CN117874178A - Method, device, equipment and medium for determining medical response text data - Google Patents
Method, device, equipment and medium for determining medical response text data Download PDFInfo
- Publication number
- CN117874178A CN117874178A CN202311425588.0A CN202311425588A CN117874178A CN 117874178 A CN117874178 A CN 117874178A CN 202311425588 A CN202311425588 A CN 202311425588A CN 117874178 A CN117874178 A CN 117874178A
- Authority
- CN
- China
- Prior art keywords
- text data
- medical
- question
- language model
- medical response
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 230000004044 response Effects 0.000 title claims abstract description 242
- 238000000034 method Methods 0.000 title claims abstract description 62
- 230000015654 memory Effects 0.000 claims description 25
- 238000004590 computer program Methods 0.000 claims description 12
- 238000011156 evaluation Methods 0.000 claims description 7
- 230000008569 process Effects 0.000 abstract description 11
- 238000012216 screening Methods 0.000 abstract description 9
- 238000003058 natural language processing Methods 0.000 abstract description 5
- 201000010099 disease Diseases 0.000 description 27
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 description 27
- 238000003745 diagnosis Methods 0.000 description 17
- 230000006870 function Effects 0.000 description 12
- 239000000463 material Substances 0.000 description 8
- 238000012545 processing Methods 0.000 description 8
- 230000036541 health Effects 0.000 description 6
- 238000010801 machine learning Methods 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000010586 diagram Methods 0.000 description 4
- 230000003993 interaction Effects 0.000 description 4
- 230000008878 coupling Effects 0.000 description 3
- 238000010168 coupling process Methods 0.000 description 3
- 238000005859 coupling reaction Methods 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 239000000203 mixture Substances 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000013473 artificial intelligence Methods 0.000 description 2
- 238000013500 data storage Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 206010012601 diabetes mellitus Diseases 0.000 description 2
- 238000005516 engineering process Methods 0.000 description 2
- NOESYZHRGYRDHS-UHFFFAOYSA-N insulin Chemical compound N1C(=O)C(NC(=O)C(CCC(N)=O)NC(=O)C(CCC(O)=O)NC(=O)C(C(C)C)NC(=O)C(NC(=O)CN)C(C)CC)CSSCC(C(NC(CO)C(=O)NC(CC(C)C)C(=O)NC(CC=2C=CC(O)=CC=2)C(=O)NC(CCC(N)=O)C(=O)NC(CC(C)C)C(=O)NC(CCC(O)=O)C(=O)NC(CC(N)=O)C(=O)NC(CC=2C=CC(O)=CC=2)C(=O)NC(CSSCC(NC(=O)C(C(C)C)NC(=O)C(CC(C)C)NC(=O)C(CC=2C=CC(O)=CC=2)NC(=O)C(CC(C)C)NC(=O)C(C)NC(=O)C(CCC(O)=O)NC(=O)C(C(C)C)NC(=O)C(CC(C)C)NC(=O)C(CC=2NC=NC=2)NC(=O)C(CO)NC(=O)CNC2=O)C(=O)NCC(=O)NC(CCC(O)=O)C(=O)NC(CCCNC(N)=N)C(=O)NCC(=O)NC(CC=3C=CC=CC=3)C(=O)NC(CC=3C=CC=CC=3)C(=O)NC(CC=3C=CC(O)=CC=3)C(=O)NC(C(C)O)C(=O)N3C(CCC3)C(=O)NC(CCCCN)C(=O)NC(C)C(O)=O)C(=O)NC(CC(N)=O)C(O)=O)=O)NC(=O)C(C(C)CC)NC(=O)C(CO)NC(=O)C(C(C)O)NC(=O)C1CSSCC2NC(=O)C(CC(C)C)NC(=O)C(NC(=O)C(CCC(N)=O)NC(=O)C(CC(N)=O)NC(=O)C(NC(=O)C(N)CC=1C=CC=CC=1)C(C)C)CC1=CN=CN1 NOESYZHRGYRDHS-UHFFFAOYSA-N 0.000 description 2
- 238000011160 research Methods 0.000 description 2
- 102000004877 Insulin Human genes 0.000 description 1
- 108090001061 Insulin Proteins 0.000 description 1
- 238000012550 audit Methods 0.000 description 1
- 238000013475 authorization Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 229940079593 drug Drugs 0.000 description 1
- 239000003814 drug Substances 0.000 description 1
- 239000000945 filler Substances 0.000 description 1
- 239000011521 glass Substances 0.000 description 1
- 229940125396 insulin Drugs 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 230000003287 optical effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Landscapes
- Information Retrieval, Db Structures And Fs Structures Therefor (AREA)
Abstract
The embodiment of the specification provides a method for determining medical response text data and a related device. The method comprises the following steps: in the case of receiving a medical response text data acquisition request accompanied by medical question text data, a target medical response text data corresponding to the medical question text data is determined from the plurality of medical response text data by recalling the plurality of medical response text data matching the medical question text data in a specified medical knowledge database and calling a generic large language model based on the plurality of medical response text data, using natural language processing capabilities of the generic large language model. According to the embodiment of the specification, the determination process of the medical response text data corresponding to the medical question text data is completed by replacing manual work with the universal large language model, and the manpower resources and time occupied by manually screening a plurality of medical response text data are saved.
Description
Technical Field
The embodiment in the specification relates to the technical field of artificial intelligence, in particular to a method, a device, equipment and a medium for acquiring medical response text data.
Background
Along with the continuous development of artificial intelligence technology, intelligent question-answering systems have been widely used in the service fields of voice assistants, intelligent customer service, online consultation and the like. Currently, intelligent question-answering systems can match answer text data for question text data entered by a user in a large-scale question-answering database based on semantic relevance analysis.
In the related art, in the case where the question text data matches a plurality of answer text data, in order to improve the accuracy of the answer text data and the association between the answer text data and the question text data, it is necessary to manually screen the plurality of answer text data to determine the answer text data having the highest matching degree with the question text data.
After the universal large language model is put into use, the capability of the intelligent question-answering system is rapidly improved. The existing general large language model can quickly generate response text data aiming at question text data. However, the general large language model is not strict enough for the generated response text data aiming at the problems in some specific fields. However, for some technical fields, the accuracy requirement for the content expressed by the response text data is high, especially for the medical field, and if the content expressed by the response text data is not accurate enough, some health risks may be brought about.
Disclosure of Invention
In view of this, various embodiments of the present disclosure are directed to providing a method, apparatus, device, and medium for acquiring medical response text data, so as to provide response text data for medical problems more quickly and accurately.
One embodiment of the present disclosure provides a method for determining text data of a medical response, which is applied to a server, and the method includes: recall, in a specified medical knowledge database, a plurality of medical response text data matching the medical question text data upon receiving a medical response text data acquisition request accompanied by the medical question text data; wherein the specified medical knowledge database comprises a plurality of medical text data pairs; the medical text data pair comprises corresponding medical question text data and medical response text data; and calling a universal large language model based on the plurality of medical response text data, and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
One embodiment of the present disclosure provides a method for determining response text data, applied to a server, where the method includes: recall, in a specified knowledge database, a plurality of response text data matching the question text data upon receiving a response text data acquisition request accompanied by the question text data; wherein the specified knowledge database comprises a plurality of text data pairs; the text data pair comprises corresponding question text data and response text data; and calling a universal large language model based on the plurality of response text data, and determining target response text data corresponding to the question text data in the plurality of response text data through the universal large language model.
One embodiment of the present specification provides a medical response text data determining apparatus, applied to a server, the apparatus including: a recall module for recalling a plurality of medical response text data matched with the medical question text data in a designated medical knowledge database under the condition that a medical response text data acquisition request accompanied by the medical question text data is received; wherein the specified medical knowledge database comprises a plurality of medical text data pairs; the medical text data pair comprises corresponding medical question text data and medical response text data; and the calling module is used for calling a universal large language model based on the plurality of medical response text data and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
The present description embodiment proposes a computer device comprising a memory storing a computer program and a processor implementing the method according to the above embodiment when the processor executes the computer program.
The present description provides a computer-readable storage medium having stored thereon computer program instructions which, when executed by a processor, implement the method described in the above embodiments.
In the embodiments provided in the present specification, when a medical response text data acquisition request accompanied by medical question text data is received, a plurality of medical response text data matching the medical question text data are recalled in a designated medical knowledge database, and a generic large language model is called based on the plurality of medical response text data, so that target medical response text data corresponding to the medical question text data is determined from the plurality of medical response text data by using natural language processing capability of the generic large language model. Because the medical response text data is formulated in advance, the medical response text data can be in a content-controllable state, and the most appropriate target medical response text data is selected by recalling a plurality of medical text data pairs and utilizing the text understanding capability of the universal large language model, so that the target medical response text data corresponding to the medical question text data can be provided more quickly and accurately.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for the description of the embodiments will be briefly described below, and it is apparent that the drawings in the following description are only some embodiments of the present invention, and other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art.
Fig. 1 is a multi-terminal interaction schematic diagram of a system for determining medical response text data according to an embodiment of the present disclosure.
Fig. 2 is a schematic architecture diagram of a system for determining medical response text data according to an embodiment of the present disclosure.
Fig. 3 is a flow chart of a method for determining medical response text data according to an embodiment of the present disclosure.
Fig. 4 is a schematic diagram of a device for determining medical response text data according to an embodiment of the present disclosure.
Fig. 5 is a schematic diagram of a computer device according to an embodiment of the present disclosure.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is evident that the embodiments described are only some, but not all, of the embodiments of the present application. All other embodiments, which can be made by those skilled in the art based on the embodiments herein without making any inventive effort, are intended to be within the scope of the present application.
In the description of the embodiments of the present application, it should be understood that the terms "first," "second," and the like are used for descriptive purposes only and are not to be construed as indicating or implying a relative importance or an implicit indication of the number of features indicated. Thus, a feature defining "a first" or "a second" may explicitly or implicitly include one or more of the described features. In the description of the embodiments of the present application, the meaning of "a plurality" is two or more, unless explicitly defined otherwise.
The large-scale question-answer database may store text data pairs composed of corresponding question text data and answer text data. After receiving the question text data input by the user, the intelligent question-answering system can perform matching operation on the question text data and question text data in the large-scale question-answering database based on a text similarity algorithm so as to obtain response text data corresponding to the question text data according to the question text data matched with the question text data.
In the case where the question text data matches a plurality of question text data, the question text data may have a plurality of corresponding answer text data. Since intelligent question-answering systems typically provide only one answer text data to a user, multiple answer text data needs to be screened.
In the related art, in order to enhance the use experience of the user, the operator of the intelligent question-answering system may measure the matching degree between the plurality of answer text data and the question text data from two dimensions of accuracy and relevance with the question text data, determine answer text data with the highest matching degree with the question text data from the plurality of answer text data, and provide the answer text data corresponding to the question text data to the user.
However, in the case where the intelligent question-answering system needs to serve more users, a large number of operators need to manually screen a plurality of answer text data, and human resources are occupied more. Meanwhile, because knowledge levels of knowledge fields to which different operators belong to the questioning text data are uneven, when the operators are not familiar enough with the knowledge fields to which the questioning text data belong, manual screening can take a long time, and further use experience of users is affected.
Along with popularization and application of the universal large language model, in order to improve efficiency of the intelligent question-answering system, after text data pairs matched with question text data are obtained from a large-scale question-answering database, the universal large language model can be called according to the question text data and the text data, so that answer text data corresponding to the question text data can be output through semantic understanding capability of the universal large language model. Universal large language models typically learn an excessive amount of world knowledge and are trained on an excessive amount of samples. So that the general large language model covers the relatively comprehensive knowledge field to a certain extent. The popularization and application of the universal large language model greatly improve the working efficiency of personnel in each industry and reduce the time length of knowledge acquisition.
However, in practice it has been found that the generic large language model is somewhat inaccurate for some specific professional field answers. Specifically, in the training process of the general large language model, the used training samples may not be all very accurate, and certain semantic deviation exists, so that the trained general large language model can compile correct seemingly wrong answers aiming at some problems beyond the capability category. This phenomenon may be referred to as "AI illusion problem". In order to reduce the "AI illusion problem", technical means that may be employed include: (1) training is performed with more correct sample data to allow for better capabilities of the generic large language model. (2) The prompt instruction (prompt) input to the generic large language model is adjusted. (3) The additional needed knowledge corpus of the questions is acquired and is input to the universal large language model together with the questions text, and the universal large language model is required to answer the questions based on the knowledge corpus.
However, the adoption of the technical means (1) requires a lot of cost and time, consumes a lot of resources, and is difficult for a general developer to bear. The technical means (2) has a certain effect, but the AI illusion problem exists in the whole, even if the ' when you don't know the answer, he does not need to edit the answer, and the answer is directly answered without knowledge ' is added in the prompt instruction. In the answer obtained, there will normally be some content compiled. By adopting the technical means (3), some knowledge corpora are additionally infused into the general large language model, but some compiled contents of the general large language model in the generated answers cannot be completely eradicated.
For some technical fields, there may be specificity in the field, and the accuracy requirement for knowledge representation is high. At this time, since the generic large language model has the "AI illusion problem", the answer generated directly using the generic large language model may have some serious consequences. For example, in the medical field, the questioning text data may be a question for the medical field and may relate to the health-related content of the patient, at which time, the medical knowledge expressed by the response text data fed back to the patient may have a great influence on the physical health of the patient.
Therefore, it is necessary to provide a method for determining medical response text data, in the case of receiving a request for obtaining medical response text data with medical question text data, recall a plurality of medical response text data matched with the medical question text data in a designated medical knowledge database, and call a general big language model based on the plurality of medical response text data, so as to determine target medical response text data corresponding to the medical question text data from the plurality of medical response text data by using the natural language processing capability of the general big language model, realize that the general big language model replaces the manual completion of the determination process of the medical response text data corresponding to the medical question text data, save the manpower resources and time occupied by manual screening of the plurality of medical response text data, and further, the medical response text data is not directly generated by the general big language model, so that the content of the medical response text data can be determined after being audited by professional staff, select the most suitable medical response text data from the plurality of medical response text data as final target medical response text data by using the semantic understanding capability of the general big language model, realize that the general response text data has high accuracy, overcome the problem of the general language model is realized, and the high accuracy of the medical response text data is realized.
An application scenario example of an intelligent question-answering system is provided in the present specification. The intelligent question-answering system may include a client, a server, and a generic large language model. Wherein the client may be used to receive medical question text data entered by a user operation, the server may be used to perform a method of determining medical answer text data, and the generic large language model may be used to determine target medical answer text data corresponding to the medical question text data from a plurality of medical answer text data. Taking an on-line inquiry as an example, a method for determining medical response text data provided in the embodiment of the present specification will be described.
Please refer to fig. 1. The intelligent question-answering system may be configured as an online question-asking platform. The client of the intelligent question-answering system can be a question client of an online question-asking platform. After the user enters the inquiry client provided by the online inquiry platform, medical inquiry text data, such as inquiry data, can be input in an inquiry interface on the inquiry client. After receiving the inquiry data input by the user, the inquiry client can add a disease identifier to the inquiry data to form a medical response text data acquisition request and send the medical response text data acquisition request to the server.
When the server receives the medical response text data acquisition request, a designated medical knowledge database corresponding to the disease identification representing the disease may be first determined in a plurality of medical knowledge databases, then a plurality of medical response text data matching the inquiry data are recalled in the designated medical knowledge database, and then the universal large language model is called based on the plurality of medical response text data. In particular, each knowledge database may have a disease identifier and may include a plurality of medical question text data pairs consisting of corresponding medical question text data and medical response text data, wherein the medical question text data and the corresponding medical response text data may identify medical question text data and corresponding medical response text data for a corresponding disease for the disease. For example, the medical question text data may be "what drugs should be used to treat diabetes", and the corresponding medical response text data may be "insulin should be used to treat diabetes". The server may first determine a medical knowledge database having the same disease identification as the disease identification of the inquiry data as the specified medical knowledge database. After determining the appointed medical knowledge database, the server can access the appointed medical knowledge database, perform text similarity matching on the inquiry data and the medical question text data in the appointed medical knowledge database, obtain a plurality of medical question text data with the text similarity meeting the appointed condition, and use the medical response text data corresponding to the plurality of medical question text data as a plurality of diagnosis and treatment data matched with the inquiry data. After obtaining the plurality of diagnosis and treatment data, the server can generate a first response prompt instruction according to the plurality of diagnosis and treatment data and the first prompt instruction template, call an access interface of the universal large language model and input the first response prompt instruction into the universal large language model.
After receiving the first response prompt instruction, the universal large language model can combine the pre-learned corpus and knowledge base to understand, analyze and infer a plurality of diagnosis and treat data, determine target diagnosis and treat data corresponding to the inquiry data from the plurality of diagnosis and treat data and send the target diagnosis and treat data to the server.
After receiving the target diagnosis and treatment data, the server can send the target diagnosis and treatment data to the inquiry client, so that a patient can obtain the target diagnosis and treatment data through the inquiry client.
Under the condition that a diagnosis and treatment data acquisition request with inquiry data transmitted by an inquiry client is received each time, the server recalls a plurality of diagnosis and treatment data matched with the inquiry data from a specified knowledge data set, calls a universal large language model based on the plurality of diagnosis and treatment data, determines target diagnosis and treatment data corresponding to the inquiry data from the plurality of diagnosis and treatment data by the universal large language model, feeds the target diagnosis and treatment data back to the server, and pushes the target diagnosis and treatment data to the inquiry client by the server.
Please refer to fig. 2. The embodiment of the specification provides an intelligent question-answering system. The intelligent question-answering system can include a client, a server, and a generic large language model. The number of clients and servers is not particularly limited in the present embodiment. The client and the server may be connected to each other via a wired or wireless network.
The client may be an electronic device having a network access function, a data storage function, and a display function. Specifically, for example, the client may be a desktop computer, a tablet computer, a notebook computer, a smart phone, a digital assistant, a smart wearable device, a smart television, or the like. Wherein, intelligent wearable device includes but is not limited to intelligent bracelet, intelligent wrist-watch, intelligent glasses, intelligent helmet etc.. Alternatively, the client may be a computer program with a man-machine interaction interface capable of running in the electronic device.
The server may be an electronic device with some arithmetic processing power and data transmission capability. In particular, the server may have a network communication module, a processor, a memory, and the like. A server may also refer to a computer program running in the electronic device. The server may also be a distributed server, or may be a system having multiple processors, memories, network communications modules, etc. operating in concert. Alternatively, the server may be a server cluster formed for several servers. Or, with the development of science and technology, the server may also be a new technical means capable of realizing the corresponding functions of the embodiment of the specification. For example, a new form of "server" based on quantum computing implementation may be possible.
The generic large language model may be a machine learning model with deep learning capabilities and natural language processing capabilities. In particular, the generic large language model may have semantic understanding, analysis, reasoning, natural language generation capabilities.
Please refer to fig. 3. One embodiment of the present specification provides a method of determining medical response text data. The method of determining medical response text data may be applied to a server, and the method may include the following steps.
Step S110: in the case of receiving a medical response text data acquisition request accompanied by medical question text data, a plurality of medical response text data matching the medical question text data are recalled in a specified medical knowledge database.
In a specific application scenario, knowledge accuracy of medical response text data corresponding to medical question text data may have a large influence on personal health of a user. In order to control the medical response text data to improve the knowledge accuracy of the medical response text data, a medical expert in the medical knowledge field may pre-audit the medical text data pairs in the designated medical knowledge database and store the medical text data pairs in the server. And recalling a plurality of medical response text data matched with the medical question text data from the audited appointed medical knowledge database by the server, thereby realizing the controllability of the medical response text data.
In this embodiment, the medical question text data may include text data input by a user received by the medical response text data acquisition request sender. Specifically, for example, the sender may respond to the data acquisition request by being a client having a man-machine interaction interface, where the man-machine interaction interface may be provided with an input box, or may be preset with a general question text and a selection button. The medical question text data may include data that a user directly performs text input operation input through an input box, may include data that a user performs voice-to-text operation input through an input box, and may include data that a user performs selection operation selection through a selection button.
In this embodiment, the specified medical knowledge database may include a plurality of medical text data pairs, each of which may include corresponding medical question text data and medical answer text data. In particular, for example, the data sources of the medical text data pairs in the knowledge database may be different for different application scenarios. For on-line inquiry intelligent customer service products, because knowledge accuracy of the medical response text data can have a great influence on life health of users, accurate knowledge is required to be provided by controlling the medical response text data. Therefore, the data sources of the medical text data pairs in the knowledge database can be preset question-answer data and historical question-answer data which are subjected to the examination of medical experts in the medical knowledge field.
In this embodiment, when a medical response text data acquisition request accompanied by medical question text data is received, a plurality of medical response text data matching the medical question text data is recalled in a specified medical knowledge database, which may be implemented based on keyword matching. Specifically, for example, keywords of the medical question text data may be extracted, and the extracted keywords are matched with the medical text data pairs in the specified medical knowledge database to obtain a plurality of medical response text data satisfying specified matching conditions with the medical question text data.
In some embodiments, in the event that a medical answer text data acquisition request accompanied by medical question text data is received, recalling a plurality of medical answer text data matching the medical question text data in a specified medical knowledge database may include: vectorizing the medical question text data to obtain question vector data; performing matching operation on the question vector data and the question vector data in the appointed knowledge vector database to obtain a plurality of question vector data meeting appointed matching conditions with the question vector data; and acquiring medical response text data from the medical text data pairs to which the medical question text data corresponding to the plurality of question vector data belong respectively.
To improve knowledge accuracy of the medical response text data, the data size of the medical text data pairs in the specified medical knowledge database may be large. In this case, implementing recall of a plurality of medical answer text data based on keyword matching may have problems of slower operation speed and longer operation time. In addition, the keyword-based matching rule may be limited to word or character matching, ignoring semantic information contained in the medical question text data, resulting in a poor actual association of the recalled plurality of medical answer text data with the medical question text data, or a smaller number of recalled medical answer text data. In order to solve the problems, the operation speed of recalling a plurality of medical response text data is improved, the overall time for providing the medical response text data corresponding to the medical question text data for a user is shortened, meanwhile, the relevance between the recalled plurality of medical response text data and the medical question text data is enhanced, vectorization processing can be carried out on the medical question text data and the medical question text data in a designated medical knowledge database, a plurality of medical question text data matched with the medical question text data are obtained based on a vector similarity algorithm, and a plurality of medical response text data matched with the medical question text data are obtained according to the plurality of medical question text data.
In this embodiment, the specified medical knowledge database may correspond to a specified knowledge vector database. Wherein the specified knowledge vector database may comprise a plurality of question vector data corresponding to the medical question text data. Specifically, for example, for each medical text data pair in the specified knowledge vector database, the medical problem text data may be vectorized to obtain problem vector data corresponding to the medical problem text data, and the problem vector data may be stored in the specified knowledge vector database. Correspondence between the medical question text data and the question vector data may be achieved by data identification or data storage rules. For example, the corresponding medical question text data and question vector data may have the same question data identification. Corresponding medical problem text data and problem vector data may also be stored correspondingly to corresponding addresses of the specified medical knowledge database and the specified knowledge vector database.
In this embodiment, the specified matching condition may be a numerical condition or a ranking condition that needs to be satisfied by the vector similarity between the question vector data and the question vector data. Specifically, in the case where the operation result of the matching operation is the value of the vector similarity, the specified matching condition may be the range of the value of the vector similarity. For example, the specified matching condition may be that the numerical value of the vector similarity falls within the ranges of [ -1, -0.8] and [0.8,1 ]. In the case where the operation result of the matching operation is the ordering of the vector similarities, the specified matching condition may be the rank range of the vector similarities. For example, the specified matching condition may be that the vector similarity is located in the first 5 bits.
In the present embodiment, the medical question text data is vectorized, and the medical question text data is vector-coded according to a vector coding rule, whereby the result of the coding can be used as question vector data. For example, vector coding rules may include One-Hot coding, bag of words model coding, TF-IDF (Term Frequency-Inverse Document Frequency) coding, and so forth.
In some embodiments, the medical question text data is vectorized, and the output result of the model can be used as question vector data by inputting the medical question text data into a text vectorization model. The text vectorization model may be a pre-trained machine learning model for converting text data into vector data. For example, word2Vec model and Doc2Vec model.
In this embodiment, the method for vectorizing the medical question text data may be the same as the method for vectorizing the medical question text data. Specifically, the dimensionality of question vector data and the dimensionality of question vector data obtained by vectorization processing by adopting the same method can be the same, so that the subsequent matching operation of the question vector data and the question vector data is facilitated.
In some embodiments, the method employed to vectorize the medical question text data may be different from the method employed to vectorize the medical question text data. When the dimensions of question vector data and the dimensions of question vector data are different, the dimensions of question vector data and question vector data may be subjected to dimension matching processing so that the dimensions of both vector data are identical.
In this embodiment, matching operation is performed on question vector data and question vector data in a specified knowledge vector database to obtain a plurality of question vector data satisfying a specified matching condition with the question vector data, and vector similarity between each question vector data in the specified knowledge vector database and the question vector data may be calculated based on a vector similarity algorithm to obtain a plurality of question vector data whose vector similarity satisfies a specified numerical condition or a specified sorting condition. Wherein the vector similarity algorithm may comprise at least one of the following algorithms: cosine similarity, euclidean distance, manhattan distance, and Minkowski distance. The embodiment of the present specification is not particularly limited to the vector similarity algorithm.
In this embodiment, the medical response text data is acquired from the pair of medical text data to which the medical question text data corresponding to the plurality of question vector data belongs, and the pair of medical text data may be determined according to the correspondence between the question vector data and the medical question text data, and then the medical response text data may be acquired from the pair of medical text data.
Step S120: and calling a universal large language model based on the plurality of medical response text data, and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
In some cases, after receiving the medical question text data input by the user, to reduce interference caused by the plurality of medical answer text data to the user, the intelligent question and answer system may provide a small amount of medical answer text data corresponding to the medical question text data to the user. Therefore, after recalling a plurality of medical response text data matching the medical question text data, it is necessary to determine target medical response text data having the highest matching degree with the medical question text data from the plurality of medical response text data. Therefore, on the basis of solving the problems that manual screening of a plurality of medical response text data consumes more manpower resources and time in the related art, the content of the medical response text data can be guaranteed to have better accuracy.
Furthermore, by calling the universal large language model and utilizing the semantic understanding capability of the universal large language model, the target medical response text data can be rapidly determined from the plurality of medical response text data. As the answer range serving as the target medical question text data is appointed, the problem of compiling the general large language model in the process of generating the answer is avoided, and the problem of AI illusion of the general large language model in the use process is overcome.
In this embodiment, different general large language models may be called in different ways for medical answer text data of different contents. Specifically, for example, in the case that the content of the medical response text data cannot be fully disclosed, the universal large language model can be deployed locally on the server, and the universal large language model is called through the local interface; in the case where the content of the medical answer text data may be fully disclosed, the generic large language model may be invoked directly through the network interface. In some embodiments, invoking a generic large language model based on the plurality of medical response text data for determining target medical response text data corresponding to the medical question text data from the plurality of medical response text data by the generic large language model may include: generating a first response prompt instruction according to the medical question text data, the plurality of medical response text data and the first prompt instruction template; the first response prompting instruction is input to the universal large language model, so that the universal large language model can be used for determining the target medical response text data in the plurality of medical response text data according to the first response prompting instruction.
In some cases, to control the generic large language model to determine an output of the target medical answer text data process from the plurality of medical answer text data, the understanding ambiguity of the input of the generic large language model is reduced, a first answer alert instruction for directing and controlling the output of the generic large language model may be generated from the medical answer text data, the plurality of medical answer text data, and the first alert instruction template, the first answer alert instruction being used as an input of the generic large language model.
In this embodiment, the first hint instruction template may include a rule hint word for providing explicit indications for the generic large language model and a filler template for filling with medical question text data and a plurality of medical answer text data. Specifically, the rule prompt word may be used to instruct the general large language model to determine target medical response text data corresponding to the medical question text data from among the plurality of medical response text data, and to instruct the format and information output by the general large language model. The filling template may be used to represent a filling format of the medical question text data and the plurality of medical answer text data. For example, a rule prompt may be "which piece of material may answer better for a given question? Please extract the following information from the corresponding material text: id, query, answer, tags. And finally returning the JSON format of the information. The filling template may be "question: {0}; material 1: {1}; material 2: {2}; … …; material n: { n }. In addition, since the setting of the prompt instruction templates has an important influence on the output of the universal large language model, the first prompt instruction template can be a preset prompt instruction template.
In this embodiment, the first response prompting instruction is generated according to the medical question text data, the plurality of medical response text data and the first prompting instruction template, and the first response prompting instruction with complete information can be formed by filling the medical question text data and the plurality of medical response text data into the designated position of the first prompting instruction template. Specifically, for example, in filling templates as "problem: {0}; material 1: {1}; material 2: {2}; … …; material n: in the case of { n } ", the medical question text data may be filled to {0}, and the plurality of medical answer text data may be respectively filled to {1}, {2} … { n }. Because the first prompt instruction template includes rule prompt words, the first response prompt instruction may also include rule prompt words for the universal large language model.
In this embodiment, the first hint instruction template may include a feedback requirement hint word that indicates a data format of the feedback content of the generic large language model. The first response prompting instruction can be input into the universal big language model to be used for indicating the universal big language model to feed back result data belonging to the data format according to the requirement of the feedback requirement prompting word expression. The requirements for the universal large language model can be conveyed through the feedback requirement prompting words, so that the universal large language model outputs result data meeting the requirements according to the feedback requirement prompting words. Specifically, for example, the feedback requirement prompting word may be "JSON format that finally returns the above information", and at this time, the feedback of the universal large language model may feedback the result data in JSON format, so as to more facilitate the subsequent use of the result data to meet the subsequent data processing requirement.
In this embodiment, the first answer prompting instruction may be input into the general large language model, so that the general large language model may be used to screen the plurality of medical answer text data according to the rule prompting word, and a screening result of the general large language model may be used as target medical answer text data.
In some cases, the plurality of medical response text data can have non-editable properties, the rule prompt word can clearly indicate that the task to be completed of the universal large language model is screening of the medical response text data, and then the universal large language model can directly execute the screening process of the plurality of medical response text data according to the rule prompt word and output screening results of the original text of the plurality of medical response text data according to the format and information requirements of the rule prompt word, so that the output control of the universal large language model is realized, and the output accuracy of the universal large language model is improved.
In this embodiment, the content that the rule hint word can express includes a specified evaluation rule. In particular, specifying the evaluation rules may include evaluating the accuracy and/or relevance of the respective medical response text data relative to the medical question text data. For example, the specified evaluation rule may be that the accuracy degree and the association degree of each piece of medical response text data with respect to the medical question text data are scored according to a preset scoring rule, then the score of each piece of medical response text data is calculated according to the preset scoring rule, the score of each piece of medical response text data is ranked, and the medical response text data with the highest score is used as the target medical response text data.
In this embodiment, the first response prompting instruction may be input into the generic large language model for the generic large language model to determine the target medical response text data from the plurality of medical response text data according to the specified evaluation rule.
In the embodiment, when a medical response text data acquisition request with medical question text data is received, a plurality of medical response text data matched with the medical question text data are recalled in a designated medical knowledge database, and then a universal large language model is called based on the plurality of medical response text data, so that target medical response text data corresponding to the medical question text data is determined from the plurality of medical response text data by utilizing the natural language processing capability of the universal large language model, the process of manually completing the determination of the medical response text data corresponding to the medical question text data by using the universal large language model instead of manual work is realized, and the manpower resources and time occupied by manually screening the plurality of medical response text data are saved. In some embodiments, the number of medical knowledge databases may be multiple, each medical knowledge database may correspond to a disease; a medical knowledge database of the plurality of medical knowledge databases that relates to the same disease as the medical question text data may be used as the specified medical knowledge database.
In some cases, to increase the scope of medical text data in a medical knowledge database over the covered medical knowledge field, the number of medical knowledge databases may be large, and determining a given medical knowledge database from a plurality of medical knowledge databases may take a long time, thereby affecting the overall time to determine the target medical response text data. In order to shorten the time for determining the appointed medical knowledge database and improve the operation speed of determining the target medical response text data, the appointed medical knowledge database can be rapidly determined from a plurality of knowledge databases according to diseases corresponding to the knowledge databases and diseases related to the medical question text data.
In this embodiment, the correspondence between the medical knowledge database and the disease and the correspondence between the medical question text data and the disease may be achieved by the disease type identification. In particular, for example, each medical knowledge database may have a first disease type identification. The medical question text data may have a second disease category identification. The first disease type identifier and the second disease type identifier may have a specified mapping relationship therebetween. The first disease type identifier and the second disease type identifier may be the same for the same disease. At this time, when receiving the medical question text data accompanied by the second disease type identifier, the second disease type identifier may be compared with the first disease type identifiers of the respective knowledge databases, and the knowledge database having the first disease type identifier identical to the second disease type identifier may be used as the specified medical knowledge database.
In some embodiments, in a case that the medical response text data in the specified medical knowledge database is not suitable as the target medical response text data, invoking a specified online knowledge database to obtain knowledge text data corresponding to the medical question text data fed back by the specified online knowledge database; generating a second response prompt instruction according to the medical question text data, the knowledge text data and a second prompt instruction template; and inputting the second response prompting instruction into the universal large language model so as to be used for generating target medical response text data of the medical question text data according to the knowledge text data by the universal large language model according to the instruction of the second response prompting instruction.
In some cases, it may not be possible to recall target medical response text data suitable as response medical question text data from the pair of medical text data included in the specified medical knowledge database. Specifically, for example, no medical question text data matching the medical question text data exists in the specified medical knowledge database. Alternatively, after the plurality of medical response text data is input into the common large language model, the common large language model recognizes that none of the plurality of medical response text data is suitable as the target medical response text data. At this time, in order to avoid that the target medical response text data is not fed back finally, a mode of calling a specified online knowledge base can be adopted as an alternative to generate the final target medical response text data.
In this embodiment, the knowledge text data corresponding to the medical question text data may be obtained by calling a specified online knowledge base. The knowledge text data may be a definition specifying medical concepts associated with the medical question text data maintained in an online knowledge base. Alternatively, the knowledge text data may be answers to medical question text data specified in an online knowledge base.
In this embodiment, the second prompt instruction template may be used to form the second response prompt instruction after filling in the medical question text data and the knowledge text data. The second answer prompt instructions may be for instructing the generic large language model to generate target medical answer text data for the medical question text data based on the knowledge text data. Specifically, for example, the second hint instruction template may include: "known information { A }, judge whether the known information has description misplace according to the above-mentioned known information, if the description is incorrect please point out the concrete error point. If the description is correct, please briefly and professionally answer the user's question. If an answer cannot be obtained from the answer, please say "the question cannot be answered according to the known information", the addition of the composition to the answer is not allowed, and the answer is made using Chinese. The problems are: { Q } ", knowledge text data replace { A }, and medical question text data replace { Q }, in the second prompt instruction template, so as to obtain a second response prompt instruction.
In some cases, the generic large language model may include some content beyond the semantics of the knowledge text data in the target medical response text data generated based on the knowledge text data, so that there may be some inaccuracy in the result data output by the generic large language model. As the medical field has a high accuracy requirement for knowledge representation, otherwise it may negatively affect the physical health of the user. In order to avoid adding self-built content to the output result data by the general large language model, a content range prompt word can be included in the second prompt instruction template. The content scope prompt word can be used for indicating the content of the target medical response text data expression generated by the general large language model and cannot exceed the semantic content of the knowledge text data expression. In this way, the second answer prompting word including the content range prompting word can be used for indicating the general large language model to generate target medical answer text data corresponding to the medical question text data in the category that the knowledge text data expresses semantic content according to the requirement of the content range prompting word expression. Specifically, for example, the second hint instruction template may include a content-scope hint word "do not allow addition of a composition to the answer". As such, the second hint instruction template may include: the second hint instruction template may include: "known information { A }, judge whether the known information has description misplace according to the above-mentioned known information, if the description is incorrect please point out the concrete error point. If the description is correct, please briefly and professionally answer the user's question. If an answer cannot be obtained from the answer, please say "the question cannot be answered according to the known information", the addition of the composition to the answer is not allowed, and the answer is made using Chinese. The problems are: { Q }).
In some embodiments, after recalling the plurality of medical response text data matching the medical question text data from the specified medical knowledge database, the plurality of medical response text data may be preprocessed using a machine learning model with a simple question-and-answer function to reject medical response text data of the plurality of medical response text data that matches less to the medical question text data. Specifically, for example, after recall, from a specified medical knowledge database, a plurality of initial medical response text data preliminarily matched with the medical question text data, the plurality of initial medical response text data may be input into a machine learning model having a simple question-and-answer function, and an output result of the machine learning model having the simple question-and-answer function may be taken as a plurality of medical response text data matched with the medical question text data.
One embodiment of the present specification provides a method for determining response text data, applied to a server, the method including: recall, in a specified knowledge database, a plurality of response text data matching the question text data upon receiving a response text data acquisition request accompanied by the question text data; wherein the specified knowledge database comprises a plurality of text data pairs; the text data pair comprises corresponding question text data and response text data; and calling a universal large language model based on the plurality of response text data, and determining target response text data corresponding to the question text data in the plurality of response text data through the universal large language model.
In this embodiment, the method for determining the response text data may be applied not only to the medical field but also to the financial field, or to other scientific research fields, etc., where the accuracy of knowledge is required to be relatively high. Accordingly, the specified knowledge database may not be limited to the medical field, but may be a financial field, or other scientific research field. With regard to the functions and effects of the respective implementations, reference may be made to the explanations given above with reference to the description.
Referring to fig. 4, an embodiment of the present disclosure provides a device for determining text data of a medical response. The determination means of the medical response text data is applied to a server. The medical response text data determining means may include: the recall module and the call module.
A recall module for recalling a plurality of medical response text data matched with the medical question text data in a designated medical knowledge database under the condition that a medical response text data acquisition request accompanied by the medical question text data is received; wherein the specified medical knowledge database comprises a plurality of medical text data pairs; the medical text data pair includes corresponding medical question text data and medical answer text data.
And the calling module is used for calling a universal large language model based on the plurality of medical response text data and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
The specific functions and effects achieved by the determination device of the medical response text data may be explained in reference to other embodiments of the present specification, and are not described herein. The individual modules in the determination means of the medical response text data may be realized in whole or in part by software, hardware and combinations thereof. The modules can be embedded in hardware or independent of a processor in the computer equipment, and can also be stored in a memory in the computer equipment in a software mode, so that the processor can call and execute the operations corresponding to the modules.
Please refer to fig. 5. The embodiment of the present specification also provides a computer device, including a memory and a processor, where the memory stores a computer program, and the processor executes the computer program to implement the method for determining medical response text data in any of the foregoing embodiments.
The present specification embodiment also provides a computer-readable storage medium having stored thereon a computer program which, when executed by a computer, causes the computer to perform the method of determining medical response text data in any of the above embodiments.
The present description also provides a computer program product comprising instructions which, when executed by a computer, cause the computer to perform the method of determining medical response text data in any of the above embodiments.
The data (including but not limited to data for analysis, stored data, presented data, etc.) referred to in various embodiments of the present disclosure are information and data that is authorized by the user or sufficiently authorized by the parties, and the collection, use, and processing of the relevant data entails compliance with relevant legal regulations and standards of the relevant country and region, and is provided with corresponding access to the user for selection of authorization or denial.
It will be appreciated that the specific examples herein are intended only to assist those skilled in the art in better understanding the embodiments of the present disclosure and are not intended to limit the scope of the present invention.
It should be understood that, in various embodiments of the present disclosure, the sequence number of each process does not mean that the execution sequence of each process should be determined by the function and the internal logic, and should not constitute any limitation on the implementation process of the embodiments of the present disclosure.
It will be appreciated that the various embodiments described in this specification may be implemented either alone or in combination, and are not limited in this regard.
Unless defined otherwise, all technical and scientific terms used in the embodiments of this specification have the same meaning as commonly understood by one of ordinary skill in the art to which this specification belongs. The terminology used in the description is for the purpose of describing particular embodiments only and is not intended to limit the scope of the description. The term "and/or" as used in this specification includes any and all combinations of one or more of the associated listed items. As used in this specification and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.
It will be appreciated that the processor of the embodiments of the present description may be an integrated circuit chip having signal processing capabilities. In implementation, the steps of the above method embodiments may be implemented by integrated logic circuits of hardware in a processor or instructions in software form. The processor may be a general purpose processor, a digital signal processor (Digital Signal Processor, DSP), an application specific integrated circuit (Application Specific Integrated Circuit, ASIC), an off-the-shelf programmable gate array (Field Programmable Gate Array, FPGA) or other programmable logic device, discrete gate or transistor logic device, discrete hardware components. The methods, steps and logic blocks disclosed in the embodiments of the present specification may be implemented or performed. A general purpose processor may be a microprocessor or the processor may be any conventional processor or the like. The steps of a method disclosed in connection with the embodiments of the present specification may be embodied directly in hardware, in a decoded processor, or in a combination of hardware and software modules in a decoded processor. The software modules may be located in a random access memory, flash memory, read only memory, programmable read only memory, or electrically erasable programmable memory, registers, etc. as well known in the art. The storage medium is located in a memory, and the processor reads the information in the memory and, in combination with its hardware, performs the steps of the above method.
It will be appreciated that the memory in the embodiments of this specification may be either volatile memory or nonvolatile memory, or may include both volatile and nonvolatile memory. The nonvolatile memory may be a read-only memory (ROM), a Programmable ROM (PROM), an Erasable Programmable ROM (EPROM), an Electrically Erasable Programmable ROM (EEPROM), or a flash memory, among others. The volatile memory may be Random Access Memory (RAM). It should be noted that the memory of the systems and methods described herein is intended to comprise, without being limited to, these and any other suitable types of memory.
Those of ordinary skill in the art will appreciate that the elements and algorithm steps described in connection with the embodiments disclosed herein can be implemented as electronic hardware, or combinations of computer software and electronic hardware. Whether such functionality is implemented as hardware or software depends upon the particular application and design constraints imposed on the solution. Skilled artisans may implement the described functionality in varying ways for each particular application, but such implementation decisions should not be interpreted as causing a departure from the scope of the present specification.
It will be clear to those skilled in the art that, for convenience and brevity of description, specific working procedures of the above-described system, apparatus and unit may refer to corresponding procedures in the foregoing method embodiments, and are not repeated herein.
In the several embodiments provided in this specification, it should be understood that the disclosed systems, apparatuses, and methods may be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative, e.g., the division of the units is merely a logical function division, and there may be additional divisions when actually implemented, e.g., multiple units or components may be combined or integrated into another system, or some features may be omitted or not performed. Alternatively, the coupling or direct coupling or communication connection shown or discussed with each other may be an indirect coupling or communication connection via some interfaces, devices or units, which may be in electrical, mechanical or other form.
The units described as separate units may or may not be physically separate, and units shown as units may or may not be physical units, may be located in one place, or may be distributed on a plurality of network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the embodiment.
In addition, each functional unit in each embodiment of the present specification may be integrated into one processing unit, each unit may exist alone physically, or two or more units may be integrated into one unit.
The functions, if implemented in the form of software functional units and sold or used as a stand-alone product, may be stored in a computer-readable storage medium. Based on this understanding, the technical solutions of the present specification may be essentially or portions contributing to the prior art or portions of the technical solutions may be embodied in the form of a software product stored in a storage medium, including several instructions to cause a computer device (which may be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in the embodiments of the present specification. And the aforementioned storage medium includes: a usb disk, a removable hard disk, a read-only memory (ROM), a random-access memory (RAM), a magnetic disk, or an optical disk, etc.
The foregoing is merely specific embodiments of the present disclosure, but the scope of the present disclosure is not limited thereto, and any person skilled in the art can easily think about changes or substitutions within the technical scope disclosed in the present disclosure, and should be covered by the scope of the present disclosure. Therefore, the protection scope of the present invention shall be subject to the protection scope of the claims.
Claims (10)
1. A method of determining medical response text data, applied to a server, the method comprising:
recall, in a specified medical knowledge database, a plurality of medical response text data matching the medical question text data upon receiving a medical response text data acquisition request accompanied by the medical question text data; wherein the specified medical knowledge database comprises a plurality of medical text data pairs; the medical text data pair comprises corresponding medical question text data and medical response text data;
and calling a universal large language model based on the plurality of medical response text data, and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
2. The method of claim 1, wherein the specified medical knowledge database corresponds to a specified knowledge vector database; wherein the specified knowledge vector database includes a plurality of question vector data corresponding to the medical question text data; in the case of receiving a medical response text data acquisition request accompanied by medical question text data, recalling a plurality of medical response text data matching the medical question text data in a specified medical knowledge database, comprising:
Vectorizing the medical question text data to obtain question vector data;
performing matching operation on the question vector data and the question vector data in the appointed knowledge vector database to obtain a plurality of question vector data meeting appointed matching conditions with the question vector data;
and acquiring medical response text data from the medical text data pairs to which the medical question text data corresponding to the plurality of question vector data belong respectively.
3. The method of claim 1, wherein the step of invoking a generic large language model based on the plurality of medical response text data for determining target medical response text data corresponding to the medical question text data from the plurality of medical response text data by the generic large language model comprises:
generating a first response prompt instruction according to the medical question text data, the plurality of medical response text data and the first prompt instruction template;
the first response prompting instruction is input to the universal large language model, so that the universal large language model can be used for determining the target medical response text data in the plurality of medical response text data according to the first response prompting instruction.
4. A method according to claim 3, wherein the first hint instruction template includes a rule hint word for expressing a specified evaluation rule; the step of inputting the first answer prompting instruction to the universal large language model for the universal large language model to determine the target medical answer text data from the plurality of medical answer text data according to the first answer prompting instruction includes:
inputting the first response prompting instruction into the universal large language model for the universal large language model to determine the target medical response text data from the plurality of medical response text data according to the specified evaluation rule; wherein the specified evaluation rule comprises evaluating the accuracy and/or relevance of the medical response text data relative to the medical question text data.
5. The method of claim 3, wherein the first hinting instructions template comprises feedback requirement hinting words for indicating a data format of the generic large language model feedback content; the step of inputting the first answer prompting instruction to the universal large language model for the universal large language model to determine the target medical answer text data from the plurality of medical answer text data according to the first answer prompting instruction includes:
And inputting the first response prompting instruction into the universal big language model to be used for indicating the universal big language model to feed back the result data belonging to the data format according to the requirement of the feedback requirement prompting word expression.
6. The method according to claim 1, wherein the method further comprises:
when the medical response text data in the appointed medical knowledge database is not suitable for serving as the target medical response text data, an appointed online knowledge database is called, and knowledge text data corresponding to the medical question text data and fed back by the appointed online knowledge database is obtained;
generating a second response prompt instruction according to the medical question text data, the knowledge text data and a second prompt instruction template;
and inputting the second response prompting instruction into the universal large language model so as to be used for generating target medical response text data of the medical question text data according to the knowledge text data by the universal large language model according to the instruction of the second response prompting instruction.
7. The method of claim 6, wherein the second hint instruction template includes content that indicates a target medical response text data representation generated by the generic large language model that cannot exceed content-scope hint words of semantic content of the knowledge text data representation; the step of inputting the second answer prompting instruction to the universal big language model for the universal big language model to generate target medical answer text data of the medical question text data according to the knowledge text data according to the instruction of the second answer prompting instruction, comprises the following steps:
Inputting the second response prompting instruction into the universal large language model to be used for indicating the universal large language model to generate target medical response text data corresponding to the medical question text data in the range of the semantic content expressed by the knowledge text data according to the requirement of the content range prompting word expression;
and maintaining the original text of the medical response text data serving as the target medical response text data in the feedback result data.
8. A device for determining text data of a medical response, applied to a server, the device comprising:
a recall module for recalling a plurality of medical response text data matched with the medical question text data in a designated medical knowledge database under the condition that a medical response text data acquisition request accompanied by the medical question text data is received; wherein the specified medical knowledge database comprises a plurality of medical text data pairs; the medical text data pair comprises corresponding medical question text data and medical response text data;
and the calling module is used for calling a universal large language model based on the plurality of medical response text data and determining target medical response text data corresponding to the medical question text data in the plurality of medical response text data through the universal large language model.
9. A computer device comprising a memory storing a computer program and a processor implementing the method of any one of claims 1 to 7 when the computer program is executed by the processor.
10. A computer readable storage medium having stored thereon a computer program which, when executed by a processor, implements the method of any of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311425588.0A CN117874178A (en) | 2023-10-30 | 2023-10-30 | Method, device, equipment and medium for determining medical response text data |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311425588.0A CN117874178A (en) | 2023-10-30 | 2023-10-30 | Method, device, equipment and medium for determining medical response text data |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117874178A true CN117874178A (en) | 2024-04-12 |
Family
ID=90595639
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311425588.0A Pending CN117874178A (en) | 2023-10-30 | 2023-10-30 | Method, device, equipment and medium for determining medical response text data |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117874178A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118093789A (en) * | 2024-04-22 | 2024-05-28 | 阿里健康科技(杭州)有限公司 | Medical text error correction system, medical query prompt text display method and device |
Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113505243A (en) * | 2021-07-29 | 2021-10-15 | 深圳万海思数字医疗有限公司 | Intelligent question-answering method and device based on medical knowledge graph |
| CN116662518A (en) * | 2023-06-16 | 2023-08-29 | 河南云迹智能技术有限公司 | Question answering method, question answering device, electronic equipment and readable storage medium |
-
2023
- 2023-10-30 CN CN202311425588.0A patent/CN117874178A/en active Pending
Patent Citations (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113505243A (en) * | 2021-07-29 | 2021-10-15 | 深圳万海思数字医疗有限公司 | Intelligent question-answering method and device based on medical knowledge graph |
| CN116662518A (en) * | 2023-06-16 | 2023-08-29 | 河南云迹智能技术有限公司 | Question answering method, question answering device, electronic equipment and readable storage medium |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN118093789A (en) * | 2024-04-22 | 2024-05-28 | 阿里健康科技(杭州)有限公司 | Medical text error correction system, medical query prompt text display method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111709233B (en) | Intelligent diagnosis guiding method and system based on multi-attention convolutional neural network | |
| US20200242140A1 (en) | Method, apparatus, device and medium for determining text relevance | |
| US10565533B2 (en) | Systems and methods for similarity and context measures for trademark and service mark analysis and repository searches | |
| US20180225363A1 (en) | System and Methods for Automating Trademark and Service Mark Searches | |
| CN112786194A (en) | Medical image diagnosis guide inspection system, method and equipment based on artificial intelligence | |
| US20200365239A1 (en) | System and method for generating clinical trial protocol design document with selection of patient and investigator | |
| CN106933806A (en) | The determination method and apparatus of medical synonym | |
| CN111310440B (en) | Text error correction method, device and system | |
| CN112015917A (en) | Data processing method and device based on knowledge graph and computer equipment | |
| CN113707300B (en) | Search intention recognition method, device, equipment and medium based on artificial intelligence | |
| CN111144120A (en) | Training sentence acquisition method and device, storage medium and electronic equipment | |
| US20220075955A1 (en) | Neural network training method and apparatus, semantic classification method and apparatus and medium | |
| CN117874178A (en) | Method, device, equipment and medium for determining medical response text data | |
| CN115292457A (en) | Knowledge question answering method and device, computer readable medium and electronic equipment | |
| CN111125295A (en) | Method and system for obtaining food safety question answers based on LSTM | |
| CN117076688A (en) | Knowledge question-answering method and device based on domain knowledge graph and electronic equipment | |
| CN114153994A (en) | Medical insurance information question-answering method and device | |
| CN109299467A (en) | Medicine text recognition method and device, sentence identification model training method and device | |
| CN116975218A (en) | Text processing method, device, computer equipment and storage medium | |
| CN111553140A (en) | Data processing method, data processing apparatus, and computer storage medium | |
| CN111353290B (en) | Method and system for automatically responding to user inquiry | |
| CN112071431A (en) | Clinical path automatic generation method and system based on deep learning and knowledge graph | |
| Du et al. | Cocoqa: Question answering for coding conventions over knowledge graphs | |
| CN111324722B (en) | Method and system for training word weight model | |
| US20230052022A1 (en) | Systems and Methods for Dynamic Charting |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination |