CN112201359A

CN112201359A - Artificial intelligence-based critical illness inquiry data identification method and device

Info

Publication number: CN112201359A
Application number: CN202011065413.XA
Authority: CN
Inventors: 满晏松; 柳恭; 李响
Original assignee: Ping An Technology Shenzhen Co Ltd
Current assignee: Ping An Technology Shenzhen Co Ltd
Priority date: 2020-09-30
Filing date: 2020-09-30
Publication date: 2021-01-08
Anticipated expiration: 2040-09-30
Also published as: CN112201359B; WO2022068160A1

Abstract

The application relates to artificial intelligence and provides a critical illness inquiry data identification method and device based on artificial intelligence. The critical illness inquiry data identification method based on artificial intelligence comprises the following steps: acquiring inquiry session data corresponding to the target user identification; inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model; determining an expert identification result corresponding to the inquiry session data according to the tags hit by the inquiry session data in the expert knowledge base; and combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data. By adopting the method, the identification accuracy of the critical illness inquiry data can be improved. In addition, the invention also relates to a block chain technology, and the inquiry session data of the user can be stored in the block chain.

Description

Artificial intelligence-based critical illness inquiry data identification method and device

Technical Field

The application relates to the technical field of artificial intelligence, in particular to a critical illness inquiry data identification method and device based on artificial intelligence.

Background

With the development of internet technology and medical technology, the application of internet technology in the medical industry is more and more common. For example, the user may self-describe symptoms, consult symptoms, learn about medications, seek medical advice, and the like through an online inquiry application or an online inquiry website. The doctor can judge whether the current inquiry data is critical inquiry data or not by inquiring the etiology, symptoms and the like in multiple rounds.

However, as the number of users initiating online inquiry increases, in order to reduce the heavy manual identification of doctors, the medical field begins to adopt a mode of discrimination by an expert system. Although expert systems can reduce the workload of the relevant personnel, they present a great challenge to the accuracy of the discrimination of whether current interrogation data is critical interrogation data due to the diversity of on-line user interrogation data.

Disclosure of Invention

In view of the above, it is necessary to provide an artificial intelligence based critical care data identification method, an apparatus, a computer device, and a storage medium, which can improve the accuracy of identifying critical care data based on artificial intelligence.

An artificial intelligence based critical illness data identification method, characterized in that the method comprises:

acquiring inquiry session data corresponding to the target user identification;

inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extracting according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data;

determining an expert identification result corresponding to the inquiry session data according to the label hit by the inquiry session data in an expert knowledge base;

and combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data.

In one embodiment, the inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model includes:

inputting the inquiry session data into a prediction model, and processing the inquiry session data through a plurality of neurons included in the prediction model to obtain a model identification result corresponding to the inquiry session data;

outputting the model identification result through the prediction model;

the threshold value of each neuron and the weight of the connection relation between the neurons are determined when the prediction model is trained through multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, and the multi-dimensional feature data further comprise intention feature data.

In one embodiment, the multi-dimensional feature data further comprises intent feature data; the method further comprises the following steps:

collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not;

extracting entity characteristic data, entity relation characteristic data and intention characteristic data from the historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data;

inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result;

and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the extracting entity feature data, entity relationship feature data, and intention feature data from the historical inquiry data to generate multidimensional feature data corresponding to the historical inquiry data includes:

adopting a rule engine to drive and extract entity characteristic data corresponding to the historical inquiry data;

carrying out named entity identification on the historical inquiry data by adopting a sequence labeling model, and extracting entity relation characteristic data among the named entities based on the identified named entities;

extracting intention characteristic data in the historical inquiry data by adopting a semantic model;

and generating multi-dimensional characteristic data corresponding to the historical inquiry data according to the entity characteristic data, the entity relation characteristic data and the intention characteristic data.

In one embodiment, the extracting, by using a rule engine driver, entity feature data corresponding to the historical inquiry data includes:

performing word segmentation on the historical inquiry data to obtain word segmentation results;

based on the entity labels provided by the expert knowledge base, adopting a rule engine to drive and extract the entity labels according to the word segmentation result to obtain entity characteristic data corresponding to the historical inquiry data;

the extraction of the intention characteristic data in the historical inquiry data by adopting the semantic model comprises the following steps:

and extracting the intention labels corresponding to the historical inquiry data by adopting a semantic model based on the intention labels provided by the expert knowledge base to obtain intention characteristic data corresponding to the historical inquiry data.

In one embodiment, the method further comprises:

screening out historical critical illness inquiry data from the historical inquiry data;

counting the occurrence frequency of entity keywords in the historical critical inquiry data;

screening entity keywords with the occurrence frequency higher than a preset threshold value to serve as entity labels to be added into an expert knowledge base;

determining an inquiry intention corresponding to the historical critical inquiry data to obtain an intention label;

adding the intention tag to the expert knowledge base.

In one embodiment, the interrogation session data is stored in a blockchain; the method further comprises the following steps:

when the obtained inquiry session data is a target identification result of the critical inquiry data, accessing the inquiry session to which the inquiry session data belongs to a doctor terminal;

and when the obtained inquiry session data is not the target identification result of the critical inquiry data, continuing to advance the inquiry session to which the inquiry session data belongs.

An artificial intelligence based critical illness data identification apparatus, the apparatus comprising:

the acquisition module is used for acquiring the inquiry session data corresponding to the target user identification;

the first identification module is used for inputting the inquiry session data into a prediction model and outputting a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extracting according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data;

the second identification module is used for determining an expert identification result corresponding to the inquiry session data according to the label hit by the inquiry session data in an expert knowledge base;

and the decision module is used for combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data.

A computer device comprising a memory and a processor, the memory storing a computer program, the processor implementing the following steps when executing the computer program:

acquiring inquiry session data corresponding to the target user identification;

A computer storage medium having a computer program stored thereon, the computer program when executed by a processor implementing the steps of:

acquiring inquiry session data corresponding to the target user identification;

After the critical illness inquiry data identification method, the critical illness inquiry data identification device, the computer equipment and the storage medium based on the artificial intelligence are obtained, on one hand, the inquiry conversation data are input into a prediction model to obtain a model identification result, on the other hand, the inquiry conversation data are identified based on an expert knowledge base to obtain an expert identification result, and then the model identification result and the expert identification result are combined to obtain a final identification result. Because the prediction model is obtained based on multi-dimensional characteristic data training, and the multi-dimensional characteristic data comprises entity characteristic data and entity relation characteristic data, the prediction model can learn information with different dimensions in the training process, and better understand the logic of language by combining the context semantic environment, thereby improving the recognition capability of the prediction model on the data of the symptom inquiry; the method combines model prediction and an expert system to identify the critical illness inquiry data, can make up for the defect of only depending on the expert system, and improves the accuracy of identification of the critical illness inquiry data.

Drawings

FIG. 1 is a diagram illustrating an example of an application of the method for identifying critical illness data based on artificial intelligence;

FIG. 2 is a schematic flow chart illustrating an embodiment of a method for identifying critical care data based on artificial intelligence;

FIG. 3 is a block flow diagram of the use of a predictive model in one embodiment;

FIG. 4 is a block diagram of a process for training a predictive model according to one embodiment;

FIG. 5 is a block diagram of an example of an apparatus for identifying critical care data based on artificial intelligence;

FIG. 6 is a block diagram showing an example of an apparatus for identifying critical care data based on artificial intelligence in another embodiment;

FIG. 7 is a diagram illustrating an internal structure of a computer device according to an embodiment;

fig. 8 is an internal structural view of a computer device in another embodiment.

Detailed Description

In order to make the objects, technical solutions and advantages of the present application more apparent, the present application is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the present application and are not intended to limit the present application.

The critical illness inquiry data identification method based on artificial intelligence can be applied to the application environment shown in fig. 1. Wherein the terminal 102 communicates with the server 104 via a network. Specifically, the terminal 102 may obtain the inquiry session data corresponding to the target user identifier, send the inquiry session data to the server 104, and the server 104 inputs the inquiry session data into the prediction model and outputs the model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extraction according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data; the server 104 then determines an expert identification result corresponding to the inquiry session data according to the tags hit by the inquiry session data in the expert knowledge base; the server 104 combines the model identification result and the expert identification result to obtain a target identification result of whether the inquiry session data is critical inquiry data. The terminal 102 may be, but not limited to, various personal computers, notebook computers, smart phones, tablet computers, and portable wearable devices, and the server 104 may be implemented by an independent server or a server cluster formed by a plurality of servers.

In another embodiment, the terminal 102 or the server 104 may be used alone to perform the artificial intelligence-based critical care data identification method. The present application is not limited thereto.

In one embodiment, as shown in fig. 2, an artificial intelligence-based method for identifying critical care inquiry data is provided, which is described by taking the method as an example applied to a computer device, which may be specifically a terminal or a server in fig. 1. The critical illness inquiry data identification method based on artificial intelligence specifically comprises the following steps:

step 202, obtaining the inquiry session data corresponding to the target user identifier.

Wherein the target user identification is used to uniquely identify a user. Such as an application account number or hospital visit card number, etc. An interrogation session is the process of conducting an interrogation interaction between at least two users. The at least two users include a user corresponding to a patient role and a user corresponding to a doctor role. The interrogation session data is data generated during an interrogation interaction. The user corresponding to the doctor role can be the doctor himself or the artificial intelligence robot.

Specifically, an on-line inquiry application or an on-line inquiry website can be operated on the terminal, and the on-line inquiry application or the on-line inquiry website can provide an inquiry entrance. The user inputs the inquiry session data based on the inquiry portal through the terminal to perform online inquiry.

In one embodiment, the interrogation session data may be voice data, text data, or image data, among others.

In one embodiment, the interrogation session data may include interrogation session data corresponding to a patient role and interrogation session data corresponding to a physician role. Wherein, the inquiry session data corresponding to the patient role, such as user basic information, symptom description information, symptom photo, medical examination report or past history information, etc. It is understood that the inquiry user may or may not be the patient himself, such as a scene for inquiring for children or the elderly. And inquiry session data corresponding to the doctor role, such as disease description information, symptom analysis information, etiology analysis information, or response information for the user inquiry data.

In one embodiment, the interrogation session data may be session data for one or more rounds of interrogation during a single interrogation. Therefore, the inquiry session data provided by the user for the first time can be recognized as early as possible to perform corresponding processing when the inquiry session data is the critical inquiry data; when the information quantity is insufficient, the user is guided to provide more information so as to combine the information to more accurately identify the critical inquiry data.

Wherein the critical care inquiry data is inquiry session data relating to critical care. "critically ill" is a medical term that generally indicates that the patient is suffering from some urgent, endangered condition that should be medically managed as early as possible, otherwise serious harm to the patient's body or death may result. For example, the data of the inquiry session includes critical clinical manifestations and symptoms. The clinical manifestations of critical illness include "syncope" and "dyspnea".

Step 204, inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extraction according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data.

The prediction model is a machine learning model which is trained in advance and used for identifying whether the inquiry session data are critical inquiry data or not. The machine learning model may employ a neural network model, a support vector machine, a logistic regression model, or the like. Neural network models such as convolutional neural networks, back propagation neural networks, feedback neural networks, radial basis neural networks, or self-organizing neural networks, among others.

An entity characteristic is data that reflects the characteristics of the entity itself. For example, the inquiry session data "alcohol consumption during dinner, and abdominal pain and patience now" includes two entities, the first entity being "alcohol consumption" and the second entity being "abdominal pain". An entity relationship characteristic refers to data reflecting a relationship between at least two entities. For example, the inquiry conversation data "drinking during supper", the physical relationship between "middle" drinking "and" abdominal pain "is now" causal relationship ", i.e. the cause of abdominal pain is drinking. Here, on one hand, it is considered that the entities in the inquiry session data are important basis for identifying the critical inquiry data, and on the other hand, it is also considered that the relationship between different entities also affects the identification result, and even the same entity has different semantics in different contexts and further affects the identification result. For example, the inquiry session data "aunt, aunt intolerance to abdominal pain" aunt "refers to menstruation, not to the name of relatives. In this case, the computer device may fuse information in multiple aspects when designing input data of the prediction model, for example, fuse two characteristic dimensions, namely an entity dimension and an entity relationship dimension, of the data to be used as input data of the prediction model, so that the prediction model can learn effective information of the two characteristic dimensions during training, and the recognition capability of the model for critical care data is improved.

Specifically, the computer device may input the inquiry session data into the prediction model, process the inquiry session data through a plurality of neurons included in the prediction model to obtain a model identification result corresponding to the inquiry session data, and output the model identification result through the prediction model. The neuron is the most basic structure in the neural network, and generally, most neurons are in an inhibitory state, but when the neuron receives input information, and the potential of the neuron exceeds a threshold value, the neuron is activated and is in an "excitatory" state, and output information is transmitted to other neurons. The connecting lines between connecting neurons correspond to a weight (the value of which is called the weight), and usually different connecting lines correspond to different weights. The threshold value of each neuron and the weight of the connection relation between the neurons are determined when the prediction model is trained through the multidimensional characteristic data and the training labels corresponding to the multidimensional characteristic data. Neurons include input neurons, output neurons, and hidden neurons.

In one embodiment, the method for identifying critical illness inquiry data based on artificial intelligence further comprises a training step of a prediction model, wherein the training step specifically comprises the following steps: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data and entity relation characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the multi-dimensional feature data further comprises intent feature data. At this time, the training step of the prediction model specifically includes: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

Wherein the intention characteristics are data reflecting an expression intention. The intention characteristics include characteristic data reflecting the doctor's inquiry or response intention and the user's inquiry or response intention. For example, the inquiry session data "you are now not colic intolerant" to express symptom confirmation intent; the inquiry session data "yes, i now abdominal pain intolerance" expresses the intention of symptom confirmation, and the like.

In this way, the intentions of inquiry and answer of doctors and users can also be used as the basis for identifying the critical care data, so that intention characteristics can be introduced in the input of the prediction model, the three characteristic dimensions of the entity dimension, the entity relationship dimension and the intention dimension are fused, the prediction model can learn the effective information of the three characteristic dimensions in training, the identification capability of the model on the critical care data is improved, and the identification direction of the critical care data can be expanded.

The details of the predictive model training step can be referred to the details in the following embodiments.

In one embodiment, the computer device may convert the interrogation session data into a data format that the predictive model is capable of processing, and then input the converted data into the predictive model. The prediction model can process data in a format such as a vector format or a matrix format.

And step 206, determining an expert identification result corresponding to the inquiry session data according to the tags hit by the inquiry session data in the expert knowledge base.

The knowledge base is a rule set applied by expert system design, and comprises facts and data related to the rules, and the knowledge base is formed by the facts and the data. The knowledge base is associated with a specific expert system, in this application the expert knowledge base is associated with an expert system for medical interrogation. The rules in the expert knowledge base are obtained by extracting data which appears frequently in the critical inquiry data in the historical inquiry process, the fact related to the rules includes whether the inquiry session data in the historical inquiry process are critical inquiry data or not, and the data related to the rules includes labels of frequently appearing disease symptoms and the like in the inquiry session data in the historical inquiry process.

In particular, the computer device may employ a rules engine driver to determine the tags hit by the interrogation session data in the expert knowledge base based on a set of rules of the expert knowledge base. When the hit label is the label corresponding to the critical interrogation data, obtaining an expert identification result corresponding to the interrogation session data, wherein the interrogation session data is the critical interrogation data; if the hit tag is not the tag corresponding to the critical care data, the expert identification result corresponding to the inquiry session data is obtained as the inquiry session data is not the critical care data.

And step 208, combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data.

Specifically, the model identification result includes two identification results that the inquiry session data is critical inquiry data and that the inquiry session data is not critical inquiry data. The expert recognition results also include recognition results that the inquiry session data is critical inquiry data and that the inquiry session data is not critical inquiry data. When the model identification result and the expert identification result are both the inquiry session data and the critical inquiry data, the target identification result that the inquiry session data is the critical inquiry data can be obtained. If at least one of the model identification result and the expert identification result is that the inquiry session data is not critical inquiry data, then a target identification result can be obtained that the inquiry session data is not critical inquiry data.

In one embodiment, the expert identification result may also include a case where it is not identified whether the inquiry session data is critical inquiry data. At this time, the model recognition result may be used as the target recognition result.

After the inquiry session data of the user are obtained, on one hand, the inquiry session data are input into a prediction model to obtain a model identification result, on the other hand, the inquiry session data are identified based on an expert knowledge base to obtain an expert identification result, and then the model identification result and the expert identification result are combined to obtain a final identification result. Because the prediction model is obtained based on multi-dimensional characteristic data training, and the multi-dimensional characteristic data comprises entity characteristic data and entity relation characteristic data, the prediction model can learn information with different dimensions in the training process, and better understand the logic of language by combining the context semantic environment, thereby improving the recognition capability of the prediction model on the data of the symptom inquiry; the method combines model prediction and an expert system to identify the critical illness inquiry data, can make up for the defect of only depending on the expert system, and improves the accuracy of identification of the critical illness inquiry data.

In one embodiment, after obtaining the target identification result of whether the inquiry session data is critical inquiry data, the computer device may execute an operation corresponding to the target identification result according to the target identification result.

In a specific embodiment, the method for identifying critical illness inquiry data based on artificial intelligence further comprises the following steps: when the obtained inquiry session data is the target identification result of the critical inquiry data, accessing the inquiry session to which the inquiry session data belongs to a doctor terminal; and when the obtained inquiry session data is not the target identification result of the critical inquiry data, continuing to advance the inquiry session to which the inquiry session data belongs.

In particular, FIG. 3 illustrates a block flow diagram of the use of a predictive model in one embodiment. Referring to the figure, after acquiring the inquiry session data, the computer device can input the inquiry session data into the prediction model and the expert system in parallel, on one hand, the computer device identifies whether the inquiry session data is critical inquiry data through the prediction model to obtain a model identification result, and on the other hand, the computer device identifies whether the inquiry session data is critical inquiry data through the expert system to obtain an expert identification result. And then, combining the model identification result and the expert identification result through a decision maker to obtain whether the inquiry session data is the target identification result of the critical inquiry data. When the obtained inquiry session data is the target identification result of the critical inquiry data, the inquiry session to which the inquiry session data belongs is accessed to the doctor terminal, the doctor manually intervenes through the doctor terminal to examine the final identification result of the critical inquiry data, and when the identification is correct, the doctor can timely perform further processing, such as giving a diagnosis suggestion and the like. When the obtained inquiry session data is not the target identification result of the critical inquiry data, the inquiry session to which the inquiry session data belongs is continuously advanced, for example, the inquiry session is continuously interacted with the user through the artificial intelligent robot to conduct inquiry.

In this embodiment, when different target recognition results are obtained, corresponding next operation is immediately performed, so that a user who needs help urgently can be effectively responded when the inquiry session data is critical inquiry data, and online inquiry can be orderly continued when the inquiry session data is not critical inquiry data.

In one embodiment, the interrogation session data is stored in a blockchain. It is emphasized that, to further ensure privacy and security of the interrogation session data, the interrogation session data may also be stored in a blockchain node.

With regard to the details of the training steps of the prediction model involved in the foregoing embodiments, reference may be made to the detailed description in the following embodiments.

In one embodiment, the method for identifying critical illness inquiry data based on artificial intelligence further comprises the following steps: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data and entity relation characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the multi-dimensional feature data further comprises intent feature data; the critical illness inquiry data identification method based on artificial intelligence further comprises the following steps: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

Specifically, the computer device may collect historical interrogation data and training labels corresponding to the historical interrogation data. The training label can be a result of manual labeling, and represents whether the historical inquiry data is critical inquiry data or not. The computer equipment can extract entity characteristic data and entity relation characteristic data from the historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data, wherein the multi-dimensional characteristic data at least comprises two characteristic dimensions, and the multi-dimensional characteristic data and the historical inquiry data are jointly used as input data of a prediction model to be trained. The computer equipment can also extract entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data, wherein the multi-dimensional characteristic data at least comprises three characteristic dimensions, and the multi-dimensional characteristic data and the historical inquiry data are jointly used as input data of a prediction model to be trained. Wherein, generating the multi-dimensional feature data based on the plurality of feature data may be stitching or fusing the plurality of feature data.

And then, the computer equipment can obtain a prediction recognition result output by the prediction model to be trained, then a training loss function is constructed according to the difference between the prediction recognition result and the training label, parameters of the prediction model are optimized by adopting a back propagation algorithm according to the direction of minimizing the training loss function, and the weights and threshold values of all the obtained hidden neurons and input/output neurons can be obtained after training is finished, so that a model parameter file of the trained prediction model is obtained. The computer device may store the model parameter file in an expert knowledge base.

In this embodiment, when designing input data of a prediction model, a computer device fuses multiple aspects of information, for example, two feature dimensions, namely an entity dimension and an entity relationship dimension, of the data are fused to be used as input data of the prediction model, so that the prediction model can learn at least effective information of the two feature dimensions during training, and the recognition capability of the model for critical care data is improved.

In addition, the intentions of inquiry and answer of doctors and users can also be taken as the basis for identifying the critical care data, so that intention characteristics can be introduced when input data of a prediction model are designed, and the three characteristic dimensions of the entity dimension, the entity relationship dimension and the intention dimension are fused, so that the prediction model can learn at least effective information of the three characteristic dimensions in training, the identification capability of the model on the critical care data is improved, and the identification direction of the critical care data can be expanded.

It can be understood that in the actual use process, the feature extraction mode is many, and mainly proceeds from the following aspects:

in one embodiment, extracting entity feature data, entity relationship feature data and intention feature data from historical inquiry data to generate multi-dimensional feature data corresponding to the historical inquiry data includes: adopting a rule engine to drive and extract entity characteristic data corresponding to historical inquiry data; carrying out named entity identification on historical inquiry data by adopting a sequence labeling model, and extracting entity relation characteristic data among the named entities based on the identified named entities; extracting intention characteristic data in historical inquiry data by adopting a semantic model; and generating multi-dimensional characteristic data corresponding to the historical inquiry data according to the entity characteristic data, the entity relation characteristic data and the intention characteristic data.

(1) In the aspect of extracting the entity feature data, in a specific embodiment, the extracting the entity feature data corresponding to the historical inquiry data by using the rule engine drive comprises: performing word segmentation on the historical inquiry data to obtain word segmentation results; and based on the entity labels provided by the expert knowledge base, adopting a rule engine drive to extract the entity labels according to the word segmentation result so as to obtain entity characteristic data corresponding to the historical inquiry data.

Specifically, the computer device can perform word segmentation on the historical inquiry data to obtain word segmentation results, and the entity feature data is obtained by adopting a rule engine to extract entity labels from the word segmentation results according to the entity labels provided by the expert knowledge base according to rules in a rule set of the expert knowledge base. Such as: entity label: entity feature data → patient: infant, symptoms: vomiting, no weight gain, etc. A rule engine is a component embedded in an application that accepts data input, interprets rules, and makes decisions based on the rules.

Therefore, when the entity feature data are extracted, the effective data in the expert knowledge base are effectively utilized, the efficiency and effectiveness of feature extraction are improved, and unnecessary training time for long-tailed words is avoided.

(2) In terms of entity relationship feature data extraction, in a particular embodiment, the computer device may group historical inquiry data into inquiry sessions, generate an inquiry data sequence from the historical inquiry data of one inquiry session in response order, then perform named entity recognition on the inquiry data sequence using a sequence tagging model, and extract entity relationship feature data between the named entities based on the recognized named entities. Such as naming the entity (alcohol, cause, abdominal pain), the entity relationship is extracted: the cause of abdominal pain is alcohol consumption. Therefore, when named entity identification and entity relation extraction are carried out, the processing mode based on the sequence can effectively combine context information, more accurate relation extraction between entities can be carried out, and the method has wide practical significance.

(3) In the aspect of intention feature data extraction, in a specific embodiment, the intention feature data in the historical inquiry data is extracted by adopting a semantic model, and the method comprises the following steps: and extracting the intention labels corresponding to the historical inquiry data by adopting a semantic model based on the intention labels provided by the expert knowledge base to obtain intention characteristic data corresponding to the historical inquiry data.

In particular, the computer device may also employ a semantic model in the third aspect to extract intent features of historical interrogation data during the stage of training the predictive model. The computer equipment can acquire historical inquiry data as a sample, manually label intention category labels (intention labels provided by an expert knowledge base), supervise and train the semantic model, and then extract intention characteristics of the historical inquiry data by utilizing the trained semantic model.

In one embodiment, the method for identifying critical illness inquiry data based on artificial intelligence further comprises the following steps: screening out historical critical illness inquiry data from the historical inquiry data; counting the occurrence frequency of entity keywords in historical critical inquiry data; screening entity keywords with the occurrence frequency higher than a preset threshold value to serve as entity labels to be added into an expert knowledge base; determining an inquiry intention corresponding to the historical critical inquiry data to obtain an intention label; the intention tag is added to the expert knowledge base.

In particular, the computer device may also establish an expert knowledge base. The computer equipment can firstly acquire critical illness inquiry data identified in the historical inquiry data to obtain a critical illness sample data set; and then counting the occurrence frequency of entity keywords such as causes, objects, symptoms and the like in the severe sample data set, and screening the entity keywords with the occurrence frequency higher than a preset threshold value to be used as entity labels to be added into an expert knowledge base. In addition, the computer equipment can also determine the inquiry intention corresponding to the historical critical inquiry data to obtain an intention label; the intention tag is added to the expert knowledge base.

Therefore, when the multidimensional characteristics of the training samples are extracted during the training of the prediction model, the multidimensional characteristics of the training samples can be extracted by accurately summarizing and summarizing the labels by using an expert system, unnecessary training time consumed for long-tailed words is avoided, and the training time of machine learning model training is greatly prolonged.

For example, FIG. 4 shows a block flow diagram of training a predictive model in one embodiment. Referring to the figure, it can be seen that the computer device may first perform data preparation, i.e., collect historical inquiry data and training labels corresponding to the historical inquiry data, and then extract entity feature data from the historical inquiry data based on the expert system and various data provided by the expert knowledge base, extract entity relationship feature data from the historical inquiry data based on semantic understanding, and extract intention feature data from the historical inquiry data in combination with various data and semantic understanding provided by the expert knowledge base. And the computer equipment fuses the feature data extracted from the three aspects, and then constructs a neural network structure by combining historical inquiry data serving as input data, and trains to obtain a prediction model for identifying the critical inquiry data.

It should be understood that, although the steps in the flowcharts of the above embodiments are shown in order as indicated by the arrows, the steps are not necessarily performed in order as indicated by the arrows. The steps are not performed in the exact order shown and described, and may be performed in other orders, unless explicitly stated otherwise. Moreover, at least a part of the steps in the flowcharts of the above embodiments may include multiple steps or multiple stages, which are not necessarily performed at the same time, but may be performed at different times, and the order of performing the steps or stages is not necessarily sequential, but may be performed alternately or alternately with other steps or at least a part of the steps or stages in other steps.

In one embodiment, as shown in fig. 5, there is provided an artificial intelligence based critical care data identification apparatus comprising: an obtaining module 501, a first identifying module 502, a second identifying module 503 and a decision module 504, wherein:

an obtaining module 501, configured to obtain inquiry session data corresponding to a target user identifier;

a first identification module 502, configured to input the inquiry session data into a prediction model, and output a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extracting according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data;

a second identification module 503, configured to determine, according to a tag hit by the inquiry session data in an expert knowledge base, an expert identification result corresponding to the inquiry session data;

and the decision module 504 is configured to obtain, by combining the model identification result and the expert identification result, a target identification result of whether the inquiry session data is critical inquiry data.

In one embodiment, the first identification module 502 is further configured to input the inquiry session data into a prediction model, and process the inquiry session data through a plurality of neurons included in the prediction model to obtain a model identification result corresponding to the inquiry session data; outputting a model identification result through a prediction model; the threshold value of each neuron and the weight of the connection relation between the neurons are determined when the prediction model is trained through the multi-dimensional feature data and the training labels corresponding to the multi-dimensional feature data, and the multi-dimensional feature data further comprises intention feature data.

As FIG. 6, in one embodiment, the multi-dimensional feature data also includes intent feature data; severe inquiry data recognition device based on artificial intelligence still includes: a training module 505, configured to collect historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the training module 505 is further configured to use a rule engine to drive and extract entity feature data corresponding to historical inquiry data; carrying out named entity identification on historical inquiry data by adopting a sequence labeling model, and extracting entity relation characteristic data among the named entities based on the identified named entities; extracting intention characteristic data in historical inquiry data by adopting a semantic model; and generating multi-dimensional characteristic data corresponding to the historical inquiry data according to the entity characteristic data, the entity relation characteristic data and the intention characteristic data.

In one embodiment, the training module 505 is further configured to perform word segmentation on the historical inquiry data to obtain word segmentation results; based on the entity labels provided by the expert knowledge base, adopting a rule engine to drive and extract the entity labels according to the word segmentation result so as to obtain entity characteristic data corresponding to the historical inquiry data; and extracting the intention labels corresponding to the historical inquiry data by adopting a semantic model based on the intention labels provided by the expert knowledge base to obtain intention characteristic data corresponding to the historical inquiry data.

In one embodiment, the training module 505 is further configured to screen historical critical care data from historical interrogation data; counting the occurrence frequency of entity keywords in historical critical inquiry data; screening entity keywords with the occurrence frequency higher than a preset threshold value to serve as entity labels to be added into an expert knowledge base; determining an inquiry intention corresponding to the historical critical inquiry data to obtain an intention label; the intention tag is added to the expert knowledge base.

In one embodiment, the interrogation session data is stored in a blockchain; the decision module 504 is further configured to access the inquiry session to which the inquiry session data belongs to the doctor terminal when the obtained inquiry session data is the target identification result of the critical care inquiry data; and when the obtained inquiry session data is not the target identification result of the critical inquiry data, continuing to advance the inquiry session to which the inquiry session data belongs.

After the critical inquiry data identification device based on artificial intelligence obtains inquiry session data of a user, the inquiry session data is input into a prediction model to obtain a model identification result, the inquiry session data is identified based on an expert knowledge base to obtain an expert identification result, and a final identification result is obtained by combining the model identification result and the expert identification result. Because the prediction model is obtained based on multi-dimensional characteristic data training, and the multi-dimensional characteristic data comprises entity characteristic data and entity relation characteristic data, the prediction model can learn information with different dimensions in the training process, and better understand the logic of language by combining the context semantic environment, thereby improving the recognition capability of the prediction model to the critical illness inquiry data based on artificial intelligence; the method combines model prediction and an expert system to identify the critical illness inquiry data, can make up for the defect of only depending on the expert system, and improves the accuracy of the critical illness inquiry data identification based on artificial intelligence.

For specific limitations of the intensive care inquiry data identification device based on artificial intelligence, reference may be made to the above limitations of the intensive care inquiry data identification method based on artificial intelligence, and details thereof are not repeated here. All or part of the modules in the above-mentioned intensive care inquiry data identification device based on artificial intelligence can be realized by software, hardware and combination thereof. The modules can be embedded in a hardware form or independent from a processor in the computer device, and can also be stored in a memory in the computer device in a software form, so that the processor can call and execute operations corresponding to the modules.

In one embodiment, a computer device is provided, which may be a server, the internal structure of which may be as shown in fig. 7. The computer device includes a processor, a memory, a network interface, and a database connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system, a computer program, and a database. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The database of the computer device is used for storing critical care inquiry data identification data based on artificial intelligence. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method for identifying critical care data based on artificial intelligence.

In one embodiment, a computer device is provided, which may be a terminal, and its internal structure diagram may be as shown in fig. 8. The computer device includes a processor, a memory, a network interface, a display screen, and an input device connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device comprises a nonvolatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of an operating system and computer programs in the non-volatile storage medium. The network interface of the computer device is used for communicating with an external terminal through a network connection. The computer program is executed by a processor to implement a method for identifying critical care data based on artificial intelligence. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, a key, a track ball or a touch pad arranged on the shell of the computer equipment, an external keyboard, a touch pad or a mouse and the like.

It will be appreciated by those skilled in the art that the configurations shown in fig. 7 or 8 are only block diagrams of some of the configurations relevant to the present application, and do not constitute a limitation on the computer apparatus to which the present application is applied, and a particular computer apparatus may include more or less components than those shown in the drawings, or may combine some components, or have a different arrangement of components.

In one embodiment, there is provided a computer device comprising a memory storing a computer program and a processor implementing the following steps when the processor executes the computer program: acquiring inquiry session data corresponding to the target user identification; inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extraction according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data; determining an expert identification result corresponding to the inquiry session data according to the tags hit by the inquiry session data in the expert knowledge base; and combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data.

In one embodiment, inputting the data of the inquiry session into a prediction model, and outputting a model identification result corresponding to the data of the inquiry session through the prediction model, comprises: inputting the inquiry session data into a prediction model, and processing the inquiry session data through a plurality of neurons included in the prediction model to obtain a model identification result corresponding to the inquiry session data; outputting a model identification result through a prediction model; the threshold value of each neuron and the weight of the connection relation between the neurons are determined when the prediction model is trained through the multi-dimensional feature data and the training labels corresponding to the multi-dimensional feature data, and the multi-dimensional feature data further comprises intention feature data.

In one embodiment, the multi-dimensional feature data further comprises intent feature data. The processor, when executing the computer program, further performs the steps of: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the method for extracting entity feature data corresponding to historical inquiry data by adopting rule engine driving comprises the following steps: performing word segmentation on the historical inquiry data to obtain word segmentation results; and based on the entity labels provided by the expert knowledge base, adopting a rule engine drive to extract the entity labels according to the word segmentation result so as to obtain entity characteristic data corresponding to the historical inquiry data. Adopting a semantic model to extract intention characteristic data in historical inquiry data, comprising the following steps: and extracting the intention labels corresponding to the historical inquiry data by adopting a semantic model based on the intention labels provided by the expert knowledge base to obtain intention characteristic data corresponding to the historical inquiry data.

In one embodiment, the processor, when executing the computer program, further performs the steps of: screening out historical critical illness inquiry data from the historical inquiry data; counting the occurrence frequency of entity keywords in historical critical inquiry data; screening entity keywords with the occurrence frequency higher than a preset threshold value to serve as entity labels to be added into an expert knowledge base; determining an inquiry intention corresponding to the historical critical inquiry data to obtain an intention label; the intention tag is added to the expert knowledge base.

In one embodiment, the interrogation session data is stored in a blockchain; the processor, when executing the computer program, further performs the steps of: when the obtained inquiry session data is the target identification result of the critical inquiry data, accessing the inquiry session to which the inquiry session data belongs to a doctor terminal; and when the obtained inquiry session data is not the target identification result of the critical inquiry data, continuing to advance the inquiry session to which the inquiry session data belongs.

In one embodiment, a computer storage medium is provided, having a computer program stored thereon, the computer program, when executed by a processor, implementing the steps of: acquiring inquiry session data corresponding to the target user identification; inputting the inquiry session data into a prediction model, and outputting a model identification result corresponding to the inquiry session data through the prediction model; the prediction model is obtained by training according to multi-dimensional feature data and training labels corresponding to the multi-dimensional feature data, the multi-dimensional feature data is obtained by extraction according to historical inquiry data, and the multi-dimensional feature data comprises entity feature data and entity relation feature data; determining an expert identification result corresponding to the inquiry session data according to the tags hit by the inquiry session data in the expert knowledge base; and combining the model identification result and the expert identification result to obtain whether the inquiry session data is the target identification result of the critical inquiry data.

In one embodiment, the multi-dimensional feature data further comprises intent feature data; the computer program when executed by the processor further realizes the steps of: collecting historical inquiry data and training labels corresponding to the historical inquiry data; the training labels corresponding to the historical inquiry data are used for indicating whether the historical inquiry data are critical inquiry data or not; extracting entity characteristic data, entity relation characteristic data and intention characteristic data from historical inquiry data to generate multi-dimensional characteristic data corresponding to the historical inquiry data; inputting the historical inquiry data and the multi-dimensional characteristic data corresponding to the historical inquiry data into a prediction model to be trained together to obtain a prediction recognition result; and training the prediction model based on the prediction recognition result of the prediction model and the training label.

In one embodiment, the computer program when executed by the processor further performs the steps of: screening out historical critical illness inquiry data from the historical inquiry data; counting the occurrence frequency of entity keywords in historical critical inquiry data; screening entity keywords with the occurrence frequency higher than a preset threshold value to serve as entity labels to be added into an expert knowledge base; determining an inquiry intention corresponding to the historical critical inquiry data to obtain an intention label; the intention tag is added to the expert knowledge base.

In one embodiment, the interrogation session data is stored in a blockchain; the computer program when executed by the processor further realizes the steps of: when the obtained inquiry session data is the target identification result of the critical inquiry data, accessing the inquiry session to which the inquiry session data belongs to a doctor terminal; and when the obtained inquiry session data is not the target identification result of the critical inquiry data, continuing to advance the inquiry session to which the inquiry session data belongs.

The block chain is a novel application mode of computer technologies such as distributed data storage, point-to-point transmission, a consensus mechanism, an encryption algorithm and the like. A block chain (Blockchain), which is essentially a decentralized database, is a series of data blocks associated by using a cryptographic method, and each data block contains information of a batch of network transactions, so as to verify the validity (anti-counterfeiting) of the information and generate a next block. The blockchain may include a blockchain underlying platform, a platform product service layer, an application service layer, and the like.

It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by hardware instructions of a computer program, which can be stored in a non-volatile computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. Any reference to memory, storage, database, or other medium used in the embodiments provided herein may include non-volatile and/or volatile memory, among others. Non-volatile memory can include read-only memory (ROM), Programmable ROM (PROM), Electrically Programmable ROM (EPROM), Electrically Erasable Programmable ROM (EEPROM), or flash memory. Volatile memory can include Random Access Memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in a variety of forms such as Static RAM (SRAM), Dynamic RAM (DRAM), Synchronous DRAM (SDRAM), Double Data Rate SDRAM (DDRSDRAM), Enhanced SDRAM (ESDRAM), Synchronous Link DRAM (SLDRAM), Rambus Direct RAM (RDRAM), direct bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).

The technical features of the above embodiments can be arbitrarily combined, and for the sake of brevity, all possible combinations of the technical features in the above embodiments are not described, but should be considered as the scope of the present specification as long as there is no contradiction between the combinations of the technical features.

The above-mentioned embodiments only express several embodiments of the present application, and the description thereof is more specific and detailed, but not construed as limiting the scope of the invention. It should be noted that, for a person skilled in the art, several variations and modifications can be made without departing from the concept of the present application, which falls within the scope of protection of the present application. Therefore, the protection scope of the present patent shall be subject to the appended claims.

Claims

1. An artificial intelligence based critical illness data identification method, characterized in that the method comprises:

acquiring inquiry session data corresponding to the target user identification;

2. The method of claim 1, wherein inputting the interrogation session data into a predictive model and outputting a model identification corresponding to the interrogation session data via the predictive model comprises:

outputting the model identification result through the prediction model;

3. The method of claim 1, wherein the multi-dimensional feature data further comprises intent feature data; the method further comprises the following steps:

4. The method according to claim 3, wherein the extracting entity feature data, entity relationship feature data and intention feature data from the historical inquiry data to generate multi-dimensional feature data corresponding to the historical inquiry data comprises:

5. The method according to claim 4, wherein the extracting entity feature data corresponding to the historical inquiry data by using a rule engine driver comprises:

6. The method of claim 5, further comprising:

adding the intention tag to the expert knowledge base.

7. The method of any one of claims 1-6, wherein the interrogation session data is stored in a blockchain;

the method further comprises the following steps:

8. An apparatus for identifying critical illness data based on artificial intelligence, the apparatus comprising:

9. A computer device comprising a memory and a processor, the memory storing a computer program, wherein the processor when executing the computer program implements the steps of the method of any one of claims 1 to 6 or 7.

10. A computer storage medium on which a computer program is stored, which computer program, when being executed by a processor, carries out the steps of the method of any one of claims 1 to 6 or 7.