Auxiliary method and auxiliary device for scientific research in medical specialized field

Abstract

The invention provides an auxiliary method and an auxiliary device for scientific research in the field of medical specialist, and the auxiliary method comprises the steps: analyzing query statement information to determine entity query information when the query statement information for performing literature retrieval on a target field of medical specialist is received, retrieving in the knowledge vector database by using a mixed retrieval mechanism to generate a retrieval document; when instruction information for analyzing and summarizing the target literature is received, the target literature is interpreted to determine element information, and a research comparison matrix and research trend summarization information are generated; and when instruction information for carrying out scientific research thought mining on the target medical specialized field is received, mining a research blank area in the knowledge vector database, and generating scientific research thought auxiliary information by utilizing a feasibility verification auxiliary mechanism based on a thinking reasoning chain of the research blank area. By means of the method, the accuracy and correlation of document retrieval in the medical specialized field and the efficiency, logicality and reliability of researching the medical specialized field are improved.

Description

Technical Field

[0001] This application relates to the field of medical research support technology, and in particular to an auxiliary method and device for medical research. Background Technology

[0002] Currently, clinicians struggle to systematically learn the latest literature and conduct efficient scientific research, while the volume of medical literature is growing exponentially. Traditional keyword retrieval systems suffer from low recall and poor precision, failing to meet the research needs of medical specialties.

[0003] Existing AI-assisted systems are limited to simple content generation, which can lead to problems such as generating illusions and difficulties in tracing sources, and they lack rigorous evidence-based logic support. In addition, existing AI-assisted systems do not distinguish the characteristics of medical specialties, resulting in subject-specific biases in literature search results. The evidence-based reasoning process lacks adaptation to specialty guidelines and cannot meet the differentiated scientific research evaluation standards of different medical specialties. Summary of the Invention

[0004] In view of this, the purpose of this application is to provide an auxiliary method and device for scientific research in the medical specialty field. By using a hybrid retrieval mechanism to search in the knowledge vector database corresponding to the medical specialty field, the accuracy and relevance of literature retrieval in the medical specialty field are improved. The literature is deeply interpreted to determine the element information and generate the corresponding research comparison matrix, and then the corresponding research trend summary information is generated. When mining research ideas in the medical specialty field, based on the reasoning chain corresponding to each research gap area, the feasibility verification auxiliary mechanism is used to generate auxiliary information for research ideas, thereby improving the efficiency, logic and reliability of research in the medical specialty field.

[0005] This application provides an auxiliary method for scientific research in the field of medical specialties, the auxiliary method including: In response to receiving a query statement from a user representing a literature search for a target medical specialty, the system uses a preset hybrid retrieval mechanism to search a preset knowledge vector database corresponding to the target medical specialty based on the entity query information determined by parsing the query statement, and generates the retrieved literature corresponding to the query statement. In response to receiving a first instruction message representing the user's analysis and summary of target documents determined from the retrieved documents and selected instruction documents, a research comparison matrix corresponding to the target documents is generated based on the element information determined by in-depth interpretation of the target documents, and research trend summary information corresponding to the target documents is generated based on the research comparison matrix. In response to receiving a second instruction representing the user's research ideas for the target medical specialty, at least one research gap area is mined from the knowledge vector database corresponding to the target medical specialty, and based on the reasoning chain corresponding to each research gap area, a preset feasibility verification assistance mechanism is used to generate research idea assistance information corresponding to the target medical specialty.

[0006] Furthermore, the knowledge vector database is pre-constructed through the following steps: Document data is acquired in batches from a preset database and knowledge base using full-text acquisition and abstract acquisition methods. The document data is then cleaned and standardized to obtain the first document data to be processed. The first document data to be processed is converted into a preset document format, and the first document data to be processed converted into the preset document format is structured using a rule engine to obtain the second document data to be processed. The second document data to be processed is segmented according to preset chapters and semantic paragraphs to obtain the third document data to be processed; The third document data to be processed is domain-specific encoded to generate an enhancement vector, and the third document data to be processed is semantically sliced ​​encoded to generate semantic paragraph encoded data, so as to determine the enhancement vector and the semantic paragraph encoded data as the fourth document data to be processed. The fourth document data to be processed is optimized by sparse indexing, and the fourth document data to be processed is scored and marked using a preset scoring rule to obtain target document data, so as to construct the knowledge vector database composed of the target document data.

[0007] Furthermore, based on the entity query information determined by parsing the query statement information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to generate the retrieval documents corresponding to the query statement information, including: The query statement information is parsed to determine the entity information and query intent information corresponding to the query statement information; Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data; A pre-defined language fine-tuning model is used to score the relevance of the documents in the candidate document data, and the score result corresponding to each document in the candidate document data is obtained. The candidate literature data is sorted and labeled with confidence level according to the scoring results to generate the retrieval literature corresponding to the query statement information.

[0008] Furthermore, based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data, including: Based on the entity information and the query intent information, vector retrieval is performed in the preset knowledge vector database corresponding to the target medical specialty field to generate vector retrieval literature data. Simultaneously, based on the entity information and the query intent information, ES keyword retrieval is performed in the knowledge vector database to generate keyword retrieval document data; Duplicate documents are removed from the vector-retrieval document data and the keyword-retrieval document data to determine candidate document data.

[0009] Furthermore, the step of generating a research comparison matrix corresponding to the target document based on the element information determined through in-depth interpretation of the target document includes: The target literature is analyzed in depth to determine the corresponding element information; wherein, the element information includes at least research design information, endpoint information, sample size, research background information, research purpose information, statistical results, and core content information; Based on the aforementioned element information, the target documents are compared and interpreted, and parameters are compared to generate a research comparison matrix corresponding to the target documents.

[0010] Furthermore, the step of generating a summary of research trends corresponding to the target literature based on the research comparison matrix includes: Based on the research comparison matrix, generate consistency analysis conclusions for the target literature. The consistency analysis conclusions are aggregated and statistically analyzed to generate a summary of research trends corresponding to the target literature.

[0011] Furthermore, at least one research gap area is mined from the knowledge vector database corresponding to the target medical specialty, and based on the reasoning chain corresponding to each research gap area, a preset feasibility verification assistance mechanism is used to generate research idea assistance information corresponding to the target medical specialty, including: In the knowledge vector database corresponding to the target medical specialty, blank path nodes that have not been sufficiently studied in the target medical specialty are detected to determine at least one research blank area corresponding to the target medical specialty. Based on the research gaps and the pre-defined research hotspots corresponding to the target medical specialty, potential research directions for the target medical specialty are determined. Construct a thought process chain corresponding to each research gap area, and based on the thought process chain and the potential research direction information, generate research idea suggestions for the target medical specialty field; The feasibility of the proposed research ideas is assessed and the trend distribution is analyzed using a pre-set feasibility verification auxiliary mechanism. The assessment and analysis results are obtained, and the proposed research ideas are adjusted based on the assessment and analysis results to generate the proposed research ideas for the target medical specialty. Based on the proposed research ideas, corresponding recommended literature is retrieved from the knowledge vector database, and the proposed research ideas and the recommended literature are identified as auxiliary information for research ideas in the target medical specialty.

[0012] This application also provides an auxiliary device for scientific research in the field of medical specialties, the auxiliary device comprising: The literature retrieval module is used to respond to the user's input query statement information representing a literature retrieval for a target medical specialty field, and based on the entity query information determined by parsing the query statement information, to perform a retrieval in the preset knowledge vector database corresponding to the target medical specialty field using a preset hybrid retrieval mechanism, and generate the retrieved literature corresponding to the query statement information. The document interpretation module is used to respond to a first instruction message received that represents the user's analysis and summary of the target document determined in the searched documents and selected instruction documents, generate a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document, and generate research trend summary information corresponding to the target document based on the research comparison matrix. The research mining module is used to respond to receiving a second instruction information representing the user's research ideas mining in the target medical specialty field, to mine at least one research gap area in the knowledge vector database corresponding to the target medical specialty field, and to generate research idea auxiliary information corresponding to the target medical specialty field based on the thought reasoning chain corresponding to each research gap area using a preset feasibility verification auxiliary mechanism.

[0013] Furthermore, when the document retrieval module is used to pre-build the knowledge vector database, the document retrieval module is used to: Document data is acquired in batches from a preset database and knowledge base using full-text acquisition and abstract acquisition methods. The document data is then cleaned and standardized to obtain the first document data to be processed. The first document data to be processed is converted into a preset document format, and the first document data to be processed converted into the preset document format is structured using a rule engine to obtain the second document data to be processed. The second document data to be processed is segmented according to preset chapters and semantic paragraphs to obtain the third document data to be processed; The third document data to be processed is domain-specific encoded to generate an enhancement vector, and the third document data to be processed is semantically sliced ​​encoded to generate semantic paragraph encoded data, so as to determine the enhancement vector and the semantic paragraph encoded data as the fourth document data to be processed. The fourth document data to be processed is optimized by sparse indexing, and the fourth document data to be processed is scored and marked using a preset scoring rule to obtain target document data, so as to construct the knowledge vector database composed of the target document data.

[0014] Furthermore, when the document retrieval module is used to retrieve entity query information determined by parsing the query statement information, and to use a preset hybrid retrieval mechanism to search the preset knowledge vector database corresponding to the target medical specialty field, and generate the retrieved documents corresponding to the query statement information, the document retrieval module is used to: The query statement information is parsed to determine the entity information and query intent information corresponding to the query statement information; Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data; A pre-defined language fine-tuning model is used to score the relevance of the documents in the candidate document data, and the score result corresponding to each document in the candidate document data is obtained. The candidate literature data is sorted and labeled with confidence level according to the scoring results to generate the retrieval literature corresponding to the query statement information.

[0015] Furthermore, when the literature retrieval module is used to retrieve candidate literature data based on the entity information and the query intent information using a preset hybrid retrieval mechanism in the preset knowledge vector database corresponding to the target medical specialty, the literature retrieval module is used to: Based on the entity information and the query intent information, vector retrieval is performed in the preset knowledge vector database corresponding to the target medical specialty field to generate vector retrieval literature data. Simultaneously, based on the entity information and the query intent information, ES keyword retrieval is performed in the knowledge vector database to generate keyword retrieval document data; Duplicate documents are removed from the vector-retrieval document data and the keyword-retrieval document data to determine candidate document data.

[0016] Furthermore, when the document interpretation module generates a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document, the document interpretation module is used to: The target literature is analyzed in depth to determine the corresponding element information; wherein, the element information includes at least research design information, endpoint information, sample size, research background information, research purpose information, statistical results, and core content information; Based on the aforementioned element information, the target documents are compared and interpreted, and parameters are compared to generate a research comparison matrix corresponding to the target documents.

[0017] Furthermore, when the literature interpretation module generates a summary of research trends corresponding to the target literature based on the research comparison matrix, the literature interpretation module is used to: Based on the research comparison matrix, generate consistency analysis conclusions for the target literature. The consistency analysis conclusions are aggregated and statistically analyzed to generate a summary of research trends corresponding to the target literature.

[0018] Furthermore, when the research mining module mines at least one research gap area from the knowledge vector database corresponding to the target medical specialty, and generates research idea support information corresponding to the target medical specialty based on the reasoning chain corresponding to each research gap area using a preset feasibility verification support mechanism, the research mining module is used to: In the knowledge vector database corresponding to the target medical specialty, blank path nodes that have not been sufficiently studied in the target medical specialty are detected to determine at least one research blank area corresponding to the target medical specialty. Based on the research gaps and the pre-defined research hotspots corresponding to the target medical specialty, potential research directions for the target medical specialty are determined. Construct a thought process chain corresponding to each research gap area, and based on the thought process chain and the potential research direction information, generate research idea suggestions for the target medical specialty field; The feasibility of the proposed research ideas is assessed and the trend distribution is analyzed using a pre-set feasibility verification auxiliary mechanism. The assessment and analysis results are obtained, and the proposed research ideas are adjusted based on the assessment and analysis results to generate the proposed research ideas for the target medical specialty. Based on the proposed research ideas, corresponding recommended literature is retrieved from the knowledge vector database, and the proposed research ideas and the recommended literature are identified as auxiliary information for research ideas in the target medical specialty.

[0019] This application also provides an electronic device, including: a processor, a memory, and a bus. The memory stores machine-readable instructions executable by the processor. When the electronic device is running, the processor communicates with the memory via the bus. When the machine-readable instructions are executed by the processor, they perform the steps of the above-described auxiliary method for scientific research in the medical specialty field.

[0020] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, performs the steps of the above-described auxiliary method for scientific research in the field of medical specialties.

[0021] The embodiments of this application provide an auxiliary method and apparatus for scientific research in the medical specialty field. The auxiliary method includes: responding to receiving query information input by a user representing a literature search for a target medical specialty field, and based on entity query information determined by parsing the query information, performing a search in a preset knowledge vector database corresponding to the target medical specialty field using a preset hybrid search mechanism, and generating searched literature corresponding to the query information; responding to receiving a first instruction information representing the user's analysis and summary of target literature determined from the searched literature and selected indicator literature, generating a research comparison matrix corresponding to the target literature based on element information determined by in-depth interpretation of the target literature, and generating research trend summary information corresponding to the target literature based on the research comparison matrix; responding to receiving a second instruction information representing the user's research idea mining in the target medical specialty field, mining at least one research gap area in the knowledge vector database corresponding to the target medical specialty field, and generating research idea auxiliary information corresponding to the target medical specialty field based on the reasoning chain corresponding to each research gap area using a preset feasibility verification auxiliary mechanism.

[0022] Compared to traditional keyword retrieval systems and existing AI-assisted systems that are limited to simple content generation and suffer from problems such as generation illusion and difficulty in tracing sources, this hybrid retrieval mechanism improves the accuracy and relevance of literature retrieval in the medical specialty field by searching knowledge vector databases corresponding to the literature. It deeply interprets the literature to determine key information and generates a research comparison matrix corresponding to the literature, thereby generating a summary of research trends. When exploring research ideas in the medical specialty field, it uses a feasibility verification mechanism to generate research idea support information based on the reasoning chain corresponding to each research gap area, thus improving the efficiency, logic, and reliability of research in the medical specialty field.

[0023] To make the above-mentioned objectives, features and advantages of this application more apparent and understandable, preferred embodiments are described below in detail with reference to the accompanying drawings. Attached Figure Description

[0024] To more clearly illustrate the technical solutions of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly introduced below. It should be understood that the following drawings only show some embodiments of this application and should not be regarded as a limitation of the scope. For those skilled in the art, other related drawings can be obtained based on these drawings without creative effort.

[0025] Figure 1 A flowchart illustrating an auxiliary method for scientific research in the field of medical specialties, provided as an embodiment of this application; Figure 2 A schematic diagram of the structure of an auxiliary device for scientific research in the field of medical specialty provided in an embodiment of this application; Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Detailed Implementation

[0026] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations. Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely represents selected embodiments of this application. Based on the embodiments of this application, every other embodiment obtained by those skilled in the art without inventive effort falls within the scope of protection of this application.

[0027] Research has found that clinicians currently struggle to systematically learn the latest literature and conduct efficient research, while the volume of medical literature is growing exponentially. Traditional keyword retrieval systems suffer from low recall and poor precision, failing to meet the research needs of medical specialties.

[0028] Existing AI-assisted systems are limited to simple content generation, which can lead to problems such as generating illusions and difficulties in tracing sources, and they lack rigorous evidence-based logic support. In addition, existing AI-assisted systems do not distinguish the characteristics of medical specialties, resulting in subject-specific biases in literature retrieval results. For example, oncology-specific terms are mixed with endocrinology literature in the recall, and the evidence-based reasoning process lacks adaptation to specialty guidelines, failing to meet the differentiated scientific research evaluation standards of different medical specialties.

[0029] Based on this, this application provides an auxiliary method for scientific research in the field of medical specialties. By using a hybrid retrieval mechanism to search in the knowledge vector database corresponding to the medical specialty, the accuracy and relevance of literature retrieval in the medical specialty are improved. The literature is deeply interpreted to determine the element information and generate the research comparison matrix corresponding to the literature, thereby generating the research trend summary information corresponding to the literature. When mining research ideas in the medical specialty, based on the reasoning chain corresponding to each research gap area, a feasibility verification auxiliary mechanism is used to generate auxiliary information for research ideas, thereby improving the efficiency, logic and reliability of research in the medical specialty.

[0030] Please see Figure 1 , Figure 1 A flowchart illustrating an auxiliary method for scientific research in the field of medical specialty, provided as an embodiment of this application. Figure 1 As shown in the embodiments of this application, the auxiliary method for scientific research in the field of medical specialties includes: S101. In response to receiving a query statement information input by a user representing a literature search for a target medical specialty, based on the entity query information determined by parsing the query statement information, a preset hybrid search mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty, and the searched literature corresponding to the query statement information is generated.

[0031] It should be noted that medical specialties refer to the specific medical professional directions that doctors choose to study and train in after basic medical education. These cover various directions, from internal medicine and surgery to more specific ones such as cardiology and neurology; for example, ophthalmology, otolaryngology, dermatology, radiology, and pathology.

[0032] Here, the target medical specialty is a medical specialty that the user selects from a variety of medical specialty fields.

[0033] In this embodiment of the application, the query statement information is a natural language query statement input by the user that represents a literature search for a target medical specialty, such as "second-line treatment strategies after failure of immunotherapy for advanced liver cancer".

[0034] Here, when the user inputs query information, the query information is encoded and vectorized in real time using a preset MiniLM model, for example, "PD-L1 negative gastric cancer treatment strategy"; furthermore, the synonyms corresponding to the query terms in the query information are dynamically expanded based on the preset MeSH vocabulary, for example, "PD-1 inhibitor" is expanded to "pembrolizumab".

[0035] In this application embodiment, the hybrid retrieval mechanism includes, but is not limited to, vector retrieval and ES keyword retrieval, and also includes using a preset language fine-tuning model to score the relevance of documents in order to rearrange the retrieved documents according to relevance.

[0036] In this embodiment, the pre-defined knowledge vector database may include a Milvus vector database. The knowledge vector database stores literature data in dense vector form and employs a hierarchical navigable small world (HNSW) index, supporting millisecond-level approximate nearest neighbor (ANN) retrieval. The HNSW index achieves fast approximate search through a multi-layer graph structure, with nodes randomly connected at each layer. Higher layers are used for coarse screening, and lower layers for fine screening. The partitioning strategy includes: dividing the database by specialty field, with each partition independently indexed to reduce cross-partition search overhead. The data format includes: each record contains a literature ID, literature data stored in dense vector form, and a pointer to associated metadata. When new literature data is added, only a local index is built for the new data, and the partition-level index is updated to avoid global reconstruction. It also supports periodic index merging to balance query efficiency and storage cost.

[0037] In one possible implementation of this application, the step of pre-constructing the knowledge vector database in step S101 may include: S101A. In a preset database and knowledge base, literature data is acquired in batches using full-text acquisition and abstract acquisition methods. The literature data is then cleaned and standardized to obtain the first literature data to be processed.

[0038] In the embodiments of this application, the preset databases include, but are not limited to, authoritative domestic and foreign databases (e.g., PubMed, EMBASE, CNKI, and Wanfang, etc.), and the preset knowledge bases include, but are not limited to, clinical guidelines, expert consensus, and textbook content.

[0039] In this step, query-based methods (e.g., MeSH thesaurus and subject terms) are used to batch retrieve the full text, abstract, and metadata (e.g., author, institution, and publication date) of documents from a preset database and knowledge base. Document data is collected through both full text and abstract channels, and the document data is cleaned and standardized to remove redundant information, including headers and footers, to obtain the first document data to be processed.

[0040] S101B. Convert the first document data to be processed into a preset document format, and use a rule engine to perform structured processing on the first document data to be processed converted into the preset document format to obtain the second document data to be processed.

[0041] In the embodiments of this application, the preset document format may include a Markdown format. Markdown is a lightweight markup language that allows documents to be written in plain text format and then converted into structured HTML (Hypertext Markup Language) documents.

[0042] In this step, when converting the first literature data to be processed into a preset document format, statistical significance markers are preserved and LaTeX rendering of charts and formulas is supported; then, a rule engine (e.g., a regular expression engine) is used to perform structured processing on the first literature data to be processed converted into the preset document format to obtain the second literature data to be processed.

[0043] S101C. The second document data to be processed is segmented according to preset chapters and semantic paragraphs to obtain the third document data to be processed.

[0044] In this step, the second literature data to be processed is divided according to a preset chapter to preserve the contextual coherence through logical block division, and the long text in the second literature data to be processed is divided into semantic paragraphs (for example, each paragraph describes a research conclusion) using the TextTiling algorithm to obtain the third literature data to be processed.

[0045] S101D. Perform domain-specific encoding on the third document data to be processed to generate an enhancement vector, and perform semantic segment encoding on the third document data to be processed to generate semantic paragraph encoding data, so as to determine the enhancement vector and the semantic paragraph encoding data as the fourth document data to be processed.

[0046] In this step, the SciBERT model is used to perform domain-specific encoding on the third literature data to be processed, generating multi-dimensional vectors and attaching research type labels (e.g., RCT=0, cohort study=1, etc.), generating augmented vectors, and semantic slicing encoding is performed independently on each semantic paragraph in the third literature data to support fine-grained retrieval (e.g., retrieving only the "Methods" section), so that the augmented vectors and semantic paragraph encoded data are identified as the fourth literature data to be processed.

[0047] S101E: Perform sparse index optimization on the fourth document data to be processed, and use preset scoring rules to score and mark the fourth document data to be processed to obtain target document data, so as to construct the knowledge vector database composed of the target document data.

[0048] In this step, the fourth document data to be processed is optimized with sparse index to support Boolean logic, phrase retrieval and fuzzy matching (e.g., “immunotherap*” matches “immunotherapy”), and the fourth document data to be processed is scored and marked using preset scoring rules (e.g., the title field is marked with a weight of 3, the abstract field with a weight of 2, and the full text field with a weight of 1) to obtain the target document data.

[0049] Furthermore, a knowledge vector database is constructed based on the target literature data.

[0050] In one possible implementation of this application, in specific implementation, the step S101, based on the entity query information determined by parsing the query statement information, uses a preset hybrid retrieval mechanism to search the preset knowledge vector database corresponding to the target medical specialty field, and generates the retrieval literature corresponding to the query statement information, may include: S1011. Parse the query statement information to determine the entity information and query intent information corresponding to the query statement information.

[0051] In this step, a pre-set large model algorithm is used to accurately semantically identify the query information in order to avoid incomplete literature retrieval caused by missed keyword detection, and to determine the entity information and query intent information corresponding to the query information.

[0052] For example, entity information may include information such as advanced liver cancer, immunotherapy, and second-line treatment; query intent information may include strategy exploration information.

[0053] S1012. Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data.

[0054] In the embodiments of this application, the hybrid retrieval mechanism can utilize the concatenated retrieval enhancement generation technology (RAG) and ultra-long text summarization combined with sliding window and attention mechanisms to perform retrieval in a preset knowledge vector database corresponding to the target medical specialty field. While suppressing the reasoning "illusion" of the large language model, it extracts the core content of the knowledge vector database and supports the original text tracing of the literature.

[0055] In one possible implementation of this application, step S1012 may include: S10121. Based on the entity information and the query intent information, perform vector retrieval in the preset knowledge vector database corresponding to the target medical specialty field to generate vector retrieval literature data.

[0056] In this step, the vector database attributes of the preset knowledge vector database corresponding to the target medical specialty are used to perform vector retrieval in the knowledge vector database according to the entity information and query intent information corresponding to the query statement information, and generate vector retrieval literature data.

[0057] When performing vector retrieval, a corresponding cosine similarity threshold is set. For example, the cosine similarity threshold can be set to 0.75.

[0058] For example, vector retrieval in a knowledge vector database can retrieve multiple relevant document data. The top 100 documents, ranked from highest to lowest relevance, are then selected to generate corresponding vector retrieval document data.

[0059] S10122. Simultaneously, based on the entity information and the query intent information, perform ES keyword retrieval in the knowledge vector database to generate keyword retrieval document data.

[0060] In this step, while performing vector retrieval, based on the entity information corresponding to the query statement information, ES keyword retrieval is performed in the knowledge vector database to match information such as title, abstract and keywords through Boolean logic.

[0061] For example, the search query for ES keyword search could include: title: "immunotherapy" AND abstract: "gastric cancer".

[0062] For example, performing an ES keyword search in a knowledge vector database can retrieve multiple relevant document data. The top 200 documents, ranked from highest to lowest relevance, are then selected to generate the corresponding keyword-based document search data.

[0063] S10123. Remove duplicate documents from the vector retrieval document data and the keyword retrieval document data to determine candidate document data.

[0064] In this step, the vector search literature data and the keyword search literature data are integrated, and duplicate literature is removed to determine the candidate literature data.

[0065] S1013. Use a preset language fine-tuning model to score the relevance of the documents in the candidate document data, and obtain the score result corresponding to each document in the candidate document data.

[0066] In this embodiment of the application, the preset language fine-tuning model may include a lightweight LoRA-tuned DeBERTa model.

[0067] S1014. Sort and mark the documents in the candidate document data according to the scoring results to generate the search documents corresponding to the query statement information.

[0068] In this step, the candidate literature data are sorted from high to low according to the score results corresponding to each literature, and the literature is marked with confidence level (e.g., high, medium, low) to perform semantic reordering of relevance and generate the retrieval literature corresponding to the query statement information.

[0069] Furthermore, users can further filter the retrieved literature based on parameters such as research design, level of evidence, and sample size.

[0070] Furthermore, long-term memory can be used to store users' historical behavioral data (e.g., high-frequency search terms and writing style preferences), and personalized model parameters for the knowledge vector database can be generated through knowledge distillation; for example, if users prefer to use "total lifetime" instead of "OS", the knowledge vector database will automatically replace the terminology.

[0071] Furthermore, short-term memory (SRM) caches can be used to maintain the context of retrieval interaction sessions (e.g., retrieval results and generated content), and an LRU strategy can be employed to manage the SRM cache.

[0072] Furthermore, it can analyze users' frequently searched topics and generate trend charts; for example, the search volume for "CAR-T therapy" has increased by 30% in the past month.

[0073] S102. In response to receiving a first instruction message representing the user's analysis and summary of the target document determined from the retrieved documents and selected instruction documents, a research comparison matrix corresponding to the target document is generated based on the element information determined by in-depth interpretation of the target document, and research trend summary information corresponding to the target document is generated based on the research comparison matrix.

[0074] In this embodiment of the application, the target literature includes retrieved literature from a preset knowledge vector database corresponding to the target medical specialty and / or indicated literature selected by the user.

[0075] Here, the first instruction information refers to the instruction statement information that analyzes and summarizes the user's expectations for the target document identified in the searched and selected instruction documents.

[0076] In one possible implementation of this application, in specific implementation, the step S102 of generating a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document may include: S1021. Perform in-depth analysis of the target document to determine the element information corresponding to the target document.

[0077] The information elements include at least the research design information, endpoint information, sample size, research background information, research objective information, statistical results, and core content information.

[0078] In this step, after conducting an in-depth interpretation of the target document to determine the corresponding element information, the structured tables (e.g., JSON and tabular formats) in the element information are extracted.

[0079] In this way, by analyzing the core content of the literature, users can grasp the key information of the article and judge the value and importance of the literature in a short time, thereby improving the efficiency of literature screening.

[0080] S1022. Based on the element information, the target documents are compared and interpreted and parameters are compared to generate a research comparison matrix corresponding to the target documents.

[0081] In the embodiments of this application, the parameters for comparison may include, but are not limited to, efficacy indicators, P-values, and confidence intervals.

[0082] Here, the comparison matrix is ​​a tool used to systematically list and compare the differences between different objects, helping users evaluate each option through a set of predefined criteria or characteristics, thus making it easier to identify the advantages and disadvantages of each option.

[0083] In one possible implementation of this application, in specific implementation, step S102, which generates the research trend summary information corresponding to the target literature based on the research comparison matrix, may include: S1023. Based on the research comparison matrix, generate consistency analysis conclusion information corresponding to the target literature.

[0084] For example, the consistency analysis conclusion information may include the majority of studies supporting conclusion A, individual studies having objections and an analysis of the reasons, and may also include points of evidence conflict and possible reasons, such as differences in inclusion criteria and follow-up time.

[0085] S1024. Aggregate and statistically analyze the consistency analysis conclusions to generate a summary of research trends corresponding to the target literature.

[0086] In this step, the consistency analysis conclusions are aggregated and statistically analyzed according to time dimension, study area, and research method to discover the trend information corresponding to the target literature, and then trend charts and research density heat maps are drawn, and trend summaries and suggestions for future potential research directions are determined.

[0087] Here, the research trend summary information includes, but is not limited to, information on changing trends, trend summaries, suggestions for potential future research directions, trend charts, and research density heatmaps.

[0088] S103. In response to receiving a second instruction message representing the user's research ideas mining for the target medical specialty, at least one research gap area is mined from the knowledge vector database corresponding to the target medical specialty, and based on the thought reasoning chain corresponding to each research gap area, a preset feasibility verification auxiliary mechanism is used to generate research idea auxiliary information corresponding to the target medical specialty.

[0089] Here, the second instruction information refers to the instruction statements used by the user to explore research ideas in the target medical specialty field.

[0090] In one possible implementation of this application, step S103 may include: S1031. Detect blank path nodes in the knowledge vector database corresponding to the target medical specialty field that have not been sufficiently studied, so as to determine at least one research blank area corresponding to the target medical specialty field.

[0091] In this step, based on the reasoning of the knowledge vector database corresponding to the target medical specialty, path nodes of the target medical specialty that have not been sufficiently studied are detected (Research Gap Detection), and a domain overview corresponding to the target medical specialty is generated through intelligent search, thereby identifying at least one research gap area corresponding to the target medical specialty.

[0092] S1032. Based on the research gap area and the preset research hotspot information corresponding to the target medical specialty field, determine the potential research direction information corresponding to the target medical specialty field.

[0093] For example, the pre-set research hotspot information may include information such as new targets and combination therapies, and the potential research direction information may include a potentially valuable research topic.

[0094] S1033. Construct a thought process chain corresponding to each research gap area, and based on the thought process chain and the potential research direction information, generate research idea suggestions for the target medical specialty field.

[0095] In this application embodiment, the reasoning chain corresponding to the research blank area includes background questions, clinical needs, existing evidence gaps and potential hypotheses, and the research idea suggestion information includes a systematic research idea suggestion, the research value of the topic and future research directions.

[0096] S1034. Using a preset feasibility verification auxiliary mechanism, the feasibility of the research idea suggestion information is assessed and trend distribution analysis is performed to obtain the assessment analysis results. Based on the assessment analysis results, the research idea suggestion information is adjusted to generate target research idea suggestion information corresponding to the target medical specialty field.

[0097] In this step, the pre-set feasibility verification assistance mechanism can automatically retrieve relevant model research, animal experiments, and early clinical trial data to assist in the feasibility assessment of topic selection, obtain assessment and analysis results, and adjust the research idea suggestions based on the assessment and analysis results to generate target research idea suggestions corresponding to the target medical specialty field.

[0098] For example, if a user specifies the target medical specialty as "strategies after failure of immunotherapy for liver cancer", literature analysis based on reasoning through thought chain and knowledge vector databases will generate research idea suggestions as "rapid growth in research on PD-1 resistance mechanisms". Based on the evaluation results of the feasibility assessment and trend distribution analysis of the research idea suggestions using a pre-set feasibility verification auxiliary mechanism, the research idea suggestions will be adjusted, and the target research idea suggestions corresponding to the target medical specialty may include "exploration of PD-1 resistance mechanisms in liver cancer based on the immune spectrum of the tumor microenvironment".

[0099] S1035. Based on the target research idea suggestion information, retrieve the corresponding recommended literature from the knowledge vector database, and determine the target research idea suggestion information and the recommended literature as the research idea auxiliary information corresponding to the target medical specialty field.

[0100] In this step, based on the suggested information of the target research ideas, the corresponding recommended literature is retrieved from the knowledge vector database, and the trend distribution analysis results are output, including time distribution, journal distribution, scholar distribution, regional distribution, etc., so as to determine the auxiliary information of the research ideas corresponding to the target medical specialty field.

[0101] The research assistance method for medical specialties provided in this application improves the accuracy and relevance of literature retrieval in the medical specialty field by using a hybrid retrieval mechanism to search the knowledge vector database corresponding to the medical specialty field. It performs in-depth interpretation of the literature to determine the element information and generates the corresponding research comparison matrix, thereby generating the corresponding research trend summary information. When mining research ideas in the medical specialty field, it generates research idea assistance information based on the reasoning chain corresponding to each research gap area, using a feasibility verification assistance mechanism, thereby improving the efficiency, logic and reliability of research in the medical specialty field.

[0102] Please see Figure 2 , Figure 2 This is a schematic diagram of the structure of an auxiliary device for scientific research in the field of medical specialty, provided as an embodiment of this application. Figure 2 As shown, the auxiliary device 200 includes: The literature retrieval module 210 is used to respond to the user input of a query statement representing a literature retrieval for a target medical specialty, and based on the entity query information determined by parsing the query statement, to perform a retrieval in the preset knowledge vector database corresponding to the target medical specialty using a preset hybrid retrieval mechanism, and generate the retrieved literature corresponding to the query statement. The document interpretation module 220 is used to respond to receiving a first instruction message that represents the user's analysis and summary of the target document determined in the searched documents and selected instruction documents, generate a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document, and generate research trend summary information corresponding to the target document based on the research comparison matrix. The research mining module 230 is used to respond to receiving a second instruction information representing the user's research ideas mining for the target medical specialty field, to mine at least one research blank area in the knowledge vector database corresponding to the target medical specialty field, and to generate research idea auxiliary information corresponding to the target medical specialty field based on the thought reasoning chain corresponding to each research blank area using a preset feasibility verification auxiliary mechanism.

[0103] Furthermore, when the document retrieval module 210 is used to pre-build the knowledge vector database, the document retrieval module 210 is used to: Document data is acquired in batches from a preset database and knowledge base using full-text acquisition and abstract acquisition methods. The document data is then cleaned and standardized to obtain the first document data to be processed. The first document data to be processed is converted into a preset document format, and the first document data to be processed converted into the preset document format is structured using a rule engine to obtain the second document data to be processed. The second document data to be processed is segmented according to preset chapters and semantic paragraphs to obtain the third document data to be processed; The third document data to be processed is domain-specific encoded to generate an enhancement vector, and the third document data to be processed is semantically sliced ​​encoded to generate semantic paragraph encoded data, so as to determine the enhancement vector and the semantic paragraph encoded data as the fourth document data to be processed. The fourth document data to be processed is optimized by sparse indexing, and the fourth document data to be processed is scored and marked using a preset scoring rule to obtain target document data, so as to construct the knowledge vector database composed of the target document data.

[0104] Furthermore, when the document retrieval module 210 is used to retrieve the entity query information determined by parsing the query statement information, and to generate the retrieved documents corresponding to the query statement information by using a preset hybrid retrieval mechanism in the preset knowledge vector database corresponding to the target medical specialty field, the document retrieval module 210 is used to: The query statement information is parsed to determine the entity information and query intent information corresponding to the query statement information; Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data; A pre-defined language fine-tuning model is used to score the relevance of the documents in the candidate document data, and the score result corresponding to each document in the candidate document data is obtained. The candidate literature data is sorted and labeled with confidence level according to the scoring results to generate the retrieval literature corresponding to the query statement information.

[0105] Furthermore, when the literature retrieval module 210 is used to retrieve candidate literature data based on the entity information and the query intent information using a preset hybrid retrieval mechanism in the preset knowledge vector database corresponding to the target medical specialty, the literature retrieval module 210 is used to: Based on the entity information and the query intent information, vector retrieval is performed in the preset knowledge vector database corresponding to the target medical specialty field to generate vector retrieval literature data. Simultaneously, based on the entity information and the query intent information, ES keyword retrieval is performed in the knowledge vector database to generate keyword retrieval document data; Duplicate documents are removed from the vector-retrieval document data and the keyword-retrieval document data to determine candidate document data.

[0106] Furthermore, when the document interpretation module 220 generates a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document, the document interpretation module 220 is used to: The target literature is analyzed in depth to determine the corresponding element information; wherein, the element information includes at least research design information, endpoint information, sample size, research background information, research purpose information, statistical results, and core content information; Based on the aforementioned element information, the target documents are compared and interpreted, and parameters are compared to generate a research comparison matrix corresponding to the target documents.

[0107] Furthermore, when the literature interpretation module 220 generates research trend summary information corresponding to the target literature based on the research comparison matrix, the literature interpretation module 220 is used to: Based on the research comparison matrix, generate consistency analysis conclusions for the target literature. The consistency analysis conclusions are aggregated and statistically analyzed to generate a summary of research trends corresponding to the target literature.

[0108] Furthermore, when the research mining module 230 mines at least one research gap area from the knowledge vector database corresponding to the target medical specialty, and generates research idea support information corresponding to the target medical specialty based on the reasoning chain corresponding to each research gap area using a preset feasibility verification support mechanism, the research mining module 230 is used to: In the knowledge vector database corresponding to the target medical specialty, blank path nodes that have not been sufficiently studied in the target medical specialty are detected to determine at least one research blank area corresponding to the target medical specialty. Based on the research gaps and the pre-defined research hotspots corresponding to the target medical specialty, potential research directions for the target medical specialty are determined. Construct a thought process chain corresponding to each research gap area, and based on the thought process chain and the potential research direction information, generate research idea suggestions for the target medical specialty field; The feasibility of the proposed research ideas is assessed and the trend distribution is analyzed using a pre-set feasibility verification auxiliary mechanism. The assessment and analysis results are obtained, and the proposed research ideas are adjusted based on the assessment and analysis results to generate the proposed research ideas for the target medical specialty. Based on the proposed research ideas, corresponding recommended literature is retrieved from the knowledge vector database, and the proposed research ideas and the recommended literature are identified as auxiliary information for research ideas in the target medical specialty.

[0109] The auxiliary device for scientific research in the medical specialty field provided in this application embodiment searches the knowledge vector database corresponding to the medical specialty field through a hybrid retrieval mechanism, which improves the accuracy and relevance of literature retrieval in the medical specialty field. It performs in-depth interpretation of the literature to determine the element information and generates the research comparison matrix corresponding to the literature, and then generates the research trend summary information corresponding to the literature. When mining research ideas in the medical specialty field, it generates research idea auxiliary information based on the reasoning chain corresponding to each research gap area mined, using a feasibility verification auxiliary mechanism, thereby improving the efficiency, logic and reliability of research in the medical specialty field.

[0110] Please see Figure 3 , Figure 3 This is a schematic diagram of the structure of an electronic device provided in an embodiment of this application. Figure 3 As shown, the electronic device 300 includes a processor 310, a memory 320, and a bus 330.

[0111] The memory 320 stores machine-readable instructions executable by the processor 310. When the electronic device 300 is running, the processor 310 and the memory 320 communicate via the bus 330. When the machine-readable instructions are executed by the processor 310, they can perform the operations described above. Figure 1 The steps of the auxiliary method for scientific research in the medical specialty field shown in the method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0112] This application also provides a computer-readable storage medium storing a computer program, which, when executed by a processor, can perform the above-described actions. Figure 1 The steps of the auxiliary method for scientific research in the medical specialty field shown in the method embodiment can be found in the method embodiment for specific implementation, and will not be repeated here.

[0113] Those skilled in the art will understand that, for the sake of convenience and brevity, the specific working processes of the systems, devices, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here.

[0114] In the several embodiments provided in this application, it should be understood that the disclosed systems, apparatuses, and methods can be implemented in other ways. The apparatus embodiments described above are merely illustrative. For example, the division of units is only a logical functional division, and in actual implementation, there may be other division methods. Furthermore, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Additionally, the shown or discussed mutual couplings, direct couplings, or communication connections may be through some communication interfaces; indirect couplings or communication connections between devices or units may be electrical, mechanical, or other forms.

[0115] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0116] In addition, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit.

[0117] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0118] Finally, it should be noted that the above-described embodiments are merely specific implementations of this application, used to illustrate the technical solutions of this application, and not to limit them. The scope of protection of this application is not limited thereto. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that any person skilled in the art can still modify or easily conceive of changes to the technical solutions described in the foregoing embodiments, or make equivalent substitutions for some of the technical features, within the scope of the technology disclosed in this application. Such modifications, changes, or substitutions do not cause the essence of the corresponding technical solutions to deviate from the spirit and scope of the technical solutions of the embodiments of this application, and should all be covered within the scope of protection of this application. Therefore, the scope of protection of this application should be determined by the scope of the claims.

Claims

1. A method for assisting scientific research in a medical specialty field, characterized in that, The auxiliary method includes: In response to receiving a query statement from a user representing a literature search for a target medical specialty, the system uses a preset hybrid retrieval mechanism to search a preset knowledge vector database corresponding to the target medical specialty based on the entity query information determined by parsing the query statement, and generates the retrieved literature corresponding to the query statement. In response to receiving a first instruction message representing the user's analysis and summary of target documents determined from the retrieved documents and selected instruction documents, a research comparison matrix corresponding to the target documents is generated based on the element information determined by in-depth interpretation of the target documents, and research trend summary information corresponding to the target documents is generated based on the research comparison matrix. In response to receiving a second instruction representing the user's research ideas for the target medical specialty, at least one research gap area is mined from the knowledge vector database corresponding to the target medical specialty, and based on the reasoning chain corresponding to each research gap area, a preset feasibility verification assistance mechanism is used to generate research idea assistance information corresponding to the target medical specialty.

2. The method according to claim 1, characterized in that, The knowledge vector database is pre-built using the following steps: Document data is acquired in batches from a preset database and knowledge base using full-text acquisition and abstract acquisition methods. The document data is then cleaned and standardized to obtain the first document data to be processed. The first document data to be processed is converted into a preset document format, and the first document data to be processed converted into the preset document format is structured using a rule engine to obtain the second document data to be processed. The second document data to be processed is segmented according to preset chapters and semantic paragraphs to obtain the third document data to be processed; The third document data to be processed is domain-specific encoded to generate an enhancement vector, and the third document data to be processed is semantically sliced ​​encoded to generate semantic paragraph encoded data, so as to determine the enhancement vector and the semantic paragraph encoded data as the fourth document data to be processed. The fourth document data to be processed is optimized by sparse indexing, and the fourth document data to be processed is scored and marked using a preset scoring rule to obtain target document data, so as to construct the knowledge vector database composed of the target document data.

3. The method according to claim 1, characterized in that, The entity query information determined by parsing the query statement information is used to retrieve the corresponding documents in the preset knowledge vector database of the target medical specialty field using a preset hybrid retrieval mechanism, including: The query statement information is parsed to determine the entity information and query intent information corresponding to the query statement information; Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data; A pre-defined language fine-tuning model is used to score the relevance of the documents in the candidate document data, and the score result corresponding to each document in the candidate document data is obtained. The candidate literature data is sorted and labeled with confidence level according to the scoring results to generate the retrieval literature corresponding to the query statement information.

4. The method according to claim 3, characterized in that, Based on the entity information and the query intent information, a preset hybrid retrieval mechanism is used to search the preset knowledge vector database corresponding to the target medical specialty field to determine candidate literature data, including: Based on the entity information and the query intent information, vector retrieval is performed in the preset knowledge vector database corresponding to the target medical specialty field to generate vector retrieval literature data. Simultaneously, based on the entity information and the query intent information, ES keyword retrieval is performed in the knowledge vector database to generate keyword retrieval document data; Duplicate documents are removed from the vector-retrieval document data and the keyword-retrieval document data to determine candidate document data.

5. The method according to claim 1, characterized in that, The step of generating a research comparison matrix corresponding to the target document based on the element information determined through in-depth interpretation of the target document includes: The target literature is analyzed in depth to determine the corresponding element information; wherein, the element information includes at least research design information, endpoint information, sample size, research background information, research purpose information, statistical results, and core content information; Based on the aforementioned element information, the target documents are compared and interpreted, and parameters are compared to generate a research comparison matrix corresponding to the target documents.

6. The method according to claim 1, characterized in that, The process of generating a summary of research trends for the target literature based on the research comparison matrix includes: Based on the research comparison matrix, generate consistency analysis conclusions for the target literature. The consistency analysis conclusions are aggregated and statistically analyzed to generate a summary of research trends corresponding to the target literature.

7. The method according to claim 1, characterized in that, The process involves identifying at least one research gap in the knowledge vector database corresponding to the target medical specialty, and generating research idea support information for the target medical specialty based on the reasoning chain corresponding to each research gap using a pre-set feasibility verification assistance mechanism. This includes: In the knowledge vector database corresponding to the target medical specialty, blank path nodes that have not been sufficiently studied in the target medical specialty are detected to determine at least one research blank area corresponding to the target medical specialty. Based on the research gaps and the pre-defined research hotspots corresponding to the target medical specialty, potential research directions for the target medical specialty are determined. Construct a thought process chain corresponding to each research gap area, and based on the thought process chain and the potential research direction information, generate research idea suggestions for the target medical specialty field; The feasibility of the proposed research ideas is assessed and the trend distribution is analyzed using a pre-set feasibility verification auxiliary mechanism. The assessment and analysis results are obtained, and the proposed research ideas are adjusted based on the assessment and analysis results to generate the proposed research ideas for the target medical specialty. Based on the proposed research ideas, corresponding recommended literature is retrieved from the knowledge vector database, and the proposed research ideas and the recommended literature are identified as auxiliary information for research ideas in the target medical specialty.

8. An auxiliary device for scientific research in a medical specialty field, characterized in that, The auxiliary device includes: The literature retrieval module is used to respond to the user's input query statement information representing a literature retrieval for a target medical specialty field, and based on the entity query information determined by parsing the query statement information, to perform a retrieval in the preset knowledge vector database corresponding to the target medical specialty field using a preset hybrid retrieval mechanism, and generate the retrieved literature corresponding to the query statement information. The document interpretation module is used to respond to a first instruction message received that represents the user's analysis and summary of the target document determined in the searched documents and selected instruction documents, generate a research comparison matrix corresponding to the target document based on the element information determined by in-depth interpretation of the target document, and generate research trend summary information corresponding to the target document based on the research comparison matrix. The research mining module is used to respond to receiving a second instruction information representing the user's research ideas mining in the target medical specialty field, to mine at least one research gap area in the knowledge vector database corresponding to the target medical specialty field, and to generate research idea auxiliary information corresponding to the target medical specialty field based on the thought reasoning chain corresponding to each research gap area using a preset feasibility verification auxiliary mechanism.

9. An electronic device, characterized in that, include: The device includes a processor, a memory, and a bus, wherein the memory stores machine-readable instructions executable by the processor, and when the electronic device is running, the processor communicates with the memory via the bus, and the machine-readable instructions are executed by the processor to perform the steps of the auxiliary method for scientific research in the medical specialty field as described in any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, performs the steps of the auxiliary method for scientific research in the medical specialty field as described in any one of claims 1 to 7.

Images