Visiting service implementation method and system based on AI social worker

Abstract

The invention relates to the technical field of artificial intelligence and social services, and discloses an AI social worker-based visit service implementation method and system, and the method comprises the steps: constructing a resident portrait vector, and calculating a visit priority score through a graph attention network; a Q-learning algorithm is adopted to optimize a visiting path according to the spatial distance; a conversation text and an emotional pressure index are obtained through voice recognition and BERT semantic analysis; performing structured archiving on the visit data; and dynamically updating the resident portrait. In the prior art, a social worker service mode of fixed schedule visiting and artificial experience judgment is mostly adopted, and especially under the conditions of service crowd diversification and risk emotion recessive expression, accurate visiting service scheduling is difficult to realize. Due to the fact that resident portrait priorities are modeled through the graph neural network and path planning is driven through reinforcement learning, efficient scheduling of social worker tasks is achieved, and the personalized coverage capacity and the intelligent decision-making capacity of social worker services are improved.

Description

Technical Field

[0001] The present invention belongs to the technical field of artificial intelligence and social services, and particularly relates to a method and system for realizing a home visit service based on AI social workers. Background Art

[0002] Currently, social worker home visit services play an important role in scenarios such as urban and rural community governance and special population care. However, their scheduling and management methods mostly rely on manual experience, fixed schedules, and paper records, resulting in problems such as low efficiency, slow response, and weak risk identification capabilities, and it is difficult to meet the service needs of diverse populations. For example, in the face of high-risk groups such as the elderly living alone, those with abnormal emotions, and psychological disorders, the existing technology cannot achieve accurate identification and priority response based on big data, leading to a serious lag in resource allocation and intervention timing. In addition, the information value of conversation content is difficult to systematically extract, historical records are scattered across multiple terminals, and there is a lack of a unified filing and portrait update mechanism, which also makes it difficult to implement long-term service optimization. The existing technology cannot fully meet the complex requirements of "multi-dimensional information integration, implicit risk identification, intelligent task scheduling, and service closed-loop feedback" in social worker services. Therefore, there is an urgent need for an AI-driven method for realizing home visit services that can still dynamically identify high-risk populations, intelligently schedule task paths, structurally file interaction information, and continuously update resident portraits in large-scale distributed service scenarios, so as to improve the intelligence level and dynamic adaptability of social worker services. Summary of the Invention

[0003] In view of the above technical deficiencies, the purpose of the present invention is to propose a method for realizing a home visit service based on AI social workers, aiming to solve the technical problem that the existing technology mostly uses fixed-schedule home visits and manual experience judgment in social worker service methods, and it is difficult to achieve accurate home visit service scheduling especially under the conditions of diverse service populations and implicit expression of risk emotions.

[0004] To solve the above technical problems, the present invention adopts the following technical solutions: The present invention provides a method for realizing a home visit service based on AI social workers,

[0005] The method for realizing a home visit service based on AI social workers includes:

[0006] Step S10: Collect basic attribute data, historical service record data, and neighborhood interaction relationship data of community residents and establish a resident portrait vector; pre-construct and train a graph attention network model, input the resident portrait vector into the graph attention network model, and output the home visit priority score of the resident;

[0007] Step S20: Generate a set of high-priority residents according to the home visit priority score; obtain the actual spatial distance, and use the Q-learning reinforcement algorithm to calculate and optimize the home visit path according to the set of high-priority residents and the actual spatial distance;

[0008] Step S30: Execute the visit process according to the optimized visit path; in real time, convert the conversation content during the visit process into conversation text data through speech recognition; calculate the resident emotional stress index based on the conversation text data combined with the BERT semantic analysis method;

[0009] Step S40: Use the graph database Neo4j to perform information structured archiving according to the conversation text data and the resident emotional stress index, and construct a visit data record unit;

[0010] Step S50: Output the set of residents with emotional risks according to the visit data record unit, and update the resident portrait vector according to the visit data record unit.

[0011] Preferably, in step S10, the basic attribute data includes age data, gender data, living status, marital status, and health status; the historical service record data includes the number of services, the problem closed-loop rate, and the service satisfaction; the data structure of the neighborhood interaction relationship data is a graph structure, and the neighborhood interaction relationship data includes the adjacency relationship of residents in space, the community event collaboration relationship, the community risk propagation factor, and the adjacency node clustering coefficient.

[0012] Preferably, in step S10, the structure of the graph attention network model specifically includes: an input layer for receiving the resident portrait vector; two hidden layers, including a first hidden layer with 64 dimensions and a second hidden layer with 32 dimensions; a multi-head graph attention layer for performing attention-weighted aggregation on each resident portrait vector and the resident portrait vectors of its neighbors; a resident priority regression output layer, and the resident priority regression output layer uses the Sigmoid activation function to output the visit priority score for the resident.

[0013] Preferably, in the training process of the graph attention network model in step S10, a loss function is constructed by introducing the neighborhood interaction relationship graph structure consistency constraint, and the formula of the loss function is: ;

[0014] where is the loss function; is the total number of residents in the training set during the training process; is the true priority label of resident i; is the predicted priority label of resident i predicted by the graph attention network model; is the neighborhood consistency regularization factor for controlling the distribution smoothing under the guidance of the graph structure; is the set of adjacent edges in the neighborhood interaction relationship data; is the true priority label of resident j.

[0015] Preferably, in step S20, a high-priority resident set is generated according to the visit priority score; obtaining the actual spatial distance, and calculating the optimized visit path according to the high-priority resident set and the actual spatial distance by using the Q-learning reinforcement algorithm specifically includes:

[0016] Step S201: Preset a visit priority score threshold, and screen out a high-priority resident set whose visit priority score is higher than the visit priority score threshold ;

[0017] Step S202: Define the state space S of the Q-learning reinforcement algorithm. The state space S includes the resident path sequence that has completed the visit, the current social worker's position, and the remaining residents to be visited; define the action space A of the Q-learning reinforcement algorithm. The action space A includes the set of selecting the next target resident from the current position; obtain the actual spatial distance, and design the reward function R of the Q-learning reinforcement algorithm according to the actual spatial distance and the high-priority resident set;

[0018] Step S203: Perform iterative update according to the state space S, the action space A, and the reward function R by using the Q-learning reinforcement learning algorithm, and output the optimized visit path.

[0019] Preferably, in step S30, the visit process is executed according to the optimized visit path; the conversation content during the visit process is real-time transcribed into conversation text data through speech recognition; the steps of calculating the resident emotional stress index based on the conversation text data in combination with the BERT semantic analysis method specifically include:

[0020] Step S301: Execute the visit process according to the optimized visit path. During the visit process, the conversation content during the visit is synchronously recorded by using a mobile terminal, and the speech stream data of the conversation content is real-time transcribed through the integrated speech recognition engine Wav2Vec 2.0 to obtain the conversation text data;

[0021] Step S302: Input the conversation text data into the pre-trained Chinese combined with the BERT semantic analysis method for semantic embedding encoding to obtain the semantic vector of each round of conversation; set a semantic risk keyword template set, and calculate the semantic similarity score for the semantic vector of each round of conversation by using the cosine similarity method;

[0022] Step S303: When the semantic similarity score exceeds the preset semantic similarity score threshold, use a convolutional neural network to extract features from the semantic vector of this round of conversation to obtain emotion word feature recognition, emotional intensity feature, language repetition feature, speech rate change feature, and vocabulary jump frequency feature;

[0023] Step S304: Output the resident emotional stress index through weighted linear combination based on emotional word features recognition, emotional intensity features, language repetition features, speech rate change features, and lexical jump frequency features.

[0024] Preferably, in step S50, the steps of outputting the set of residents with emotional risks according to the visit data recording unit and updating the resident portrait vector according to the visit data recording unit specifically include:

[0025] Step S501: Extract the sequence data of the resident emotional stress index and the sequence data of the conversation text data from the visit data recording unit;

[0026] Step S502: Extract risk features from the sequence data of the resident emotional stress index and the sequence data of the conversation text data to obtain a risk feature vector, where the risk feature vector includes an emotional feature sub-vector, a text feature sub-vector, and a service feedback feature sub-vector;

[0027] Step S503: Input the risk feature vector into a lightweight classifier, and the lightweight classifier outputs the resident emotional risk level. The residents with a resident emotional risk level greater than the preset resident emotional risk level threshold are output as the set of residents with emotional risks;

[0028] Step S504: Finally, update the resident portrait vector according to the visit data recording unit.

[0029] The present invention also provides a visit service implementation system based on an AI social worker, including:

[0030] A portrait construction module, configured to collect the basic attribute data, historical service record data, and neighborhood interaction relationship data of community residents and establish a resident portrait vector; pre-construct and train a graph attention network model, input the resident portrait vector into the graph attention network model, and output the visit priority score of the resident;

[0031] A path scheduling module, configured to generate a set of high-priority residents according to the visit priority score; obtain the actual spatial distance, and calculate and optimize the visit path by using the Q-learning reinforcement algorithm according to the set of high-priority residents and the actual spatial distance;

[0032] A visit execution and emotion analysis module, configured to execute the visit process according to the optimized visit path; convert the conversation content during the visit process into conversation text data in real time through speech recognition; calculate the resident emotional stress index based on the conversation text data in combination with the BERT semantic analysis method;

[0033] A structured archiving module, configured to perform information structured archiving by using the graph database Neo4j according to the conversation text data and the resident emotional stress index, and construct a visit data recording unit;

[0034] A risk identification and portrait update module, configured to output a set of residents with emotional risks according to the visit data recording unit, and update the resident portrait vector according to the visit data recording unit.

[0035] The present invention also provides a computer program product, including a program for implementing the visit service based on AI social workers. When the program for implementing the visit service based on AI social workers is executed by a processor, the method for implementing the visit service based on AI social workers as described above is implemented.

[0036] The beneficial effects of the present invention are as follows: Compared with the existing social work service methods that mostly conduct visits according to fixed schedules and judge by manual experience, especially under the conditions of diverse service populations and implicit expression of risk emotions, it is difficult to achieve precise visit service scheduling. Since the present application models the priority of resident portraits through a graph neural network and drives path planning through reinforcement learning, it realizes high-efficiency scheduling of social work tasks and improves the personalized coverage ability and intelligent decision-making ability of social work services. BRIEF DESCRIPTION OF THE DRAWINGS

[0037] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only some embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained based on these drawings without creative efforts.

[0038] Figure 1 It is a schematic flowchart of the first embodiment of a method for implementing a visit service based on AI social workers of the present invention.

[0039] Figure 2 It is a schematic diagram of the device for a method for implementing a visit service based on AI social workers of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0040] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, rather than all embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative efforts fall within the protection scope of the present invention.

[0041] Embodiment 1: As Figure 1 shown, it is a schematic flowchart of the first embodiment of the method for implementing a visit service based on AI social workers of the present invention, and the first embodiment of the method for implementing a visit service based on AI social workers of the present invention is proposed.

[0042] In the first embodiment, the method for implementing a visit service based on AI social workers includes:

[0043] Step S10: Collect the basic attribute data, historical service record data, and neighborhood interaction relationship data of community residents and establish a resident portrait vector; pre-construct and train a graph attention network model, input the resident portrait vector into the graph attention network model, and output the visit priority score of the residents;

[0044] It should be noted that in step S10, the basic attribute data includes age data, gender data, living status, marital status, and health status; the historical service record data includes the number of services, the problem closed-loop rate, and service satisfaction; the data structure of the neighborhood interaction relationship data is a graph structure, and the neighborhood interaction relationship data includes the adjacency relationship of residents in space, the community event collaboration relationship, the community risk propagation factor, and the adjacency node clustering coefficient. In step S10, during the training process of the graph attention network model, a loss function is constructed by introducing the neighborhood interaction relationship graph structure consistency constraint, and the formula of the loss function is: ;

[0045] where, is the loss function; is the total number of residents in the training set during the training process; is the true priority label of resident i; is the predicted priority label of resident i predicted by the graph attention network model; is the neighborhood consistency regularization factor, which is used to control the distribution smoothing under the guidance of the graph structure; is the set of adjacent edges in the neighborhood interaction relationship data; is the true priority label of resident j.

[0046] It can be understood that the neighborhood interaction relationship graph structure consistency constraint regular term introduced in the above training of the graph attention network model enables the model to not only learn the attribute characteristics of each resident node itself, but also jointly model information propagation and visit importance according to the high-risk commonalities existing in the neighborhood relationship, thereby improving the model's ability to identify potential hidden high-risk residents.

[0047] It should be understood that traditional scoring methods are mostly based on independent resident information and do not have the ability to identify spatial propagation and community patterns, which are extremely likely to cause missed judgments of residents "hiding in marginal areas" or "not labeled but with concentrated neighborhood risks". And this method automatically introduces social collaboration signals through the edge connection relationship in the graph structure, enabling the model to still produce a scoring result with propagation consistency when the resident labels are incomplete or the service history is insufficient.

[0048] For example, in a certain community, residents A and B are neighbors living upstairs and downstairs. Resident A has been identified as a resident with psychological risks. Although there is no clear service record for their adjacent resident B, the model captures their high connectivity with A through the regularization term. Finally, the model predicts that B's priority score is 0.87, while the traditional model only gets 0.42. Actual tests found that B is indeed a real potential risk object due to mood swings caused by living alone, verifying the risk reasoning ability and identification accuracy of the model driven by neighborhood relationships.

[0049] Step S20: Generate a set of high-priority residents based on the visit priority scores; obtain the actual spatial distances, and use the Q-learning reinforcement algorithm to calculate and optimize the visit path according to the set of high-priority residents and the actual spatial distances.

[0050] It should be noted that in step S20, the steps of generating a set of high-priority residents based on the visit priority scores; obtaining the actual spatial distances, and using the Q-learning reinforcement algorithm to calculate and optimize the visit path according to the set of high-priority residents and the actual spatial distances specifically include:

[0051] Step S201: Preset a visit priority score threshold, and screen out a set of high-priority residents whose visit priority scores are higher than the visit priority score threshold ;

[0052] Step S202: Define the state space S of the Q-learning reinforcement algorithm. The state space S includes the sequence of resident paths that have been visited, the current social worker's location, and the set of remaining residents to be visited; define the action space A of the Q-learning reinforcement algorithm. The action space A includes the set of selecting the next target resident from the current location; obtain the actual spatial distances, and design the reward function R of the Q-learning reinforcement algorithm according to the actual spatial distances and the set of high-priority residents.

[0053] Step S203: Perform iterative updates using the Q-learning reinforcement learning algorithm according to the state space S, the action space A, and the reward function R, and output the optimized visit path.

[0054] It can be understood that by introducing the reinforcement learning mechanism, the social worker's path planning is not only based on the resident priority scores, but also can dynamically adapt to the spatial distribution and resource costs, achieving a balance of "maximizing service value / minimizing distance". Compared with the static scheduling algorithm, Q-learning can optimize the strategy according to real-time feedback and adapt to uncertainties such as personnel changes and resident non-visits.

[0055] It should be understood that traditional path optimization methods (such as TSP and greedy algorithms) usually only minimize the total path length, ignoring the priority differences among residents, and it is difficult to accurately serve high-risk groups. Q-learning can build a "state-action-reward" system, incorporate the resident portrait scores, service frequencies, and spatial distances into the learning framework, so as to realize the intelligence and value-driven of path scheduling.

[0056] For example, in a community pilot project, residents A, B, C, and D are located in different units respectively, among which A and D have the highest psychological risk scores, but are relatively far away on the map. The traditional greedy algorithm selects the order of B→C→D→A. Although the path is short, high-risk objects are not covered in the early stage; while the strategy trained by Q-learning gives priority to visiting A and D. Although the path length increases by 8%, the average response time of high-risk objects is shortened by 42%, and the service quality is significantly improved, which proves the practical value and strategy superiority of the reinforcement learning method in this scenario.

[0057] Step S30: Execute the visit process according to the optimized visit path; convert the conversation content during the visit process into conversation text data in real time through speech recognition; calculate the resident emotional stress index based on the conversation text data combined with the BERT semantic analysis method;

[0058] It should be noted that in step S30, the steps of executing the visit process according to the optimized visit path, converting the conversation content during the visit process into conversation text data in real time through speech recognition, and calculating the resident emotional stress index based on the conversation text data combined with the BERT semantic analysis method specifically include:

[0059] Step S301: Execute the visit process according to the optimized visit path. During the visit process, use a mobile terminal to synchronously record the conversation content during the visit, and transcribe the speech stream data of the conversation content in real time through the integrated speech recognition engine Wav2Vec 2.0 to obtain the conversation text data;

[0060] Step S302: Input the conversation text data into the pre-trained Chinese combined with the BERT semantic analysis method for semantic embedding encoding to obtain the semantic vector of each round of conversation; set a semantic risk keyword template set, and calculate the semantic similarity score for the semantic vector of each round of conversation using the cosine similarity method;

[0061] Step S303: When the semantic similarity score exceeds the preset semantic similarity score threshold, use a convolutional neural network to extract features from the semantic vector of this round of conversation to obtain emotional word feature recognition, emotional intensity feature, language repetition feature, speech rate change feature, and lexical jump frequency feature;

[0062] Step S304: Based on the emotional word features recognition, emotional intensity features, language repetition features, speech rate change features, and lexical jump frequency features, output the resident emotional stress index through weighted linear combination.

[0063] It can be understood that by combining speech-to-text conversion, deep semantic analysis, and emotional feature modeling, it is possible to automatically detect potential emotional changes such as anxiety, depression, and resistance in the conversation of residents, and output the quantitative index of the resident emotional stress index PI, realizing the effective transformation of "unstructured conversation → structured risk judgment", and significantly enhancing the emotional recognition ability and decision-making assistance ability of social workers during home visits.

[0064] It should be understood that traditional home visit service records often rely on subjective written records or coarse-grained emotional labels, lacking real-time and quantification. Through the context understanding ability of the BERT model and the emotional feature extraction ability of the CNN, combined with a custom keyword template, the present invention can effectively identify emotional fluctuations of "indirect language expression", such as emotional expressions like "euphemistic refusal" and "indirect negation", significantly improving the recognition accuracy of hidden risk groups.

[0065] For example, during a home visit, the resident's expression was: "There really is no problem. I'm used to it. There's no need to trouble you." Traditional records would only classify it as "no service need"; but in this system, the BERT model calculates the semantic similarity between it and "being ignored" and "being left unattended" as 0.82, and the convolutional neural network extracts significant language repetition (the phrase "I'm used to it" is repeated 3 times) and lexical jump features. Finally, the resident emotional stress index is calculated as 0.81, automatically listing the resident as an "object under high emotional stress observation". Subsequent interviews confirmed that the resident had persistent loneliness, corroborating the effective recognition ability of this method for hidden emotional signals.

[0066] Step S40: According to the conversation text data and the resident emotional stress index, use the graph database Neo4j for information structured archiving and construct a home visit data record unit;

[0067] It can be understood that compared with traditional record-keeping methods mainly based on table storage, the graph database supports expressing multi-source data involved in social workers' home visits in the structure of "nodes + relationships", and retains the natural connection logic between various types of information, thus enabling visual tracking and semantic path mining of "service process", "emotional changes", "keyword trends", and "risk behavior chains".

[0068] It should be understood that traditional on-site visit data records often face problems such as fragmented information, redundant fields, and difficulty in global association. Especially in the process of multiple on-site visits or multi-person collaborative services, the lack of a unified data archiving standard leads to insufficient risk identification and trend judgment capabilities. Through the introduction of a graph database, the present invention constructs a multi-layer relationship network of "residents - events - semantics - emotions", realizing the structured organization, long-term storage, and intelligent retrieval of data, providing high-quality basic data for subsequent portrait updates and risk warnings.

[0069] For example, during an on-site visit, the conversation text of resident B, "Living alone for too long, I don't want to care about many things anymore", is automatically transcribed, and keyword tags such as "lonely" and "mentally exhausted" are extracted. Combining the semantic score and the result of the resident's emotional stress index of 0.84, the conversation text data and the resident's emotional stress index node are automatically generated and classified into the Visit path of this resident. When the resident's emotional stress index is higher than 0.75 in three consecutive on-site visits, an abnormal trend can be found through path query, and it is clustered and analyzed in the graph database that the frequency of the theme of "psychological counseling" increases. Automatically list it as a "resident at emotional risk" and successfully trigger early intervention.

[0070] Step S50: Output a set of residents at emotional risk according to the on-site visit data recording unit, and update the resident portrait vector according to the on-site visit data recording unit.

[0071] It should be noted that in step S50, the steps of outputting a set of residents at emotional risk according to the on-site visit data recording unit and updating the resident portrait vector according to the on-site visit data recording unit specifically include:

[0072] Step S501: Extract the sequence data of the resident's emotional stress index and the sequence data of the conversation text data from the on-site visit data recording unit;

[0073] Step S502: Extract risk characteristics from the sequence data of the resident's emotional stress index and the sequence data of the conversation text data to obtain a risk feature vector, which includes an emotional feature sub-vector, a text feature sub-vector, and a service feedback feature sub-vector;

[0074] Step S503: Input the risk feature vector into a lightweight classifier, and the lightweight classifier outputs the resident's emotional risk level. Residents with a resident's emotional risk level greater than the preset resident's emotional risk level threshold are output as a set of residents at emotional risk;

[0075] Step S504: Finally, update the resident portrait vector according to the on-site visit data recording unit.

[0076] It is understandable that through the construction of a risk identification feature structure of "multi-round dialogue + emotional change + service response", this step can quantitatively identify hidden psychological risk groups without relying on the participation of doctors or psychological experts. By adaptively updating the portrait vector, it can provide more accurate basic information support for the next round of social worker scheduling, forming a data-driven closed-loop feedback mechanism.

[0077] It should be understood that traditional portrait updates mostly rely on static attributes (such as age, family structure), with low update frequency, coarse granularity, and fixed strategies, making it difficult to reflect the dynamic changes of residents' states. In contrast, the present invention regards "residents' states as time-varying variables" and conducts fine-grained modeling by combining continuous visit data. In particular, it can output the risk level in real time through a lightweight model, with the advantages of strong real-time performance, easy model deployment, and timely feedback, and is suitable for high-frequency execution scenarios of daily social worker services.

[0078] For example, during three visits, resident C repeatedly expressed emotions such as "annoyed", "helpless", and "don't want to talk". The sequence of its emotional stress index is 0.62, 0.75, 0.84. The model identified that its emotional fluctuations were intense, with many service requests but no closed-loop. Its risk score was 0.89, higher than the set threshold of 0.8. Based on this, the system listed it as a "high-emotion-risk group" and enhanced the dimensions of "psychological counseling intention" and "loneliness tendency" in the portrait. Subsequently, social workers gave priority to arranging psychological service follow-ups to intervene in potential crises in advance, demonstrating the effectiveness and practicality of this method in identifying hidden groups.

[0079] Embodiment 2: In addition, a visit service implementation system based on AI social workers provided by the present invention adopts the method for implementing a visit service based on AI social workers in the above embodiment, and can solve the technical problems of implementing a visit service based on AI social workers. Compared with the prior art, the beneficial effects of the visit service implementation system based on AI social workers provided by the present invention are the same as those of the method for implementing a visit service based on AI social workers provided by the above embodiment, and the other technical features in the visit service implementation system based on AI social workers are the same as the features disclosed in the method of the above embodiment, and will not be elaborated here.

[0080] Embodiment 3: The present invention provides a visit service implementation device based on AI social workers. Please refer to Figure 2, A device for implementing a home visit service based on AI social workers includes: at least one processor; and a memory communicatively connected to the at least one processor; wherein, the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to execute a method for implementing a home visit service based on AI social workers in the first embodiment above. A device for implementing a home visit service based on AI social workers in the embodiments of the present invention may include, but is not limited to, mobile terminals such as mobile phones, laptop computers, digital broadcast receivers, PDAs (Personal Digital Assistant), PADs (Portable Application Description: tablet computers), PMPs (Portable Media Player: portable multimedia players), in-vehicle terminals (such as in-vehicle navigation terminals), etc., and fixed terminals such as digital TVs, desktop computers, etc. A device for implementing a home visit service based on AI social workers is merely an example and should not impose any limitations on the functions and scope of use of the embodiments of the present invention. A device for implementing a home visit service based on AI social workers may include a processing device 1001 (such as a central processing unit, a graphics processing unit, etc.), which may perform various appropriate actions and processes according to a program stored in a read-only memory (ROM: Read-Only Memory) 1002 or a program loaded from a storage device 1003 into a random access memory (RAM: Random Access Memory) 1004. In the RAM 1004, various programs and data required for the operation of a device for implementing a home visit service based on AI social workers are also stored. The processing device 1001, the ROM 1002, and the RAM 1004 are connected to each other through a bus 1005. An input / output (I / O) interface 1006 is also connected to the bus. Generally, the following systems may be connected to the I / O interface 1006: an input device 1007 including, for example, a touch screen, a touchpad, a keyboard, a mouse, an image sensor, a microphone, an accelerometer, a gyroscope, etc.; an output device 1008 including, for example, a liquid crystal display (LCD: Liquid Crystal Display), a speaker, a vibrator, etc.; a storage device 1003 including, for example, a magnetic tape, a hard disk, etc.; and a communication device 1009. The communication device 1009 may allow a device for implementing a home visit service based on AI social workers to communicate with other devices wirelessly or wiredly to exchange data. Although the figure shows a device for implementing a home visit service based on AI social workers having various systems, it should be understood that it is not required to implement or have all the systems shown. Instead, more or fewer systems may be implemented or had.

[0081] Embodiment 4: The present invention also provides a computer program product, including a computer program, which when executed by a processor implements the steps of a method for implementing a home visit service based on AI social workers as described above. The computer program product provided by the present invention can solve the technical problem of implementing a home visit service based on AI social workers. Compared with the prior art, the beneficial effects of the computer program product provided by the present invention are the same as those of the method for implementing a home visit service based on AI social workers provided in the above embodiment, and will not be elaborated here.

[0082] Specifically, according to the embodiments disclosed by the present invention, the processes described above with reference to the flowcharts can be implemented as computer software programs. For example, an embodiment disclosed by the present invention includes a computer program product, which includes a computer program carried on a computer-readable medium, and the computer program contains program codes for executing the methods shown in the flowcharts. In such an embodiment, the computer program can be downloaded and installed from the network through a communication device, or installed from a storage device 1003, or installed from a ROM 1002. When the computer program is executed by a processing device 1001, it executes the above functions defined in the methods of the embodiments disclosed by the present invention.

[0083] It should be understood that various parts of the present invention can be implemented by hardware, software, firmware, or a combination thereof. In the description of the above embodiments, specific features, structures, materials, or characteristics can be combined in a suitable manner in any one or more embodiments or examples.

[0084] Obviously, those skilled in the art can make various changes and modifications to the present invention without departing from the spirit and scope of the present invention. Thus, if these modifications and variations of the present invention fall within the scope of the claims of the present invention and their equivalent technologies, the present invention also intends to include these changes and modifications.

Claims

1. A method for implementing a home visit service based on AI social workers, characterized in that, The method includes: Step S10: Collect the basic attribute data, historical service record data, and neighborhood interaction relationship data of community residents and establish a resident portrait vector; pre-construct and train a graph attention network model, input the resident portrait vector into the graph attention network model, and output the visit priority score of the residents; Step S20: Generate a set of high-priority residents according to the visit priority score; obtain the actual spatial distance, and use the Q-learning reinforcement algorithm to calculate and optimize the visit path according to the set of high-priority residents and the actual spatial distance; Step S30: Execute the visit process according to the optimized visit path; use speech recognition to convert the conversation content during the visit process into conversation text data in real time; calculate the resident emotion stress index based on the conversation text data combined with the BERT semantic analysis method; Step S40: Use the graph database Neo4j to perform information structured archiving according to the conversation text data and the resident emotion stress index, and construct a visit data record unit; Step S50: Output a set of residents with emotion risks according to the visit data record unit, and update the resident portrait vector according to the visit data record unit.

2. The implementation method of a home visit service based on AI social workers according to claim 1, characterized in that In step S10, the basic attribute data includes age data, gender data, living status, marital status, and health status; the historical service record data includes the number of services, the problem closed-loop rate, and the service satisfaction; the data structure of the neighborhood interaction relationship data is a graph structure, and the neighborhood interaction relationship data includes the spatial adjacency relationship of residents, the community event collaboration relationship, the community risk propagation factor, and the adjacency node clustering coefficient.

3. The implementation method of a home visit service based on AI social workers as claimed in claim 1, wherein, In step S10, the structure of the graph attention network model specifically includes: an input layer for receiving the resident portrait vector; two hidden layers, including a first hidden layer with 64 dimensions and a second hidden layer with 32 dimensions; a multi-head graph attention layer for performing attention-weighted aggregation on each resident portrait vector and the resident portrait vectors of its neighbors; a resident priority regression output layer, and the resident priority regression output layer uses a Sigmoid activation function to output the visit priority score of the residents.

4. The implementation method of a home visit service based on AI social workers according to claim 1, characterized in that, In step S10, during the training process of the graph attention network model, a loss function is constructed by introducing the consistency constraint of the neighborhood interaction relationship graph structure, and the formula of the loss function is: ； Among them, is the loss function; is the total number of residents in the training set during the training process; is the true priority label of resident i; is the predicted priority label of resident i predicted by the graph attention network model; is the neighborhood consistency regularization factor, which is used to control the distribution smoothing guided by the graph structure; is the set of adjacent edges in the neighborhood interaction relationship data; is the true priority label of resident j.

5. The implementation method of a home visit service based on AI social workers as described in claim 1, characterized in that In step S20, generate a set of high-priority residents according to the visit priority score; The steps of obtaining the actual spatial distance and using the Q-learning reinforcement algorithm to calculate and optimize the visit path according to the set of high-priority residents and the actual spatial distance specifically include: Step S201: Preset a threshold for the visit priority score, and filter out the set of high-priority residents whose visit priority scores are higher than the threshold for the visit priority score ; Step S202: Define the state space S of the Q-learning reinforcement algorithm. The state space S includes the resident path sequence that has completed the visit, the current social worker's position, and the remaining set of residents to be visited; define the action space A of the Q-learning reinforcement algorithm. The action space A includes the set of selecting the next target resident from the current position; Obtain the actual spatial distance, and design the reward function R of the Q-learning reinforcement algorithm according to the actual spatial distance and the set of high-priority residents; Step S203: Perform iterative updates using the Q-learning reinforcement learning algorithm based on the state space S, action space A, and reward function R, and output an optimized visit path.

6. The implementation method of a home visit service based on AI social workers according to claim 1, characterized in that, In step S30, perform the visit process according to the optimized visit path; convert the conversation content during the visit process into conversation text data in real time through speech recognition; the steps of calculating the resident emotion stress index based on the conversation text data in combination with the BERT semantic analysis method specifically include: Step S301: Perform the visit process according to the optimized visit path. During the visit process, use a mobile terminal to synchronously record the conversation content during the visit process, and transcribe the voice stream data of the conversation content in real time through the integrated speech recognition engine Wav2Vec 2.0 to obtain conversation text data; Step S302: Input the conversation text data into the pre-trained Chinese combined with the BERT semantic analysis method for semantic embedding encoding to obtain the semantic vector of each round of conversation; set a semantic risk keyword template set, and calculate the semantic similarity score for the semantic vector of each round of conversation using the cosine similarity method; Step S303: When the semantic similarity score exceeds the preset semantic similarity score threshold, use a convolutional neural network to extract features from the semantic vector of this round of conversation to obtain emotion word feature recognition, emotional intensity feature, language repetition feature, speech rate change feature, and vocabulary jump frequency feature; Step S304: Output the resident emotion stress index based on the emotion word feature recognition, emotional intensity feature, language repetition feature, speech rate change feature, and vocabulary jump frequency feature through weighted linear combination.

7. The implementation method of a home visit service based on AI social workers according to claim 1, characterized in that, In step S50, the steps of outputting a set of residents with emotion risks based on the visit data recording unit and updating the resident portrait vector according to the visit data recording unit specifically include: Step S501: Extract the sequence data of the resident emotion stress index and the sequence data of the conversation text data from the visit data recording unit; Step S502: Extract risk features from the sequence data of the resident emotion stress index and the sequence data of the conversation text data to obtain a risk feature vector, where the risk feature vector includes an emotion feature sub-vector, a text feature sub-vector, and a service feedback feature sub-vector; Step S503: Input the risk feature vector into a lightweight classifier, and the lightweight classifier outputs the resident emotion risk level. The residents with a resident emotion risk level greater than the preset resident emotion risk level threshold are output as a set of residents with emotion risks; Step S504: Finally, update the resident portrait vector according to the visit data recording unit.

8. A visit service implementation system based on AI social workers, which is applied to a visit service implementation method based on AI social workers according to any one of claims 1-7, and is characterized in that, The visit service implementation system based on AI social workers includes: A portrait construction module, which is used to collect the basic attribute data, historical service record data, and neighborhood interaction relationship data of community residents and establish a resident portrait vector; pre-construct and train a graph attention network model, input the resident portrait vector into the graph attention network model, and output the visit priority score of the resident; A path scheduling module, configured to generate a high-priority resident set according to the visit priority score; obtain the actual spatial distance, and calculate and optimize the visit path by using the Q-learning reinforcement algorithm based on the high-priority resident set and the actual spatial distance; A visit execution and emotion analysis module, configured to execute the visit process according to the optimized visit path; convert the conversation content during the visit into conversation text data in real time through speech recognition; calculate the resident emotion stress index based on the conversation text data in combination with the BERT semantic analysis method; A structured archiving module, configured to perform information structured archiving by using the graph database Neo4j according to the conversation text data and the resident emotion stress index, and construct a visit data record unit; A risk identification and portrait update module, configured to output a set of residents with emotion risks according to the visit data record unit, and update the resident portrait vector according to the visit data record unit.

9. An on-site visit service implementation device based on AI social workers, characterized in that, The visit service implementation device based on AI social workers includes: a memory, a processor, and a visit service implementation program based on AI social workers that is stored on the memory and can run on the processor. When the visit service implementation program based on AI social workers is executed by the processor, it implements the method for implementing a visit service based on AI social workers according to any one of claims 1 to 7.

10. A computer program product, characterized in that, The computer program product includes a visit service implementation program based on AI social workers. When the visit service implementation program based on AI social workers is executed by a processor, it implements the method for implementing a visit service based on AI social workers according to any one of claims 1 to 7.

Images