AIDS unmanned consultation self-examination system, method and device based on large language model

Abstract

The embodiment of the invention discloses an AIDS unmanned consultation self-examination system, method and device based on a large language model.The system comprises a user interaction processing unit, a dialogue management unit, a risk assessment unit and a data security unit, and the user interaction processing unit is used for receiving and recognizing interaction information and user information input by a user; the dialogue management unit is used for acquiring key information in the interaction information, selecting matched answer information based on the key information and feeding back the answer information to the user; and the risk assessment unit is used for acquiring feature information of the user according to the interaction information, the feature information comprises behavior information, disease information and time information, and generating a risk index according to the feature information. According to the embodiment, key information is extracted from interaction information of the user through the large language model, the intention of the user is deeply understood in combination with interaction instructions input by the user for multiple times, more accurate solutions are provided for the user, AIDS related knowledge meeting the individual requirements of the user is provided, and therefore high-individuation individualized user interaction experience is achieved.

Description

Technical Field

[0001] The present invention relates to the field of medical consultation technology, and in particular to an unmanned AIDS consultation self-examination system, method and device based on a large language model. Background Art

[0002] At present, the field of Internet medical technology usually involves online medical consultation and answering. The existing technology often obtains query information, processes the query information, obtains the user's personal profile, associates the personal profile with each key word, searches for the question information corresponding to the associated content in the medical knowledge graph, calculates the similarity between the word vector and each question vector, and determines the target question vector with the highest similarity to the word vector, and finally outputs the medical answer corresponding to the target question vector to the user. This interactive method cannot achieve personalized instant interaction, nor can it deeply understand the user's intention. The patent with publication number CN111538913A discloses a personalized recommendation method for AIDS prevention and control knowledge education. Its main innovation is to provide a method for AIDS prevention and control knowledge education, especially in personalized recommendation. The method is based on big data processing, uses a collaborative filtering recommendation algorithm, and solves the data sparsity problem through KL divergence. In addition, the method effectively distinguishes different education contents when selecting cluster centers, and adopts a distance model that considers the influence of different rating numbers between education contents to improve the accuracy of recommendations. Although the patent provides a personalized recommendation method, it does not mention the application of a large language model. The problems faced by existing HIV consultation and risk assessment systems mainly focus on privacy protection, data quality, user acceptance, algorithm bias, coverage, technical barriers, complexity of language understanding, and risk of misdiagnosis. These problems have resulted in the inadequacy of existing model scales and algorithms to cope with the complexity, dynamic changes, resource constraints, and diversity of user groups of HIV information, thus affecting the accuracy and widespread application of the system. In particular, when dealing with large amounts of complex medical information, existing models and algorithms often fail to meet the needs of real-time updates and personalized services, resulting in inaccurate information transmission and difficulty in effectively supporting HIV publicity, risk assessment, and prevention and control. Summary of the invention

[0003] In response to the above-mentioned defects, the embodiments of the present invention disclose an unmanned AIDS consultation and self-examination system, method and device based on a large language model, which can provide deeply personalized intelligent dialogue and provide more accurate risk assessment.

[0004] The first aspect of an embodiment of the present invention discloses an unmanned AIDS consultation and self-examination system based on a large language model, including a user interaction processing unit, a dialogue management unit, a risk assessment unit and a data security unit. The user interaction processing unit is used to receive and identify interaction information and user information input by a user, and send the interaction information to the dialogue management unit and the risk assessment unit. The dialogue management unit is used to process the interaction information to obtain key information in the interaction information, and select matching answer information based on the key information to feed back to the user; the risk assessment unit is used to obtain the user's characteristic information based on the interaction information, and the characteristic information includes behavior information, disease information, and time information, and generate a risk index based on the characteristic information; the data security unit is used to encrypt and store the user information and interaction information

[0005] The second aspect of the embodiment of the present invention discloses an unmanned AIDS consultation self-examination method based on a large language model, comprising:

[0006] In response to an interaction instruction of a user, identifying user information and interaction information in the interaction instruction;

[0007] Acquire historical interaction information of the user according to the user information, acquire key information according to the interaction information, and select matching answer information according to the historical interaction information and the key information;

[0008] The characteristic information of the user is acquired according to the interaction information, and a risk index is generated according to the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

[0009] As an optional implementation manner, in the second aspect of the embodiment of the present invention, it also includes:

[0010] Collect historical text data and pre-process the historical text data; the pre-processing includes classifying, cleaning, segmenting, and part-of-speech tagging the historical text data;

[0011] A neural network model is constructed, and the historical text data is input into the neural network model for training to obtain a large language neural network model.

[0012] As an optional implementation manner, in the second aspect of the embodiment of the present invention, the acquiring key information according to the interaction information includes:

[0013] The interactive information is preprocessed, and the preprocessed interactive information is input into a large language neural network model to process the interactive information and output key information.

[0014] As an optional implementation manner, in the second aspect of the embodiment of the present invention, performing information processing on the interaction information includes:

[0015] Divide the long sentences in the interactive information into several short sentences according to the preset length;

[0016] The short sentences are matched with the keywords stored in the database to extract the keywords in each short sentence, and all the keywords are combined into key information.

[0017] As an optional implementation manner, in the second aspect of the embodiment of the present invention, selecting matching answer information according to the historical interaction information and the key information includes:

[0018] Match at least one answer data according to the key information;

[0019] The key information and the answer data are converted into a key information vector and an answer data vector respectively, and the cosine similarity between the key information vector and the answer data vector is calculated;

[0020] The matching degree between the answer data and the key information is verified based on the cosine similarity, and the answer data with the greatest matching degree is selected as the answer information.

[0021] As an optional implementation manner, in a second aspect of the embodiment of the present invention, generating a risk index according to the characteristic information includes:

[0022] Cleaning the feature information to remove invalid features and redundant features, and standardizing the cleaned feature information;

[0023] The calculation weight coefficients corresponding to the behavior information, disease information, and time information in the feature information are assigned;

[0024] The behavior information, the disease information, and the time information are all input into a pre-trained evaluation model to output risk scores corresponding to the behavior information, the disease information, and the time information, respectively;

[0025] The risk index is calculated based on the calculated weight coefficients and risk scores corresponding to the behavior information, disease information and time information respectively.

[0026] The third aspect of the embodiment of the present invention discloses an unmanned AIDS consultation and self-examination device based on a large language model, comprising:

[0027] Interaction response module: used to respond to the user's interaction instruction and identify the user information and interaction information in the interaction instruction;

[0028] User question and answer module: used to obtain the user's historical interaction information according to the user information, obtain key information according to the interaction information, and select matching answer information according to the historical interaction information and the key information;

[0029] Risk assessment module: used to obtain the user's characteristic information according to the interactive information, and generate a risk index according to the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

[0030] As an optional implementation manner, in the third aspect of the embodiment of the present invention, it also includes:

[0031] Collect historical text data and pre-process the historical text data; the pre-processing includes classifying, cleaning, segmenting, and part-of-speech tagging the historical text data;

[0032] A neural network model is constructed, and the historical text data is input into the neural network model for training to obtain a large language neural network model.

[0033] As an optional implementation manner, in a third aspect of the embodiment of the present invention, the acquiring key information according to the interaction information includes:

[0034] The interactive information is preprocessed, and the preprocessed interactive information is input into a large language neural network model to process the interactive information and output key information.

[0035] As an optional implementation manner, in a third aspect of the embodiment of the present invention, performing information processing on the interaction information includes:

[0036] Divide the long sentences in the interactive information into several short sentences according to the preset length;

[0037] The short sentences are matched with the keywords stored in the database to extract the keywords in each short sentence, and all the keywords are combined into key information.

[0038] As an optional implementation manner, in the third aspect of the embodiment of the present invention, selecting matching answer information according to the historical interaction information and the key information includes:

[0039] Match at least one answer data according to the key information;

[0040] The key information and the answer data are converted into a key information vector and an answer data vector respectively, and the cosine similarity between the key information vector and the answer data vector is calculated;

[0041] The matching degree between the answer data and the key information is verified based on the cosine similarity, and the answer data with the greatest matching degree is selected as the answer information.

[0042] As an optional implementation manner, in a third aspect of the embodiment of the present invention, generating a risk index according to the characteristic information includes:

[0043] Cleaning the feature information to remove invalid features and redundant features, and standardizing the cleaned feature information;

[0044] The calculation weight coefficients corresponding to the behavior information, disease information, and time information in the feature information are assigned;

[0045] The behavior information, the disease information, and the time information are all input into a pre-trained evaluation model to output risk scores corresponding to the behavior information, the disease information, and the time information, respectively;

[0046] The risk index is calculated based on the calculated weight coefficients and risk scores corresponding to the behavior information, disease information and time information respectively.

[0047] The fourth aspect of an embodiment of the present invention discloses an electronic device, comprising: a memory storing executable program code; a processor coupled to the memory; the processor calls the executable program code stored in the memory to execute the unmanned AIDS consultation self-examination method based on a large language model disclosed in the second aspect of an embodiment of the present invention.

[0048] A fifth aspect of an embodiment of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute the unmanned AIDS consultation and self-examination method based on a large language model disclosed in the second aspect of an embodiment of the present invention.

[0049] Compared with the prior art, the embodiments of the present invention have the following beneficial effects:

[0050] In the embodiment of the present invention, in response to the user's interactive instructions, the user's interactive information is identified and key information is extracted based on the large language model constructed in advance. The user's intention is deeply understood in combination with the interactive instructions input by the user for multiple times, and the user is given a more accurate answer. AIDS-related knowledge that meets the individual needs of the user is provided, thereby achieving a highly personalized personalized user interactive experience, and providing comprehensive risk assessment and personalized suggestions based on the user's personal situation. The user's disease risk is assessed through algorithm analysis in combination with the personal information and symptom description provided by the user. Due to the combination of big data analysis, the risk assessment is more accurate; and the present invention has data security and system performance optimization mechanisms. In order to ensure the privacy of users, security processing mechanisms such as data encryption and Mi Minghua processing are adopted to strictly protect user data and continuously optimize system performance, ensuring that users obtain efficient and accurate health consultation services in a safe environment. BRIEF DESCRIPTION OF THE DRAWINGS

[0051] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the drawings required for use in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0052] Figure 1 It is a schematic diagram of the module structure of the AIDS unmanned consultation and self-examination system based on a large language model disclosed in an embodiment of the present invention;

[0053] Figure 2 It is a flowchart of an unmanned AIDS consultation self-examination method based on a large language model disclosed in an embodiment of the present invention;

[0054] Figure 3 It is a structural schematic diagram of an unmanned AIDS consultation and self-examination device based on a large language model provided by an embodiment of the present invention;

[0055] Figure 4 It is a structural schematic diagram of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0056] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0057] It should be noted that the terms "first", "second", "third", "fourth", etc. in the specification and claims of the present invention are used to distinguish different objects rather than to describe a specific order. The terms "including" and "having" in the embodiments of the present invention and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product or device including a series of steps or units is not necessarily limited to those steps or units clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0058] The embodiment of the present invention discloses an unmanned AIDS consultation and self-examination system, method, device, electronic device and storage medium based on a large language model. In the embodiment, in response to the user's interactive instructions, the user's interactive information is identified and key information is extracted based on the large language model constructed in advance, and the user's intention is deeply understood in combination with the interactive instructions input by the user for multiple times, and a more accurate answer is given to the user, and AIDS-related knowledge that meets the individual needs of the user is provided, thereby achieving a highly personalized personalized user interaction experience, and providing comprehensive risk assessment and personalized suggestions for the user's personal situation, combining the personal information and symptom description provided by the user, and evaluating the user's disease risk through algorithm analysis. Due to the combination of big data analysis, the risk assessment is more accurate; and the present invention has a data security and system performance optimization mechanism. In order to ensure the privacy of the user, security processing mechanisms such as data encryption and Mi Minghua processing are adopted, strict protection of user data and continuous optimization of system performance are achieved, ensuring that users obtain efficient and accurate health consultation services in a safe environment.

[0059] Embodiment 1

[0060] See also Figure 1 , Figure 1 FIG. 1 shows a module structure diagram of an unmanned AIDS consultation and self-examination system based on a large language model disclosed in an embodiment of the present invention. Figure 1 As shown, the AIDS unmanned consultation and self-examination system includes a user interaction processing unit, a dialogue management unit, a risk assessment unit and a data security unit. Among them, the user interaction processing unit is used to receive and identify the interaction information and user information input by the user, and send the interaction information to the dialogue management unit and the risk assessment unit. The dialogue management unit is used to process the interaction information to obtain key information in the interaction information, and select matching answer information based on the key information to feed back to the user; the risk assessment unit is used to obtain the user's characteristic information based on the interaction information, and the characteristic information includes behavior information, disease information, time information, and generate a risk index based on the characteristic information; the data security unit is used to encrypt and store the user information and interaction information.

[0061] The user interaction processing unit of the embodiment does not receive the user's interaction information once, but can receive the interaction information input by the user multiple times, and then conduct multiple question-and-answer interactions with the user. In this unit, an interface layout, an interaction method, a feedback mechanism, and dialogue management are provided. The interface provides text and voice input methods, which can support multilingual communication for use without user groups. And optimize touch or click responses to ensure that users can easily interact with the system. The specific interaction processing involves tracking the dialogue state and processing multiple rounds of dialogue. In the natural language nursing method, the system parses the user's input through word segmentation, part-of-speech tagging, user intent, etc., and uses dialogue state tracking technology to manage the dialogue process. In this process, sequence labeling algorithms such as conditional random fields CRF or bidirectional LSTM+CRF are used for word segmentation and part-of-speech tagging, and entity recognition and intent recognition may use neural network-based methods, such as BiLSTM+CRF or BERT models. Dialogue state tracking combines a recurrent neural network RNN ​​or Transformer architecture to update the dialogue state, while question generation may use template matching, decision trees, or seq2seq models based on generative models. These algorithms and technologies work together to extract key information such as the departure place and destination based on the user's previous conversation content, identify the user's booking intention, and generate the next question based on this, such as asking about the departure date, to guide the user to further provide complete information. This series of operations ensures that the system can ask appropriate questions based on the context of the conversation, thereby gradually guiding the conversation, avoiding mechanical repetition, and maintaining the natural flow of the conversation.

[0062] The implementation example uses a pre-trained large language model, such as BERT or GLM, and fine-tunes and customizes it according to AIDS-related fields. It uses machine learning frameworks such as TensorFlow and PyTorch to implement NLP algorithms, and adopts quantitative methods to optimize algorithms and models to improve response speed and computing resource utilization efficiency.

[0063] In the dialogue management unit, we first collect and organize AIDS-related professional knowledge, including the latest research results, treatment methods, prevention measures, and epidemiological data. This information is generally collected by the original CDCs in various places. When building the knowledge base, the knowledge is divided into different categories and subcategories, such as disease overview, transmission routes, symptom identification, detection methods, treatment guidelines, etc., and a unique identifier is assigned to each knowledge point to facilitate system retrieval and update. At the same time, the construction of the knowledge base also takes into account the source and quality control of the knowledge to ensure that all information is based on authoritative and latest medical research. In addition, in order to keep the knowledge base up to date, the content will be reviewed and updated regularly to reflect the latest medical advances and policy changes. This unit focuses on how to guide the dialogue, maintain the coherence of the dialogue, and provide personalized recommendations when appropriate. This involves the management of the dialogue state, that is, the system tracks the dialogue history, understands the user's intentions, and decides the next action based on this information. Dialogue management should also include an error handling mechanism so that when the user input is unclear or wrong, the system can appropriately request more information or correct misunderstandings. In addition, in order to improve the user experience, the dialogue management module should be able to handle multiple rounds of dialogue and guide users in a timely manner during the dialogue to obtain complete and necessary information. This unit is able to retrieve the most relevant information from the knowledge base based on the user's input and the dialogue context, and present the information to the user in an easy-to-understand way through natural language generation technology. In this process, the design of the algorithm should take into account the different backgrounds and needs of users to achieve truly personalized services.

[0064] Specifically, use deep learning models such as BERT to understand user intent, use RNN or Transformer to track conversation context, apply TF-IDF, BM25 or DPR for information retrieval, find matching knowledge base entries through similarity calculation, and combine collaborative filtering or content recommendation for personalized recommendation. The implementation steps include receiving user input, performing intent recognition and context analysis, building queries, performing information retrieval and semantic matching, applying personalized recommendations, extracting and organizing key information, and finally generating answers through natural language generation algorithms such as seq2seq or GPT. This process also requires a feedback loop to optimize the user experience and ensure that truly personalized services are provided based on the different backgrounds and needs of users.

[0065] In the risk assessment unit, a large amount of medical data is collected and sorted, including the epidemiological characteristics of AIDS, the correlation between symptoms and risk factors, and the accuracy of various detection methods. Then, the personal information and symptom descriptions provided by the user are analyzed to identify possible AIDS risk factors, and the probability of the user's illness is calculated based on this information to divide the risk level. Then, personalized suggestions are generated for users based on the risk assessment results, such as taking HIV tests, taking preventive measures, etc., and warning reminders are provided to high-risk users. Finally, the assessment results are compared with the opinions of clinical experts to ensure the accuracy of the assessment, and medical data and models are updated regularly to reflect the latest research and epidemiological information. Through these steps, the disease risk assessment and suggestion generation module can assess the user's risk of contracting AIDS through algorithmic analysis based on the information provided by the user, and provide corresponding suggestions and warnings to help users take appropriate preventive measures or seek medical treatment in time. The model is trained to identify and weight different risk factors. In order to ensure the accuracy of risk assessment, the algorithm needs to be rigorously trained and verified. This involves selecting appropriate training data sets, which should contain enough positive and negative cases, as well as detailed risk factor records. During the model training process, cross-validation and other techniques are used to avoid overfitting and ensure that the model has good generalization capabilities. The professional knowledge mastered by the large model is used to provide customized suggestions for different risk assessment results.

[0066] In addition, building a data security and system optimization module is the key to ensuring the reliability and user trust of the AIDS unmanned consultation and risk assessment system. The data security strategy covers all aspects of data collection, storage, processing and transmission to ensure that all user data is properly protected. The system should use the latest encryption technology to protect data. Whether it is static storage or dynamic transmission, strong encryption standards should be implemented. Access control mechanism is also indispensable. It ensures that only authorized personnel can access sensitive data. At the same time, the system needs to record and audit all access activities for easy tracking and monitoring. In order to further strengthen data security, the system is anonymized and data is desensitized. In particular, for users' personal information, such as name, address, etc., it should be desensitized during storage and processing to ensure that even in the event of data leakage, it cannot be traced back to a specific individual. At the same time, the system will conduct security assessments and vulnerability scans regularly to promptly discover and repair potential security risks. At the same time, in order to ensure the long-term stable operation of the system, reliable servers and high-performance computing resources should be selected in hardware to support the operation and data processing needs of the system. In terms of software, optimization algorithms and database queries are performed to reduce unnecessary calculations and I / O operations and improve the processing efficiency of the system. In addition, system optimization also includes continuous improvement of the user interface and interaction process to enhance the user experience. By collecting user feedback and system usage data, user behavior patterns can be analyzed, and then the interface layout and interaction logic can be adjusted to make it more intuitive and convenient.

[0067] Embodiment 2

[0068] See also Figure 2 , Figure 2 It is a flow chart of an unmanned AIDS consultation and self-examination method based on a large language model disclosed in an embodiment of the present invention. Among them, the execution subject of the method described in the embodiment of the present invention is an execution subject composed of software and / or hardware, and the execution subject can receive relevant information by wired or / and wireless means, and can send certain instructions. Of course, it can also have certain processing functions and storage functions. The execution subject can control multiple devices, such as a remote physical server or cloud server and related software, or it can be a local host or server and related software that performs related operations on a device placed somewhere. In some scenarios, multiple storage devices can also be controlled, and the storage devices can be placed in the same place or different places as the devices. For example Figure 2 As shown, the unmanned AIDS self-examination method based on a large language model includes the following steps:

[0069] 201. In response to an interaction instruction from a user, identify user information and interaction information in the interaction instruction.

[0070] When a user initiates a query or consultation, the system first identifies the user's interactive instructions, which includes parsing the user's input text, voice or other forms of instructions to extract user information and specific interactive content.

[0071] Furthermore, the embodiment also includes collecting historical text data and preprocessing the historical text data; the preprocessing includes classifying, cleaning, segmenting, and part-of-speech tagging the historical text data; building a neural network model, and inputting the historical text data into the neural network model for training to obtain a large language neural network model. That is, the embodiment first builds a large language model through a large amount of data, and then uses the large voice model to perform an interactive processing process with the user, and performs personalized question-and-answer interaction with the user.

[0072] 202. Acquire historical interaction information of the user according to the user information, acquire key information according to the interaction information, and select matching answer information according to the historical interaction information and the key information.

[0073] Based on user information (such as age, gender, region, etc.), the system can obtain the user's past interaction records with the system, that is, historical interaction information. At the same time, the system will extract key information from the current interaction information. This key information is usually related to the user's health concerns, symptom descriptions, or specific consultation questions. Combining historical interaction information and current key information, the system can more accurately select or generate answer information that matches the user's question.

[0074] In this step, the interactive information is preprocessed, the preprocessed interactive information is processed, and key information is output. The information processing of the interactive information includes: dividing the long sentences in the interactive information into a number of short sentences according to a preset length; matching the short sentences with the keywords stored in the database to extract the keywords in each short sentence, and forming all the keywords into key information.

[0075] When the user inputs interactive information through the user interaction processing unit, this information will be transmitted to the dialogue management unit. At this time, the interactive information is presented in text form and contains various related contents. First, data preprocessing is performed to remove noise and irrelevant characters and segment words so that the model can better understand it. Then the preprocessed information is input into the large language neural network model. The model is trained with a large amount of relevant data and can deeply analyze the information. Through semantics, context and knowledge application, it can extract key information, such as user behavior characteristics, symptom description, medical history, etc. These key information are transmitted to the dialogue management unit, and as the user inputs interactive instructions many times, the system continuously accumulates information and understands the intention more deeply. Each new information is processed and integrated with the previous key information for optimization. In terms of long sentence splitting, the system first uses the syntactic analysis algorithm to analyze the grammatical structure of the long sentence, such as identifying the subject, predicate and object of the sentence "I have had many unprotected sexual behaviors in the past three months, and I feel a little weak recently". At the same time, semantic coherence is fully considered, and the content describing different aspects but related is reasonably split based on the complete semantic unit, such as splitting the sentence into "I have had many unprotected sexual behaviors in the past three months" and "I feel a little weak recently". A preset length limit is also set. When the character length or number of words in a long sentence exceeds the threshold, the split is triggered, but in practice, it is still prioritized according to sentence structure and semantics. In terms of keyword extraction, the system builds a special AIDS-related keyword database, covering various terms, symptom words, behavior words, etc. For the split short sentences, a combination of exact matching and fuzzy matching is used, such as exact matching of "fever" and fuzzy matching of "fever" to match it with "fever". After extracting the keywords, they are combined and screened to remove those that are meaningless or repeated, retain the core keywords, and finally form key information for subsequent operations.

[0076] Furthermore, matching answer information is selected based on the historical interaction information and the key information, including: matching at least one answer data based on the key information; converting the key information and the answer data into a key information vector and an answer data vector, respectively, and calculating the cosine similarity between the key information vector and the answer data vector; verifying the matching degree between the answer data and the key information based on the cosine similarity, and selecting the answer data with the largest matching degree as the answer information.

[0077] The word embedding method is used to convert each keyword in the key information into a high-dimensional vector to capture its semantic and contextual information. At the same time, the historical interactive information is also processed in the same way to convert it into a sequence vector. For each possible answer data, word embedding is also used to convert it into a vector representation. The similarity is measured by calculating the cosine similarity between the key information vector and the answer data vector. The formula is cosθ=(A·B) / (|A|×|B|), where A is the key information vector and B is the answer data vector. Considering the influence of historical interactive information, its vector is weighted, and the weighted historical interactive information vector and the key information vector are merged into a comprehensive vector, and then the cosine similarity is calculated with the answer data vector. In addition, in addition to cosine similarity, factors such as the length of the answer data, the number of keywords, and semantic relevance are also considered, and weights are assigned to each factor. The weighted sum is used to obtain the final matching score.

[0078] 203. Acquire characteristic information of the user according to the interaction information, and generate a risk index according to the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

[0079] The system also further extracts user characteristic information from the interactive information, including but not limited to behavioral information (such as sexual behavior habits, drug use, etc.), disease information (such as past medical history, family genetic diseases, etc.) and time information (such as the time of symptom onset, duration, etc.). Based on this characteristic information, the system can generate a risk index to assess the user's possibility or risk level of HIV infection.

[0080] In the specific process, the feature information is cleaned to remove invalid features and redundant features, and the cleaned feature information is standardized; calculation weight coefficients corresponding to the behavior information, disease information, and time information in the feature information are assigned; the behavior information, disease information, and time information are input into a pre-trained evaluation model to output risk scores corresponding to the behavior information, disease information, and time information, respectively; and the risk index is calculated based on the calculation weight coefficients and risk scores corresponding to the behavior information, disease information, and time information, respectively.

[0081] The embodiment uses a large language model to deeply understand and process AIDS-related knowledge and information, and provides personalized consulting services through intelligent dialogue. This innovation is reflected in the system's ability to dynamically adjust the content of the conversation according to the user's input and conversation context, and provide AIDS-related knowledge, prevention measures, detection methods, etc. that meet the individual needs of the user, thereby achieving a highly personalized user interaction experience. Not only does it provide AIDS-related knowledge consultation, but it also evaluates the user's risk of illness through algorithm analysis based on the personal information and symptom description provided by the user, and provides corresponding suggestions and warnings. This innovation lies in the combination of big data analysis and machine learning technology, making risk assessment more accurate, and the suggestions and warnings provided are more targeted and practical. To ensure user privacy and data security, the system adopts multiple data protection measures, including data encryption, anonymization processing, access control, etc., and establishes a system optimization mechanism to optimize algorithm performance through continuous feedback and learning mechanisms, and improve system response speed and stability. This innovation is reflected in the system's strict protection of user data and continuous optimization of system performance, ensuring that users can obtain efficient and accurate health consulting services in a safe environment.

[0082] Embodiment 3

[0083] See also Figure 3 , Figure 3 Schematic diagram of the structure of an unmanned AIDS consultation and self-examination device based on a large language model disclosed in an embodiment of the present invention. Figure 3 As shown, the unmanned AIDS consultation and self-examination device based on a large language model may include: an interactive response module 301, a user question and answer module 302 and a risk assessment module 303, wherein the interactive response module 301 is used to respond to the user's interactive instructions and identify the user information and interactive information in the interactive instructions; the user question and answer module 302 is used to obtain the user's historical interactive information based on the user information, obtain key information based on the interactive information, and select matching answer information based on the historical interactive information and key information; the risk assessment module 303 is used to obtain the user's characteristic information based on the interactive information, and generate a risk index based on the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

[0084] The embodiment may also include a model building module for collecting historical text data and preprocessing the historical text data; the preprocessing includes classifying, cleaning, segmenting, and part-of-speech tagging the historical text data; building a neural network model, and inputting the historical text data into the neural network model for training to obtain a large language neural network model.

[0085] Based on the large language neural network model, in the user question and answer module 302, the key information is obtained according to the interactive information, including: inputting the interactive information into the large language neural network model to process the interactive information and output the key information. The interactive information is processed, including: dividing the long sentences in the interactive information into a number of short sentences according to a preset length; matching the short sentences with the keywords stored in the database to extract the keywords in each short sentence, and forming all the keywords into key information. Selecting matching answer information according to the historical interactive information and key information includes: matching at least one answer data according to the key information; verifying the matching degree between the answer data and the key information based on the historical interactive information, and selecting the answer data with the largest matching degree as the answer information.

[0086] In the risk assessment module 303, a risk index is generated based on the feature information, including: cleaning the feature information to remove invalid features and redundant features, and standardizing the cleaned feature information; assigning calculation weight coefficients corresponding to the behavior information, disease information, and time information in the feature information; inputting the behavior information, disease information, and time information into a pre-trained assessment model to output risk scores corresponding to the behavior information, disease information, and time information, respectively; and calculating the risk index based on the calculation weight coefficients and risk scores corresponding to the behavior information, disease information, and time information, respectively.

[0087] Embodiment 4

[0088] See also Figure 4 , Figure 4 Schematic diagram of the structure of an electronic device disclosed in an embodiment of the present invention. The electronic device may be a computer, a server, etc. Of course, in certain circumstances, it may also be a smart device such as a mobile phone, a tablet computer, a monitoring terminal, and an image acquisition device with processing functions. Figure 4 As shown, the electronic device may include:

[0089] A memory 401 storing executable program codes;

[0090] a processor 402 coupled to the memory 401;

[0091] The processor 402 calls the executable program code stored in the memory 401 to execute some or all steps of the unattended AIDS self-examination method based on a large language model in the second embodiment.

[0092] An embodiment of the present invention discloses a computer-readable storage medium storing a computer program, wherein the computer program enables a computer to execute part or all of the steps in the unmanned AIDS consultation self-examination method based on a large language model in Embodiment 2.

[0093] The embodiment of the present invention further discloses a computer program product, wherein when the computer program product runs on a computer, the computer executes part or all of the steps in the unmanned AIDS consultation and self-examination method based on a large language model in the second embodiment.

[0094] An embodiment of the present invention also discloses an application publishing platform, wherein the application publishing platform is used to publish a computer program product, wherein when the computer program product runs on a computer, the computer executes some or all of the steps in the unmanned AIDS consultation and self-examination method based on a large language model in Example 2.

[0095] In various embodiments of the present invention, it should be understood that the size of the serial numbers of the processes does not necessarily mean the order of execution. The execution order of each process should be determined by its function and internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present invention.

[0096] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, i.e., they may be located in one place or distributed over multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the present embodiment.

[0097] In addition, each functional unit in each embodiment of the present invention may be integrated into one processing unit, or each unit may exist physically separately, or two or more units may be integrated into one unit. The integrated unit may be implemented in the form of hardware or in the form of software functional units.

[0098] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product, which is stored in a memory and includes several requests for a computer device (which can be a personal computer, a server or a network device, etc., specifically a processor in a computer device) to perform some or all of the steps of the method described in each embodiment of the present invention.

[0099] In the embodiments provided by the present invention, it should be understood that "B corresponding to A" means that B is associated with A, and B can be determined according to A. However, it should also be understood that determining B according to A does not mean determining B only according to A, and B can also be determined according to A and / or other information.

[0100] A person of ordinary skill in the art can understand that some or all of the steps in the various methods of the embodiments can be completed by instructing related hardware through a program, and the program can be stored in a computer-readable storage medium, and the storage medium includes a read-only memory (ROM), a random access memory (RAM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), a one-time programmable read-only memory (OTPROM), an electronically erasable rewritable read-only memory (EEPROM), a compact disc (CD-ROM) or other optical disc storage, magnetic disk storage, magnetic tape storage, or any other computer-readable medium that can be used to carry or store data.

[0101] The above is a detailed introduction to the unmanned AIDS consultation and self-examination system, method, device, electronic device and storage medium based on a large language model disclosed in the embodiments of the present invention. Specific examples are used in this article to illustrate the principles and implementation methods of the present invention. The description of the above embodiments is only used to help understand the method of the present invention and its core idea; at the same time, for general technical personnel in this field, according to the idea of ​​the present invention, there will be changes in the specific implementation method and application scope. In summary, the content of this specification should not be understood as a limitation on the present invention.

Claims

1. An unmanned AIDS consultation and self-examination system based on a large language model, characterized in that: It includes a user interaction processing unit, a dialogue management unit, a risk assessment unit and a data security unit. The user interaction processing unit is used to receive and identify interaction information and user information input by the user, and send the interaction information to the dialogue management unit and the risk assessment unit. The dialogue management unit is used to process the interaction information to obtain key information in the interaction information, and select matching answer information based on the key information to feed back to the user; the risk assessment unit is used to obtain characteristic information of the user according to the interaction information, and the characteristic information includes behavior information, disease information, and time information, and generate a risk index according to the characteristic information; The data security unit is used to encrypt and store user information and interaction information.

2. A method for self-examination of AIDS without consultation based on a large language model, characterized in that: include: In response to an interaction instruction of a user, identifying user information and interaction information in the interaction instruction; Acquire historical interaction information of the user according to the user information, acquire key information according to the interaction information, and select matching answer information according to the historical interaction information and the key information; The characteristic information of the user is acquired according to the interaction information, and a risk index is generated according to the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

3. The AIDS self-examination method without consultation according to claim 2, characterized in that: Also includes: Collect historical text data and pre-process the historical text data; the pre-processing includes classifying, cleaning, segmenting, and part-of-speech tagging the historical text data; A neural network model is constructed, and the historical text data is input into the neural network model for training to obtain a large language neural network model.

4. The AIDS self-examination method without consultation according to claim 3, characterized in that: The acquiring key information according to the interaction information includes: The interactive information is preprocessed, and the preprocessed interactive information is input into a large language neural network model to process the interactive information and output key information.

5. The AIDS self-examination method without consultation according to claim 4, characterized in that: Performing information processing on the interaction information includes: Divide the long sentences in the interactive information into several short sentences according to the preset length; The short sentences are matched with the keywords stored in the database to extract the keywords in each short sentence, and all the keywords are combined into key information.

6. The AIDS self-examination method without consultation according to claim 5, characterized in that: Selecting matching answer information according to the historical interaction information and key information includes: Match at least one answer data according to the key information; The key information and the answer data are converted into a key information vector and an answer data vector respectively, and the cosine similarity between the key information vector and the answer data vector is calculated; The matching degree between the answer data and the key information is verified based on the cosine similarity, and the answer data with the greatest matching degree is selected as the answer information.

7. The AIDS self-examination method without consultation according to claim 2, characterized in that: Generating a risk index according to the characteristic information includes: Cleaning the feature information to remove invalid features and redundant features, and standardizing the cleaned feature information; The calculation weight coefficients corresponding to the behavior information, disease information, and time information in the feature information are assigned; The behavior information, the disease information, and the time information are all input into a pre-trained evaluation model to output risk scores corresponding to the behavior information, the disease information, and the time information, respectively; The risk index is calculated based on the calculated weight coefficients and risk scores corresponding to the behavior information, disease information and time information respectively.

8. An unmanned AIDS consultation and self-examination device based on a large language model, characterized in that: include: Interaction response module: used to respond to the user's interaction instruction and identify the user information and interaction information in the interaction instruction; User question-and-answer module: used to obtain historical interaction information of the user according to the user information, obtain key information according to the interaction information, and select matching answer information according to the historical interaction information and the key information; Risk assessment module: used to obtain the user's characteristic information based on the interactive information, and generate a risk index based on the characteristic information, wherein the characteristic information includes behavior information, disease information, and time information.

9. An electronic device, characterized in that: include: A memory storing executable program code; a processor coupled to the memory; The processor calls the executable program code stored in the memory to execute the unmanned AIDS consultation self-examination method based on a large language model as described in any one of claims 2 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, wherein the computer program enables a computer to execute the unmanned AIDS consultation self-examination method based on a large language model as described in any one of claims 2 to 7.

Images