the expected consultation time is within the time range of the service time period information, and mark the service path whose time difference is within the time range as a spatiotemporally reachable path. The department location nodes and medical process nodes in the spatiotemporal reachable path are extracted as service nodes. The service nodes, the department information corresponding to the spatiotemporal reachable path, the service time period information, and the cognitive load assessment value are structured and encapsulated to form the candidate medical service solution set.

[0011] Based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, including slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. Adaptive weight allocation is performed on the slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule, including: Extract the time urgency identifier and symptom description features from the structured medical needs description; Obtain the slope information of the path segment and establish a slope load mapping relationship with the preset slope tolerance threshold for the elderly population, generate a slope penalty coefficient, and set a passage constraint weight for the path segment that exceeds the preset slope tolerance threshold for the elderly population to form the slope constraint. The spatial distribution information of rest facility nodes is obtained, the required interval distance of rest nodes is calculated based on a preset continuous walking distance threshold for the elderly population, the density coefficient of rest nodes is calculated according to the required interval distance and the spatial distribution information, and a rest node compensation weight is set for path segments below the preset density threshold to form the rest node distribution constraint. Obtain the location information of emergency aid points, calculate the shortest distance from any location node to the nearest emergency aid point, and set an aid accessibility penalty weight for location nodes whose shortest distance exceeds the preset aid coverage radius to form the emergency aid point coverage constraint. The slope constraint, the rest node distribution constraint, and the emergency aid point coverage constraint are structurally combined to form the multi-dimensional path optimization rule; The weight reduction factor of the slope penalty coefficient is proportional to the time urgency indicator, and the weight amplification factor of the rest node density coefficient is proportional to the severity of symptoms.

[0012] Furthermore, by combining the internal spatial topology of the medical institution, path solving is performed on the service nodes in the candidate medical service solution set to generate a navigation solution, including: Obtain the location coordinates of each service node and the access order constraints between service nodes; Based on the location coordinate information, the set of connecting paths between nodes is queried from the internal spatial topology of the medical institution. For each path in the set of connecting paths, the corresponding constraint cost is calculated based on the slope constraint, rest node distribution constraint and emergency rescue point coverage constraint in the multi-dimensional path optimization rules. The constraint cost is then weighted and summed with the physical distance of the path to obtain the comprehensive path cost. The traversal sequence of service nodes is determined according to the access order constraint relationship. For the path selection between adjacent service nodes in the traversal sequence, the path with the minimum comprehensive path cost is selected from the connected path set as the optimal path segment between nodes. All optimal path segments between nodes are spliced ​​together according to the traversal sequence to form a complete navigation path. For each path segment in the complete navigation path, extract the starting coordinates, ending coordinates, direction of travel, and estimated travel time of the path segment, and generate path segment navigation instructions. The navigation scheme is generated by arranging all navigation instructions for all path segments in the complete navigation path according to the traversal sequence and combining the department information and service time information corresponding to the service nodes.

[0013] A second aspect of the present invention provides an age-friendly medical navigation service system based on voice interaction, comprising: The semantic parsing unit is used to perform semantic parsing on medical demand voice information, extract medical intent type and health status description features, construct a temporal association graph based on the fragmentation features and semantic jump features of elderly users’ speech, identify breakpoint nodes and isolated nodes in the temporal association graph, perform missing dimension reasoning and semantic link repair on the medical intent type and health status description features, and generate a structured medical demand description. The path planning unit is used to map the structured medical demand description to a medical knowledge graph, identify the set of related departments and cross-diagnosis and treatment paths, and perform complexity quantification to obtain a path complexity index value. Based on the path complexity index value, it constructs service paths and performs cognitive load assessment, filters out the set of service paths with assessment values ​​lower than the preset cognitive load boundary, and performs spatiotemporal accessibility matching with service time information in the medical resource database to obtain a set of candidate medical service solutions. The navigation generation unit is used to construct a multi-dimensional path optimization rule that includes slope constraints, rest node distribution constraints, and emergency aid point coverage constraints based on the structured medical demand description. It adaptively assigns weights to the slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule, and solves the path for the service nodes in the candidate medical service scheme set by combining the internal spatial topology of the medical institution, thereby generating a navigation scheme.

[0014] A third aspect of the embodiments of the present invention, An electronic device is provided, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0015] Fourth aspect of the present invention, A computer-readable storage medium is provided, having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0016] The beneficial effects of this application are as follows: This method effectively improves the accuracy and adaptability of medical navigation services for elderly users through voice interaction. By semantically parsing the voice information of medical needs, extracting the characteristics of medical intent type and health status description, and constructing a temporal correlation graph, it can accurately capture the fragmented and semantic jump characteristics commonly found in elderly users' speech. Identifying breakpoints and isolated nodes in the temporal correlation graph, and then performing missing dimension reasoning and semantic link repair, can significantly improve the understanding of users' true and complete medical needs, generate high-quality structured medical need descriptions, and lay a solid foundation for subsequent processing.

[0017] By mapping structured medical needs descriptions to a medical knowledge graph, it is possible to intelligently identify related departments and cross-diagnosis pathways, and objectively assess the cognitive load of different pathways through complex quantification. Based on this, service pathways with low cognitive load are selected and matched with the real-time service hours of medical resources for spatiotemporal accessibility. This ensures that the generated set of candidate medical service solutions not only conforms to medical logic but also fully considers the cognitive limitations and actual travel feasibility of elderly users, thereby improving the feasibility and user acceptance of the solutions.

[0018] Based on the physiological and safety needs of elderly users, multi-dimensional path optimization rules are constructed, integrating key factors such as slope constraints, rest node distribution, and emergency aid point coverage. By adaptively weighting the slope penalty coefficient and rest node density coefficient in the rules, personalized adjustments can be made according to the specific health status descriptions of different users. Combining the precise spatial topology within medical institutions for path solving, the final navigation scheme can effectively avoid physical obstacles and safety risks during the journey, planning more effortless, safer, and more humane in-hospital movement routes for elderly users, comprehensively improving the age-friendliness, safety, and user experience of medical navigation services. Attached Figure Description

[0019] Figure 1 This is a flowchart illustrating the age-friendly medical navigation service method based on voice interaction according to an embodiment of the present invention. Figure 2 This is a schematic diagram of the process for constructing a temporal correlation graph and repairing semantic links. Detailed Implementation

[0020] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0021] The technical solution of the present invention will be described in detail below with reference to specific embodiments. These specific embodiments can be combined with each other, and the same or similar concepts or processes may not be described again in some embodiments.

[0022] Figure 1 This is a flowchart illustrating the age-friendly medical navigation service method based on voice interaction according to an embodiment of the present invention. Figure 1 As shown, the method includes: Semantic analysis is performed on medical demand voice information to extract medical intent type and health status description features. A temporal association graph is constructed based on the fragmentation features and semantic jump features of elderly users’ speech. By identifying breakpoint nodes and isolated nodes in the temporal association graph, missing dimension reasoning and semantic link repair are performed on the medical intent type and health status description features to generate a structured medical demand description. The structured medical demand description is mapped using a medical knowledge graph. The associated department set and cross-diagnosis and treatment path are identified and their complexity is quantified to obtain the path complexity index value. Based on the path complexity index value, service paths are constructed and cognitive load is assessed. A set of service paths with assessment values ​​lower than the preset cognitive load boundary is selected and matched with the service time period information in the medical resource database for spatiotemporal accessibility to obtain a set of candidate medical service solutions. Based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, which includes slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. The slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule are adaptively weighted and assigned. The path is solved for the service nodes in the candidate medical service scheme set by combining the internal spatial topology of the medical institution, and a navigation scheme is generated.

[0023] In one optional implementation, semantic parsing is performed on the medical need voice information to extract medical intent type and health status description features, including: The medical demand voice information is subjected to acoustic feature extraction and speech recognition conversion to obtain a textualized expression sequence. Dialect feature recognition is performed on the textual expression sequence, and dialect words in the textual expression sequence are converted into standard medical terminology by constructing a dialect word mapping relationship for elderly users, thus obtaining a standardized text sequence; Syntactic structure analysis and semantic role labeling are performed on the standardized text sequence to identify symptom description components, time description components, and medical visit purpose description components in the standardized text sequence. Based on the medical symptom ontology knowledge base, the non-standard symptom expressions of the symptom description components are identified and mapped to standard symptom classifications. The non-standard expressions in the symptom description components are converted into standard symptom classification identifiers, and the time description components are converted into time urgency identifiers. Semantic association analysis is performed between the standard symptom classification identifier and the description component of the purpose of medical visit to extract the medical intent type; The standard symptom classification identifier is combined with the time urgency identifier to extract the health status description features.

[0024] Elderly users input their medical needs via voice. The system extracts Mel-frequency cepstral coefficients, fundamental frequency variation curves, and energy distribution characteristics from the collected voice signals. These acoustic features reflect the pronunciation characteristics and speech rate variations of elderly users. Speech recognition is then performed on the extracted acoustic features, adapting to elderly pronunciation characteristics such as vibrato and frequent pauses. A preliminary textual representation sequence is output, which can include dialectal vocabulary, colloquial expressions, and non-standard terminology.

[0025] To address the dialectal features present in the textualized expression sequences, a mapping table was established that corresponds common dialectal words to standard medical terms. The mapping table is divided according to regional characteristics; for example, "dizziness" is mapped to "vertigo," "chest tightness" to "chest tightness," and "stomach ache" to "abdominal pain." By scanning the textualized expression sequences character by character, when a fragment matching a dialectal word in the mapping table is detected, it is automatically replaced with the corresponding standard medical term. Simultaneously, adverbs of degree and time modifiers in the original expressions are preserved to ensure the standardized text sequence maintains the integrity of the original semantics.

[0026] Dependency parsing is performed on standardized text sequences to identify subject-verb-object structures and modification relationships within sentences. Semantic role labeling technology is used to categorize phrases in the text as symptom descriptions, temporal descriptions, or descriptions of the purpose of seeking medical attention. Symptom descriptions include expressions of physical abnormalities such as pain and discomfort; temporal descriptions cover time information such as "started yesterday" and "lasted for three days"; and descriptions of the purpose of seeking medical attention include expressions of intent to see a doctor such as "want to see a doctor" and "need to get medication."

[0027] Symptom descriptions are standardized using a medical symptom ontology knowledge base. This knowledge base stores standard symptom classifications and their corresponding non-standard expressions. For example, "heart racing" is identified as "palpitations," and "insomnia" is identified as "insomnia." The semantic similarity between symptom descriptions and standard symptoms in the knowledge base is calculated, and the standard symptom with the highest similarity is selected as the classification identifier. Each standard symptom classification identifier is accompanied by an ICD code and a severity level.

[0028] After time-descriptive components are identified and normalized using time expressions, they are converted into time urgency indicators. Based on the duration of symptoms, frequency of attacks, and worsening trend, the time urgency is divided into three levels: urgent, moderate, and general. For example, "I just suddenly fainted" corresponds to the urgent level, and "I've had recurring headaches for the past week" corresponds to the moderate level.

[0029] By analyzing the semantic relationship between standard symptom classification identifiers and the descriptive components of the purpose of medical visit, the medical intent type is extracted. When the purpose of medical visit is "want to have a physical examination" and the symptoms are mild, the medical intent type is health check-up; when there are clear symptoms and the purpose of medical visit is "seek medical treatment," the medical intent type is disease diagnosis and treatment. The standard symptom classification identifiers, symptom severity levels, and time urgency identifiers are combined to form a health status description feature vector containing symptom category, severity, and time dimensions. This vector serves as the core input for subsequent medical service matching.

[0030] In one optional implementation, a temporal correlation graph is constructed based on the fragmented and semantic jump features of elderly users' speech. By identifying breakpoints and isolated nodes in the temporal correlation graph, missing dimension inference and semantic link repair are performed on the medical intent type and the health status description features to generate a structured medical need description, including: The medical intent type and the health status description feature are timestamped to obtain a time-series labeled information sequence; The semantic jump position is identified by calculating the semantic association strength between adjacent information segments, and the temporal annotation information sequence is divided into multiple discourse segment units based on the semantic jump position; The temporal association graph is constructed using the utterance fragment units as nodes and the temporal relationships marked by timestamps as edges; In the temporal correlation graph, the isolated node is identified by identifying the node with an edge degree of zero, and the breakpoint node is identified by identifying the node where the temporal relationship between the preceding and following nodes is interrupted. Semantic content analysis is performed on the isolated nodes and the breakpoint nodes to identify the missing information dimensions in the medical intent type and the health status description features; Based on medical domain knowledge reasoning rules, the missing information dimension is filled in to obtain the reasoning completion information; The inference completion information is semantically associated with the isolated node and the breakpoint node, and a new semantic link connection is established in the temporal association graph to complete the semantic link repair. By integrating all node information in the repaired temporal correlation graph, the structured medical demand description is generated.

[0031] like Figure 2 As shown, the method includes: During voice interaction, elderly users often exhibit characteristics such as incoherent speech, omission of key information, and frequent topic switching. To address these characteristics, after semantic analysis of the collected medical request voice information, in-depth processing is needed to address the fragmentation and jumpiness issues. After completing the initial semantic analysis to obtain the medical intent type and health status description features, each parsed information segment is assigned a timestamp accurate to the millisecond level, forming a time-series labeled information sequence. This sequence records each information segment according to the chronological order of the voice input, providing a time reference for subsequent analysis.

[0032] For adjacent information segments, the cosine similarity of semantic vectors is used to calculate their association strength. When the semantic similarity between adjacent segments is lower than a preset threshold (e.g., 0.4), it is determined that a semantic jump has occurred, and this position is marked as a semantic jump position. Based on the identified semantic jump positions, the entire temporal labeled information sequence is divided into several discourse segment units, each containing a set of semantically coherent information.

[0033] Each discourse fragment is mapped to a node in the graph. Node attributes include the semantic content, timestamp, and information type label of the fragment. Directed edges are established between nodes based on the temporal relationship indicated by the timestamps, with the edges pointing to later nodes in time. The edge weights reflect the time interval and semantic continuity. By traversing the graph structure, the in-degree and out-degree of each node are counted. When both the in-degree and out-degree of a node are zero, the node is identified as an isolated node, indicating that the information fragment lacks temporal and semantic connections with other fragments. For nodes with zero in-degree or out-degree but not completely isolated, their predecessor or successor nodes are checked for temporal breaks. If the time interval exceeds three standard deviations of the normal dialogue rhythm, it is determined to be a breakpoint node.

[0034] For identified isolated and breakpoint nodes, their semantic content is extracted and dimensional integrity is checked. Medical needs descriptions should typically include symptom description, time, location, and severity dimensions. By comparing with a standard dimensional framework, the missing information dimensions of the current node are determined. For example, if an isolated node only contains the symptom description of "headache" but lacks information on the time of occurrence and the severity of pain, then the time and severity dimensions are marked as missing dimensions.

[0035] The system applies medical reasoning rules to complete missing dimensions. These rules are built upon a symptom-disease association knowledge base, which includes typical time patterns and severity grading standards for common symptoms. For the missing time dimension, inferences are made based on consultation times or the development patterns of related symptoms mentioned in other known fragments. For the missing severity dimension, a default severity level is set based on medical common sense about symptom types. The reasoning process generates reasoning completion information, which is labeled as the source of the reasoning to distinguish it from the user's original input.

[0036] The semantic association matching of the inferred completion information with isolated and broken nodes is performed to calculate the semantic relevance between the completion information and each node. When the relevance exceeds the association threshold, new edge connections are established in the temporal association graph, connecting the originally isolated or broken nodes to the main semantic link. The repaired graph forms a connected semantic network, with all nodes connected as a whole through direct or indirect edges. The repaired temporal association graph is traversed, and the semantic content of all nodes is extracted according to temporal relationships, integrating them into a structured medical demand description containing complete medical intent type, symptom description, time information, location information, and severity information. This description is organized in key-value pair format, facilitating subsequent medical knowledge graph mapping and service matching processing.

[0037] In one optional implementation, the structured medical needs description is mapped using a medical knowledge graph to identify related department sets and cross-diagnosis paths, and a path complexity index value is obtained by performing complexity quantification, including: The health status description features are extracted from the structured medical needs description. After extracting symptom entities from the health status description features, semantic matching is performed with disease entity nodes in the medical knowledge graph to identify a set of candidate disease entities associated with the health status description features. The department nodes associated with each disease entity node in the candidate disease entity set are extracted along the department affiliation edges in the medical knowledge graph. The extracted department nodes are deduplicated to form the associated department set. For each department node in the associated department set, query the medical knowledge graph for the diagnosis and treatment relationship path between the department node and the disease entity node in the candidate disease entity set; Identify department nodes that are simultaneously associated with multiple disease entity nodes as cross-department nodes, extract all diagnosis and treatment relationship paths from the cross-department nodes to different disease entity nodes, and summarize all diagnosis and treatment relationship paths to form the cross-diagnosis and treatment path; The path complexity index value is obtained by weighting the path length, the number of intermediate nodes, and the number of disease entities involved in each treatment relationship path in the cross-treatment path.

[0038] A conditional random field-based sequence labeling model was used for text segmentation and part-of-speech tagging. For colloquial expressions commonly used by elderly users, such as "very dizzy" and "chest tightness," these were standardized into standard symptom terms using a pre-defined symptom dictionary and fuzzy matching rules. The symptom entity extraction process employed a BiLSTM-CRF neural network model. During the training phase, this model used an annotated corpus containing 100,000 medical records of elderly patients, enabling it to identify core entities and their modifying attributes in symptom descriptions.

[0039] The medical knowledge graph stores over 8,000 disease entity nodes, each containing attributes such as a standard disease name, ICD code, symptom set, and typical signs. Semantic matching employs a word vector-based similarity calculation method. The extracted symptom entities are converted into 300-dimensional vectors, and then cosine similarity is calculated between these vectors and the symptom vectors of disease nodes in the knowledge graph. When the similarity exceeds 0.75, the corresponding disease is included in the candidate disease entity set. To address incomplete symptom descriptions, a symptom combination inference mechanism is introduced. By analyzing the co-occurrence patterns of identified symptoms in the knowledge graph, missing associated symptoms are inferred.

[0040] Departmental affiliation edges are stored in the knowledge graph as "disease-affiliated department" triples. Each node in the candidate disease entity set is traversed, and its directly associated department nodes are retrieved through relation queries. Considering that some diseases involve multiple departments, such as "diabetes" which can be associated with endocrinology and nutrition, all relevant departments are collected during the traversal. A deduplication operation based on department codes is performed on the extracted list of department nodes to ensure that each department appears only once, forming the final set of associated departments.

[0041] The diagnosis-treatment relationship path query is based on a depth-first traversal algorithm using a knowledge graph. Starting from the department node, it expands towards the disease entity node along relationship edges such as "diagnosis-treatment process," "examination items," and "treatment plan," with the search depth limited to no more than 5 layers. Intermediate nodes in the path include test nodes, imaging examination nodes, or consultation nodes. The traversal process records the node sequence and edge type of each complete path, generating a path set.

[0042] Cross-department nodes are identified by constructing a department-disease association matrix. Rows represent department nodes, and columns represent disease entity nodes. A department is marked as 1 when it can diagnose and treat a disease. The number of non-zero elements in each row is counted; if the number is greater than 1, the department is considered a cross-department node. For each cross-department node, the diagnostic and treatment paths to different disease entity nodes are retrieved. For example, if the respiratory medicine department can treat both "pneumonia" and "asthma," then all paths from this department to both disease nodes are extracted.

[0043] The path length is calculated by accumulating along the path edges, with the weight of each edge set according to the actual treatment time. The number of intermediate nodes reflects the complexity of the treatment process, including necessary examination and consultation nodes. The number of disease entities involved represents the uncertainty of the diagnosis. In the specific calculation, the path length weight coefficient is set to 0.4, the intermediate node number weight coefficient is set to 0.35, and the disease entity number weight coefficient is set to 0.25. The quantified values ​​of each path are normalized to the range of 0 to 1; the larger the value, the more complex the path, thus obtaining the path complexity index value.

[0044] In one optional implementation, service paths are constructed based on the path complexity index value, and cognitive load is assessed. A set of service paths with assessment values ​​lower than a preset cognitive load boundary is selected and matched with service time period information in the medical resource database for spatiotemporal accessibility, resulting in a candidate medical service solution set including: Based on the path complexity index value, the department nodes in the associated department set are prioritized and the corresponding department location information and treatment process information are extracted, and department location node sequence and treatment process node sequence are constructed respectively to form a service path; For each path in the service path, the spatial transfer distance and the number of spatial transfers are calculated based on the sequence of department location nodes, and the number of process steps and the number of process decision branches are calculated based on the sequence of medical process nodes. The spatial transfer distance, the number of spatial transfers, the number of process steps, and the number of process decision branches are weighted and summed to obtain the cognitive load assessment value. The service path set is formed by filtering service paths whose cognitive load assessment values ​​are lower than the preset cognitive load boundary. Extract the service time information of the corresponding department for each service path from the medical resource database; Obtain the current geographical location and expected consultation time of elderly users, calculate the arrival time to the corresponding department of each service path in the service path set based on the current geographical location, determine whether the time difference between the arrival time and the expected consultation time is within the time range of the service time period information, and mark the service path whose time difference is within the time range as a spatiotemporally reachable path. The department location nodes and medical process nodes in the spatiotemporal reachable path are extracted as service nodes. The service nodes, the department information corresponding to the spatiotemporal reachable path, the service time period information, and the cognitive load assessment value are structured and encapsulated to form the candidate medical service solution set.

[0045] After obtaining the path complexity index, each department node in the associated department set is sorted in descending order based on this index value. Department nodes with higher index values ​​are assigned lower access priority, and those with lower index values ​​are assigned higher access priority. Floor information, clinic number, and spatial coordinates corresponding to each department node are extracted from the medical resource database, and these spatial attributes are organized into a department location node sequence according to priority. Simultaneously, the operational steps for each department, such as registration, waiting, consultation, examination, and medication dispensing, are extracted, and these steps and their sequential relationships are constructed into a patient flow node sequence. The department location node sequence is then associated and bound with the patient flow node sequence to form a complete service path structure.

[0046] For each service path, the spatial coordinates of adjacent nodes in the department location node sequence are extracted, the Euclidean distance between adjacent nodes is calculated and summed to obtain the spatial transfer distance. The number of spatial transfers is obtained by subtracting 1 from the number of nodes in the department location node sequence. The total number of process nodes in the patient visit flow node sequence is extracted to obtain the number of process steps. Nodes in the patient visit flow node sequence with multiple subsequent optional paths are identified, and the number of these decision branch points is counted to obtain the number of process decision branches. The weighting coefficients for spatial transfer distance, spatial transfer count, process steps, and process decision branches are set to 0.4, 0.3, 0.2, and 0.1 respectively. The cognitive load assessment value of the service path is obtained by multiplying the values ​​of the four dimensions by their corresponding weighting coefficients and summing the results.

[0047] Set the preset cognitive load boundary value to 6.5. Iterate through the cognitive load assessment values ​​of all service paths, filter out service paths with assessment values ​​less than 6.5, and store these paths in the service path set.

[0048] From the service time information table in the medical resource database, query the service start time and service end time of each department involved in the service path set, and extract the lunch break time identifier and special closure identifier of each department.

[0049] The system obtains the latitude and longitude coordinates of the elderly user's current location using a positioning module, and extracts the date and specific time from the user's input of their expected appointment time. Based on the current geographic location coordinates and the spatial coordinates of each department in the service path set, a shortest path algorithm is used to calculate the travel time. The arrival time is obtained by adding the current time to the travel time. The time difference between the arrival time and the expected appointment time is calculated. It is determined whether this time difference is greater than or equal to 0 and less than or equal to 60 minutes, and whether the arrival time falls between the service start time and service end time of the corresponding department, and is not during the lunch break. When all the above conditions are met, the service path is marked as a spatiotemporally reachable path.

[0050] Traverse all spatiotemporally reachable paths, extracting department location nodes and patient flow nodes from each path as service nodes. Encapsulate the node identifier, node type, spatial coordinates, and flow description information of each service node. Associate the department name, department code, and specialist scheduling information corresponding to the spatiotemporally reachable path. Associate the service time period information for this path, including service start time, service end time, and available time window. Associate the cognitive load assessment value for this path. Construct the above information into a structured data object and store it in a candidate medical service solution set.

[0051] In one optional implementation, based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, including slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. Adaptive weight allocation is performed on the slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule, including: Extract the time urgency identifier and symptom description features from the structured medical needs description; Obtain the slope information of the path segment and establish a slope load mapping relationship with the preset slope tolerance threshold for the elderly population, generate a slope penalty coefficient, and set a passage constraint weight for the path segment that exceeds the preset slope tolerance threshold for the elderly population to form the slope constraint. The spatial distribution information of rest facility nodes is obtained, the required interval distance of rest nodes is calculated based on a preset continuous walking distance threshold for the elderly population, the density coefficient of rest nodes is calculated according to the required interval distance and the spatial distribution information, and a rest node compensation weight is set for path segments below the preset density threshold to form the rest node distribution constraint. Obtain the location information of emergency aid points, calculate the shortest distance from any location node to the nearest emergency aid point, and set an aid accessibility penalty weight for location nodes whose shortest distance exceeds the preset aid coverage radius to form the emergency aid point coverage constraint. The slope constraint, the rest node distribution constraint, and the emergency aid point coverage constraint are structurally combined to form the multi-dimensional path optimization rule; The weight reduction factor of the slope penalty coefficient is proportional to the time urgency indicator, and the weight amplification factor of the rest node density coefficient is proportional to the severity of symptoms.

[0052] Feature extraction is performed on the structured medical needs descriptions to locate the time urgency identifier field, which includes preset classification labels such as emergency, general outpatient, and chronic follow-up. At the same time, symptom description features are extracted, including key health indicators such as pain intensity score and degree of mobility limitation.

[0053] Topographic data of the internal path network of medical institutions is acquired, and the slope value is calculated for each path segment. This value is then compared with a preset slope tolerance threshold for the elderly population, which is set at 5° based on geriatric medicine research. A piecewise linear function is used to establish the slope load mapping relationship: the slope penalty coefficient is 1.0 when the slope is below 5°, and increases by 0.3 for every 1° increase in slope. For steep path segments exceeding 8°, a passage constraint weight of 5.0 is set, amplifying their cost value in path calculation and effectively preventing elderly users from choosing steep paths, thus creating slope constraints.

[0054] Spatial coordinate data of rest facility nodes within medical institutions are acquired, including available rest locations such as seats, waiting areas, and service counters. Based on geriatric medicine literature, a preset threshold for continuous walking distance for the elderly population is set at 80 meters. The actual path distance between any two adjacent rest nodes is calculated, and the ratio of this distance to the 80-meter threshold is used as the base value for the rest node density coefficient. When the interval between adjacent rest nodes exceeds 100 meters, the density coefficient decreases to below 0.6. For path segments with a density coefficient below the preset density threshold of 0.7, a rest node compensation weight is set, adding a virtual penalty term to the path cost calculation to encourage the algorithm to prioritize paths with sufficient rest facilities, thus forming a rest node distribution constraint.

[0055] The system acquires location information of emergency aid points within medical institutions, including first aid stations, medical staff duty rooms, and AED device installation points. Dijkstra's algorithm is used to calculate the shortest path distance from any node in the path network to all emergency aid points, and the minimum of these distances is taken as the reachability distance for that node. A preset aid coverage radius of 150 meters is set. For nodes whose reachability exceeds this radius, an aid accessibility penalty weight is applied, with the penalty value proportional to the square of the excess distance, forming an emergency aid point coverage constraint.

[0056] The three types of constraints mentioned above are combined in a structured manner to construct a comprehensive path cost function. This function includes a basic distance cost term, a slope penalty term, a rest node compensation term, and a rescue accessibility penalty term. These terms are weighted and summed to form a multi-dimensional path optimization rule.

[0057] In the adaptive weight allocation phase, the slope penalty coefficient weight is adjusted based on the time urgency indicator. When the time urgency indicator is "emergency," the slope penalty coefficient weight is reduced to 0.4 times its original value, allowing the selection of some routes with steeper slopes but shorter distances; when it is "general outpatient," the weight remains at 1.0 times; and when it is "chronic follow-up," the weight is increased to 1.2 times. Simultaneously, the rest node density coefficient weight is adjusted based on symptom severity. Pain scores are extracted from symptom description features, ranging from 0 to 10. When the pain score exceeds 7 or mobility is severely limited, the rest node density coefficient weight is increased to 1.5 times to ensure sufficient rest points along the route; when the score is below 3, the weight remains at 1.0 times. Through this dynamic weight allocation mechanism, the route optimization rules are adaptively adjusted to different medical demand scenarios.

[0058] In one optional implementation, the navigation scheme is generated by solving the path between service nodes in the candidate medical service scheme set and combining the internal spatial topology of the medical institution: Obtain the location coordinates of each service node and the access order constraints between service nodes; Based on the location coordinate information, the set of connecting paths between nodes is queried from the internal spatial topology of the medical institution. For each path in the set of connecting paths, the corresponding constraint cost is calculated based on the slope constraint, rest node distribution constraint and emergency rescue point coverage constraint in the multi-dimensional path optimization rules. The constraint cost is then weighted and summed with the physical distance of the path to obtain the comprehensive path cost. The traversal sequence of service nodes is determined according to the access order constraint relationship. For the path selection between adjacent service nodes in the traversal sequence, the path with the minimum comprehensive path cost is selected from the connected path set as the optimal path segment between nodes. All optimal path segments between nodes are spliced ​​together according to the traversal sequence to form a complete navigation path. For each path segment in the complete navigation path, extract the starting coordinates, ending coordinates, direction of travel, and estimated travel time of the path segment, and generate path segment navigation instructions. The navigation scheme is generated by arranging all navigation instructions for all path segments in the complete navigation path according to the traversal sequence and combining the department information and service time information corresponding to the service nodes.

[0059] When obtaining the location coordinates of each service node and the access order constraints between service nodes, the three-dimensional spatial coordinates (x, y, z) of various service nodes, including registration windows, consultation rooms, examination departments, and pharmacies, are extracted from the candidate medical service solution set. The z-axis represents floor information. Access order constraints are determined based on the characteristics of the medical process; for example, the registration node has a priority of 1, the consultation room node 2, the examination department node 3, and the pharmacy node 4, forming a priority matrix P. When certain examinations can be completed before diagnosis, the corresponding node pairs are marked as commutative.

[0060] When querying the set of connected paths between nodes in the internal spatial topology of a medical institution based on location coordinates, the spatial topology is stored in a graph structure. Nodes represent key spatial points such as corridor intersections, elevator entrances, and stairwells, while edges represent passable road segments with associated slope, length, and facility markers. A depth-first search algorithm is used to traverse the graph structure and obtain all physically reachable paths between any pair of service nodes. For each path in the connected path set, when calculating the slope constraint cost, the cumulative length L of road segments with a slope greater than 5 degrees is counted. slope Multiplying by the slope penalty coefficient α yields the slope cost C. slope =α×L slope When calculating the cost of rest node distribution constraints, rest facilities such as seats and waiting areas are detected along the path. When no rest node appears after a continuous walking distance of more than 50 meters, the excess distance is multiplied by the rest node density coefficient β to obtain the rest cost C. rest=β×L excess When calculating the coverage constraint cost of emergency aid points, it is determined whether the distance from any point on the path to the nearest AED device or medical station is less than 30 meters. The length of the road segment that does not meet the coverage requirement is multiplied by the aid point penalty coefficient γ to obtain the aid cost C. emergency =γ×L uncovered The cost of the three types of constraints is correlated with the physical distance D of the path. phys We perform a weighted summation to obtain the comprehensive path cost C. total =w1×D phys +w2×C slope +w3×C rest +w4×C emergency The weighting coefficients are dynamically adjusted based on the user's action capability assessment results.

[0061] When determining the traversal sequence of service nodes based on access order constraints, the service nodes are arranged in ascending order of their values ​​in the priority matrix P. For node pairs marked as commutative, the total path lengths corresponding to the two access orders are compared, and the shorter path is selected. For path selection between adjacent service nodes in the traversal sequence, the path with the minimum overall path cost is selected from the set of connected paths as the optimal path segment between nodes. When concatenating all optimal path segments between nodes according to the traversal sequence, it is checked whether the endpoints and starting points of adjacent path segments coincide. If spatial discontinuities exist, supplementary connecting path segments are inserted to form a complete navigation path.

[0062] Extract the starting coordinates (x, y) of each path segment in the complete navigation path. start y start , z start ), Path segment endpoint coordinates (x end y end , z end The system calculates the walking direction of each path segment using vector calculations and converts the direction angle into natural language descriptions such as "go straight ahead," "turn left," "turn right," and "go upstairs." The estimated travel time for each path segment is calculated by dividing the path segment length by the average walking speed of elderly users (0.8 meters per second) and adding delays caused by obstacles such as slope and stairs. The generated navigation instructions use a structured format such as "go straight ahead for 20 meters in corridor Y on floor X, then turn right."

[0063] The navigation instructions for all route segments in the complete navigation path are arranged according to the traversal sequence. Department information and service time information are inserted at each service node, forming a node description such as "Arriving at the cardiology clinic on the 3rd floor at 09:30, estimated waiting time 15 minutes". The final navigation plan includes three components: step-by-step navigation instruction sequence, key node service information, and total estimated time.

[0064] A third aspect of the present invention provides an electronic device, comprising: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the aforementioned method.

[0065] A fourth aspect of the present invention provides a computer-readable storage medium having stored thereon computer program instructions that, when executed by a processor, implement the aforementioned method.

[0066] This invention can be a method, apparatus, system, and / or computer program product. The computer program product may include a computer-readable storage medium having computer-readable program instructions loaded thereon for performing various aspects of the invention.

[0067] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and not to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features; and these modifications or substitutions 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 invention.

Claims

1. A method for providing age-friendly medical navigation services based on voice interaction, characterized in that: include: Semantic analysis is performed on medical demand voice information to extract medical intent type and health status description features. A temporal association graph is constructed based on the fragmentation features and semantic jump features of elderly users’ speech. By identifying breakpoint nodes and isolated nodes in the temporal association graph, missing dimension reasoning and semantic link repair are performed on the medical intent type and health status description features to generate a structured medical demand description. The structured medical demand description is mapped using a medical knowledge graph. The associated department set and cross-diagnosis and treatment path are identified and their complexity is quantified to obtain the path complexity index value. Based on the path complexity index value, service paths are constructed and cognitive load is assessed. A set of service paths with assessment values ​​lower than the preset cognitive load boundary is selected and matched with the service time period information in the medical resource database for spatiotemporal accessibility to obtain a set of candidate medical service solutions. Based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, which includes slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. The slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule are adaptively weighted and assigned. The path is solved for the service nodes in the candidate medical service scheme set by combining the internal spatial topology of the medical institution, and a navigation scheme is generated.

2. The method according to claim 1, characterized in that, Semantic analysis of medical demand voice information is performed to extract features related to medical intent type and health status description, including: The medical demand voice information is subjected to acoustic feature extraction and speech recognition conversion to obtain a textualized expression sequence. Dialect feature recognition is performed on the textual expression sequence, and dialect words in the textual expression sequence are converted into standard medical terminology by constructing a dialect word mapping relationship for elderly users, thus obtaining a standardized text sequence; Syntactic structure analysis and semantic role labeling are performed on the standardized text sequence to identify symptom description components, time description components, and medical visit purpose description components in the standardized text sequence. Based on the medical symptom ontology knowledge base, the non-standard symptom expressions of the symptom description components are identified and mapped to standard symptom classifications. The non-standard expressions in the symptom description components are converted into standard symptom classification identifiers, and the time description components are converted into time urgency identifiers. Semantic association analysis is performed between the standard symptom classification identifier and the description component of the purpose of medical visit to extract the medical intent type; The standard symptom classification identifier is combined with the time urgency identifier to extract the health status description features.

3. The method according to claim 1, characterized in that, Based on the fragmented and semantic jump features of elderly users' speech, a temporal correlation graph is constructed. By identifying breakpoints and isolated nodes in the temporal correlation graph, missing dimension inference and semantic link repair are performed on the medical intent type and the health status description features to generate a structured medical need description, including: The medical intent type and the health status description features are timestamped to obtain a time-series labeled information sequence; the semantic jump position is identified by calculating the semantic association strength between adjacent information segments, and the time-series labeled information sequence is divided into multiple discourse segment units based on the semantic jump position; The temporal association graph is constructed using the utterance fragment units as nodes and the temporal relationships marked by timestamps as edges; In the temporal correlation graph, the isolated node is identified by identifying the node with an edge degree of zero, and the breakpoint node is identified by identifying the node where the temporal relationship between the preceding and following nodes is interrupted. Semantic content analysis is performed on the isolated nodes and the breakpoint nodes to identify the missing information dimensions in the medical intent type and the health status description features; Based on medical domain knowledge reasoning rules, the missing information dimension is filled in to obtain reasoning completion information; the reasoning completion information is semantically associated with the isolated node and the breakpoint node, and a new semantic link connection is established in the temporal association graph to complete the semantic link repair. By integrating all node information in the repaired temporal correlation graph, the structured medical demand description is generated.

4. The method according to claim 1, characterized in that, The structured medical needs description is mapped using a medical knowledge graph to identify related department sets and cross-diagnosis paths, and the path complexity index is obtained by performing complexity quantification, including: The health status description features are extracted from the structured medical needs description. After extracting symptom entities from the health status description features, semantic matching is performed with disease entity nodes in the medical knowledge graph to identify a set of candidate disease entities associated with the health status description features. The department nodes associated with each disease entity node in the candidate disease entity set are extracted along the department affiliation edges in the medical knowledge graph. The extracted department nodes are deduplicated to form the associated department set. For each department node in the associated department set, query the medical knowledge graph for the diagnosis and treatment relationship path between the department node and the disease entity node in the candidate disease entity set; Identify department nodes that are simultaneously associated with multiple disease entity nodes as cross-department nodes, extract all diagnosis and treatment relationship paths from the cross-department nodes to different disease entity nodes, and summarize all diagnosis and treatment relationship paths to form the cross-diagnosis and treatment path; The path complexity index value is obtained by weighting the path length, the number of intermediate nodes, and the number of disease entities involved in each treatment relationship path in the cross-treatment path.

5. The method according to claim 1, characterized in that, Based on the path complexity index, service paths are constructed and cognitive load is assessed. A set of service paths with assessment values ​​below a preset cognitive load boundary is selected and matched with service time period information in the medical resource database for spatiotemporal accessibility. This yields a candidate medical service solution set, including: Based on the path complexity index value, the department nodes in the associated department set are prioritized and the corresponding department location information and treatment process information are extracted, and department location node sequence and treatment process node sequence are constructed respectively to form a service path; For each path in the service path, the spatial transfer distance and the number of spatial transfers are calculated based on the sequence of department location nodes, and the number of process steps and the number of process decision branches are calculated based on the sequence of medical process nodes. The spatial transfer distance, the number of spatial transfers, the number of process steps, and the number of process decision branches are weighted and summed to obtain the cognitive load assessment value. The service path set is formed by filtering service paths whose cognitive load assessment values ​​are lower than the preset cognitive load boundary. Extract the service time information of the corresponding department for each service path from the medical resource database; Obtain the current geographical location and expected consultation time of elderly users, calculate the arrival time to the corresponding department of each service path in the service path set based on the current geographical location, determine whether the time difference between the arrival time and the expected consultation time is within the time range of the service time period information, and mark the service path whose time difference is within the time range as a spatiotemporally reachable path. The department location nodes and medical process nodes in the spatiotemporal reachable path are extracted as service nodes. The service nodes, the department information corresponding to the spatiotemporal reachable path, the service time period information, and the cognitive load assessment value are structured and encapsulated to form the candidate medical service solution set.

6. The method according to claim 2, characterized in that, Based on the structured medical needs description, a multi-dimensional path optimization rule is constructed, including slope constraints, rest node distribution constraints, and emergency aid point coverage constraints. Adaptive weight allocation is performed on the slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule, including: Extract the time urgency identifier and symptom description features from the structured medical needs description; Obtain the slope information of the path segment and establish a slope load mapping relationship with the preset slope tolerance threshold for the elderly population, generate a slope penalty coefficient, and set a passage constraint weight for the path segment that exceeds the preset slope tolerance threshold for the elderly population to form the slope constraint. The spatial distribution information of rest facility nodes is obtained, the required interval distance of rest nodes is calculated based on a preset continuous walking distance threshold for the elderly population, the density coefficient of rest nodes is calculated according to the required interval distance and the spatial distribution information, and a rest node compensation weight is set for path segments below the preset density threshold to form the rest node distribution constraint. Obtain the location information of emergency aid points, calculate the shortest distance from any location node to the nearest emergency aid point, and set an aid accessibility penalty weight for location nodes whose shortest distance exceeds the preset aid coverage radius to form the emergency aid point coverage constraint. The slope constraint, the rest node distribution constraint, and the emergency aid point coverage constraint are structurally combined to form the multi-dimensional path optimization rule; The weight reduction factor of the slope penalty coefficient is proportional to the time urgency indicator, and the weight amplification factor of the rest node density coefficient is proportional to the severity of symptoms.

7. The method according to claim 1, characterized in that, Furthermore, by combining the internal spatial topology of the medical institution, path solving is performed on the service nodes in the candidate medical service solution set to generate a navigation solution, including: Obtain the location coordinates of each service node and the access order constraints between service nodes; Based on the location coordinate information, the set of connecting paths between nodes is queried from the internal spatial topology of the medical institution. For each path in the set of connecting paths, the corresponding constraint cost is calculated based on the slope constraint, rest node distribution constraint and emergency rescue point coverage constraint in the multi-dimensional path optimization rules. The constraint cost is then weighted and summed with the physical distance of the path to obtain the comprehensive path cost. The traversal sequence of service nodes is determined according to the access order constraint relationship. For the path selection between adjacent service nodes in the traversal sequence, the path with the minimum comprehensive path cost is selected from the connected path set as the optimal path segment between nodes. All optimal path segments between nodes are spliced ​​together according to the traversal sequence to form a complete navigation path. For each path segment in the complete navigation path, extract the starting coordinates, ending coordinates, direction of travel, and estimated travel time of the path segment, and generate path segment navigation instructions. The navigation scheme is generated by arranging all navigation instructions for all path segments in the complete navigation path according to the traversal sequence and combining the department information and service time information corresponding to the service nodes.

8. A voice-interactive age-friendly medical navigation service system, used to implement the method as described in any one of claims 1-7, characterized in that, include: The semantic parsing unit is used to perform semantic parsing on medical demand voice information, extract medical intent type and health status description features, construct a temporal association graph based on the fragmentation features and semantic jump features of elderly users’ speech, identify breakpoint nodes and isolated nodes in the temporal association graph, perform missing dimension reasoning and semantic link repair on the medical intent type and health status description features, and generate a structured medical demand description. The path planning unit is used to map the structured medical demand description to a medical knowledge graph, identify the set of related departments and cross-diagnosis and treatment paths, and perform complexity quantification to obtain a path complexity index value. Based on the path complexity index value, it constructs service paths and performs cognitive load assessment, filters out the set of service paths with assessment values ​​lower than the preset cognitive load boundary, and performs spatiotemporal accessibility matching with service time information in the medical resource database to obtain a set of candidate medical service solutions. The navigation generation unit is used to construct a multi-dimensional path optimization rule that includes slope constraints, rest node distribution constraints, and emergency aid point coverage constraints based on the structured medical demand description. It adaptively assigns weights to the slope penalty coefficient and rest node density coefficient in the multi-dimensional path optimization rule, and solves the path for the service nodes in the candidate medical service scheme set by combining the internal spatial topology of the medical institution, thereby generating a navigation scheme.

9. An electronic device, characterized in that, include: processor; Memory used to store processor-executable instructions; The processor is configured to invoke instructions stored in the memory to execute the method according to any one of claims 1 to 7.

10. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, they implement the method described in any one of claims 1 to 7.

Images