CN113051905A - Medical named entity recognition training model and medical named entity recognition method - Google Patents
Medical named entity recognition training model and medical named entity recognition method Download PDFInfo
- Publication number
- CN113051905A CN113051905A CN201911382710.4A CN201911382710A CN113051905A CN 113051905 A CN113051905 A CN 113051905A CN 201911382710 A CN201911382710 A CN 201911382710A CN 113051905 A CN113051905 A CN 113051905A
- Authority
- CN
- China
- Prior art keywords
- named entity
- entity recognition
- medical named
- text data
- doctor
- 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
- 238000000034 method Methods 0.000 title claims abstract description 94
- 238000012549 training Methods 0.000 title claims abstract description 73
- 239000013598 vector Substances 0.000 claims abstract description 87
- 238000003745 diagnosis Methods 0.000 claims abstract description 46
- 230000011218 segmentation Effects 0.000 claims abstract description 25
- 201000010099 disease Diseases 0.000 claims description 38
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims description 38
- 238000003058 natural language processing Methods 0.000 claims description 16
- 238000004590 computer program Methods 0.000 claims description 15
- 238000002372 labelling Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 8
- 238000010276 construction Methods 0.000 claims description 8
- 238000012163 sequencing technique Methods 0.000 claims description 7
- 238000007635 classification algorithm Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 description 18
- 208000024891 symptom Diseases 0.000 description 16
- 238000007781 pre-processing Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 9
- 238000000605 extraction Methods 0.000 description 8
- 238000012545 processing Methods 0.000 description 7
- 238000013528 artificial neural network Methods 0.000 description 6
- 238000004891 communication Methods 0.000 description 6
- 238000012360 testing method Methods 0.000 description 6
- 238000013135 deep learning Methods 0.000 description 5
- 230000007246 mechanism Effects 0.000 description 5
- 230000002457 bidirectional effect Effects 0.000 description 4
- 238000013461 design Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 4
- 238000001914 filtration Methods 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000004092 self-diagnosis Methods 0.000 description 4
- 230000008961 swelling Effects 0.000 description 4
- 230000009471 action Effects 0.000 description 3
- 238000004364 calculation method Methods 0.000 description 3
- 238000012986 modification Methods 0.000 description 3
- 230000004048 modification Effects 0.000 description 3
- 239000003814 drug Substances 0.000 description 2
- 229940079593 drug Drugs 0.000 description 2
- 230000010365 information processing Effects 0.000 description 2
- 210000003141 lower extremity Anatomy 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 238000006467 substitution reaction Methods 0.000 description 2
- 206010030113 Oedema Diseases 0.000 description 1
- 238000007792 addition Methods 0.000 description 1
- 230000005540 biological transmission Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000012937 correction Methods 0.000 description 1
- 238000000586 desensitisation Methods 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 239000000835 fiber Substances 0.000 description 1
- 208000010729 leg swelling Diseases 0.000 description 1
- 208000014987 limb edema Diseases 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000002093 peripheral effect Effects 0.000 description 1
- 238000011084 recovery Methods 0.000 description 1
- 238000012216 screening Methods 0.000 description 1
- 239000004065 semiconductor Substances 0.000 description 1
- 210000002784 stomach Anatomy 0.000 description 1
- 238000001356 surgical procedure Methods 0.000 description 1
- 238000012546 transfer Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H15/00—ICT specially adapted for medical reports, e.g. generation or transmission thereof
Landscapes
- Health & Medical Sciences (AREA)
- Engineering & Computer Science (AREA)
- Epidemiology (AREA)
- General Health & Medical Sciences (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Public Health (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The embodiment of the invention discloses a medical named entity recognition training model and a medical named entity recognition method. The method for training the medical named entity recognition model comprises the following steps: acquiring historical patient complaint text data and doctor diagnosis text data corresponding to the historical patient complaint text data; performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain each word vector and part-of-speech information and label information corresponding to each word vector; and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model. According to the embodiment of the invention, a more accurate medical named entity recognition model can be trained, and then the medical named entity can be recognized accurately.
Description
Technical Field
The invention belongs to the related fields of natural language processing, deep learning and recommendation, and particularly relates to a method and a device for training a medical named entity recognition model, a medical named entity recognition method and device based on the medical named entity recognition model, electronic equipment and a computer storage medium.
Background
Medical named entity recognition aims at extracting medical entities from medical texts and classifying their categories, such as drugs, surgery, symptoms, diseases and body parts. For example, given the sentence "patient had lower limb edema before May", the goal of medical named entity recognition is to extract "lower limb" and "edema" from this sentence and classify them as body part entities and disease entities, respectively. Medical named entity identification is an important task in intelligent healthcare and is an important prerequisite for many downstream tasks, such as drug relocation, entity linking and clinical decision support systems. Therefore, medical named entity identification has become an increasing concern in recent years.
At present, the traditional medical named entity identification mainly comprises the following methods: (1) the traditional retrieval method is directly used for disease pairing, and the effect is poor. (2) Text entity recognition based on deep learning, and an Attention mechanism is not added; deep learning based classification considers only whole sentences and does not incorporate the overall importance of words. (3) The existing Attention classification method does not add part-of-speech semantic information, is used for classifying the whole text and has poor interpretability. (4) The input part of the neural network directly inputs the whole sentence, does not process partial redundant input in advance and has larger complexity.
Therefore, how to train a more accurate medical named entity recognition model and further accurately recognize a medical named entity is a technical problem that needs to be solved urgently by those skilled in the art.
Disclosure of Invention
The embodiment of the invention provides a method and a device for training a medical named entity recognition model, a medical named entity recognition method and device based on the medical named entity recognition model, electronic equipment and a computer storage medium, which can train a more accurate medical named entity recognition model so as to accurately recognize a medical named entity.
In a first aspect, a method for training a medical named entity recognition model is provided, which includes:
acquiring historical patient complaint text data and doctor diagnosis text data corresponding to the historical patient complaint text data;
performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain each word vector and part-of-speech information and label information corresponding to each word vector;
and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
Optionally, performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain each word vector and part-of-speech information and label information corresponding to each word vector, including:
and performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm to obtain each word vector and part-of-speech information and label information corresponding to each word vector.
Optionally, training a target network structure by using each word vector and the part-of-speech information and the tag information corresponding to each word vector to obtain a medical named entity recognition model, including:
obtaining a BI-GRU network and a CRF network;
constructing a target network structure based on an Attention classification algorithm, a BI-GRU network and a CRF network;
and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
In a second aspect, a medical named entity recognition method based on a medical named entity recognition model is provided, where the medical named entity recognition model is obtained by using the method for training a medical named entity recognition model of the first aspect, and includes:
acquiring patient complaint text data;
and inputting the patient complaint text data into the medical named entity recognition model, and outputting the medical named entity recognition result.
Optionally, after outputting the medical named entity recognition result, the method further includes:
constructing a disease condition characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result;
acquiring a doctor dictionary set;
and determining a doctor sequencing result corresponding to the disease condition characteristic value based on the disease condition characteristic value and the doctor dictionary set.
Optionally, obtaining a set of physician dictionaries, comprising:
acquiring medical text information of any doctor;
determining the diagnosed patient complaint text data of any doctor and the doctor diagnosis text data corresponding to the diagnosed patient complaint text data based on the medical text information;
determining a doctor dictionary of any doctor based on the diagnosed patient complaint text data and the doctor diagnosis text data;
and summarizing the doctor dictionaries of all doctors to obtain a doctor dictionary set.
In a third aspect, an apparatus for training a medical named entity recognition model is provided, including:
the acquisition module is used for acquiring the historical patient complaint text data and the doctor diagnosis text data corresponding to the historical patient complaint text data;
the labeling module is used for performing word segmentation and part-of-speech labeling on the historical patient chief complaint text data and the doctor diagnosis text data to obtain each word vector and part-of-speech information and label information corresponding to each word vector;
and the model training module is used for training a target network structure by utilizing each word vector and the part of speech information and the label information corresponding to each word vector to obtain the medical named entity recognition model.
Optionally, the tagging module is configured to perform word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm, so as to obtain each word vector and part-of-speech information and tag information corresponding to each word vector.
Optionally, the model training module is used for acquiring a BI-GRU network and a CRF network; constructing a target network structure based on an Attention classification algorithm, a BI-GRU network and a CRF network; and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
In a fourth aspect, there is provided a medical named entity recognition apparatus based on a medical named entity recognition model, where the medical named entity recognition model is obtained by using the method for training a medical named entity recognition model of the first aspect, and the method includes:
the data acquisition module is used for acquiring the patient complaint text data;
and the output module is used for inputting the patient complaint text data into the medical named entity recognition model and outputting the medical named entity recognition result.
Optionally, the medical named entity recognition apparatus further includes:
the construction module is used for constructing a disease condition characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result;
the acquisition module is used for acquiring a doctor dictionary set;
and the determining module is used for determining a doctor sequencing result corresponding to the illness state characteristic value based on the illness state characteristic value and the doctor dictionary set.
Optionally, the obtaining module is configured to: acquiring medical text information of any doctor; determining the diagnosed patient complaint text data of any doctor and the doctor diagnosis text data corresponding to the diagnosed patient complaint text data based on the medical text information; determining a doctor dictionary of any doctor based on the diagnosed patient complaint text data and the doctor diagnosis text data; and summarizing the doctor dictionaries of all doctors to obtain a doctor dictionary set.
In a fifth aspect, an electronic device is provided, the device comprising: a processor and a memory storing computer program instructions;
the processor, when executing the computer program instructions, implements the method of training a medical named entity recognition model of the first aspect or any of the optional implementations of the first aspect; or a processor implementing the medical named entity recognition method based on the medical named entity recognition model of the second aspect or any alternative of the second aspect when executing the computer program instructions.
A sixth aspect provides a computer storage medium having computer program instructions stored thereon, the computer program instructions, when executed by a processor, implementing the method of training a medical named entity recognition model of the first aspect or any of the optional implementations of the first aspect; or computer program instructions which, when executed by a processor, implement a method for medical named entity recognition based on a medical named entity recognition model according to the second aspect or any alternative of the second aspect.
The method and the device for training the medical named entity recognition model, the medical named entity recognition method and device based on the medical named entity recognition model, the electronic equipment and the computer storage medium can train a more accurate medical named entity recognition model so as to accurately recognize the medical named entity. The method for training the medical named entity recognition model comprises the steps of obtaining historical patient chief complaint text data and doctor diagnosis text data corresponding to the historical patient chief complaint text data, then carrying out word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors and part-of-speech information and label information corresponding to each word vector, and finally training a target network structure by using the word vectors and the part-of-speech information and label information corresponding to each word vector, so that the more accurate medical named entity recognition model can be trained, and further the medical named entity can be recognized accurately.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required to be used in the embodiments of the present invention will be briefly described below, and for those skilled in the art, other drawings can be obtained according to the drawings without creative efforts.
FIG. 1 is a schematic flow chart diagram illustrating a method for training a medical named entity recognition model according to an embodiment of the present invention;
FIG. 2 is a flow chart of a method for medical named entity recognition based on a medical named entity recognition model according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of system components provided by an embodiment of the present invention;
FIG. 4 is a schematic overall flow chart provided by the embodiment of the present invention;
FIG. 5 is a flowchart illustrating a method for training a medical entity recognition model according to an embodiment of the present invention;
FIG. 6 is a flowchart illustrating a method for constructing a physician dictionary according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of a GRU network according to an embodiment of the present invention;
FIG. 8 is a schematic overall flow chart of a medical entity identification module according to an embodiment of the present invention;
FIG. 9 is a schematic structural diagram of an apparatus for training a medical named entity recognition model according to an embodiment of the present invention;
FIG. 10 is a schematic structural diagram of an apparatus for medical named entity recognition based on a medical named entity recognition model according to an embodiment of the present invention;
fig. 11 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
Detailed Description
Features and exemplary embodiments of various aspects of the present invention will be described in detail below, and in order to make objects, technical solutions and advantages of the present invention more apparent, the present invention will be further described in detail below with reference to the accompanying drawings and specific embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not to be construed as limiting the invention. It will be apparent to one skilled in the art that the present invention may be practiced without some of these specific details. The following description of the embodiments is merely intended to provide a better understanding of the present invention by illustrating examples of the present invention.
It is noted that, herein, relational terms such as first and second, and the like may be used solely to distinguish one entity or action from another entity or action without necessarily requiring or implying any actual such relationship or order between such entities or actions. Also, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but may include other elements not expressly listed or inherent to such process, method, article, or apparatus. Without further limitation, an element defined by the phrase "comprising … …" does not exclude the presence of other identical elements in a process, method, article, or apparatus that comprises the element.
At present, the traditional medical named entity identification mainly comprises the following methods: (1) the traditional retrieval method is directly used for disease pairing, and the effect is poor. (2) Text entity recognition based on deep learning, and an Attention mechanism is not added; deep learning based classification considers only whole sentences and does not incorporate the overall importance of words. (3) The existing Attention classification method does not add part-of-speech semantic information, is used for classifying the whole text and has poor interpretability. (4) The input part of the neural network directly inputs the whole sentence, does not process partial redundant input in advance and has larger complexity.
In order to solve the problems of the prior art, embodiments of the present invention provide a method and an apparatus for training a medical named entity recognition model, a medical named entity recognition method and apparatus based on the medical named entity recognition model, an electronic device, and a computer storage medium. First, a method for training a medical named entity recognition model according to an embodiment of the present invention is described below.
Fig. 1 is a flowchart illustrating a method for training a medical named entity recognition model according to an embodiment of the present invention. As shown in fig. 1, the method for training the medical named entity recognition model includes the following steps:
s101, obtaining historical patient complaint text data and doctor diagnosis text data corresponding to the historical patient complaint text data.
S102, performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors and part-of-speech information and label information corresponding to the word vectors.
In order to obtain more accurate part-of-speech information and tag information, in an embodiment, the word segmentation and part-of-speech tagging are performed on the historical patient chief complaint text data and the doctor diagnosis text data to obtain each word vector and part-of-speech information and tag information corresponding to each word vector, which may generally include: and performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm to obtain each word vector and part-of-speech information and label information corresponding to each word vector.
S103, training a target network structure by using each word vector and the part of speech information and the label information corresponding to each word vector to obtain a medical named entity recognition model.
In order to obtain a more accurate medical named entity recognition model, in an embodiment, the training of the target network structure by using each word vector and the part-of-speech information and the tag information corresponding to each word vector to obtain the medical named entity recognition model may generally include: obtaining a BI-GRU network and a CRF network; constructing a target network structure based on an Attention classification algorithm, a BI-GRU network and a CRF network; and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
The embodiment of the invention also provides a medical named entity recognition method based on the medical named entity recognition model, the medical named entity recognition model is a model obtained by using the method for training the medical named entity recognition model shown in fig. 1, and as shown in fig. 2, the medical named entity recognition method based on the medical named entity recognition model comprises the following steps:
s201, obtaining patient complaint text data.
S202, inputting the patient complaint text data into the medical named entity recognition model, and outputting a medical named entity recognition result.
In one embodiment, after outputting the medical named entity recognition result, the method may further generally include: constructing a disease condition characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result; acquiring a doctor dictionary set; and determining a doctor sequencing result corresponding to the disease condition characteristic value based on the disease condition characteristic value and the doctor dictionary set.
In one embodiment, obtaining a set of physician dictionaries may generally include: acquiring medical text information of any doctor; determining the diagnosed patient complaint text data of any doctor and the doctor diagnosis text data corresponding to the diagnosed patient complaint text data based on the medical text information; determining a doctor dictionary of any doctor based on the diagnosed patient complaint text data and the doctor diagnosis text data; and summarizing the doctor dictionaries of all doctors to obtain a doctor dictionary set.
The above contents are described below with a specific embodiment, which is as follows:
1. establishing a data set;
in one embodiment, historical patient complaint text and physician diagnostic text data can be obtained, including relevant labeling of the disease site, symptoms, and disease name. All data are processed to form corresponding relations and are properly stored.
2. The medical named entity is recognition model training;
in one embodiment, medical entity recognition may include text preprocessing, model building, and training processes.
Optionally, in the text preprocessing process, the text data may be subjected to preliminary word segmentation processing and part-of-speech tagging, and then the text data is converted into word vectors and corresponding to part-of-speech and tag result information.
Optionally, the model building and training comprises: the framework structure of the design model, the design in-layer relation and the calculation method. In one embodiment, the model may be trained using the processed word vectors and labels, and then medical entity recognition may be performed on the new text data using the trained model.
3. Constructing characteristics;
optionally, in the feature construction process, the entity identification result of each medical text may be converted to construct a corresponding feature value.
4. Constructing a dictionary;
alternatively, a dictionary set may be constructed for each doctor, including feature values and frequencies corresponding to the feature values.
5. Matching doctors;
in an embodiment, the feature values of the text in step 3 may be matched with the dictionary set of the doctor in step 4 to obtain the best doctor ranking result corresponding to the current condition feature value.
6. And 5, displaying the doctor information acquired in the step 5 to a user to finish the whole process.
As shown in fig. 3, the embodiment of the present invention may implement the above-mentioned process through 6 modules, namely, an input component, an NLP module, a feature extraction module, a doctor dictionary construction module, a doctor matching module, and a result display module. The respective modules will be described in detail below:
(1) the input component is used for inputting entries of patient complaint texts and doctor diagnosis texts, can support text and voice input, and is stored in a text form.
(2) And the NLP module is used for realizing text preprocessing and medical entity naming identification. In the text preprocessing, word segmentation and part-of-speech tagging can be performed by using a natural language processing method, and a word is converted into a word vector. In the medical entity naming recognition, an Attention-based BI-GRU-CRF model is constructed, and the model can convert the contents of patient chief complaint texts and doctor diagnosis texts into three entity types of disease parts, symptoms and disease names. Optionally, in view of context, embodiments of the present invention may use Bi-GRUs in the design to simplify the training of the model; in addition, considering the input global weight, an Attention mechanism can be added to improve the precision; optionally, word part-of-speech features may also be added to further improve recognition rates.
(3) And the feature extraction module can be used for extracting the result of the text passing through the NLP module and screening the disease part, symptom and disease name of each text. Alternatively, the disease site and symptom may be constructed as a < disease site and symptom > entity combination, so as to obtain the characteristics that the entity combination, disease name, site and symptom are main bond types.
(4) And the doctor dictionary building module is used for summarizing all the chief complaint texts of patients diagnosed by doctors and the self diagnosis texts. In one embodiment, an own dictionary set can be constructed for each doctor, wherein keys are characteristics and the value is the support degree, namely the occurrence frequency.
(5) And the doctor matching module is used for performing matching retrieval on the features obtained by the feature extraction module and the doctor dictionary constructed in the doctor dictionary module, and retrieving the diagnosis linear weighted sum of each doctor on the features. And sequencing the association degrees of doctors, and finally returning the doctor information of the Top-N, the information of the department where the doctor is located and the extracted features of the patient.
(6) And the result display module is used for displaying the final result to the patient and specifically displaying the result retrieved by the doctor matching module and the features constructed by the feature extraction module.
As shown in fig. 4, the NLP is an overall flowchart, for example, in an embodiment, after receiving the chief complaint text sent by the patient and the diagnosis text sent by the doctor, the NLP module may perform preprocessing and medical body recognition on the chief complaint text and the diagnosis text, and then send the processing results to the patient feature extraction module and the doctor dictionary set construction module, where the patient feature extraction module may send the extracted patient features to the doctor dictionary set construction module. The doctor recommending module can recommend a TopN doctor for the patient according to the patient characteristics sent by the patient characteristic extracting module and the doctor dictionary set constructing module result.
As shown in fig. 5, the figure is a schematic flow chart of the medical entity recognition model training, and includes the following steps: obtaining and preprocessing the corpus; constructing a training/testing data set based on the preprocessed corpus; inputting the constructed training/testing data set into an Attention-BI-GUR network for training so as to obtain a usable model.
As shown in fig. 6, the method is a flowchart for constructing a doctor dictionary, and includes the following steps:
acquiring medical doctor text information, and then acquiring a chief complaint text and a self-diagnosis text of a diagnosed patient based on the medical doctor text information; inputting the chief complaint text of the diagnosed patient and the self-diagnosis text into an NLP module to realize text preprocessing and medical entity naming identification; the NLP module can perform word segmentation and part-of-speech tagging on the diagnosed patient chief complaint text and the self-diagnosed text by using a natural language processing method, and sends a processed result to the feature extraction module; the feature extraction module receives the diagnosed patient chief complaint text and the self-diagnosis text after word segmentation and part-of-speech tagging, can extract features, and arranges, corresponds and stores the extracted features.
The method and flow of implementation of the technical details will be specifically listed below.
1. Preprocessing, vectorizing the text;
firstly, the text is subjected to word segmentation, entity type labeling and related processing of stop words and punctuation marks. The short text is subjected to word segmentation and part-of-speech tagging through a THULAC tool and is marked as jt. The main types of parts of speech are shown in table 1.
TABLE 1 Primary part-of-speech types
| Noun/n | Pronouns/r | Adjective/a | Adverb/d | Verb/v |
| Preposition/p | Conjunction word/c | Time word/t | Punctuation/w | Time word/t |
| Number/m | Help/u | Orientation word/f | Term/s | Other/x |
Each text segment is divided and labeled and then consists of two parts, namely Chinese words and parts of speech. Wherein the Chinese word is converted into a word vector form using word2vec tools. word2vec is trained in the presence of a large amount of text data by predicting the occurrence of the selected word in its context, where we can set the size range of the vector.
After text word segmentation and vectorization processing, patient complaints and doctor diagnosis texts can be effectively converted into numerical vectors, and later-stage model building is facilitated.
2. Building an algorithm model;
2.1 bidirectional GRU neural network
The core of the bidirectional GRU network is that under the condition of taking a word sequence as input, the context information of the current vocabulary is fully considered, and then the knot of the current vocabulary is output, the specific structure is shown as figure 7, and an updating gate z in figure 7 selects whether the hidden state needs to be a new hidden state or not
The reset gate r decides whether the previous hidden state h has been ignored. Wherein, the GRU model has fewer parameters and can better train the model. Vector output of each word through bidirectional GRU neural network structureAnd other word information in long distance is fully considered in the result.
2.2 Attention mechanism
The Attention model can effectively highlight the importance degree of key entity words in the complete text. By calculating the similarity between the current target word and the word of the whole input text and integrating the results of all words, the attention weight vector s of the whole sentence can be finally obtained. Suppose now a sentence in the main complaint text with a total length of L, xt represents the t-th word in the sentence. After filtering through the input layer, calculating the output h of each Chinese word GRU network according to the word vectort、
Calculating the output mu of each Chinese character by combining the results of the front part and the rear parttWherein W is a weight matrix and b is an offset value; a istCalculating an attention weight value of each Chinese word, namely global weight probability; stAnd the method is based on the weight values corresponding to all the words of the tth word in the whole sentence text and is used for being combined with the later prediction classification.
μt＝tanh(Wht+b)
2.3 vector binding and CRF tag prediction
In the embodiment of the invention, the label part of the module can be classified into three types: location of disease, symptoms, disease name. It can be found that the part-of-speech types of an entity are dominated by nouns, adjectives and verbs.
In the embodiment of the invention, the Attention is used to obtain the global vector s of each Chinese word corresponding to a sentencetAnd a bidirectional GRU model output vector htCombined and additionally added with word part of speech information jtWherein nouns, adjectives and verbs are 1, and others are 0. A new feature vector [ j ] is obtained by assembling the three vectorst:ht:st]. Then, after the tanh function is carried out, the output z after the attention layer is obtainedt＝tanh(wu[jt:ht:st]). And finally, performing entity prediction by using a CRF model.
In the embodiment of the invention, three different optimizations, namely a front and a rear, global similarity and part of speech, can be added into the vector input into the GRU model in the Attention-BI-GRU-CRF model. The CRF is run to consider neighboring labels and eventually obtain the final prediction result for each word.
2.4 medical entity recognition model Overall diagrams and explanations
In the embodiment of the invention, all modules built by the module summary algorithm model are shown in fig. 8.
The main details are as follows:
(1) inputting a text sequence; through preprocessing, the text data is converted into Chinese words and is labeled with the form of part of speech, such as the Chinese word 'cold' is labeled as a noun 'n'; for example, "recent", "some", "cold", "gum" and "sore" as shown in fig. 8 are input text sequences, and optionally after preprocessing, "cold" and "gum" can be labeled as the term "n"; "most recent" may be labeled as the time word "t"; "soreness" may be labeled as the adjective "a"; "dotted" is labeled as adverb "d".
(2) Word embedding and part-of-speech tagging; carrying out feature vector coding on the Chinese words by combining a word2vec method, and keeping part of speech tagging unchanged;
(3) filtering an input layer; performing early input information processing through the result of part-of-speech tagging, wherein the early input information processing comprises removing relevant contents such as prepositions, conjunctions, auxiliary words, punctuations and the like, and improving a prediction result;
for example, as shown in FIG. 8, "f" indicates input layer filtering to remove prepositions, conjunctions, co-words, punctuation, and the like.
(4) A BI-GRU layer; obtaining the context characteristics of each Chinese word by using a BI-GRU layer; for example, as shown in FIG. 8, h1、h2……hn， 1h、 2h…… nhRespectively representing the corresponding context characteristics of the Chinese words.
(5) An Attention layer; and introducing an Attention mechanism to calculate the relevance importance between the input and the output of the BI-GRU model, and acquiring the overall characteristics of the sentence according to the importance. As shown in FIG. 8, ztThe calculation comprises three parts of the part of speech of the tth Chinese word, the output of the BI-GRU of the tth Chinese word and the integral characteristics of the attribute corresponding to the tth Chinese word, and the specific operation is zt＝tanh(wu[jt:ht:st])；
(6) A Tanh layer; a score of the label used to predict the neural network output; for example, as shown in fig. 8, e1, e2 … … en represent scores of z1, z2 … … zn, respectively.
(7) A CRF prediction layer; obtaining the label result corresponding to the final Chinese word, wherein '0' represents a stop word, 'DISE' represents a disease name, 'BODY' represents a disease position, and 'SIGN' represents a symptom. For example, as shown in fig. 8, the label results corresponding to e1 and e2 are both 0, and it can be understood that the first Chinese word "nearest" and the second Chinese word "somewhat" are all used words. The result of the label corresponding to et is BODY, and the corresponding label can be understood as that 'gum' is the disease part; the label result for en is SIGN, and a symptom can be characterized as "soreness".
3. Construction of a doctor dictionary
According to the embodiment of the invention, the disease part, symptom and disease name can be respectively obtained through the chief complaint text and the diagnosis text data set of the patient to be treated by each doctor through the medical entity recognition module.
According to the embodiment of the invention, the nearest disease parts and symptoms in the text are integrated separately, and the entity combination of the disease parts and symptoms is obtained. The dictionary with entity combination, disease name, part and symptom as main keys is constructed by a dictionary method in a data structure, and the value corresponding to each main key is the frequency of occurrence, namely the support degree. Wherein the weight setting of the entity combination and the disease name is large.
4. New patient text processing and recommendation
For new patient complaint text data, the disease part, symptom and disease name are respectively obtained through a medical entity recognition module. And (4) constructing the user characteristics of the patient according to the method in the step (4), sequentially searching the patient characteristics and keys of each doctor dictionary, and adding the support degree results of the key values corresponding to all the characteristics of the patient to obtain the matching score of each doctor for the patient. Finally returning the matching doctor list of Top-N through a sorting method.
The invention is described in further detail below with reference to the figures and the embodiments.
Step one, collecting relevant data, collecting historical patient chief complaint text data and doctor diagnosis data, wherein all collected contents need desensitization treatment in order to consider privacy problems of patients.
And step two, data labeling, namely labeling the data collected in the step one, wherein the labeling process is carried out in a mode of automatic labeling, manual labeling and correction, and four types of labels including 0, BODY, SIGN and DISE are decomposed and set. If "I have swollen legs today", the statement is marked as:
| i am | 0 |
| Today's |
0 |
| Leg part | BODY |
| Swelling of the stomach | SIGN |
Considering that parameters need to be tuned and optimized according to test results in the process of training the model, the marked data set can be divided into a training data set and a test set.
And step three, the medical entity recognition module utilizes the medical text data marked in the step two, utilizes the training set to train the neural network, sends the test set into the trained model, and adjusts the training parameters according to the test result until the accuracy rate reaches 90%.
And step four, constructing a doctor dictionary, identifying medical entities according to the patient chief complaint text and the diagnosis text data of each doctor, and constructing the dictionary according to the method 3 in the technical route.
And step five, when a new patient visits a hospital to issue a chief complaint text, obtaining the visiting characteristics (disease part, symptom and disease name) of the chief complaint text through a medical entity recognition module, inquiring a doctor dictionary according to the visiting characteristics in a matching way, and returning the list of the former N doctors through the ordering result of the support degree of each doctor dictionary.
The process flow for "i today's leg swelling" as the main complaint text is as follows: (1) the word segmentation results are 'I-r, today-t, leg-n and swelling-v'; (2) the results of the entity recognition after word embedding were "I-0, leg-BODY, swelling SIGN" (notably, "today" is filtered in the input layer filtering for time words); (3) constructing and processing the diagnosis characteristics to obtain { leg, swelling, < leg, swelling > }; (4) and (3) searching a Doctor dictionary, performing searching and weighted calculation according to the features in the step (3) to obtain { vector 1:264, vector 2:241, vector 3:185, … }, and outputting a Top-N Doctor list after ordering the support degree.
And step six, displaying to the user according to the Top-N doctor list obtained in the step five, wherein the display mode comprises but is not limited to various modes such as lists, pictures and the like.
The embodiment of the invention has the following advantages:
(1) the patient complains about the text information more comprehensively; the input layer in the design of the model is filtered, and medical entity prediction can be better performed by considering the front and back words, the part of speech, the Attention global similarity and the like.
(2) The interpretability is stronger; partial entity combinations and disease names recommended at the patient can help the patient to see a doctor accurately and improve the reliability of diagnosis.
(3) The efficiency is higher and the automation is higher; the whole process of the system directly processes the main complaint text of the new patient, and doctor recommendation can be quickly carried out.
An embodiment of the present invention further provides a device for training a medical named entity recognition model, as shown in fig. 9, the device for training a medical named entity recognition model includes:
an obtaining module 901, configured to obtain historical patient chief complaint text data and doctor diagnosis text data corresponding to the historical patient chief complaint text data;
a labeling module 902, configured to perform word segmentation and part-of-speech labeling on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors and part-of-speech information and tag information corresponding to each word vector;
and the model training module 903 is configured to train a target network structure by using each word vector and the part-of-speech information and the label information corresponding to each word vector, so as to obtain a medical named entity recognition model.
Optionally, in an embodiment, the labeling module 902 is configured to perform word segmentation and part-of-speech labeling on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm, so as to obtain each word vector and part-of-speech information and tag information corresponding to each word vector.
Optionally, in an embodiment, the model training module 903 is configured to obtain a BI-GRU network and a CRF network; constructing a target network structure based on an Attention classification algorithm, a BI-GRU network and a CRF network; and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
Each module in the apparatus for training a medical named entity recognition model provided in fig. 9 has a function of implementing each step in the example shown in fig. 1, and achieves the same technical effect as the method for training a medical named entity recognition model shown in fig. 1, and for brevity, no further description is given here.
An embodiment of the present invention further provides a medical named entity recognition apparatus based on a medical named entity recognition model, where the medical named entity recognition model is a model obtained by using the method for training the medical named entity recognition model shown in fig. 1, and as shown in fig. 10, the medical named entity recognition apparatus based on the medical named entity recognition model includes:
a data acquisition module 1001 for acquiring patient complaint text data;
the output module 1002 is configured to input the patient complaint text data into the medical named entity recognition model, and output a medical named entity recognition result.
Optionally, in an embodiment, the medical named entity recognition apparatus based on the medical named entity recognition model further includes:
the construction module is used for constructing a disease condition characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result;
the acquisition module is used for acquiring a doctor dictionary set;
and the determining module is used for determining a doctor sequencing result corresponding to the illness state characteristic value based on the illness state characteristic value and the doctor dictionary set.
Optionally, in an embodiment, the obtaining module is configured to: acquiring medical text information of any doctor; determining the diagnosed patient complaint text data of any doctor and the doctor diagnosis text data corresponding to the diagnosed patient complaint text data based on the medical text information; determining a doctor dictionary of any doctor based on the diagnosed patient complaint text data and the doctor diagnosis text data; and summarizing the doctor dictionaries of all doctors to obtain a doctor dictionary set.
Each module in the medical named entity recognition apparatus based on the medical named entity recognition model provided in fig. 10 has a function of implementing each step in the example shown in fig. 2, and achieves the same technical effect as the medical named entity recognition method based on the medical named entity recognition model shown in fig. 2, and is not described herein again for brevity.
Fig. 11 is a schematic structural diagram of an electronic device according to an embodiment of the present invention.
The electronic device may include a processor 1101 and a memory 1102 in which computer program instructions are stored.
Specifically, the processor 1101 may include a Central Processing Unit (CPU), or an Application Specific Integrated Circuit (ASIC), or may be configured as one or more Integrated circuits implementing embodiments of the present invention.
The processor 1101 reads and executes computer program instructions stored in the memory 1102 to implement the method for training a medical named entity recognition model according to the embodiment shown in fig. 1, or the processor 1101 reads and executes computer program instructions stored in the memory 1102 to implement the medical named entity recognition method based on a medical named entity recognition model according to the embodiment shown in fig. 2.
In one example, the electronic device can also include a communication interface 1103 and a bus 1110. As shown in fig. 11, the processor 1101, the memory 1102, and the communication interface 1103 are connected via a bus 1110 to complete communication therebetween.
The communication interface 1103 is mainly used for implementing communication between modules, apparatuses, units and/or devices in the embodiment of the present invention.
In addition, embodiments of the present invention may be implemented by providing a computer storage medium. The computer storage medium having computer program instructions stored thereon; the computer program instructions, when executed by a processor, implement the method of training a medical named entity recognition model of the embodiment shown in fig. 1, or the computer program instructions, when executed by a processor, implement the medical named entity recognition method based on a medical named entity recognition model of the embodiment shown in fig. 2.
It is to be understood that the invention is not limited to the specific arrangements and instrumentality described above and shown in the drawings. A detailed description of known methods is omitted herein for the sake of brevity. In the above embodiments, several specific steps are described and shown as examples. However, the method processes of the present invention are not limited to the specific steps described and illustrated, and those skilled in the art can make various changes, modifications and additions or change the order between the steps after comprehending the spirit of the present invention.
The functional blocks shown in the above-described structural block diagrams may be implemented as hardware, software, firmware, or a combination thereof. When implemented in hardware, it may be, for example, an electronic circuit, an Application Specific Integrated Circuit (ASIC), suitable firmware, plug-in, function card, or the like. When implemented in software, the elements of the invention are the programs or code segments used to perform the required tasks. The program or code segments may be stored in a machine-readable medium or transmitted by a data signal carried in a carrier wave over a transmission medium or a communication link. A "machine-readable medium" may include any medium that can store or transfer information. Examples of a machine-readable medium include electronic circuits, semiconductor memory devices, ROM, flash memory, Erasable ROM (EROM), floppy disks, CD-ROMs, optical disks, hard disks, fiber optic media, Radio Frequency (RF) links, and so forth. The code segments may be downloaded via computer networks such as the internet, intranet, etc.
It should also be noted that the exemplary embodiments mentioned in this patent describe some methods or systems based on a series of steps or devices. However, the present invention is not limited to the order of the above-described steps, that is, the steps may be performed in the order mentioned in the embodiments, may be performed in an order different from the order in the embodiments, or may be performed simultaneously.
As described above, only the specific embodiments of the present invention are provided, and it can be clearly understood by those skilled in the art that, for convenience and brevity of description, the specific working processes of the system, the module and the unit described above may refer to the corresponding processes in the foregoing method embodiments, and are not described herein again. It should be understood that the scope of the present invention is not limited thereto, and any person skilled in the art can easily conceive various equivalent modifications or substitutions within the technical scope of the present invention, and these modifications or substitutions should be covered within the scope of the present invention.
Claims (13)
1. A method of training a medical named entity recognition model, comprising:
acquiring historical patient complaint text data and doctor diagnosis text data corresponding to the historical patient complaint text data;
performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors and part-of-speech information and label information corresponding to the word vectors;
and training a target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain a medical named entity recognition model.
2. The method for training a medical named entity recognition model according to claim 1, wherein the step of performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors and part-of-speech information and label information corresponding to each word vector comprises the steps of:
and performing word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm to obtain each word vector and part-of-speech information and label information corresponding to each word vector.
3. The method for training a medical named entity recognition model according to claim 1, wherein the training of the target network structure using the word vectors, the part-of-speech information corresponding to each word vector, and the tag information to obtain the medical named entity recognition model comprises:
obtaining a BI-GRU network and a CRF network;
constructing the target network structure based on an Attention classification algorithm, the BI-GRU network and the CRF network;
and training the target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain the medical named entity recognition model.
4. A medical named entity recognition method based on a medical named entity recognition model, wherein the medical named entity recognition model is obtained by using the method for training the medical named entity recognition model according to any one of claims 1 to 3, and comprises the following steps:
acquiring patient complaint text data;
and inputting the patient complaint text data into the medical named entity recognition model, and outputting a medical named entity recognition result.
5. The medical named entity recognition method according to claim 4, further comprising, after outputting the medical named entity recognition result:
constructing a disease condition characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result;
acquiring a doctor dictionary set;
and determining a doctor sequencing result corresponding to the condition characteristic value based on the condition characteristic value and the doctor dictionary set.
6. The medical named entity recognition method of claim 5, wherein the obtaining a set of physician dictionaries comprises:
acquiring medical text information of any doctor;
determining diagnosed patient complaint text data of any one doctor and doctor diagnosis text data corresponding to the diagnosed patient complaint text data based on the medical text information;
determining a doctor dictionary of any one doctor based on the diagnosed patient complaint text data and the doctor diagnosis text data;
and summarizing the doctor dictionaries of all doctors to obtain the doctor dictionary set.
7. An apparatus for training a medical named entity recognition model, comprising:
the acquisition module is used for acquiring historical patient complaint text data and doctor diagnosis text data corresponding to the historical patient complaint text data;
the labeling module is used for performing word segmentation and part-of-speech labeling on the historical patient chief complaint text data and the doctor diagnosis text data to obtain word vectors, and part-of-speech information and label information corresponding to the word vectors;
and the model training module is used for training a target network structure by utilizing the word vectors and the part of speech information and the label information corresponding to the word vectors to obtain a medical named entity recognition model.
8. The apparatus for training a medical named entity recognition model as claimed in claim 7, wherein the tagging module is configured to perform word segmentation and part-of-speech tagging on the historical patient chief complaint text data and the doctor diagnosis text data by using a natural language processing algorithm, so as to obtain each word vector and part-of-speech information and tag information corresponding to each word vector.
9. The apparatus for training a medical named entity recognition model of claim 7, wherein the model training module is configured to obtain a BI-GRU network and a CRF network; constructing the target network structure based on an Attention classification algorithm, the BI-GRU network and the CRF network; and training the target network structure by using each word vector and the part of speech information and label information corresponding to each word vector to obtain the medical named entity recognition model.
10. A medical named entity recognition device based on a medical named entity recognition model, wherein the medical named entity recognition model is obtained by using the method for training the medical named entity recognition model according to any one of claims 1 to 3, and comprises:
the data acquisition module is used for acquiring the patient complaint text data;
and the output module is used for inputting the patient complaint text data into the medical named entity recognition model and outputting a medical named entity recognition result.
11. The medical named entity recognition device of claim 10, further comprising:
the construction module is used for constructing an illness state characteristic value corresponding to the medical named entity identification result based on the medical named entity identification result;
the acquisition module is used for acquiring a doctor dictionary set;
and the determining module is used for determining a doctor sequencing result corresponding to the illness state characteristic value based on the illness state characteristic value and the doctor dictionary set.
12. An electronic device, characterized in that the device comprises: a processor and a memory storing computer program instructions;
the computer program instructions, when executed by the processor, implement a method of training a medical named entity recognition model according to any one of claims 1-3; or a medical named entity recognition method based on a medical named entity recognition model according to any of claims 4-6.
13. A computer storage medium having stored thereon computer program instructions which, when executed by a processor, implement a method of training a medical named entity recognition model according to any one of claims 1-3; or a medical named entity recognition method based on a medical named entity recognition model according to any of claims 4-6.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911382710.4A CN113051905A (en) | 2019-12-28 | 2019-12-28 | Medical named entity recognition training model and medical named entity recognition method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911382710.4A CN113051905A (en) | 2019-12-28 | 2019-12-28 | Medical named entity recognition training model and medical named entity recognition method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN113051905A true CN113051905A (en) | 2021-06-29 |
Family
ID=76507393
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911382710.4A Pending CN113051905A (en) | 2019-12-28 | 2019-12-28 | Medical named entity recognition training model and medical named entity recognition method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113051905A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113299375A (en) * | 2021-07-27 | 2021-08-24 | 北京好欣晴移动医疗科技有限公司 | Method, device and system for marking and identifying digital file information entity |
| CN113409907A (en) * | 2021-07-19 | 2021-09-17 | 广州方舟信息科技有限公司 | Intelligent pre-inquiry method and system based on Internet hospital |
| CN113420059A (en) * | 2021-08-23 | 2021-09-21 | 中关村科学城城市大脑股份有限公司 | Method and device for actively treating citizen hot line problem |
| CN113486178A (en) * | 2021-07-12 | 2021-10-08 | 恒安嘉新（北京）科技股份公司 | Text recognition model training method, text recognition device and medium |
| CN113537346A (en) * | 2021-07-15 | 2021-10-22 | 思必驰科技股份有限公司 | Medical field data labeling model training method and medical field data labeling method |
| CN113626429A (en) * | 2021-07-26 | 2021-11-09 | 上海齐网网络科技有限公司 | Method and system for constructing intelligent range emergency medical knowledge base based on metadata |
| CN113903422A (en) * | 2021-09-09 | 2022-01-07 | 北京邮电大学 | Medical image diagnosis report entity extraction method, device and equipment |
| CN115171835A (en) * | 2022-09-02 | 2022-10-11 | 北京智源人工智能研究院 | Case structured model training method and device and case structured method |
| CN117313733A (en) * | 2023-11-30 | 2023-12-29 | 北京航空航天大学杭州创新研究院 | Medical entity identification system |
| CN117766137A (en) * | 2024-02-22 | 2024-03-26 | 广东省人民医院 | medical diagnosis result determining method and device based on reinforcement learning |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090326923A1 (en) * | 2006-05-15 | 2009-12-31 | Panasonic Corporatioin | Method and apparatus for named entity recognition in natural language |
| CN109871538A (en) * | 2019-02-18 | 2019-06-11 | 华南理工大学 | A kind of Chinese electronic health record name entity recognition method |
| CN110298036A (en) * | 2019-06-06 | 2019-10-01 | 昆明理工大学 | A kind of online medical text symptom identification method based on part of speech increment iterative |
| CN110444261A (en) * | 2019-07-11 | 2019-11-12 | 新华三大数据技术有限公司 | Sequence labelling network training method, electronic health record processing method and relevant apparatus |
-
2019
- 2019-12-28 CN CN201911382710.4A patent/CN113051905A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20090326923A1 (en) * | 2006-05-15 | 2009-12-31 | Panasonic Corporatioin | Method and apparatus for named entity recognition in natural language |
| CN109871538A (en) * | 2019-02-18 | 2019-06-11 | 华南理工大学 | A kind of Chinese electronic health record name entity recognition method |
| CN110298036A (en) * | 2019-06-06 | 2019-10-01 | 昆明理工大学 | A kind of online medical text symptom identification method based on part of speech increment iterative |
| CN110444261A (en) * | 2019-07-11 | 2019-11-12 | 新华三大数据技术有限公司 | Sequence labelling network training method, electronic health record processing method and relevant apparatus |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113486178B (en) * | 2021-07-12 | 2023-12-01 | 恒安嘉新（北京）科技股份公司 | Text recognition model training method, text recognition method, device and medium |
| CN113486178A (en) * | 2021-07-12 | 2021-10-08 | 恒安嘉新（北京）科技股份公司 | Text recognition model training method, text recognition device and medium |
| CN113537346B (en) * | 2021-07-15 | 2023-08-15 | 思必驰科技股份有限公司 | Medical field data labeling model training method and medical field data labeling method |
| CN113537346A (en) * | 2021-07-15 | 2021-10-22 | 思必驰科技股份有限公司 | Medical field data labeling model training method and medical field data labeling method |
| CN113409907A (en) * | 2021-07-19 | 2021-09-17 | 广州方舟信息科技有限公司 | Intelligent pre-inquiry method and system based on Internet hospital |
| CN113626429A (en) * | 2021-07-26 | 2021-11-09 | 上海齐网网络科技有限公司 | Method and system for constructing intelligent range emergency medical knowledge base based on metadata |
| CN113626429B (en) * | 2021-07-26 | 2024-04-12 | 上海齐网网络科技有限公司 | Metadata-based intelligent range emergency medical knowledge base construction method and system |
| CN113299375A (en) * | 2021-07-27 | 2021-08-24 | 北京好欣晴移动医疗科技有限公司 | Method, device and system for marking and identifying digital file information entity |
| CN113420059A (en) * | 2021-08-23 | 2021-09-21 | 中关村科学城城市大脑股份有限公司 | Method and device for actively treating citizen hot line problem |
| CN113903422A (en) * | 2021-09-09 | 2022-01-07 | 北京邮电大学 | Medical image diagnosis report entity extraction method, device and equipment |
| CN115171835A (en) * | 2022-09-02 | 2022-10-11 | 北京智源人工智能研究院 | Case structured model training method and device and case structured method |
| CN117313733A (en) * | 2023-11-30 | 2023-12-29 | 北京航空航天大学杭州创新研究院 | Medical entity identification system |
| CN117766137A (en) * | 2024-02-22 | 2024-03-26 | 广东省人民医院 | medical diagnosis result determining method and device based on reinforcement learning |
| CN117766137B (en) * | 2024-02-22 | 2024-05-28 | 广东省人民医院 | Medical diagnosis result determining method and device based on reinforcement learning |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN113051905A (en) | Medical named entity recognition training model and medical named entity recognition method | |
| CN111401066B (en) | Artificial intelligence-based word classification model training method, word processing method and device | |
| Zhang et al. | Artificial intelligence–based traditional Chinese medicine assistive diagnostic system: validation study | |
| US9621601B2 (en) | User collaboration for answer generation in question and answer system | |
| CN108027823B (en) | Information processing device, information processing method, and computer-readable storage medium | |
| EP3895178A1 (en) | System and method for providing health information | |
| CN112786194A (en) | Medical image diagnosis guide inspection system, method and equipment based on artificial intelligence | |
| CN111897967A (en) | Medical inquiry recommendation method based on knowledge graph and social media | |
| US11670420B2 (en) | Drawing conclusions from free form texts with deep reinforcement learning | |
| CN112257422B (en) | Named entity normalization processing method and device, electronic equipment and storage medium | |
| Fang et al. | Feature Selection Method Based on Class Discriminative Degree for Intelligent Medical Diagnosis. | |
| US20190057773A1 (en) | Method and system for performing triage | |
| CN112885478B (en) | Medical document retrieval method, medical document retrieval device, electronic device and storage medium | |
| CN112232065A (en) | Method and device for mining synonyms | |
| US11481557B2 (en) | Clinical terminology mapping with natural language processing | |
| CN113095081A (en) | Disease identification method and device, storage medium and electronic device | |
| CN117149998B (en) | Intelligent diagnosis recommendation method and system based on multi-objective optimization | |
| Chandra et al. | Natural language Processing and Ontology based Decision Support System for Diabetic Patients | |
| CN114238639A (en) | Construction method and device of medical term standardized framework and electronic equipment | |
| CN114141384A (en) | Method, apparatus and medium for retrieving medical data | |
| CN117609635A (en) | Collaborative filtering-based data pushing method and device | |
| CN116741333B (en) | Medicine marketing management system | |
| CN115565655A (en) | Enhanced auxiliary inquiry method | |
| CN114117082A (en) | Method, apparatus, and medium for correcting data to be corrected | |
| CN115713992A (en) | Data analysis system and data analysis method |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20210629 |
|
| RJ01 | Rejection of invention patent application after publication |