Nursing question and answer method, device and equipment based on medical industry large model and medium

Through the nursing Q&A method based on the medical industry big model, the problem of insufficient language understanding of the nursing professional by artificial intelligence system is solved, and the accurate identification and response of nursing problems is achieved, which improves the efficiency and quality of nursing work.

CN118982073APending Publication Date: 2024-11-19SHANDONG LANGCHAO YUNTOU INFORMATION TECH CO LTD
View PDF 7 Cites 0 Cited by

Patent Information

Application Number
CN202411226414.6
Authority / Receiving Office
CN · China
Patent Type
Applications(China)
Current Assignee / Owner
Filing Date
2024-09-03
Publication Date
2024-11-19

AI Technical Summary

Technical Problem

Most existing artificial intelligence systems have a shallow understanding of the professional language of nursing, and it is difficult to accurately identify specific problems in nursing work and respond to them, resulting in low efficiency and quality in drug management, patient symptom monitoring, and special nursing guidance operations.

Method used

Adopt nursing Q&A method based on the medical industry big model, and obtain user questions and use language processing models to convert them into target nursing problem texts. Based on preset semantic matching technology, determine relevant knowledge points from the nursing knowledge graph, and generate templated answer structures. , and iteratively trained through a convergent question and answer model to generate accurate nursing answers.

Benefits of technology

It improves the efficiency and quality of nursing work, provides accurate Q&A services that meet medical professional standards, and supports daily decision-making of nursing staff.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN118982073A_ABST
    Figure CN118982073A_ABST
Patent Text Reader

Abstract

The invention discloses a nursing question-answering method and device based on a medical industry large model, equipment and a medium, and relates to the field of artificial intelligence, and the method comprises the steps: obtaining a nursing question sent by a user side, and converting the nursing question into a target nursing question text through a language processing model in a target medical industry large model; determining a plurality of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value of the target knowledge point and the target nursing problem text is greater than the vector similarity values of other knowledge points in the nursing knowledge graph and the target nursing problem text; and generating a nursing answer candidate set based on the templated answer structures corresponding to the plurality of target knowledge points, and performing iterative training on a convergence question and answer model in the target medical industry large model by using the nursing answer candidate set to obtain a target nursing answer corresponding to the target nursing question text. Nursing questions and answers based on the large model of the medical industry are realized.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence, and in particular to a nursing question-answering method, device, equipment and medium based on a large model of the medical industry. Background Art

[0002] At present, the introduction of artificial intelligence has brought innovation to the medical field, especially in diagnosis support, case management and other aspects. However, in the sub-field of nursing question and answer, most of the existing artificial intelligence systems have a shallow understanding of nursing professional language and find it difficult to accurately identify and respond to specific problems in nursing work, such as drug management, patient symptom monitoring, special nursing guidance operations, etc.

[0003] Therefore, how to deeply integrate the deep learning capabilities of artificial intelligence with medical care expertise to improve the efficiency and quality of nursing work is an urgent problem that needs to be solved. Summary of the invention

[0004] In view of this, the purpose of the present invention is to provide a nursing question-answering method, device, equipment and medium based on a large model of the medical industry, which can deeply integrate the deep learning capabilities of artificial intelligence with medical nursing professional knowledge to effectively support the daily decision-making of nursing staff and improve the efficiency and quality of nursing work. The specific scheme is as follows:

[0005] In a first aspect, the present application discloses a nursing question-answering method based on a large model of the medical industry, comprising:

[0006] Obtain the nursing questions sent by the user, and use the language processing model in the target medical industry big model to convert the nursing questions into target nursing question texts;

[0007] Based on a preset semantic matching technology, a plurality of target knowledge points corresponding to the target nursing problem text are determined from a pre-constructed nursing knowledge graph; a vector similarity value between the target knowledge point and the target nursing problem text is greater than a vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text;

[0008] A nursing answer candidate set is generated based on the templated answer structure corresponding to the several target knowledge points, and the convergent question-answering model in the target medical industry big model is iteratively trained using the nursing answer candidate set to obtain the target nursing answer corresponding to the target nursing question text.

[0009] Optionally, before obtaining the nursing question sent by the user terminal and converting the nursing question into a target nursing question text by using a language processing model in the target medical industry big model, the method further includes:

[0010] Performing data preprocessing and data annotation processing on the collected medical care data to obtain processed care data, and using the processed care data to fine-tune the pre-trained large model to obtain the converged question-answering model;

[0011] The language processing model is determined using natural language processing technology, and the target medical industry big model is constructed based on the language processing model and the convergent question-answering model.

[0012] Optionally, the nursing knowledge graph is a structured knowledge graph generated based on pre-acquired nursing knowledge.

[0013] Optionally, the step of acquiring the nursing question sent by the user terminal and converting the nursing question into a target nursing question text by using a language processing model in a target medical industry big model includes:

[0014] Obtain the nursing voice questions sent by the user and send them to the language processing model in the target medical industry big model;

[0015] Using the text understanding model in the language processing model to perform semantic analysis operations, entity recognition operations, and relationship extraction operations on the nursing voice question to obtain a text understanding result, and using the speech recognition technology in the language processing model to perform speech recognition on the nursing voice question to obtain a speech recognition result;

[0016] The text understanding result and the speech recognition result are semantically aligned using cross-modal semantic alignment technology to obtain the target nursing question text.

[0017] Optionally, the step of determining a plurality of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text being greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text comprising:

[0018] The target nursing question text is converted into a question vector using a text vectorization model, and a vector space model or a graph neural network is used to determine the vector similarity value between all knowledge points in the nursing knowledge graph and the question vector;

[0019] Based on the vector similarity value, several target knowledge points corresponding to the target nursing problem text are determined from a pre-constructed nursing knowledge graph; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text.

[0020] Optionally, the iterative training of the convergent question-answering model in the target medical industry big model using the nursing answer candidate set to obtain a target nursing answer corresponding to the target nursing question text includes:

[0021] Acquiring feedback information from the user terminal for the nursing answer candidate set, and iteratively training the convergent question-answering model based on the feedback information to obtain a trained convergent question-answering model;

[0022] The trained convergent question-answering model is used to generate a target nursing answer set corresponding to the target nursing question text, and a preset answer scoring method is used to determine a target nursing answer from the target nursing answer set.

[0023] Optionally, the step of determining a target nursing answer from the target nursing answer set by using a preset answer scoring method includes:

[0024] Scoring all nursing answers in the target nursing answer set based on a preset quality assessment method to obtain scoring results corresponding to all nursing answers in each preset assessment dimension; the preset assessment dimensions include answer accuracy, answer completeness and answer credibility;

[0025] Based on a preset comprehensive ranking method, the corresponding scoring results on each of the preset evaluation dimensions are comprehensively processed to obtain a target scoring result, and the nursing answer with the largest target scoring result in the target nursing answer set is determined as the target nursing answer.

[0026] In a second aspect, the present application discloses a nursing question-answering device based on a large model of the medical industry, comprising:

[0027] A question conversion module is used to obtain the nursing questions sent by the user end and convert the nursing questions into target nursing question texts by using the language processing model in the target medical industry big model;

[0028] A knowledge point determination module is used to determine a number of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text;

[0029] The target answer determination module is used to generate a nursing answer candidate set based on the templated answer structure corresponding to the several target knowledge points, and use the nursing answer candidate set to iteratively train the convergent question-answering model in the target medical industry big model to obtain the target nursing answer corresponding to the target nursing question text.

[0030] In a third aspect, the present application discloses an electronic device, comprising:

[0031] Memory, used to store computer programs;

[0032] A processor is used to execute the computer program to implement the steps of the aforementioned nursing question-and-answer method based on the big model of the medical industry.

[0033] In a fourth aspect, the present application discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the steps of the aforementioned nursing question-and-answer method based on a big model of the medical industry are implemented.

[0034] In the present application, when using the medical industry big model for nursing question and answer, the nursing questions sent by the user end are first obtained, and the language processing model in the target medical industry big model is used to convert the nursing questions into target nursing question texts; based on the preset semantic matching technology, several target knowledge points corresponding to the target nursing question texts are determined from the pre-built nursing knowledge graph; the vector similarity value between the target knowledge point and the target nursing question text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing question text; a nursing answer candidate set is generated based on the templated answer structure corresponding to the several target knowledge points, and the convergent question and answer model in the target medical industry big model is iteratively trained using the nursing answer candidate set to obtain the target nursing answer corresponding to the target nursing question text. It can be seen that this application converts the user's nursing questions into target nursing question texts using the language processing model in the target medical industry big model, so that the target medical industry big model can understand the nursing questions more deeply, and then uses the preset semantic matching technology to determine the target knowledge points corresponding to the nursing questions based on the nursing knowledge graph, and based on the target knowledge points and the templated answer structure corresponding to the target knowledge points, it can generate an accurate and standardized nursing answer candidate set for the nursing questions, and then use the convergent question and answer model to converge the nursing answers in the nursing answer candidate set to determine the most appropriate nursing answer as the target nursing answer, and finally send the target nursing answer to the user end, so as to give answers and guidance suggestions that meet medical professional standards, provide accurate question and answer services, effectively support nursing staff's daily decision-making, and improve the efficiency and quality of nursing work. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0036] Figure 1A flowchart of a nursing question-and-answer method based on a large model of the medical industry disclosed in this application;

[0037] Figure 2 A flowchart of a specific nursing question-and-answer method based on a large model of the medical industry disclosed in this application;

[0038] Figure 3 This is a schematic diagram of information flow of a nursing question-and-answer module based on a large model of the medical industry disclosed in this application;

[0039] Figure 4 This is a schematic diagram of the structure of a nursing question-and-answer device based on a large model of the medical industry disclosed in this application;

[0040] Figure 5 This is a structural diagram of an electronic device disclosed in this application. DETAILED DESCRIPTION

[0041] 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.

[0042] At present, when using artificial intelligence systems to provide professional knowledge support for nursing questions and answers, most artificial intelligence systems have a shallow understanding of nursing professional language, which makes it difficult for the medical industry big model built by artificial intelligence systems to accurately identify and respond to specific problems in nursing work, such as drug management, patient symptom monitoring, special nursing guidance operations, etc. In order to solve the above technical problems, this application discloses a nursing question and answer method, device, equipment and medium based on a medical industry big model, which can deeply integrate the deep learning capabilities of artificial intelligence with medical nursing professional knowledge to effectively support the daily decision-making of nursing staff and improve the efficiency and quality of nursing work.

[0043] See also Figure 1 As shown, the embodiment of the present invention discloses a flow chart of a nursing question-answering method based on a large model of the medical industry, including:

[0044] Step S11, obtaining the nursing questions sent by the user end, and using the language processing model in the target medical industry big model to convert the nursing questions into target nursing question texts.

[0045] In this embodiment, in order to achieve the characteristics of nursing profession, the large-scale medical nursing data is used to carefully customize and optimize the medical industry big model, so that the medical industry big model can deeply understand the nursing professional language and accurately respond to various practical problems in nursing work. Before obtaining the nursing problems sent by the user end and using the language processing model in the target medical industry big model to convert the nursing problems into target nursing problem texts, it is also necessary to perform data preprocessing operations and data annotation processing on the collected medical nursing data to obtain processed nursing data, wherein the data preprocessing operation is to perform denoising and deduplication processing on the collected data to ensure data quality, and the data annotation processing is performed by medical experts to professionally annotate the data, including but not limited to keywords, disease types, nursing operation steps, etc., to provide high-quality supervision information for model training. And the processed nursing data is used to fine-tune the pre-trained big model to obtain a convergent question-answering model, and then the language processing model is determined by NLP (Natural Language Processing, natural language processing technology), so as to use the obtained language processing model and convergent question-answering model to construct the target medical industry big model, so as to ensure that the target medical industry big model can deeply understand the nursing professional language and complex situations, and provide more accurate question-answering services for nursing staff.

[0046] After the preliminary preparations are completed, the acquired nursing questions can be processed. In a specific embodiment, the nursing questions raised by the nursing staff are in the form of voice, and the language processing model in the target medical industry big model is used to convert the nursing questions into target nursing question texts. Specifically, it may include: obtaining the nursing voice questions sent by the user end, and sending them to the language processing model in the target medical industry big model, and then using the text understanding model in the language processing model to perform semantic analysis operations, entity recognition operations and relationship extraction operations on the nursing voice questions to obtain text understanding results, and using the speech recognition technology in the language processing model to perform speech recognition on the nursing voice questions to obtain speech recognition results, and finally using the cross-modal semantic alignment technology to semantically align the text understanding results and the speech recognition results to obtain the target nursing question text.

[0047] In a specific embodiment, the text understanding model uses NLP technology to achieve semantic analysis, entity recognition and relationship extraction of natural language questions input by nursing staff. The text understanding model is constructed based on deep learning models, such as BERT (Bidirectional Encoder Representations from Transformers, a language representation model), GPT (Generative Pre-Trained, a deep learning model based on the Internet, trained with available data, and text generation) and other pre-trained models, which perform deep semantic understanding of the text and capture the contextual relationship and implicit meaning in the text. When performing entity recognition operations, NER (Named Entity Recognition) technology is used to accurately locate key entities in the text, such as names, place names, time, etc., laying the foundation for subsequent relationship extraction and knowledge graph construction. When extracting relationships for nursing questions, based on the entity recognition results, dependency syntactic analysis and relationship classifiers are used to identify the relationships between entities, such as belonging relationships, causal relationships, etc., to enhance the system's ability to understand complex contexts. When using the speech recognition technology in the language processing model to perform speech recognition on nursing voice questions, the speech recognition technology can integrate advanced ASR (Automatic Speech Recognition) to efficiently and accurately convert the oral inquiries of nursing staff into text format. In order to more accurately identify nursing voice questions, the language processing model can also use signal processing technology, such as spectral noise reduction and echo cancellation, to achieve noise suppression and speech enhancement, improve recording quality, and improve the accuracy of speech recognition.

[0048] After obtaining the text understanding results and speech recognition results using the language processing model, the cross-modal fusion module in the language processing model is used to semantically align the text understanding results and speech recognition results, and convert the nursing problem into the target nursing problem text, thereby ensuring that the target medical industry big model can uniformly process and understand the intention of the nursing staff to raise the nursing problem regardless of whether the nursing problem is input in voice or text. Specifically, after obtaining the text understanding results and speech recognition results, the pre-designed cross-modal feature fusion algorithm is used to vector-fuse the text understanding results and speech recognition results to construct a unified semantic representation space, and then the semantic representation space is used to construct a cross-modal semantic alignment model, such as a cross-modal attention mechanism, to ensure that inputs under different modalities can be accurately understood and processed by the target medical industry big model. Finally, a consistency verification mechanism is introduced for the cross-modal semantic alignment model to compare the text understanding results and speech recognition results, to ensure the consistency and reliability of the output of the target medical industry big model, thereby realizing the transformation of nursing problems into target nursing problem texts.

[0049] Step S12: determine several target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text.

[0050] In this embodiment, the target medical industry big model uses a preset semantic matching technology to determine several target knowledge points corresponding to the target nursing problem text from the pre-built nursing knowledge graph. Among them, the nursing knowledge graph is a structured knowledge graph generated based on pre-acquired nursing knowledge, and the nursing knowledge graph contains nodes such as diseases, nursing points, nursing assessments, and health guidance and their relationships, so as to more accurately determine several target knowledge points corresponding to the target nursing problem text from the nursing knowledge graph based on the preset semantic matching technology. When determining the knowledge points corresponding to the target nursing problem text from the nursing knowledge graph based on the preset semantic matching technology, a vector space model or a graph neural network can be used to search for the most relevant knowledge points in the knowledge graph according to the problem vector. In a specific implementation, a pre-trained word embedding model, such as Word2Vec (Word to Vector, a group of related models used to generate word vectors), GloVe (Global Vectors for Word Representation, a word representation tool based on global word frequency statistics) or a more advanced Transformer-based model (such as BERT) is first used to convert the user's question into a high-dimensional vector representation to obtain a question vector, and then a vector space model or a graph neural network, such as cosine similarity, is used to calculate the vector similarity value between the question vector and each node in the knowledge graph and the question vector, thereby determining the set of knowledge points most relevant to the nursing problem as the target knowledge point. It can be understood that the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text, that is, the target knowledge point is the knowledge point in the nursing knowledge graph that is most relevant to the nursing problem.

[0051] Step S13: Generate a nursing answer candidate set based on the templated answer structure corresponding to the target knowledge points, and use the nursing answer candidate set to iteratively train the convergent question-answering model in the target medical industry big model to obtain the target nursing answer corresponding to the target nursing question text.

[0052] In this embodiment, after determining the target knowledge point from the nursing knowledge graph, the target medical industry big model designs a templated answer structure for the retrieved target knowledge point, and fills the target knowledge point into the corresponding template according to the generated templated answer structure, thereby forming a more standardized answer, and all the answers generated based on the target knowledge point are used as a preliminary nursing answer candidate set, and then the nursing answer candidate set is used to iteratively train the convergent question and answer model in the target medical industry big model to obtain the target nursing answer corresponding to the target nursing question text. The specific process may include: obtaining feedback information from the user end for the nursing answer candidate set, and iteratively training the convergent question and answer model based on the feedback information to obtain a trained convergent question and answer model; using the trained convergent question and answer model to generate the target nursing answer set corresponding to the target nursing question text, and using a preset answer scoring method to determine the target nursing answer from the target nursing answer set. That is to say, in the process of interaction with the user, through the circular feedback mechanism, the target medical industry big model collects user feedback information, such as whether the nursing answer candidate set is satisfactory, whether additional information is needed, etc. After collecting user feedback, the target medical industry model uses the feedback information as new training data to fine-tune itself online or offline, gradually optimizing the model's parameters to provide more accurate answers.

[0053] In a specific implementation, if the nursing question raised by the nursing staff is which medicines should be used for colds, the target medical industry big model retrieves the relevant medicines from the nursing knowledge graph based on the two keywords of colds and medicines as the nursing answer candidate set and feeds it back to the nursing staff. Then the nursing staff asks what medicines should be used for adults with colds, then the target medical industry big model uses the newly obtained keywords to filter the knowledge points most related to adults, colds and medicines in the nursing knowledge graph, and generates answers based on the knowledge points and feeds them back to the nursing staff. At the same time, the target medical industry big model can also filter the knowledge points most related to children, colds and medicines in the nursing knowledge graph based on these keywords, and give corresponding answers when similar nursing questions are obtained next time, or directly give corresponding solutions for how to use medicines to treat colds for adults and children when the nursing staff does not indicate the object of medication. In another specific implementation, if the nursing question raised by the nursing staff is a standardized question, the rules can be set in advance, and the user can be answered directly according to the pre-set rules, such as the query of the dosage of common medicines. Through instant and accurate question-and-answer services, the time for nursing staff to query information is reduced, and the speed of nursing decision-making is accelerated, thereby significantly improving work efficiency in a busy medical environment.

[0054] In addition, the target medical industry big model can also record and analyze the conversation history between the user and the system, use the attention mechanism or RNN (Recurrent Neural Network) to capture the key information of the conversation context, and adjust the answer generation strategy based on the key information, that is, the user's specific needs or implicit information in the conversation history, so as to provide users with more personalized and coherent responses. For example, if the user asks to know what medicines should be used for a cold, and provides information about the patient's fever before asking this nursing question, the target medical industry big model can use colds and fevers as keywords to give the corresponding target nursing answers based on the nursing knowledge graph. By providing accurate nursing guidance, it helps to avoid nursing errors caused by incorrect information, improve the standardization and safety of nursing operations, and protect the health of patients.

[0055] After the target medical industry large model uses the user's feedback information to iteratively train the convergent question-answering model, the trained convergent question-answering model is used to generate a target nursing answer set corresponding to the target nursing text, and then the target nursing answer is determined from the target nursing answer set based on the preset answer scoring method. In a specific implementation, the specific process of determining the target nursing answer from the target nursing answer set may include: scoring all nursing answers in the target nursing answer set based on the preset quality evaluation method to obtain the corresponding scoring results of all the nursing answers in each preset evaluation dimension; wherein the preset evaluation dimension includes answer accuracy, answer completeness and answer credibility; then based on the preset comprehensive sorting method, the corresponding scoring results in each preset evaluation dimension are comprehensively processed to obtain the target scoring result, and the nursing answer with the largest target scoring result in the target nursing answer set is determined as the target nursing answer. That is, combined with the scores of each dimension, a weighted average or other composite sorting algorithm is used to comprehensively sort all the answers in the target nursing answer set, and then the answer with the highest score in the sorting list is selected as the final output to ensure that the answer presented to the user is strictly screened and best meets the needs.

[0056] In another specific implementation, after all answers in the target nursing answer set are comprehensively ranked based on the scoring results, all answers in the target nursing answer set and the corresponding ranking results and scoring results can be directly sent to the user end, and the user can select the answer in the target nursing answer set as the final answer according to their own needs. Then the target medical industry big model can fine-tune its own model parameters again based on the answer selected by the user, including but not limited to adjusting the weights of the corresponding scoring results on each preset evaluation dimension in the composite scoring algorithm. At the same time, after obtaining feedback on the answer selected by the user, the answer adoption rate can be combined with the nursing answer finally taken by the user for similar problems after the model is put into use, and the answer adoption rate can be added as an evaluation dimension to the preset quality assessment method.

[0057] It is understandable that users can evaluate and supplement information on the target nursing answers provided by the target medical industry big model through the user feedback interface preset on the user side, including some cutting-edge nursing knowledge, and then integrate the user feedback information into the training data to obtain the integrated training data, and regularly fine-tune the target medical industry big model based on the integrated training data to achieve self-optimization of the target medical industry big model and timely update of the nursing knowledge map, and ensure that nursing staff can obtain and apply the latest nursing knowledge and practice guidelines in a timely manner through a dynamic update mechanism, and accelerate the dissemination and application of knowledge in the nursing team. Users can also make corresponding improvement suggestions to the target medical industry big model based on their feelings when using the target medical industry big model through the user feedback interface to optimize the user experience. At the same time, the user interface of the user side can be set to a more intuitive user interface and support multiple interaction methods to ensure that users can quickly enter questions and receive answers. In addition, the user side can also provide personalized configuration options, such as language preferences, information display formats, etc., to meet the needs of different nursing staff, and through intelligent assistance, reduce the cognitive load and psychological pressure of nursing staff, improve job satisfaction, and facilitate the long-term retention and development of nursing talents.

[0058] It can be seen that this application converts the user's nursing questions into target nursing question texts using the language processing model in the target medical industry big model, so that the target medical industry big model can understand the nursing questions more deeply, and then uses the preset semantic matching technology to determine the target knowledge points corresponding to the nursing questions based on the nursing knowledge graph, and based on the target knowledge points and the templated answer structure corresponding to the target knowledge points, it can generate an accurate and standardized nursing answer candidate set for the nursing questions, and then use the convergent question and answer model to converge the nursing answers in the nursing answer candidate set to determine the most appropriate nursing answer as the target nursing answer, and finally send the target nursing answer to the user end, so as to give answers and guidance suggestions that meet medical professional standards, provide accurate question and answer services, effectively support nursing staff's daily decision-making, and improve the efficiency and quality of nursing work.

[0059] Based on the previous embodiment, it can be seen that the present application discloses a nursing question-and-answer method based on a large model of the medical industry, which can provide accurate question-and-answer services, effectively support the daily decision-making of nursing staff, and improve the efficiency and quality of nursing work. Next, the specific nursing question-and-answer process based on the large model of the medical industry will be described.

[0060] See also Figure 2As shown, before obtaining the nursing voice questions raised by the user, the collected medical nursing data is first subjected to data preprocessing and data annotation to obtain the processed nursing data, and the pre-trained big model is fine-tuned using the processed nursing data to obtain a convergent question-answering model. Then, NLP is used to determine the language processing model, and the obtained language processing model and convergent question-answering model are used to construct the target medical industry big model, thereby completing the construction, training and optimization of the target medical industry big model to ensure that the target medical industry big model can deeply understand the nursing professional language and complex situations, and provide nursing staff with more accurate question-answering services.

[0061] In addition to building a large model of the target medical industry, it is also necessary to generate a structured knowledge graph based on the pre-acquired nursing knowledge to obtain a nursing knowledge graph, which contains nodes such as diseases, nursing points, nursing assessment, health guidance and their relationships, so as to more accurately determine several target knowledge points corresponding to the target nursing problem text from the nursing knowledge graph based on the preset semantic matching technology.

[0062] After the nursing knowledge graph and the target medical industry big model are constructed, the nursing voice questions sent by the user are obtained, and then the text understanding model in the language processing model in the target medical industry big model is used to perform semantic analysis operations, entity recognition operations and relationship extraction operations on the nursing voice questions to obtain text understanding results, and the speech recognition technology in the language processing model is used to perform speech recognition on the nursing voice questions to obtain speech recognition results. After obtaining the text understanding results and speech recognition results using the language processing model, the cross-modal fusion module in the language processing model is used to semantically align the text understanding results and the speech recognition results, and convert the nursing voice questions into target nursing question texts, thereby ensuring that the target medical industry big model can process and understand the intention of the nursing staff in raising the nursing voice questions.

[0063] After obtaining the target nursing question text, the target medical industry big model uses a preset semantic matching technology to determine several target knowledge points corresponding to the target nursing question text from the pre-built nursing knowledge graph. When determining the knowledge points corresponding to the target nursing question text from the nursing knowledge graph based on the preset semantic matching technology, the user question is converted into a high-dimensional vector representation, and then a vector space model or a graph neural network is used to calculate the vector similarity value between the question vector and each node in the knowledge graph and the question vector, so as to determine the set of knowledge points most relevant to the nursing question as the target knowledge point, wherein the vector similarity value between the target knowledge point and the target nursing question text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing question text.

[0064] Finally, the target medical industry big model designs a templated answer structure for the retrieved target knowledge points, and fills the target knowledge points into the corresponding template according to the generated templated answer structure, thereby forming a more standardized answer, and uses all the answers generated based on the target knowledge points as the preliminary nursing answer candidate set, and uses the trained convergent question-answering model to generate the target nursing answer set corresponding to the target nursing question text, and uses the preset answer scoring method to determine the target nursing answer from the target nursing answer set. After outputting the target nursing answer, the target medical industry big model can achieve self-optimization and timely update of the nursing knowledge map based on user feedback information, and ensure that nursing staff can obtain and apply the latest nursing knowledge and practice guidelines in a timely manner through a dynamic update mechanism, accelerating the dissemination and application of knowledge in the nursing team.

[0065] See also Figure 3 As shown, the present application discloses a module schematic diagram of the target medical industry big model. The target medical industry big model is composed of a convergent question-answering model and a language processing model. The convergent question-answering model can retrieve knowledge points related to nursing problems based on the nursing knowledge graph. The language processing model is composed of a text understanding model, a speech recognition module and a cross-modal fusion module. The convergent question-answering model is composed of a knowledge base retrieval module, an answer generation module and an answer evaluation and sorting module. Among them, the language processing model is responsible for converting nursing problems into target nursing problem texts, so that the knowledge base retrieval module understands the nursing problems and retrieves the knowledge points in the nursing knowledge graph based on the question vector converted from the target nursing problem text. The answer generation module generates a nursing answer candidate set based on the target knowledge points retrieved by the knowledge base retrieval module and converges the nursing answer candidate set based on user feedback to obtain a more accurate target nursing answer set. The answer evaluation module scores and sorts all answers in the target nursing answer set based on a preset quality evaluation method, thereby determining the most appropriate answer as the target answer output to the user end. In addition, the target medical industry big model can also include a user feedback interface, through which users can evaluate and supplement information on the target nursing answers provided by the target medical industry big model, including some cutting-edge nursing knowledge, and then integrate the user feedback information into the training data to obtain integrated training data, and regularly fine-tune the target medical industry big model based on the integrated training data to achieve self-optimization of the target medical industry big model and timely updating of the nursing knowledge graph.

[0066] It can be seen that this application converts the user's nursing questions into target nursing question texts using the language processing model in the target medical industry big model, so that the target medical industry big model can understand the nursing questions more deeply, and then use the preset semantic matching technology to determine the target knowledge points corresponding to the nursing questions based on the nursing knowledge graph, and based on the target knowledge points and the templated answer structure corresponding to the target knowledge points, an accurate and standardized nursing answer candidate set for the nursing questions can be generated, and then the convergent question-answer model is used to converge the nursing answers in the nursing answer candidate set to determine the most appropriate nursing answer as the target nursing answer, and finally the target nursing answer is sent to the user end, so as to give answers and guidance suggestions that meet medical professional standards, provide accurate question-answering services, effectively support nursing staff's daily decision-making, and improve nursing work efficiency and quality. The implementation of this application has set a new benchmark for intelligent services in the medical care industry and provided a solid technical foundation and practical example for the development and innovation of subsequent nursing information systems.

[0067] See also Figure 4 As shown, the embodiment of the present application discloses a structural schematic diagram of a nursing question-and-answer device based on a large model of the medical industry, including:

[0068] The question conversion module 11 is used to obtain the nursing question sent by the user end, and convert the nursing question into a target nursing question text by using the language processing model in the target medical industry big model;

[0069] A knowledge point determination module 12 is used to determine a number of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text;

[0070] The target answer determination module 13 is used to generate a nursing answer candidate set based on the templated answer structure corresponding to the several target knowledge points, and use the nursing answer candidate set to iteratively train the convergent question-answering model in the target medical industry big model to obtain the target nursing answer corresponding to the target nursing question text.

[0071] It can be seen that this application converts the user's nursing questions into target nursing question texts using the language processing model in the target medical industry big model, so that the target medical industry big model can understand the nursing questions more deeply, and then uses the preset semantic matching technology to determine the target knowledge points corresponding to the nursing questions based on the nursing knowledge graph, and based on the target knowledge points and the templated answer structure corresponding to the target knowledge points, it can generate an accurate and standardized nursing answer candidate set for the nursing questions, and then use the convergent question and answer model to converge the nursing answers in the nursing answer candidate set to determine the most appropriate nursing answer as the target nursing answer, and finally send the target nursing answer to the user end, so as to give answers and guidance suggestions that meet medical professional standards, provide accurate question and answer services, effectively support nursing staff's daily decision-making, and improve the efficiency and quality of nursing work.

[0072] In a specific embodiment, the nursing question-answering device may further include:

[0073] A convergent question-answering model building module, used to perform data preprocessing and data annotation processing on the collected medical care data to obtain processed care data, and use the processed care data to fine-tune the pre-trained large model to obtain the convergent question-answering model;

[0074] The big model building module is used to determine the language processing model using natural language processing technology, and to build the target medical industry big model based on the language processing model and the convergent question-answering model.

[0075] In a specific embodiment, the problem conversion module 11 may specifically include:

[0076] A question acquisition unit is used to acquire the nursing voice questions sent by the user end and send them to the language processing model in the target medical industry big model;

[0077] A question processing unit, configured to perform semantic analysis, entity recognition and relationship extraction operations on the nursing voice question using the text understanding model in the language processing model to obtain a text understanding result, and to perform speech recognition on the nursing voice question using the speech recognition technology in the language processing model to obtain a speech recognition result;

[0078] A text acquisition unit, used for semantically aligning the text understanding result and the speech recognition result using a cross-modal semantic alignment technology to obtain a target nursing question text;

[0079] A vector similarity determination unit, used to convert the target nursing question text into a question vector using a text vectorization model, and determine the vector similarity value between all knowledge points in the nursing knowledge graph and the question vector using a vector space model or a graph neural network;

[0080] A knowledge point determination unit is used to determine a number of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on the vector similarity value; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text.

[0081] In a specific embodiment, the target answer determination module 13 may specifically include:

[0082] A candidate answer generation submodule, used to generate a nursing answer candidate set based on the templated answer structure corresponding to the plurality of target knowledge points;

[0083] A model iteration training submodule, used for obtaining feedback information of the user terminal for the nursing answer candidate set, and iteratively training the convergent question-answering model based on the feedback information to obtain a trained convergent question-answering model;

[0084] The target answer determination submodule is used to generate a target nursing answer set corresponding to the target nursing question text using the trained convergent question-answering model, and determine the target nursing answer from the target nursing answer set using a preset answer scoring method.

[0085] In a specific embodiment, the target answer determination submodule may specifically include:

[0086] A target answer set determination unit, configured to generate a target nursing answer set corresponding to the target nursing question text by using the trained convergent question-answering model;

[0087] An answer scoring unit, used for scoring all nursing answers in the target nursing answer set based on a preset quality assessment method to obtain scoring results corresponding to all nursing answers in each preset assessment dimension; the preset assessment dimensions include answer accuracy, answer completeness and answer credibility;

[0088] The target answer determination unit is used to comprehensively process the corresponding scoring results on each of the preset evaluation dimensions based on a preset comprehensive ranking method to obtain a target scoring result, and determine the nursing answer with the largest target scoring result in the target nursing answer set as the target nursing answer.

[0089] Furthermore, the present application also discloses an electronic device. Figure 5 This is a structural diagram of an electronic device 20 according to an exemplary embodiment. The content in the diagram cannot be regarded as any limitation on the scope of use of the present application.

[0090] Figure 5A schematic diagram of the structure of an electronic device 20 provided in an embodiment of the present application. The electronic device 20 may specifically include: at least one processor 21, at least one memory 22, a power supply 23, a communication interface 24, an input / output interface 25, and a communication bus 26. The memory 22 is used to store a computer program, which is loaded and executed by the processor 21 to implement the relevant steps in the nursing question-and-answer method based on the medical industry big model disclosed in any of the aforementioned embodiments. In addition, the electronic device 20 in this embodiment may specifically be an electronic computer.

[0091] In this embodiment, the power supply 23 is used to provide working voltage for each hardware device on the electronic device 20; the communication interface 24 can create a data transmission channel between the electronic device 20 and the external device, and the communication protocol it follows is any communication protocol that can be applied to the technical solution of the present application, and is not specifically limited here; the input and output interface 25 is used to obtain external input data or output data to the outside world, and its specific interface type can be selected according to specific application needs and is not specifically limited here.

[0092] In addition, the memory 22 as a carrier for resource storage may be a read-only memory, a random access memory, a disk or an optical disk, etc. The resources stored thereon may include an operating system 221, a computer program 222, etc., and the storage method may be temporary storage or permanent storage.

[0093] The operating system 221 is used to manage and control the hardware devices and computer program 222 on the electronic device 20, which can be Windows Server, Netware, Unix, Linux, etc. In addition to the computer program that can be used to complete the nursing question-and-answer method based on the medical industry big model executed by the electronic device 20 disclosed in any of the aforementioned embodiments, the computer program 222 can further include computer programs that can be used to complete other specific tasks.

[0094] Furthermore, the present application also discloses a computer-readable storage medium for storing a computer program; wherein, when the computer program is executed by a processor, the aforementioned nursing question-and-answer method based on a large model of the medical industry is implemented. The specific steps of the method can be referred to the corresponding contents disclosed in the aforementioned embodiments, and will not be repeated here.

[0095] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0096] Professionals may further appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, computer software, or a combination of the two. In order to clearly illustrate the interchangeability of hardware and software, the composition and steps of each example have been generally described in the above description according to function. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professionals and technicians may use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of this application.

[0097] The steps of the method or algorithm described in conjunction with the embodiments disclosed herein may be implemented directly using hardware, a software module executed by a processor, or a combination of the two. The software module may be placed in a random access memory (RAM), a memory, a read-only memory (ROM), an electrically programmable ROM, an electrically erasable programmable ROM, a register, a hard disk, a removable disk, a CD-ROM, or any other form of storage medium known in the art.

[0098] Finally, it should be noted that, in this article, relational terms such as first and second, etc. are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Moreover, the terms "include", "comprise" or any other variants thereof are intended to cover non-exclusive inclusion, so that a process, method, article 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, article or device. In the absence of further restrictions, the elements defined by the sentence "comprise a ..." do not exclude the presence of other identical elements in the process, method, article or device including the elements.

[0099] The technical solution provided by the present application is introduced in detail above. Specific examples are used in this article to illustrate the principles and implementation methods of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for general technicians in this field, according to the idea of ​​the present application, there will be changes in the specific implementation methods and application scope. In summary, the content of this specification should not be understood as a limitation on the present application.

Claims

1. A nursing question-answering method based on a large model of the medical industry, characterized in that: include: Obtain the nursing questions sent by the user, and use the language processing model in the target medical industry big model to convert the nursing questions into target nursing question texts; Based on a preset semantic matching technology, a plurality of target knowledge points corresponding to the target nursing problem text are determined from a pre-constructed nursing knowledge graph; a vector similarity value between the target knowledge point and the target nursing problem text is greater than a vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text; A nursing answer candidate set is generated based on the templated answer structure corresponding to the several target knowledge points, and the convergent question-answering model in the target medical industry big model is iteratively trained using the nursing answer candidate set to obtain the target nursing answer corresponding to the target nursing question text.

2. The nursing question-answering method based on the medical industry big model according to claim 1 is characterized in that: Before obtaining the nursing question sent by the user end and converting the nursing question into a target nursing question text by using the language processing model in the target medical industry big model, the method further includes: Performing data preprocessing and data annotation processing on the collected medical care data to obtain processed care data, and using the processed care data to fine-tune the pre-trained large model to obtain the converged question-answering model; The language processing model is determined using natural language processing technology, and the target medical industry big model is constructed based on the language processing model and the convergent question-answering model.

3. The nursing question-answering method based on the medical industry big model according to claim 1 is characterized in that: The nursing knowledge graph is a structured knowledge graph generated based on pre-acquired nursing knowledge.

4. The nursing question-answering method based on the medical industry big model according to claim 1 is characterized in that: The step of obtaining the nursing question sent by the user and converting the nursing question into a target nursing question text by using a language processing model in a target medical industry big model includes: Obtain the nursing voice questions sent by the user and send them to the language processing model in the target medical industry big model; Using the text understanding model in the language processing model to perform semantic analysis operations, entity recognition operations, and relationship extraction operations on the nursing voice question to obtain a text understanding result, and using the speech recognition technology in the language processing model to perform speech recognition on the nursing voice question to obtain a speech recognition result; The text understanding result and the speech recognition result are semantically aligned using cross-modal semantic alignment technology to obtain the target nursing question text.

5. The nursing question-answering method based on the medical industry big model according to claim 1 is characterized in that: The method of determining a plurality of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text, including: The target nursing question text is converted into a question vector using a text vectorization model, and a vector space model or a graph neural network is used to determine the vector similarity value between all knowledge points in the nursing knowledge graph and the question vector; Based on the vector similarity value, several target knowledge points corresponding to the target nursing problem text are determined from a pre-constructed nursing knowledge graph; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text.

6. The nursing question-answering method based on the medical industry big model according to any one of claims 1 to 5, characterized in that: The step of iteratively training the convergent question-answering model in the target medical industry big model using the nursing answer candidate set to obtain a target nursing answer corresponding to the target nursing question text includes: Acquiring feedback information from the user terminal for the nursing answer candidate set, and iteratively training the convergent question-answering model based on the feedback information to obtain a trained convergent question-answering model; The trained convergent question-answering model is used to generate a target nursing answer set corresponding to the target nursing question text, and a preset answer scoring method is used to determine a target nursing answer from the target nursing answer set.

7. The nursing question-answering method based on the medical industry big model according to claim 6 is characterized in that: The method of determining a target nursing answer from the target nursing answer set by using a preset answer scoring method comprises: Scoring all nursing answers in the target nursing answer set based on a preset quality assessment method to obtain scoring results corresponding to all nursing answers in each preset assessment dimension; the preset assessment dimensions include answer accuracy, answer completeness and answer credibility; Based on a preset comprehensive ranking method, the corresponding scoring results on each of the preset evaluation dimensions are comprehensively processed to obtain a target scoring result, and the nursing answer with the largest target scoring result in the target nursing answer set is determined as the target nursing answer.

8. A nursing question-answering device based on a large model of the medical industry, characterized in that: include: A question conversion module is used to obtain the nursing questions sent by the user end and convert the nursing questions into target nursing question texts by using the language processing model in the target medical industry big model; A knowledge point determination module is used to determine a number of target knowledge points corresponding to the target nursing problem text from a pre-constructed nursing knowledge graph based on a preset semantic matching technology; the vector similarity value between the target knowledge point and the target nursing problem text is greater than the vector similarity value between other knowledge points in the nursing knowledge graph and the target nursing problem text; The target answer determination module is used to generate a nursing answer candidate set based on the templated answer structure corresponding to the several target knowledge points, and use the nursing answer candidate set to iteratively train the convergent question-answering model in the target medical industry big model to obtain the target nursing answer corresponding to the target nursing question text.

9. An electronic device, characterized in that: include: Memory, used to store computer programs; A processor is used to execute the computer program to implement the nursing question-answering method based on the medical industry big model as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: Used to store a computer program, wherein when the computer program is executed by a processor, the nursing question-and-answer method based on the medical industry big model as described in any one of claims 1 to 7 is implemented.

Citation Information

Patent Citations

  • Question and answer processing method and system based on medical treatment, computer equipment and readable medium

    CN110390003A

  • Hypertension question-answering method and system based on deep learning

    CN116737911A

  • Interactive dialogue method, device and equipment based on artificial intelligence and storage medium

    CN116991982A

  • Knowledge graph generation type question answering method and system based on large language model

    CN117033608A

  • Question answering method and device based on artificial intelligence, electronic equipment and storage medium

    CN117725188A