Medication guidance generation system

Abstract

The invention provides a medication guidance generation system, and relates to the technical field of medical informatization. Comprising a knowledge base construction module used for constructing a medication guidance knowledge base; the graph term construction module is used for constructing a knowledge graph and a standard term set according to the medication guidance knowledge base; the index construction module is used for splitting the medication guidance knowledge base and constructing a vector index and a full-text index; the personal information processing module is used for acquiring personal information of a patient and preprocessing the personal information to obtain preprocessed personal information; the mixed detection module is used for performing retrieval by utilizing a knowledge graph, a vector index and a full-text index according to the preprocessed personal information to obtain a retrieval result set and obtain an intersection result according to the retrieval result set; and the generation module is used for generating medication guidance corresponding to the target patient according to the intersection result. The problems that traditional medication guidance mainly depends on experience and professional knowledge of medical staff, and is low in efficiency, prone to errors and difficult to personalize are solved.

Description

Technical Field

[0001] This invention relates to the field of medical information technology, and in particular to a medication guidance generation system. Background Technology

[0002] In modern medical practice, medication guidance is a crucial step in ensuring patients use medications correctly, improving treatment outcomes, and reducing adverse drug reactions. Traditional medication guidance relies primarily on the experience and expertise of healthcare professionals, but this approach suffers from inefficiency, error-proneness, and a lack of personalization. In recent years, with the development of medical information technology, medication guidance knowledge bases have been gradually established and continuously improved, providing important decision support for rational drug use in clinical practice. However, how to efficiently utilize these knowledge bases to automatically generate personalized medication guidance remains a pressing issue that needs to be addressed. Summary of the Invention

[0003] To overcome the shortcomings of existing technologies, the purpose of this invention is to provide a medication guidance generation system. This invention solves the problems of traditional medication guidance in the prior art, which mainly relies on the experience and professional knowledge of medical staff, resulting in low efficiency, easy errors, and difficulty in personalization.

[0004] To achieve the above objectives, the present invention provides the following solution:

[0005] A medication guidance generation system, comprising:

[0006] The knowledge base construction module is used to build a medication guidance knowledge base;

[0007] The knowledge graph terminology construction module is used to build a knowledge graph and standard terminology set based on the medication guidance knowledge base.

[0008] The index building module is used to split the medication guidance knowledge base and build vector indexes and full-text indexes;

[0009] The personal information processing module is used to acquire and preprocess patient personal information to obtain preprocessed personal information, which includes: standard terms corresponding to personal information keywords, vectorized data and word segmentation data corresponding to personal information;

[0010] The mixed-search module is used to perform retrieval based on the preprocessed personal information using the knowledge graph, vector index, and full-text index to obtain a first retrieval result, a second retrieval result, and a third retrieval result, and to obtain an intersection result based on the first retrieval result, the second retrieval result, and the third retrieval result;

[0011] The generation module is used to generate medication guidance for the target patient based on the intersection result.

[0012] Preferably, the knowledge base construction module includes:

[0013] Data acquisition submodule, data processing submodule, and setup submodule;

[0014] The data acquisition submodule is used to acquire medication guidance materials, the data processing submodule is used to remove duplicate information, correct erroneous data and unify the format of the medication guidance materials to obtain preprocessed medication guidance materials, and the construction submodule is used to construct a medication guidance knowledge base based on the preprocessed medication guidance materials.

[0015] Preferably, the atlas terminology construction module includes:

[0016] The module includes an entity recognition model construction submodule, an initial graph construction submodule, and a final graph construction submodule.

[0017] The entity recognition model construction submodule is used to construct an entity recognition model based on the medication guidance knowledge base and the BERT model. The initial knowledge graph construction submodule is used to complete the article entity annotation of the medication guidance knowledge base based on the entity recognition model to obtain an initial knowledge graph. The final knowledge graph construction submodule is used to fuse the initial knowledge graph and standardize the entity names into standard names to obtain a knowledge graph and a standard terminology set.

[0018] Preferably, the index building module includes:

[0019] Submodules include: text splitting, text preprocessing, vectorization, and word segmentation.

[0020] The splitting submodule is used to perform structured splitting of the drug guidance knowledge base to obtain split text data. The text preprocessing submodule is used to perform text cleaning on the split text data to obtain preprocessed split data. The vectorization submodule is used to vectorize the preprocessed split data using a vector model to obtain a vector index. The word segmentation submodule is used to segment the preprocessed split data using a word segmentation model to obtain a full-text index.

[0021] Preferably, the vector model is gte-Qwen2.

[0022] Preferably, the word segmentation model is jieba.

[0023] Preferably, the personal information processing module includes:

[0024] The module includes a personal information acquisition submodule, a first conversion submodule, a second conversion submodule, and a third conversion submodule.

[0025] The personal information acquisition submodule is used to acquire patient personal information. The first conversion submodule is used to identify entity keywords of patient personal information using the entity recognition model and convert them into standard terms. The second conversion submodule is used to vectorize patient personal information according to the vector model to obtain vectorized personal information. The third conversion submodule is used to segment patient personal information according to the word segmentation model to obtain word-segmented personal information.

[0026] Preferably, the pooled testing module includes:

[0027] The system comprises the first search submodule, the second search submodule, the third search submodule, the rearrangement submodule, and the intersection submodule.

[0028] The first retrieval submodule is used to perform a first retrieval based on the knowledge graph and standard terminology set to obtain a first retrieval result. The second retrieval submodule is used to perform a second retrieval based on the vector index to obtain a second retrieval result. The third retrieval submodule is used to perform a third retrieval based on the full-text index to obtain a third retrieval result. The rearrangement submodule is used to rearrange the second retrieval result and the third retrieval result using a rearrangement model to obtain a rearrangement result. The intersection submodule is used to determine the intersection result based on the rearrangement result and the first retrieval result.

[0029] The present invention discloses the following technical effects:

[0030] This invention provides a medication guidance generation system, comprising: a knowledge base construction module for constructing a medication guidance knowledge base; a graph terminology construction module for constructing a knowledge graph and a standard terminology set based on the medication guidance knowledge base; an index construction module for splitting the medication guidance knowledge base and constructing a vector index and a full-text index; a personal information processing module for acquiring and preprocessing patient personal information to obtain preprocessed personal information, wherein the preprocessed personal information includes: standard terms corresponding to personal information keywords, vectorized data corresponding to personal information, and word segmentation data; a mixed-search module for performing searches based on the preprocessed personal information using the knowledge graph, vector index, and full-text index to obtain a first search result, a second search result, and a third search result, and obtaining an intersection result based on the first search result, the second search result, and the third search result; and a generation module for generating medication guidance corresponding to the target patient based on the intersection result. This invention obtains patient-related medication guidance knowledge by querying a structured medication guidance knowledge base using patient personal information, and then efficiently and accurately generates medication guidance through a large model. This invention automates the generation of medication guidance, reducing the workload of medical staff; it generates personalized medication guidance based on the patient's specific condition, improving the safety and effectiveness of medication; it is based on authoritative knowledge base content, ensuring the accuracy and reliability of medication guidance; the knowledge base content can be updated at any time, and the system can adapt to new drugs and medication guidelines. Attached Figure Description

[0031] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0032] Figure 1 This is a schematic diagram of a medication guidance generation system provided in an embodiment of the present invention;

[0033] Figure 2 This is a schematic diagram of the automatic medication guidance generation system provided in an embodiment of the present invention;

[0034] Figure 3 This is a schematic diagram of the core process for automatically generating medication guidance provided in an embodiment of the present invention;

[0035] Figure 4 This is an example diagram of a knowledge graph provided in an embodiment of the present invention;

[0036] Figure 5 An example diagram of the generated medication guidance provided for an embodiment of the present invention.

[0037] Explanation of reference numerals in the attached figures:

[0038] 1-Knowledge base construction module, 2-Graph terminology construction module, 3-Index construction module, 4-Personal information processing module, 5-Mixed detection module, 6-Generation module. Detailed Implementation

[0039] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0040] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments.

[0041] like Figure 1 As shown, the present invention provides a medication guidance generation system, comprising:

[0042] Knowledge base construction module 1 is used to build a medication guidance knowledge base;

[0043] Module 2 for constructing knowledge graphs and standard terminology sets based on the medication guidance knowledge base;

[0044] Index building module 3 is used to split the medication guidance knowledge base and build vector indexes and full-text indexes;

[0045] Personal information processing module 4 is used to acquire and preprocess patient personal information to obtain preprocessed personal information, which includes: standard terms corresponding to personal information keywords, vectorized data and word segmentation data corresponding to personal information;

[0046] The mixed-search module 5 is used to perform retrieval based on the preprocessed personal information using the knowledge graph, vector index and full-text index, to obtain a first retrieval result, a second retrieval result and a third retrieval result, and to obtain an intersection result based on the first retrieval result, the second retrieval result and the third retrieval result;

[0047] The generation module 6 is used to generate medication guidance for the target patient based on the intersection result.

[0048] Specifically, such as Figure 2 As shown, the workflow steps of the system disclosed in this embodiment are as follows:

[0049] Step 1: Collect authoritative medication guidance materials, including drug instructions, clinical medication guidelines, drug interaction databases, and previous medication guidance, to form a comprehensive medication guidance knowledge base. Preprocess the collected knowledge, including removing duplicate information, correcting errors, and standardizing the format. Based on the professional knowledge base, annotate entities and relationships, train an entity recognition model, construct a medical knowledge graph, and establish a standard terminology set. Store the knowledge graph and terminology set in a relational database and create an index for the knowledge graph and terminology set.

[0050] Step 2 involves structurally splitting the medication guidance knowledge base and removing irrelevant information to reduce text clutter. Then, using vector and word segmentation models, vector indexes are created by vectorizing the text blocks, and full-text search indexes are created through word segmentation.

[0051] Step 3: Obtain the patient's personal information, including but not limited to gender, age, pregnancy / lactation status, allergy history, diseases, and medication information. This information can come from the Hospital Information System (HIS), Electronic Medical Record (EMR), or be manually entered. The patient's original personal information includes both structured and unstructured data, which requires further preprocessing.

[0052] Step 4: Using the processing results of Step 3, perform a multi-path hybrid retrieval of the medication guidance knowledge base to obtain knowledge base text blocks related to the patient's personal information;

[0053] Step 5: Pass the personal information-related text blocks, the patient's historical medication guidance records, the article search results top_kc for the drug, and the prompt words to the generative big model. The big model includes deepseek-r1, deepseek-v1, QwQ, qwen2.5, etc., so that the big model can generate medication guidance for each drug according to the requirements.

[0054] Step 6: After generating medication guidelines, the system displays them to medical staff in separate sections according to the medication. Medical staff can then add, delete, or modify the generated medication guidelines. The system updates the medication guidelines knowledge base based on the feedback from medical staff to improve the accuracy and usability of subsequent medication guidelines.

[0055] Furthermore, the knowledge base construction module 1 includes:

[0056] Data acquisition submodule, data processing submodule, and setup submodule;

[0057] The data acquisition submodule is used to acquire medication guidance materials, the data processing submodule is used to remove duplicate information, correct erroneous data and unify the format of the medication guidance materials to obtain preprocessed medication guidance materials, and the construction submodule is used to construct a medication guidance knowledge base based on the preprocessed medication guidance materials.

[0058] Furthermore, the atlas terminology construction module 2 includes:

[0059] The module includes an entity recognition model construction submodule, an initial graph construction submodule, and a final graph construction submodule.

[0060] The entity recognition model construction submodule is used to construct an entity recognition model based on the medication guidance knowledge base and the BERT model. The initial knowledge graph construction submodule is used to complete the article entity annotation of the medication guidance knowledge base based on the entity recognition model to obtain an initial knowledge graph. The final knowledge graph construction submodule is used to fuse the initial knowledge graph and standardize the entity names into standard names to obtain a knowledge graph and a standard terminology set.

[0061] Specifically, firstly, the schema of the medication guidance knowledge graph is defined using the knowledge base, including medication timing, dosage requirements, handling of missed doses, methods for spacing multiple medications, common or serious adverse drug reactions, precautions for patients driving or engaged in high-risk work, precautions for elderly patients, precautions for pregnant women, precautions for breastfeeding women, precautions for children, precautions for patients with liver dysfunction, precautions for patients with kidney dysfunction, storage methods, and lifestyle and dietary recommendations, etc. Note that in addition to professional drug instructions, clinical drug use guidelines, and drug interaction databases, the knowledge base also includes historically written, authentic medication guidance. This part of the knowledge is more concise and appropriate, and its content proportion and parsing weight can be appropriately increased. Furthermore, entity types in the knowledge base articles are manually labeled according to the defined definitions. A small portion of the articles are pre-labeled for further fine-tuning of the existing BERT model, ultimately resulting in an entity recognition model applied to medication guidance.

[0062] Complete the entity annotation of all articles, export the annotation data, and form the initial knowledge graph; then use entity name similarity calculation plus manual verification to realize the entity fusion of the knowledge graph, standardize the entity names of the same entity type with the same meaning into standard names, and thus form the final medication guidance knowledge graph and standard terminology library.

[0063] The standard terminology database includes not only the standard name of each term but also aliases from the annotated data. A mapping relationship is suggested between the aliases and the standard names for terminology conversion of keywords identified from patient personal information. Table 1 shows an example of the standard terminology database.

[0064] Table 1 Entity type Entity Name Standard Name disease Diabetes diabetes disease Type 2 diabetes diabetes disease Gestational diabetes diabetes drug Metoprolol tartrate Metoprolol drug Betaloc Metoprolol drug Meiduoxin'an Metoprolol

[0065] Create a knowledge graph and terminology set index based on the established knowledge graph and standard terminology base.

[0066] Furthermore, the index building module 3 includes:

[0067] Submodules include: text splitting, text preprocessing, vectorization, and word segmentation.

[0068] The splitting submodule is used to perform structured splitting of the drug guidance knowledge base to obtain split text data. The text preprocessing submodule is used to perform text cleaning on the split text data to obtain preprocessed split data. The vectorization submodule is used to vectorize the preprocessed split data using a vector model to obtain a vector index. The word segmentation submodule is used to segment the preprocessed split data using a word segmentation model to obtain a full-text index.

[0069] Specifically, first, the entire article content is read, and text segmentation is performed based on paragraph headings. For paragraphs without headings, a larger model can be used to summarize and extract the paragraphs; alternatively, a word segmentation model can be used to obtain high-frequency words as headings. Each segmented paragraph is then given a semantic heading, including the article title / first-level heading / second-level heading / third-level heading / fourth-level heading, and so on. Next, the paragraphs are further subdivided according to the text length limitations of the vector model. During segmentation, for tables, the integrity of the table content is maintained as much as possible based on the table's HTML tags, and redundant and useless tags and symbols are removed for text cleaning. Each segmented paragraph needs to be appended with its original semantic heading as an index.

[0070] Paragraph splitting example: Text blocks after splitting the medication instructions for "metoprolol":

[0071] Metoprolol / Precautions during medication

[0072] 1. Drinking alcohol while taking this medication may increase the blood concentration of metoprolol. It is recommended that you avoid drinking alcohol or beverages containing alcohol.

[0073] 2. If you sit or lie down and then get up quickly while taking this medication, you may experience dizziness or fainting. Please get up slowly after sitting or lying down, and be careful when going up or down stairs.

[0074] Furthermore, irrelevant information in the text blocks, especially content unrelated to medication education, should be removed to reduce text complexity, such as removing drug production quality control methods and obsolete instructions.

[0075] Using a vector model, the pre-segmented paragraphs with headings are vectorized, and the vectorization results are stored in a vector database, creating a vector index. The vector model can be something like gte-Qwen2.

[0076] Using a word segmentation model, the pre-segmented paragraphs with titles are segmented into words, and the frequency and position of each keyword are recorded to create a full-text search inverted index. The word segmentation model can be jieba, etc.

[0077] Furthermore, the personal information processing module 4 includes:

[0078] The module includes a personal information acquisition submodule, a first conversion submodule, a second conversion submodule, and a third conversion submodule.

[0079] The personal information acquisition submodule is used to acquire patient personal information. The first conversion submodule is used to identify entity keywords of patient personal information using the entity recognition model and convert them into standard terms. The second conversion submodule is used to vectorize patient personal information according to the vector model to obtain vectorized personal information. The third conversion submodule is used to segment patient personal information according to the word segmentation model to obtain word-segmented personal information.

[0080] Specifically, a trained entity recognition model is used to identify entity keywords in the patient's personal information. These keywords are then converted into standard terms using a standard terminology set. The patient's personal information to be identified includes, but is not limited to, height, weight, age, pregnancy status, breastfeeding status, education level, occupation, lifestyle habits, dietary habits, smoking history, alcohol consumption history, allergy history, medical history, family history, diagnostic information, medication information, laboratory information, examination information, and surgical information.

[0081] For example, the diagnostic information is "gestational hypertension". The JSON string of the entity recognition result is: [{"begin":1,"end":3,"entity":"pregnancy","type":"pregnancy status"}]. The entity name is "pregnancy", the entity type is "pregnancy status", and then the entity name field in the standard terminology library is matched with "pregnancy" to obtain the standard term name: "pregnancy".

[0082] Using vector models and word segmentation models, patient personal information is vectorized or segmented into words, respectively.

[0083] Furthermore, the mixed testing module 5 includes:

[0084] The system comprises the first search submodule, the second search submodule, the third search submodule, the rearrangement submodule, and the intersection submodule.

[0085] The first retrieval submodule is used to perform a first retrieval based on the knowledge graph and standard terminology set to obtain a first retrieval result. The second retrieval submodule is used to perform a second retrieval based on the vector index to obtain a second retrieval result. The third retrieval submodule is used to perform a third retrieval based on the full-text index to obtain a third retrieval result. The rearrangement submodule is used to rearrange the second retrieval result and the third retrieval result using a rearrangement model to obtain a rearrangement result. The intersection submodule is used to determine the intersection result based on the rearrangement result and the first retrieval result.

[0086] Specifically, the standard terminology for the acquired personal information first involves breaking down patient medication (the drugs targeted for generating medication guidance) and recording it as follows: <drug1> ， <drug2> ， <drug3>And other personal information, recorded as <individual>Each medicine is then compared with... <individual>The process involves merging the results, then, based on the created knowledge graph index, finding the relationship between this standard term and the article, as well as the term weight. If multiple keywords are identified, the weight of each search result is the sum of the weights of all keywords. Finally, the top_ka is obtained by sorting the results in reverse order based on the weight values, thus obtaining the top_ka for each drug. Specifically, for example... <drug1>joint <individual>Top_ka1 can be obtained. <drug2>joint <individual>Top_ka2 can be obtained. <drug3>joint <individual>You can obtain top_ka3, and so on.

[0087] The vectorized results of the obtained personal information first need to be broken down into patient medications (the drugs targeted for generating medication guidance) and recorded as follows: <drug1> ， <drug2> ， <drug3>And other personal information, recorded as <individual>Each medicine is then compared with... <individual>Merge the data based on the created vector indexes, sorting them in descending order of similarity values ​​to obtain the text blocks for each drug index.

[0088] The word segmentation results of the obtained personal information must first be broken down into patient medications (the drugs for which medication guidance is generated) and recorded as follows: <drug1> ， <drug2> ， <drug3>And other personal information, recorded as <individual>Each medicine is then compared with... <individual>Merge the data, and retrieve each drug index text block using the BM25 algorithm based on the created full-text index.

[0089] Based on the results of drug-based vector retrieval and word segmentation retrieval, the candidate results are rearranged using the bce-ranker-base_v1 reordering model. The vector retrieval weight and full-text retrieval weight can be flexibly set (e.g., 0.7:0.3) to reorder the retrieval results and extract the top_kb text blocks for each drug.

[0090] The top_ka text block retrieved by the knowledge graph retrieval and the top_kb text block retrieved by the reordering model may correspond to paragraphs at positions x and y, respectively. Taking the intersection of the paragraph titles at x and y (i.e., only text blocks whose titles match the titles of kb within the ka text block) yields the top_kc text block, which is retrieved through multiple retrieval methods. This text block can be denoted as... <content>Note that we still need to perform intersection analysis on each drug group to obtain the top_kc for each drug. <drug1>Taking the intersection of top_ka1 and top_kb1, we can obtain top_kc1, denoted as... <content1> ； <drug2>Taking the intersection of top_ka2 and top_kb2, we can obtain top_kc2, denoted as... <content2> ； <drug3>Taking the intersection of top_ka3 and top_kb3, we can obtain top_kc3, denoted as... <content3>And so on.

[0091] Furthermore, each of the filtered... <content>The system takes only the first 15 entries and combines them with the patient's historical medication guidance data. This mixture is then passed to the larger model via an interface, serving as the context for generating medication guidance. Simultaneously, the larger model, combined with pre-set prompts, generates medication guidance and transmits it to the display interface.

[0092] The organization of the prompts will vary depending on the patient's specific situation. If the patient does not have historical medication guidance data, the prompts can be organized in the following ways:

[0093] The patient's medication is <drug>The current personal information is <individual>The relevant knowledge content obtained based on the patient's personal information is <content>Based on the above information, write a medication guide for this patient and this medication, with the following specific requirements: 1. List 3 to 5 points, each no longer than 70 words. 2. This medication guide is intended for the patient, who is not a medical professional; therefore, use simple, easy-to-understand, and accurate language. 3. The content of the medication guide should focus on incorporating the patient's personal information and relevant knowledge. Irrelevant or unsuitable content should be omitted. 4. All guidance should strictly adhere to the provided knowledge.

[0094] If the patient has historical medication guidance data, the prompts can be organized in the following ways:

[0095] The patient's medication is <drug>The current personal information is <individual>The relevant knowledge content obtained based on the patient's personal information is <content>The patient has used this medication before, and the previous medication guidelines were as follows:<history_content> Based on the above information, write a medication guide for this patient and this medication, with the following specific requirements: 1. List 3 to 5 points, each preferably no longer than 70 words. 2. This medication guide is intended for the patient, who is not a medical professional; therefore, use simple, easy-to-understand, and accurate language. 3. The content of the medication guide should focus on combining the patient's personal information and relevant knowledge. Irrelevant or unsuitable content should not be included. 4. Refer to the patient's previous medication guidelines. If previous guidelines are no longer applicable to the patient's current condition, they can be omitted. 5. All guidance content should strictly adhere to the provided knowledge or be derived from previous medication guidelines.

[0096] Note the prompt words. <drug> 、 <content>,<history_content> They always appear in groups, meaning they are selected based on the patient's medication list. <drug1>At times, it is also necessary to take <drug1>Corresponding generation <content1>,as well as <drug1>Corresponding historical medication guidance records<history_content1> ;Pick <drug2>At times, it is also necessary to take <drug2>Corresponding generation <content2>,as well as <drug2>Corresponding historical medication guidance records<history_content2> And so on. <individual>Always obtain all personal information of the current patient.

[0097] Furthermore, the newly generated medication guidelines for different drugs are automatically tagged with different labels, such as drug name, potential adverse reactions, and precautions, for subsequent display by label category and for saving back to the different categories of historical medication guidance data. Based on the content of the medication guidelines and the different labels, they are displayed separately for each drug on the interactive interface for verification by medical staff. Simultaneously, medical staff can add, delete, and modify already generated medication guidelines. All added, deleted, and modified medication guidelines will replace previously generated guidelines and be saved back to the different categories of historical medication guidance data as reference knowledge for future use, further updating the medication guidance knowledge base to improve the accuracy and usability of future generated medication guidelines.

[0098] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0099] This document uses specific examples to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. Furthermore, those skilled in the art will recognize that, based on the ideas of the present invention, there will be changes in the specific implementation methods and application scope. Therefore, the content of this specification should not be construed as a limitation of the present invention.< / individual> < / content> < / drug> < / content> < / individual> < / drug> < / content> < / individual> < / drug> < / content> < / content2> < / content1> < / content> < / individual> < / individual> < / drug2> < / drug1> < / individual> < / individual> < / drug2> < / drug1> < / individual> < / individual> < / individual> < / individual> < / individual> < / drug2> < / drug1>

Claims

1. A medication guidance generation system, characterized in that, include: The knowledge base construction module is used to build a medication guidance knowledge base; The knowledge graph terminology construction module is used to build a knowledge graph and standard terminology set based on the medication guidance knowledge base. The index building module is used to split the medication guidance knowledge base and build vector indexes and full-text indexes; The personal information processing module is used to acquire and preprocess patient personal information to obtain preprocessed personal information, which includes: standard terms corresponding to personal information keywords, vectorized data and word segmentation data corresponding to personal information; The mixed-search module is used to perform retrieval based on the preprocessed personal information using the knowledge graph, vector index, and full-text index to obtain a first retrieval result, a second retrieval result, and a third retrieval result, and to obtain an intersection result based on the first retrieval result, the second retrieval result, and the third retrieval result; The generation module is used to generate medication guidance for the target patient based on the intersection result.

2. The medication guidance generation system according to claim 1, characterized in that, The knowledge base construction module includes: Data acquisition submodule, data processing submodule, and setup submodule; The data acquisition submodule is used to acquire medication guidance materials, the data processing submodule is used to remove duplicate information, correct erroneous data and unify the format of the medication guidance materials to obtain preprocessed medication guidance materials, and the construction submodule is used to construct a medication guidance knowledge base based on the preprocessed medication guidance materials.

3. The medication guidance generation system according to claim 1, characterized in that, The atlas terminology construction module includes: The module includes an entity recognition model construction submodule, an initial graph construction submodule, and a final graph construction submodule. The entity recognition model construction submodule is used to construct an entity recognition model based on the medication guidance knowledge base and the BERT model. The initial knowledge graph construction submodule is used to complete the article entity annotation of the medication guidance knowledge base based on the entity recognition model to obtain an initial knowledge graph. The final knowledge graph construction submodule is used to fuse the initial knowledge graph and standardize the entity names into standard names to obtain a knowledge graph and a standard terminology set.

4. A medication guidance generation system according to claim 3, characterized in that, The index building module includes: Submodules include: text splitting, text preprocessing, vectorization, and word segmentation. The splitting submodule is used to perform structured splitting of the drug guidance knowledge base to obtain split text data. The text preprocessing submodule is used to perform text cleaning on the split text data to obtain preprocessed split data. The vectorization submodule is used to vectorize the preprocessed split data using a vector model to obtain a vector index. The word segmentation submodule is used to segment the preprocessed split data using a word segmentation model to obtain a full-text index.

5. A medication guidance generation system according to claim 4, characterized in that, The vector model is gte-Qwen2.

6. A medication guidance generation system according to claim 5, characterized in that, The word segmentation model is jieba.

7. A medication guidance generation system according to claim 6, characterized in that, The personal information processing module includes: The module includes a personal information acquisition submodule, a first conversion submodule, a second conversion submodule, and a third conversion submodule. The personal information acquisition submodule is used to acquire patient personal information. The first conversion submodule is used to identify entity keywords of patient personal information using the entity recognition model and convert them into standard terms. The second conversion submodule is used to vectorize patient personal information according to the vector model to obtain vectorized personal information. The third conversion submodule is used to segment patient personal information according to the word segmentation model to obtain word-segmented personal information.

8. A medication guidance generation system according to claim 6, characterized in that, The pooled testing module includes: The system comprises the first search submodule, the second search submodule, the third search submodule, the rearrangement submodule, and the intersection submodule. The first retrieval submodule is used to perform a first retrieval based on the knowledge graph and standard terminology set to obtain a first retrieval result. The second retrieval submodule is used to perform a second retrieval based on the vector index to obtain a second retrieval result. The third retrieval submodule is used to perform a third retrieval based on the full-text index to obtain a third retrieval result. The rearrangement submodule is used to rearrange the second retrieval result and the third retrieval result using a rearrangement model to obtain a rearrangement result. The intersection submodule is used to determine the intersection result based on the rearrangement result and the first retrieval result.

Images