Digital intelligence pharmacist medication guidance system based on artificial intelligence

Abstract

The invention provides a digital intelligence pharmacist medication guidance system based on artificial intelligence, and relates to the technical field of medical treatment, the system comprises a user interaction module, a server and an answer generation module, the user interaction module is used for receiving and initiating a question proposed by a user to the server; the server side is used for packaging questions proposed by a user and submitting the questions to the answer generation module; and the answer generation module is used for receiving the question from the server side, retrieving information associated with the question in a pre-constructed medicine specification knowledge base by using a retrieval enhancement generation (RAG) method, integrating the retrieved information by using a pre-trained language model to generate an answer for the question, and finally, sending the answer to the server side. According to the system and the method, a more convenient, accurate and intelligent medicine information query and medication guidance mode is provided, the quality and the safety of medical services are improved, and the use experience of the user is greatly improved.

Description

Technical Field

[0001] The present invention relates to the field of medical technology, and in particular to a digital pharmacist medication guidance system based on artificial intelligence. Background Art

[0002] In today's medical field, drug instructions play a vital role in the correct use of drugs by patients and medical staff. However, the traditional way of obtaining drug instructions has many inconveniences. Paper instructions are not only small in size, but also difficult to find and save. In addition, the information is not updated in a timely manner, which easily leads to patients obtaining outdated drug information. Although the existing electronic databases facilitate the storage and query of information to a certain extent, there are often problems with inaccurate and incomplete information. The drug information in many electronic databases may not be updated in a timely manner, or lack coverage of certain special drugs or newly launched drugs, which makes it impossible for users to obtain comprehensive and accurate information when querying.

[0003] In addition, the existing drug information query system is also relatively simple in terms of information transmission. Most of them simply present text information, which cannot meet patients' needs for intuitive and easy-to-understand drug information. For some drug instructions with a lot of professional terms, patients may find it difficult to understand the meaning, which may lead to incorrect or improper medication. Moreover, these systems are usually unable to provide personalized drug information based on the specific circumstances of the patient, such as medication recommendations for special groups such as the elderly, pregnant women, and children.

[0004] There is an urgent need for a more convenient, accurate and intelligent drug instruction knowledge base that can integrate comprehensive drug information, present it to users in an intuitive and easy-to-understand manner, and provide personalized services based on the user's specific circumstances, thereby improving the quality and safety of medical services.

[0005] Therefore, it is necessary to provide an artificial intelligence-based digital pharmacist medication guidance system to solve the above technical problems. Summary of the invention

[0006] In order to solve the above technical problems, the present invention provides an artificial intelligence-based digital pharmacist medication guidance system, which provides a more convenient, accurate and intelligent drug information query and medication guidance method, which not only improves the quality and safety of medical services, but also greatly improves the user experience.

[0007] The present invention provides a digital pharmacist medication guidance system based on artificial intelligence, including a user interaction module, a server and an answer generation module.

[0008] The user interaction module is used to receive and initiate questions raised by users to the server;

[0009] The server is used to encapsulate the questions raised by the user and submit them to the answer generation module;

[0010] The answer generation module is used to receive questions from the server, use the retrieval enhancement generation RAG method to retrieve information related to the question in a pre-built drug instruction manual knowledge base, and use a pre-trained language model to integrate the retrieved information to generate an answer to the question, and finally, stream the answer to the question to the user interaction module.

[0011] Preferably, in the medication guidance system, the process of constructing the drug instructions knowledge base includes:

[0012] Collecting drug instructions through multiple channels and performing data preprocessing on the collected drug instructions, wherein the preprocessing includes removing duplicate information, correcting erroneous data and unifying the format;

[0013] Structural classification of the preprocessed drug instructions according to preset key fields to obtain content blocks corresponding to the key fields, wherein the key fields include drug information, usage and dosage, indications, adverse reactions and precautions;

[0014] Build the first question-answer pair dataset based on all structured content blocks;

[0015] The retrieval enhancement generation RAG method is used, and the selected language model is trained in combination with the first question-answer pair dataset to generate a drug instruction manual knowledge base.

[0016] Preferably, in the training process of the language model, the following is also included:

[0017] A manual review process is introduced to screen the answers generated by the language model during the training process.

[0018] Preferably, it also includes an expert think tank for providing expert-level medication recommendations when questions raised by users match questions in the expert think tank, wherein the expert think tank contains common and special medicines, different diseases and populations, and personalized medication guidance.

[0019] Preferably, the construction process of the expert think tank includes:

[0020] Collect and organize medication data provided by the expert team on common and special drugs, and medication recommendations for different conditions and populations;

[0021] Using a natural language processing method to perform structured processing on the medication data to construct a second question-answer pair data set;

[0022] Based on the second question-answer pair data set, a searchable expert think tank is established to provide matching expert-level medication recommendations when users ask questions.

[0023] Preferably, it also includes a digital human avatar module, which is used to convert the answers generated by the answer generation module into natural speech, and enhance the user's understanding and trust through the visual presentation of the digital human, so that the user can obtain medication guidance in a more intuitive and interactive way.

[0024] Preferably, the construction process of the digital human avatar module includes:

[0025] By collecting voice samples of pharmacists and using deep learning technology to perform audio analysis, we can extract voice features and create a voice model.

[0026] Image processing methods were used to analyze static photos of pharmacists, extract facial features, build a three-dimensional model of the digital human avatar, and use animation technology to show expression changes and lip sync;

[0027] The sound model and the three-dimensional model are combined, a speech synthesis method and a lip synchronization algorithm are integrated, and the sound and lip movement of the digital human avatar are synchronously processed.

[0028] Preferably, it also includes a communication module for establishing a data transmission channel between the user interaction module and the server, and between the server and the answer generation module.

[0029] Preferably, the communication module is also used to establish a data transmission channel between the digital human avatar module and the answer generation module, so as to efficiently transmit the answer generated by the answer generation module to the digital human avatar module, thereby realizing real-time speech synthesis and broadcasting.

[0030] Compared with related technologies, the AI-based digital pharmacist medication guidance system provided by the present invention has the following beneficial effects:

[0031] The present invention constructs a drug instruction knowledge base, which covers a large amount of drug information, and ensures the accuracy and timeliness of the information through strict preprocessing and manual review processes, effectively solving the problems of incomplete information and delayed updates in traditional paper instructions.

[0032] By using retrieval-augmented generation (RAG) technology and a pre-trained large-scale language model, the present invention can accurately understand the user's question intention, quickly retrieve relevant drug information, generate coherent and accurate answers, and improve the speed and quality of information retrieval.

[0033] By building an expert think tank, the present invention can provide professional medication recommendations for different groups of people (such as the elderly, pregnant women, children, etc.) and special conditions, making up for the deficiency of the existing system in providing personalized services.

[0034] With the help of realistic digital human avatar technology, the present invention can present drug information in voice and visual forms, enhancing the user's understanding and trust, making the originally complex drug information easy to understand, especially for non-professionals. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 A schematic diagram of the structure of the artificial intelligence-based digital pharmacist medication guidance system provided by the present invention;

[0036] Figure 2 A workflow diagram of the artificial intelligence-based digital pharmacist medication guidance system provided by the present invention;

[0037] Figure 3 This is a flow chart of constructing the digital human avatar module of the present invention. DETAILED DESCRIPTION

[0038] The present invention will be further described in detail below in conjunction with the accompanying drawings and embodiments. It is to be understood that the specific embodiments described herein are only used to explain the present invention, rather than to limit the present invention. It should also be noted that, for ease of description, only the parts related to the present invention, rather than all structures, are shown in the accompanying drawings. In addition, the embodiments of the present invention and the features in the embodiments may be combined with each other without conflict.

[0039] It should also be noted that, for ease of description, only the part relevant to the present invention but not all the content is shown in the accompanying drawings. It should be mentioned before discussing the exemplary embodiments in more detail that some exemplary embodiments are described as the processing or method described as a flow chart. Although the flow chart describes each operation (or step) as being processed sequentially, many operations therein can be implemented in parallel, concurrently or simultaneously. In addition, the order of each operation can be rearranged. When its operation is completed, the processing can be terminated, but it can also have additional steps not included in the accompanying drawings. The processing can correspond to a method, a function, a procedure, a subroutine, a subprogram, etc.

[0040] At present, the acquisition of drug information mainly relies on paper instructions or electronic documents. This method has problems such as insufficient vividness of information transmission and lack of personalized experience. In addition, due to the lack of accuracy and completeness of information, users may have reduced trust in drug information and affect medication compliance. Therefore, a system that can provide accurate, vivid and convenient access to drug information is needed.

[0041] Therefore, the present invention provides a digital pharmacist medication guidance system based on artificial intelligence, referring to Figures 1 to 3 As shown, it includes a user interaction module, a server, an answer generation module, an expert think tank module, a digital human avatar module and a communication module, aiming to provide accurate, vivid and convenient drug information services, ensure that users can obtain high-quality drug information and reduce the risk of medication errors.

[0042] The user interaction module is responsible for receiving questions from users and passing them to the server. The user interaction module can be implemented in the form of, but not limited to, web pages, mobile applications, etc. Users can ask questions by text input or voice input.

[0043] The server is used to encapsulate the questions raised by the user and submit them to the answer generation module. At the same time, the server is also responsible for managing the communication between the user interaction module and the answer generation module.

[0044] The answer generation module is the core part of this system. It receives questions from the server, uses the retrieval-augmented generation (RAG) method to retrieve information related to the question in the pre-built drug instructions knowledge base, and uses the pre-trained language model to integrate the retrieved information to generate answers to the questions. Finally, the answers will be returned to the user interaction module in a streaming form.

[0045] Exemplary, reference Figure 2 As shown, it is a workflow diagram of the medication guidance system, which is specifically:

[0046] The user initiates a question request through the user interaction module. At this time, the user interface displays a waiting status prompt to inform the user that the system is processing the request.

[0047] The user interaction module sends the question request to the server. After receiving the question, the server encapsulates it to ensure that the question is submitted to the back-end processing module in a unified format.

[0048] The encapsulated question is forwarded by the server to the answer generation module. The answer generation module uses the retrieval-augmented generation (RAG) technology to search the pre-built drug instructions knowledge base to find drug information related to the user's question.

[0049] After retrieving relevant information, the answer generation module uses a pre-trained language model to integrate the search results and generate answers to user questions. The generated answers are returned to the server in a streaming manner, that is, they are sent word by word as the answers are generated.

[0050] As answers are generated, the system synchronously stores them in segments, which means that the answers are divided into several small paragraphs or sentences for easy subsequent processing and management. At the same time, the system extracts the text content of each segment to prepare for the next step of speech synthesis.

[0051] After receiving the segmented answer, the server requests the Text-to-Speech (TTS) service. In this step, the system converts the extracted text content into a voice file to realize the real-time broadcast function of the digital human avatar.

[0052] The answer generation module returns the address of the voice file to the server. The server caches the received voice file to ensure that the voice file can be loaded quickly and improve the response speed.

[0053] On the front end, the server displays the voice file and the corresponding text content to the user in a progressive manner. Specifically, as the voice is played, the corresponding text content will be displayed line by line on the user interface to help users better understand and follow the information.

[0054] At the same time, the pharmacist image in the digital human avatar module will also switch and display accordingly according to the progress of voice playback. For example, when the digital human avatar announces the usage and dosage of a certain medicine, its mouth shape and expression will change synchronously with the voice to enhance the user's visual and auditory experience.

[0055] At the same time, in the design of the answer production module, the drug instructions knowledge base, as its core component, directly determines the accuracy and comprehensiveness of the output answers. The drug instructions knowledge base integrates detailed and structured drug information, which is derived from officially approved drug instructions, ensuring the authority and reliability of the data. When the answer production module is operating, the drug instructions knowledge base, as a back-end support, can quickly respond to query requests, accurately match and integrate relevant drug information, and thus generate answers that meet user needs.

[0056] In the process of building a drug instructions knowledge base, it is necessary to use advanced data processing technology and information extraction methods to efficiently and accurately parse and structure a large number of drug instructions to form a data format that is easy to query and use. In addition, the update and maintenance mechanism of the knowledge base must also be considered to ensure the timeliness and accuracy of its content.

[0057] Specifically, the construction process of the drug instructions knowledge base includes:

[0058] S11: Collect drug instructions through multiple channels and perform data preprocessing on the collected drug instructions, where the preprocessing includes removing duplicate information, correcting erroneous data and unifying the format.

[0059] In this embodiment, drug instructions are first widely collected through multiple authoritative channels to ensure the comprehensiveness and reliability of the data. These channels include but are not limited to drug manufacturers, regulatory agencies (such as official agencies such as the State Drug Administration NMPA), medical institutions (such as hospitals, clinics, etc.), and professional databases (such as PubMed, Cochrane Library, etc.). The latest drug instructions are obtained directly from drug manufacturers, which usually contain the most accurate and detailed information. The drug information released by regulatory agencies is highly authoritative and credible. The drug instructions used by medical institutions are often combined with clinical practice experience and provide specific guidance on the use of drugs. The drug information in professional databases has been strictly screened and reviewed and has high academic value.

[0060] The collected drug instructions need to be preprocessed in detail to improve data quality and lay the foundation for subsequent steps. The preprocessing process includes removing duplicate information, identifying and deleting duplicate drug instructions by comparing information such as drug name, manufacturer, and approval number to avoid data redundancy; checking for typos, format errors, logical contradictions, and other problems in the instructions, and making corrections.

[0061] For information that cannot be confirmed, it is verified by consulting other authoritative channels; drug instructions from different sources are uniformly converted into a standard format, including font, font size, paragraph format, etc., to facilitate subsequent processing and analysis; the information in the instructions is structured according to preset key fields, such as drug information, usage and dosage, indications, adverse reactions, precautions, etc., to prepare for the subsequent construction of a question-and-answer dataset.

[0062] S12: Structural classification of the preprocessed drug instructions is performed according to preset key fields to obtain content blocks corresponding to the key fields, wherein the key fields include drug information, usage and dosage, indications, adverse reactions and precautions.

[0063] In this embodiment, the pre-processed drug instructions are structured and classified according to preset key fields. These key fields include drug information (including basic information such as drug name, ingredients, dosage form, specifications, manufacturer, approval number, etc.), usage and dosage (detailed description of drug dosage, frequency of use, route of use, etc.), indications (listing which diseases or symptoms the drug is suitable for), adverse reactions (listing possible adverse reactions or side effects after using the drug), and precautions (reminding patients or medical staff of matters that need to be paid attention to during the medication process, such as contraindications, drug interactions, etc.). The information in the instructions is automatically extracted and classified into corresponding fields through natural language processing technology (NLP) and regular expressions. For information that cannot be automatically identified, it is supplemented and improved through manual review and annotation.

[0064] S13: Build the first question-answer pair dataset based on all structured content blocks.

[0065] In this embodiment, the specific steps include automatically generating corresponding questions for each key field. For example, for the "Usage and Dosage" field, the question "What is the usage and dosage of XX drug?" can be generated; for the "Indications" field, the question "What diseases or symptoms is XX drug suitable for?" can be generated. The answers corresponding to the questions are extracted from the structured content blocks to ensure the accuracy and completeness of the answers and avoid missing important information. The questions and answers are combined into question-answer pairs and stored in the database. Each question-answer pair contains a question ID, question text, answer text, and corresponding drug information. By constructing the first question-answer pair data set, rich training data is provided for the subsequent training of the language model.

[0066] S14: Utilize the retrieval enhancement generation RAG method and train the selected language model in combination with the first question-answer pair dataset to generate a drug instruction knowledge base.

[0067] In this embodiment, the specific steps include selecting a language model with strong language understanding capabilities and large-scale pre-training data, such as Alibaba's open source Tongyi Qianwen large model, which can accurately understand the user's question intention and quickly extract key information from the knowledge base.

[0068] The first question-answer pair dataset is divided into a training set, a validation set, and a test set. The training set is used to train the model, the validation set is used to adjust the model parameters, and the test set is used to evaluate the model performance.

[0069] Use pre-trained language model parameters for initialization to ensure that the model has a certain language understanding ability.

[0070] During the training process, the RAG method is used. When a user asks a question, the system first uses the retrieval module to quickly retrieve content blocks related to the question in the knowledge base, and then passes the retrieved content blocks as input to the generation module to generate a coherent and accurate answer. The performance and accuracy of the model are improved by continuously adjusting the model parameters and optimizing the training algorithm, for example, adjusting the similarity threshold of the retrieval module and the decoding strategy of the generation module.

[0071] In addition, a manual review process is introduced to screen and optimize the answers generated by the language model during the training process to ensure that the generated answers meet the review standards of professional pharmacists and improve the accuracy and reliability of the knowledge base. After sufficient training, the trained language model is deployed to the server to form a drug instructions knowledge base that can respond to user query requests in real time and provide accurate and detailed drug information.

[0072] Through the above steps, the present invention successfully constructed a drug instructions knowledge base based on RAG technology, providing powerful data support for the subsequent medication guidance system.

[0073] Specifically, it also includes an expert think tank, which is used to provide expert-level medication recommendations when the questions raised by users match the questions in the expert think tank. The expert think tank contains common and special medicines, different diseases and populations, as well as personalized medication guidance.

[0074] For example, when a user asks "What is the dosage of amoxicillin?", the drug instructions knowledge base will quickly extract relevant content from a large amount of drug information based on the trained knowledge and algorithms, and generate the answer "Adults take 0.5 grams orally each time, once every 6-8 hours, and the daily dose does not exceed 4 grams." If the user's question has a more detailed answer in the expert think tank, such as medication recommendations for specific conditions or populations, the digital human avatar module will broadcast this professional information to the patient to help the patient make more reasonable medication guidance.

[0075] Specifically, the construction process of the expert think tank includes:

[0076] S21: Collect and organize medication data provided by the expert team on common and special medicines, and medication recommendations for different conditions and populations.

[0077] In this embodiment, firstly, relevant data on common and special drugs, different conditions and population medication recommendations provided by a professional pharmacist team need to be collected and sorted. These data come from pharmacists with rich clinical experience and professional knowledge, ensuring the professionalism and practicality of the information. When sorting the data, ensure that each medication recommendation can be traced back to a specific provider or relevant research literature to ensure the authenticity and authority of the information.

[0078] S22: Use natural language processing methods to perform structured processing on medication data and construct the second question-answer pair data set.

[0079] In this embodiment, specifically, natural language processing technology (NLP) is used to clean, annotate and extract the collected medication data. First, irrelevant or redundant information is removed through text cleaning to retain useful core content. Secondly, named entity recognition (NER) and relationship extraction (RE) technology are used to automatically extract relevant entities such as drug names, indications, usage and dosage, adverse reactions and their relationships from the text. On this basis, a question-answer pair is constructed, that is, for each medication suggestion, a corresponding question is constructed, and the suggestion is used as the answer. For example, if the medication suggestion mentions "For patients with hypertension, it is recommended to use XX drug, once a day, one tablet each time", the constructed question may be "What is the recommended medication for patients with hypertension?" The answer is the above suggestion. Finally, these question-answer pairs are stored in the database in a unified format to form a second question-answer pair data set.

[0080] S23: Based on the second question-answer pair dataset, a searchable expert think tank is established to provide matching expert-level medication recommendations when users ask questions.

[0081] In this embodiment, first, the second question-answer data set is indexed so that the system can quickly locate answers related to the user's question. Then, a retrieval algorithm is designed and implemented, which can quickly find the most relevant medication recommendations in the indexed data set based on the content of the question after receiving the user's question.

[0082] In addition, considering that medication recommendations may vary due to individual differences, the system must also have a certain fuzzy matching ability so that it can find appropriate recommendations even if the user's question statement is not completely consistent with the exact question and answer pairs in the dataset.

[0083] Finally, the matched medication recommendations are broadcast to the user in the form of natural voice through the digital human avatar, ensuring that the user can clearly understand the expert's recommendations and make reasonable medication decisions. In this way, the present invention establishes a searchable expert think tank to provide users with authoritative and professional medication guidance.

[0084] The implementation steps of the retrieval algorithm include:

[0085] First, the received user questions are preprocessed, including removing stop words, punctuation marks and other unnecessary characters, and performing stemming or word form restoration to ensure the brevity and consistency of the questions. This can improve the efficiency and accuracy of subsequent retrieval.

[0086] The most important keywords are extracted from the preprocessed user questions using text feature extraction methods including but not limited to TF-IDF (TermFrequency-InverseDocumentFrequency). These keywords will be used for subsequent index matching.

[0087] An inverted index is constructed based on the second question-answer pair dataset. An inverted index is a data structure that maps words that appear in a document to all documents that contain these words. When constructing an inverted index, for each keyword, it is recorded in which question-answer pairs it appears and the frequency of its appearance.

[0088] When a user asks a question and enters the system, the inverted index is used to quickly locate the question-answer pair containing the question keywords. Then, the similarity score between the user's question and the candidate question-answer pair is calculated using metrics including but not limited to cosine similarity and Jaccard similarity. The higher the similarity score, the better the match.

[0089] The candidate question-answer pairs are sorted according to the calculated similarity scores, and the top few with the highest scores are selected as the final results. In addition, a threshold needs to be set, and only question-answer pairs with scores exceeding this threshold will be considered as part of the final results to ensure the quality of the retrieval results.

[0090] Specifically, it also includes a digital human avatar module, which is used to convert the answers generated by the answer generation module into natural speech, and enhance the user's understanding and trust through the visual presentation of the digital human, so that the user can obtain medication guidance in a more intuitive and interactive way.

[0091] Specifically, the construction process of the digital human avatar module includes:

[0092] S31: By collecting voice samples of pharmacists and using deep learning technology to perform audio analysis, voice features are extracted and a voice model is created.

[0093] In this embodiment, the specific method is to first record the pharmacist's voice sample. Generally, at least one minute of continuous recording is required to capture the pharmacist's voice characteristics, including frequency, amplitude, rhythm, etc. Then, the collected sound samples are analyzed using a deep learning framework to extract the key features of the pharmacist's voice. Based on these features, a voice model of the digital human avatar is constructed so that the digital human avatar can simulate the pharmacist's voice for broadcasting.

[0094] S32: Use image processing methods to analyze static photos of pharmacists, extract facial features, build a three-dimensional model of the digital human avatar, and use animation technology to display expression changes and lip synchronization.

[0095] In this embodiment, first, a full-body photo of the pharmacist is collected at high resolution, and the pharmacist's facial features, including eyes, nose, mouth and other facial features, are extracted using image processing technology. Then, a three-dimensional model of the digital human avatar is constructed using these facial features. On this basis, animation technology is used to simulate the pharmacist's facial expression changes, such as blinking, smiling, etc., and ensure that the digital human can accurately show the changes in the mouth shape when speaking when broadcasting.

[0096] S33: Combining the sound model and the three-dimensional model, integrating the speech synthesis method and the lip synchronization algorithm, the sound and lip shape movement of the digital human avatar are synchronously processed.

[0097] In this embodiment, the specific steps include combining the previously created sound model with the three-dimensional model, using speech synthesis technology to convert text information into speech output, and driving the mouth movement of the digital human avatar through a lip synchronization algorithm to synchronize it with the speech output. In this way, when the digital human avatar broadcasts drug information, its mouth movement can be synchronized with the pronunciation, making the broadcast more natural and smooth, and enhancing the consistency of the user's auditory and visual experience.

[0098] In addition, it is necessary to optimize speech synthesis technology so that the digital human avatar's voice broadcast is natural and smooth, with a good sense of rhythm and intonation, to further enhance the user's auditory experience. When broadcasting in real time, the digital human avatar can adjust the speed, volume and tone according to the context and the user's response to enhance the naturalness of the interaction.

[0099] Specifically, the system also includes a communication module, which is used to establish an efficient data transmission channel to ensure that data exchange between modules can be carried out quickly and stably. Specifically, the functions of the communication module are reflected in the following aspects:

[0100] First, the communication module is used to establish a data transmission channel between the user interaction module and the server. When the user asks questions through the user interaction module, the communication module is responsible for transmitting these questions to the server in real time. This process needs to ensure the speed and stability of data transmission so that the server can receive the user's questions in time and process them.

[0101] Secondly, the communication module also establishes a data transmission channel between the server and the answer generation module. When the server receives the user's question, it encapsulates the question and sends it to the answer generation module. The communication module ensures that the data transmission in this process is fast and reliable, so that the answer generation module can start processing the user's question in time, retrieve relevant information from the knowledge base and generate answers.

[0102] In addition, the communication module is also used to establish a data transmission channel between the digital human avatar module and the answer generation module. When the answer generation module generates answers to user questions, the communication module is responsible for efficiently transmitting these answers to the digital human avatar module. This process requires special attention to the delay of data transmission, because only when the answers can be transmitted to the digital human avatar module in real time can real-time speech synthesis and broadcasting be achieved, thus providing users with a smooth interactive experience.

[0103] To achieve the above functions, the communication module uses a variety of network communication protocols and technologies, such as the TCP / IP protocol stack to ensure the reliability and sequentiality of data transmission, the UDP protocol to reduce the negative impact of delays on real-time performance, and WebSocket and other technologies to maintain long connections and reduce the time overhead of establishing a connection for each request.

[0104] The working principle of the AI-based pharmacist medication guidance system provided by the present invention is as follows:

[0105] By integrating advanced technical modules, the system provides users with vivid, intuitive, personalized and efficient drug information services. First, the system uses the real-time broadcast function of the digital human avatar to present drug-related information to users in a more vivid and intuitive way. By combining voice broadcast, drug information is easier to understand and accept, thereby improving the effect of information communication. Compared with traditional text information, this method can better attract users' attention and reduce the possibility of information misunderstanding.

[0106] In addition, the system can create personalized pharmacist images in batches according to user needs and preferences. Each digital avatar can represent a different pharmacist and provide users with a unique service experience. This personalized service enhances the interaction and emotional connection between users and products, and improves user satisfaction and loyalty.

[0107] In order to enhance user trust, the broadcast of the digital human avatar is based on the review and certification of the expert think tank to ensure that the drug information provided is accurate and authoritative. The image and voice of the pharmacist give users a professional and reliable feeling, allowing users to use drugs with more confidence. At the same time, relying on a powerful knowledge base and advanced model training, the system can provide accurate and detailed drug information. Based on QAnydrugs' powerful drug information knowledge base management system, it ensures that users can obtain high-quality drug information and reduce the risk of medication errors.

[0108] This system covers 23,093 oral drugs, including almost all types of oral drugs commonly used in clinical practice, and can meet users' needs for a variety of drug information. Whether it is common drugs or rare drugs, the system can provide accurate information and provide users with comprehensive services. Users only need to ask questions, and the digital human avatar can broadcast the answers in real time without tedious operations, which greatly improves the convenience of use.

[0109] The system uses RAG technology to understand the user's question intention and extract relevant information from the knowledge base to answer. The digital human avatar can also further explain and illustrate drug information based on user feedback, providing a more intelligent interactive experience. Whether in a hospital, pharmacy or home, users can use this system to obtain drug information anytime and anywhere to meet the needs of different scenarios.

[0110] The system has strong scalability and can be easily expanded and upgraded as drug information is continuously updated and user needs change, such as adding new drug information, improving model algorithms, optimizing the performance of digital human avatars, etc., to adapt to market development and user needs.

[0111] Ultimately, this system can provide patients and medical staff with reliable drug information support, helping them make correct decisions and thus improving the safety and effectiveness of medical services.

[0112] The present application is described with reference to flowcharts and / or block diagrams of methods, devices (systems) and computer program products according to embodiments of the present application. It should be understood that each process and / or box in the flowchart and / or block diagram, as well as the combination of processes and / or boxes in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor or other programmable data processing device to generate a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0113] A person of ordinary skill in the art can understand that all or part of the steps in the various methods of the above 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 programmable read-only memory (EEPROM), a compact disc read-only memory (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.

[0114] It should also be noted that the terms "include", "comprises" or any other variations thereof are intended to cover non-exclusive inclusion, so that a process, method, commodity or device including a series of elements includes not only those elements, but also other elements not explicitly listed, or also includes elements inherent to such process, method, commodity or device. In the absence of more restrictions, the elements defined by the sentence "comprises a ..." do not exclude the existence of other identical elements in the process, method, commodity or device including the elements.

Claims

1. The AI-based digital pharmacist medication guidance system includes a user interaction module, a server and an answer generation module, and is characterized in that: The user interaction module is used to receive and initiate questions raised by users to the server; The server is used to encapsulate the questions raised by the user and submit them to the answer generation module; The answer generation module is used to receive questions from the server, use the retrieval enhancement generation RAG method to retrieve information related to the question in a pre-built drug instruction manual knowledge base, and use a pre-trained language model to integrate the retrieved information to generate an answer to the question, and finally, stream the answer to the question to the user interaction module.

2. The AI-based digital pharmacist medication guidance system according to claim 1 is characterized in that: In the medication guidance system, the construction process of the drug instructions knowledge base includes: Collecting drug instructions through multiple channels and performing data preprocessing on the collected drug instructions, wherein the preprocessing includes removing duplicate information, correcting erroneous data and unifying the format; Structural classification of the preprocessed drug instructions according to preset key fields to obtain content blocks corresponding to the key fields, wherein the key fields include drug information, usage and dosage, indications, adverse reactions and precautions; Build the first question-answer pair dataset based on all structured content blocks; The retrieval enhancement generation RAG method is used, and the selected language model is trained in combination with the first question-answer pair dataset to generate a drug instruction manual knowledge base.

3. The AI-based digital pharmacist medication guidance system according to claim 2 is characterized in that: The training process of the language model also includes: A manual review process is introduced to screen the answers generated by the language model during the training process.

4. The AI-based digital pharmacist medication guidance system according to claim 1 is characterized in that: It also includes an expert think tank, which is used to provide expert-level medication recommendations when the questions raised by the user match the questions in the expert think tank, wherein the expert think tank contains common and special medicines, different diseases and populations, and personalized medication guidance.

5. The AI-based digital pharmacist medication guidance system according to claim 4 is characterized in that: The construction process of the expert think tank includes: Collect and organize medication data provided by the expert team on common and special drugs, and medication recommendations for different conditions and populations; Using a natural language processing method to perform structured processing on the medication data to construct a second question-answer pair data set; Based on the second question-answer pair data set, a searchable expert think tank is established to provide matching expert-level medication recommendations when users ask questions.

6. The AI-based digital pharmacist medication guidance system according to claim 1 is characterized in that: It also includes a digital human avatar module, which is used to convert the answers generated by the answer generation module into natural speech, and enhance the user's understanding and trust through the visual presentation of the digital human, so that the user can obtain medication guidance in a more intuitive and interactive way.

7. The AI-based digital pharmacist medication guidance system according to claim 6 is characterized in that: The construction process of the digital human avatar module includes: By collecting voice samples of pharmacists and using deep learning technology to perform audio analysis, we can extract voice features and create a voice model. Image processing methods were used to analyze static photos of pharmacists, extract facial features, build a three-dimensional model of the digital human avatar, and use animation technology to show expression changes and lip sync; The sound model and the three-dimensional model are combined, a speech synthesis method and a lip synchronization algorithm are integrated, and the sound and lip movement of the digital human avatar are synchronously processed.

8. The AI-based digital pharmacist medication guidance system according to claim 7 is characterized in that: It also includes a communication module for establishing a data transmission channel between the user interaction module and the server, and between the server and the answer generation module.

9. The AI-based digital pharmacist medication guidance system according to claim 8 is characterized in that: The communication module is also used to establish a data transmission channel between the digital human avatar module and the answer generation module, so as to efficiently transmit the answer generated by the answer generation module to the digital human avatar module, thereby realizing real-time speech synthesis and broadcasting.

Images