State detection method and device, computer device, and storage medium

Abstract

Embodiments of the present application disclose a state detection method and device, computer equipment and a storage medium, belonging to the technical field of computers. The method comprises: obtaining description information, the description information comprising a plurality of state description information and a type text corresponding to each state description information, the type text corresponding to the state description information indicating the type to which the state description information belongs; combining each state description information and the corresponding type text respectively to obtain a text segment; splicing the obtained plurality of text segments to obtain a target text; and performing state detection on the target text to obtain a state word. The method not only realizes the fusion of a plurality of state description information into a target text, but also can distinguish different types of state description information according to the type text in the text segment. Therefore, the state detection on the target text can comprehensively consider a plurality of types of state description information, which is conducive to improving the accuracy of state detection.

Description

Technical Field

[0001] This application relates to the field of computer technology, and in particular to a state detection method, apparatus, computer device, and storage medium. Background Technology

[0002] With the development of computer technology, semantic recognition technology is being applied more and more widely. For example, in medical scenarios, semantic recognition technology is used to identify the state description information input by the user and obtain state words, which can represent the state described by the state description information.

[0003] In related technologies, a knowledge database is created, which includes the relationships between state description information and state terms. Therefore, a query can be performed in this knowledge database to find the state terms associated with a given state description, thereby determining the state described by that description. However, because this method relies heavily on the quality of the knowledge database, the state detection performance is often unsatisfactory. Summary of the Invention

[0004] This application provides a state detection method, apparatus, computer device, and storage medium, which can improve the accuracy of state detection. The technical solution is as follows:

[0005] On the one hand, a state detection method is provided, the method comprising:

[0006] Obtain description information, which includes multiple state descriptions and a type text corresponding to each state description, wherein the type text corresponding to the state description indicates the type to which the state description belongs;

[0007] Each state description and its corresponding type text are combined to obtain a text fragment;

[0008] The obtained text fragments are concatenated to obtain the target text;

[0009] The target text is subjected to state detection to obtain state words, which represent the state described by the descriptive information.

[0010] Optionally, the description information includes multiple fields, and before combining each state description information and its corresponding type text to obtain a text fragment, the method further includes:

[0011] For any one of the multiple fields, obtain the status description information in that field, and determine the field name of that field as the type text corresponding to the status description information in that field.

[0012] Optionally, the status description information includes at least one status description entry, each status description entry including physical examination item type text and physical examination result information, wherein the physical examination item type text indicates the type of physical examination item; the step of mapping the status description information to physical examination result description terms includes:

[0013] Each status description entry is mapped to a physical examination result description word to obtain at least one physical examination result description word.

[0014] Optionally, the step of performing state detection on the target text to obtain the predicted probabilities corresponding to multiple reference state words includes:

[0015] Feature extraction is performed on the target text to obtain the text features corresponding to the target text;

[0016] The text features are transformed to obtain the predicted probabilities corresponding to the multiple reference state words.

[0017] Optionally, the method further includes:

[0018] If the difference parameter is less than the second threshold, the state detection is determined to have failed.

[0019] Optionally, determining the difference parameter between the first predicted probability and the second predicted probability includes:

[0020] Determine the absolute value of the difference between the first predicted probability and the second predicted probability;

[0021] Determine the sum of the first predicted probability and the second predicted probability;

[0022] The ratio between the absolute value and the sum is determined as the difference parameter.

[0023] Optionally, determining the state word with the highest predicted probability among the plurality of reference state words includes:

[0024] Based on the confidence prediction model, the prediction probabilities corresponding to the multiple reference state words are predicted according to the target text, and the prediction results are obtained. The prediction results indicate whether the prediction probabilities corresponding to the multiple reference state words are accurate.

[0025] If the prediction result indicates that the prediction probability corresponding to the plurality of reference state words is accurate, then the state word corresponding to the maximum prediction probability is determined from the plurality of reference state words.

[0026] Optionally, the state detection model includes multiple reference state words, and the state detection of the target text based on the state detection model to obtain the state words includes:

[0027] Based on the state detection model, the target text is state detected to obtain the predicted probabilities corresponding to the multiple reference state words;

[0028] Among the plurality of reference state words, the state word corresponding to the highest predicted probability is determined.

[0029] Optionally, the step of performing state detection on the target text based on the state detection model to obtain the predicted probabilities corresponding to the plurality of reference state words includes:

[0030] Based on the feature extraction network in the state detection model, feature extraction is performed on the target text to obtain the text features corresponding to the target text.

[0031] Based on the feature transformation network in the state detection model, the text features are transformed to obtain the predicted probabilities corresponding to the multiple reference state words.

[0032] Optionally, training the state detection model based on the sample prediction probabilities corresponding to the plurality of reference state words and the sample state words includes:

[0033] The target value is determined as the true probability corresponding to the sample state word;

[0034] The loss value of the state detection model is determined based on the true probability corresponding to the sample state word and the sample prediction probability corresponding to the same reference state word.

[0035] The state detection model is trained based on the loss value.

[0036] On the other hand, a state detection device is provided, the device comprising:

[0037] The information acquisition module is used to acquire description information, which includes multiple state description information and a type text corresponding to each state description information. The type text corresponding to the state description information indicates the type to which the state description information belongs.

[0038] The combination module is used to combine each state description information and the corresponding type text to obtain a text fragment;

[0039] The concatenation module is used to concatenate multiple text fragments to obtain the target text;

[0040] The state detection module is used to perform state detection on the target text to obtain state words, which represent the state described by the descriptive information.

[0041] Optionally, the description information includes multiple fields, and the device further includes:

[0042] The information and field determination module is used to obtain the status description information of any field among the plurality of fields, and determine the field name of any field as the type text corresponding to the status description information of any field.

[0043] Optionally, the combined module includes:

[0044] A tag adding unit is used to add a target tag to the type text corresponding to each state description information, wherein the target tag is used to indicate that the text with the target tag added is type text;

[0045] The combination unit is used to combine the type text with the target mark added and the state description information to obtain the text fragment corresponding to the state description information.

[0046] Optionally, the target marker includes a start marker and an end marker, and the marker adding unit is used to add the start marker and the end marker to the type text respectively, to obtain the type text with the start marker added and the type text with the end marker added;

[0047] The combining unit is used to sequentially concatenate the type text with the start mark, the state description information, and the type text with the end mark to obtain the text fragment corresponding to the state description information.

[0048] Optionally, the combined module includes:

[0049] A parameter mapping unit is used to map the parameters into parameter description words according to the type text corresponding to the state description information when the state description information includes parameters. The parameter description words are used to describe the parameters belonging to the type.

[0050] The combination unit is used to combine the parameter description words and the type text corresponding to the state description information to obtain the text fragment corresponding to the state description information.

[0051] Optionally, the combined module includes:

[0052] The information mapping unit is used to map the status description information into physical examination result description words when the type text corresponding to the status description information represents the physical examination type. The physical examination result description words are used to describe the physical examination results.

[0053] The combination unit is used to combine the descriptive words of the physical examination results and the type text corresponding to the status description information to obtain the text fragment corresponding to the status description information.

[0054] Optionally, the status description information includes at least one status description entry, each status description entry including physical examination item type text and physical examination result information, wherein the physical examination item type text indicates the type of physical examination item; the information mapping unit is used to map each status description entry into a physical examination result description word respectively, to obtain at least one physical examination result description word.

[0055] Optionally, the state description information is information describing the state of the object, and the combining module includes:

[0056] The statement recognition unit is used to recognize the state description information when the type text corresponding to the state description information represents the symptom type, and to obtain the target symptom description statement in the state description information, wherein the target symptom description statement is used to describe the symptom that does not exist in the object;

[0057] The statement acquisition unit is used to acquire statements other than the target symptom description statement in the status description information;

[0058] The combination unit is used to combine the other statements and the type text corresponding to the state description information to obtain the text fragment corresponding to the state description information.

[0059] Optionally, the combined module further includes:

[0060] The symptom word acquisition unit is used to acquire symptom words from the target symptom description statement;

[0061] A type text determination unit is used to determine a target type text, wherein the target type text represents a symptom type in which the object does not exist;

[0062] The combination unit is used to combine the symptom words and the target type text to obtain the text fragment corresponding to the state description information.

[0063] Optionally, the state detection module includes:

[0064] The prediction probability detection unit is used to perform state detection on the target text and obtain the prediction probability corresponding to multiple reference state words. The prediction probability corresponding to each reference state word represents the probability that the reference state word is the state word corresponding to the description information.

[0065] The state word determination unit is used to determine the state word with the highest predicted probability among the plurality of reference state words.

[0066] Optionally, the prediction probability detection unit is used for:

[0067] Feature extraction is performed on the target text to obtain the text features corresponding to the target text;

[0068] The text features are transformed to obtain the predicted probabilities corresponding to the multiple reference state words.

[0069] Optionally, the state word determination unit is used to:

[0070] Determine a first prediction probability, which is the maximum prediction probability among the prediction probabilities corresponding to the plurality of reference state words;

[0071] If the first predicted probability is greater than the first threshold, the state word corresponding to the first predicted probability is determined from the plurality of reference state words.

[0072] Optionally, the device further includes:

[0073] The second prediction probability determination module is used to determine a second prediction probability when the first prediction probability is not greater than the first threshold. The second prediction probability is the maximum prediction probability among a plurality of prediction probabilities other than the first prediction probability.

[0074] The difference parameter determination module is used to determine the difference parameter between the first predicted probability and the second predicted probability;

[0075] The state detection module is further configured to determine the state word corresponding to the first predicted probability among the plurality of reference state words, provided that the difference parameter is not less than the second threshold.

[0076] Optionally, the device further includes:

[0077] The detection failure determination module is used to determine that the state detection has failed when the difference parameter is less than the second threshold.

[0078] Optionally, the difference parameter determination module includes:

[0079] An absolute value determination unit is used to determine the absolute value of the difference between the first predicted probability and the second predicted probability;

[0080] A sum determination unit is used to determine the sum of the first predicted probability and the second predicted probability;

[0081] The difference parameter determination unit is used to determine the ratio between the absolute value and the sum as the difference parameter.

[0082] Optionally, the state word determination unit is used to:

[0083] Based on the confidence prediction model, the prediction probabilities corresponding to the multiple reference state words are predicted according to the target text, and the prediction results are obtained. The prediction results indicate whether the prediction probabilities corresponding to the multiple reference state words are accurate.

[0084] If the prediction result indicates that the prediction probability corresponding to the plurality of reference state words is accurate, then the state word corresponding to the maximum prediction probability is determined from the plurality of reference state words.

[0085] Optionally, the state detection module includes:

[0086] The state detection unit is used to perform state detection on the target text based on the state detection model to obtain the state words.

[0087] Optionally, the state detection model includes multiple reference state terms, and the state detection unit is used for:

[0088] Based on the state detection model, the target text is state detected to obtain the predicted probabilities corresponding to the multiple reference state words;

[0089] Among the plurality of reference state words, the state word corresponding to the highest predicted probability is determined.

[0090] Optionally, the state detection unit is used for:

[0091] Based on the feature extraction network in the state detection model, feature extraction is performed on the target text to obtain the text features corresponding to the target text.

[0092] Based on the feature transformation network in the state detection model, the text features are transformed to obtain the predicted probabilities corresponding to the multiple reference state words.

[0093] Optionally, the device further includes:

[0094] The sample acquisition module is used to acquire sample text and corresponding sample state words, wherein the sample state words represent the state described by the sample text.

[0095] The state detection module is also used to perform state detection on the sample text based on the state detection model to obtain the sample prediction probability corresponding to multiple reference state words.

[0096] The model training module is used to train the state detection model based on the sample prediction probabilities corresponding to the multiple reference state words and the sample state words.

[0097] Optionally, the model training module includes:

[0098] The true probability determination unit is used to determine the target value as the true probability corresponding to the sample state word;

[0099] The loss value determination unit is used to determine the loss value of the state detection model based on the true probability corresponding to the sample state word and the sample prediction probability corresponding to the same reference state word as the sample state word.

[0100] The model training unit is used to train the state detection model based on the loss value.

[0101] On the other hand, a computer device is provided, the computer device including a processor and a memory, the memory storing at least one computer program, the at least one computer program being loaded and executed by the processor to perform the operations performed in the state detection method as described above.

[0102] On the other hand, a computer-readable storage medium is provided that stores at least one computer program, which is loaded and executed by a processor to perform the operations performed in the state detection method described above.

[0103] On the other hand, a computer program product or computer program is provided, the computer program product or computer program including computer program code stored in a computer-readable storage medium, a processor of a computer device reading the computer program code from the computer-readable storage medium, the processor executing the computer program code, causing the computer device to perform the operations performed in the state detection method described above.

[0104] In the methods, apparatus, computer devices, and storage media provided in this application, the description information includes multiple state description information and corresponding type text. The state description information and the corresponding type text are combined into a text fragment, and then the multiple text fragments are spliced ​​together to form a target text. This not only realizes the fusion of multiple state description information into a single target text, but also distinguishes different types of state description information based on the type text in the text fragment. Therefore, state detection of the target text can comprehensively consider multiple types of state description information, which is beneficial to improving the accuracy of state detection. Attached Figure Description

[0105] To more clearly illustrate the technical solutions in the embodiments of this application, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the embodiments of this application. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0106] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application.

[0107] Figure 2 This is a flowchart of a state detection method provided in an embodiment of this application.

[0108] Figure 3 This is a flowchart of a state detection method provided in an embodiment of this application.

[0109] Figure 4 This is a schematic diagram of a feature fusion method provided by related technologies.

[0110] Figure 5 This is a schematic diagram of an information fusion method provided in an embodiment of this application.

[0111] Figure 6 This is a flowchart of a state detection method provided in an embodiment of this application.

[0112] Figure 7 This is a schematic diagram of a model-based state detection method provided in an embodiment of this application.

[0113] Figure 8 This is a flowchart of a training method for a state detection model provided in an embodiment of this application.

[0114] Figure 9 This is a schematic diagram of the structure of a state detection device provided in an embodiment of this application.

[0115] Figure 10 This is a schematic diagram of another state detection device provided in an embodiment of this application.

[0116] Figure 11 This is a schematic diagram of the structure of a terminal provided in an embodiment of this application.

[0117] Figure 12 This is a schematic diagram of the structure of a server provided in an embodiment of this application. Detailed Implementation

[0118] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0119] It is understood that the terms "first," "second," etc., used in this application may be used to describe various concepts herein, but unless otherwise stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of this application, a first prediction probability may be referred to as a second prediction probability, and similarly, a second prediction probability may be referred to as a first prediction probability.

[0120] "At least one" refers to one or more. For example, at least one state description entry can be one state description entry, two state description entries, three state description entries, or any integer number of state description entries greater than or equal to one. "Multiple" refers to two or more. For example, multiple state description information can be two state description information, three state description information, or any integer number of state description information greater than or equal to two. "Each" refers to each of the at least one entries. For example, each state description information refers to each of the multiple state description information. If the multiple state description information consists of three state description information, then each state description information refers to each of the three state description information.

[0121] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a comprehensive technology within computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.

[0122] Artificial intelligence (AI) is a comprehensive discipline encompassing a wide range of fields, including both hardware and software technologies. Fundamental AI technologies generally include sensors, dedicated AI chips, cloud computing, distributed storage, big data processing, operating / interactive systems, and mechatronics. AI software technologies primarily include computer vision, speech processing, natural language processing, and machine learning / deep learning.

[0123] Natural Language Processing (NLP) is an important field within computer science and artificial intelligence. It studies the theories and methods for enabling effective communication between humans and computers using natural language. NLP is a science that integrates linguistics, computer science, and mathematics. Therefore, research in this field involves natural language—the language people use in daily life—and thus it has a close relationship with linguistic research. NLP techniques typically include text processing, semantic understanding, machine translation, question answering, and knowledge graphs.

[0124] Machine Learning (ML) is a multidisciplinary field involving probability theory, statistics, approximation theory, convex analysis, and algorithm complexity theory. It specifically studies how computers can simulate or implement human learning behavior to acquire new knowledge or skills and reorganize existing knowledge structures to continuously improve their performance. Machine learning is the core of artificial intelligence and the fundamental way to endow computers with intelligence; its applications span all areas of artificial intelligence. Machine learning and deep learning typically include techniques such as artificial neural networks, belief networks, reinforcement learning, transfer learning, inductive learning, and instruction-based learning.

[0125] The solution provided in this application embodiment, based on artificial intelligence technologies such as natural language processing and machine learning, can process descriptive information to obtain target text, and perform state detection on the target text to obtain the state words corresponding to the descriptive information, thereby determining the state described by the descriptive information.

[0126] The state detection method provided in this application embodiment can be used in computer devices. Optionally, the computer device is a terminal or a server. Optionally, the server is an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. Optionally, the terminal is a smartphone, tablet computer, laptop computer, desktop computer, smart speaker, smartwatch, etc., but is not limited to these.

[0127] In one possible implementation, the computer program involved in the embodiments of this application may be deployed and executed on a computer device, or executed on multiple computer devices located in one location, or executed on multiple computer devices distributed in multiple locations and interconnected through a communication network. Multiple computer devices distributed in multiple locations and interconnected through a communication network can form a blockchain system.

[0128] Blockchain is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and cryptographic algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying platform, a platform product service layer, and an application service layer.

[0129] The underlying blockchain platform can include modules for user management, basic services, smart contracts, and operational monitoring. The user management module is responsible for managing the identity information of all blockchain participants, including maintaining public and private key generation (account management), key management, and maintaining the correspondence between user identities and blockchain addresses (access management). Under authorization, it also monitors and audits transactions of certain real identities and provides risk control rule configuration (risk control audit). The basic services module is deployed on all blockchain node devices to verify the validity of business requests. After consensus is reached on valid requests, they are recorded in storage. For a new business request, the basic services first perform interface adaptation parsing and authentication (interface adaptation), and then encrypt the business information using a consensus algorithm (consensus management). After encryption, the data is transmitted completely and consistently to the shared ledger (network communication) and recorded and stored. The smart contract module is responsible for contract registration, issuance, triggering, and execution. Developers can define contract logic using a programming language and publish it to the blockchain (contract registration). According to the contract terms, the key or other events are invoked to trigger execution and complete the contract logic. It also provides functions for contract upgrades and cancellations. The operation and monitoring module is mainly responsible for deployment, configuration modification, contract settings, cloud adaptation, and real-time status visualization output during product release, such as alarms, network status monitoring, and node device health status monitoring.

[0130] The platform's product service layer provides the basic capabilities and implementation frameworks for typical applications. Developers can leverage these basic capabilities, along with the specific characteristics of their business needs, to implement blockchain-based business logic. The application service layer provides blockchain-based application services to business stakeholders.

[0131] Figure 1 This is a schematic diagram of an implementation environment provided in an embodiment of this application. The implementation environment includes multiple terminals 101 ( Figure 1 (Taking three terminals 101 as an example) and a server 102. Terminals 101 and server 102 are connected via a wireless or wired network. A target application provided by server 102 is installed on terminal 101, enabling terminal 101 to perform functions such as data transmission and message interaction. Optionally, the target application can be a target application within the operating system of terminal 101, or a target application provided by a third party. For example, the target application could be a medical consultation assistance application, which assists doctors in making diagnoses. This application could also have other functions, such as commenting, content sharing, and navigation.

[0132] Terminal 101 is used to log in to the target application based on the user identifier, obtain the description information input by the user through the target application, process the description information to obtain the target text, and then send the target text to server 102. Server 102 receives the target text sent by terminal 101, calls the state detection model to perform state detection on the target text, obtains the state words corresponding to the target text, and then returns the state words to terminal 101, so that terminal 101 obtains the state words corresponding to the description information.

[0133] The state detection method provided in this application can be applied to a variety of scenarios.

[0134] For example, in the medical field, descriptive information refers to information describing a user's state. A diagnostic assistance application is installed on the terminal, and the server corresponding to this application is configured with a state detection model. The terminal logs into the diagnostic assistance application based on the user's identifier, obtains the descriptive information input by the user, processes the state description information and its corresponding type text, obtaining the target text concatenated from the state description information and type text. This target text is then sent to the server. Upon receiving the target text, the server performs state detection on it based on the state detection model, obtaining the corresponding state words. These state words represent the state described by the descriptive information. Doctors can then use these state words, combined with other user information, to determine the user's potential illness, thus providing auxiliary decision-making support during clinical diagnosis.

[0135] Figure 2 This is a flowchart of a state detection method provided in an embodiment of this application. The execution subject of this embodiment is a computer device; see [link to relevant documentation]. Figure 2 The method includes:

[0136] 201. Computer devices obtain description information.

[0137] This application provides a method for detecting the state of an object using descriptive information, whereby the descriptive information describes the state of the object. The type of object varies depending on the application scenario. For example, in the medical field, the object is the human body; in the agricultural field, the target object is a plant; in the equipment maintenance field, the target object is electronic equipment or machinery; in the weather forecasting field, the target object is the weather; and in the medical field, the target object is the human body. In addition, the target object can be any other object.

[0138] The description information includes multiple state descriptions and a type text corresponding to each state description. The multiple state descriptions are state descriptions of different types, and the type text corresponding to each state description indicates the type to which the state description belongs.

[0139] 202. The computer device combines each state description information with the corresponding type text to obtain a text fragment.

[0140] The computer device acquires multiple state descriptions and the type text corresponding to each state description from the description information. For each state description, the computer device combines the state description and the type text corresponding to it to obtain the text fragment corresponding to that state description, thus obtaining the text fragment corresponding to each state description.

[0141] 203. The computer device splices together the multiple text fragments obtained to obtain the target text.

[0142] After receiving multiple text fragments, the computer device concatenates these fragments to obtain the target text. Therefore, the target text comprises multiple text fragments, and each text fragment includes a state description and the corresponding type text.

[0143] 204. The computer device performs state detection on the target text and obtains state words.

[0144] After obtaining the target text, the computer device performs state detection on the target text to obtain state words, which represent the state described by the descriptive information.

[0145] In the method provided in this application embodiment, the description information includes multiple state description information and corresponding type text. The state description information and the corresponding type text are combined into a text fragment, and then the multiple text fragments are spliced ​​together to form the target text. This not only realizes the fusion of multiple state description information into a single target text, but also distinguishes different types of state description information based on the type text in the text fragment. Therefore, performing state detection on the target text can comprehensively consider multiple types of state description information, which is beneficial to improving the accuracy of state detection.

[0146] Figure 3 This is a flowchart of a state detection method provided in an embodiment of this application. The execution subject of this embodiment is a computer device; see [link to relevant documentation]. Figure 3 The method includes:

[0147] 301. Computer devices obtain description information.

[0148] This application provides a method for detecting the state of an object using descriptive information, which describes the state of the object. The descriptive information includes multiple state descriptions and a type text corresponding to each state description. These multiple state descriptions represent different types of state descriptions, and the type text corresponding to each state description indicates the type to which that state description belongs.

[0149] The type of object varies depending on the application scenario, and therefore the type of descriptive information also differs. Taking the medical field as an example, if the object is a patient, then the descriptive information is the patient's medical information. Medical information includes multiple different types of status description information. For example, status description information includes status description information used to describe age, gender, chief complaint, medical history, and physical examination, etc. The corresponding type text for status description information includes "age", "gender", "chief complaint", "medical history", and "physical examination", etc.

[0150] Optionally, the computer device may obtain description information input by the user, or the computer device may obtain description information stored locally, or the computer device may obtain description information sent by other devices. This application embodiment does not limit this.

[0151] 302. For any field among multiple fields, the computer device obtains the status description information in that field and determines the field name of that field as the type text corresponding to the status description information in that field.

[0152] The description information includes multiple fields, each with a corresponding field name. The field name indicates the type of content within that field. Multiple status descriptions are stored as fields, with each field containing one status description.

[0153] Therefore, the computer device identifies the fields in the description information. For any identified field, the computer device obtains the status description information in that field and determines the type text corresponding to the status description information in that field by using the field name of that field. Thus, the computer device is able to obtain each status description information in the description information and the type text corresponding to each status description information.

[0154] It should be noted that this embodiment only illustrates the process of obtaining state description information and its corresponding type text by identifying fields, using the storage of state description information in the form of fields as an example. In another embodiment, multiple state description information can be stored in other forms, and the computer device can use other methods to obtain each state description information and its corresponding type text.

[0155] For example, the description information includes multiple information entries, each containing a status description and its corresponding type text. By extracting the information entries from the description information, the computer device can obtain the status description and its corresponding type text. As another example, the description information may include the relationships between the status description and the corresponding type text. The computer device can retrieve the status description and its corresponding type text by querying these relationships.

[0156] 303. The computer device combines each state description information with the corresponding type text to obtain a text fragment.

[0157] After the computer device obtains the status description information and the corresponding type text, it combines the status description information and the corresponding type text to obtain the text fragment corresponding to the status description information, thereby obtaining the text fragment corresponding to each status description information.

[0158] The text fragments include type text and state description information belonging to that type. Therefore, each text fragment corresponds to a type text. By combining the state description information with the corresponding type text into a text fragment, different types of state description information can be distinguished based on the text fragments.

[0159] In one possible implementation, for each state description, the computer device adds a target marker to the type text corresponding to the state description, and combines the target-marked type text with the state description to obtain the text fragment corresponding to the state description. The target marker indicates that the text with the target marker is type text.

[0160] The text fragment includes state description information and type text. To distinguish which part of the text fragment is state description information and which part is type text, the computer device first adds target tags to the type text, and then combines the target-tagged type text and the state description information into a text fragment. Therefore, the text with target tags in the text fragment is the type text. By recognizing the target tags in the text fragment, the computer device can identify the type text in the text fragment, thereby distinguishing between state description information and type text. This eliminates the need for the computer device to distinguish between state description information and type text through semantic recognition, reducing the processing complexity of the computer device and improving the efficiency of subsequent state detection.

[0161] The target mark can be any type of mark. For example, the target mark can be a target symbol in text format, and the computer device adds the target symbol to the text; another example is that the target mark can be a target color, and the computer device sets the text to the target color; yet another example is that the target mark can be a target font, and the computer device sets the text to the target font.

[0162] Optionally, the target marker includes a start marker and an end marker. The computer device adds a start marker and an end marker to the type text, respectively, to obtain type text with the start marker and type text with the end marker. The type text with the start marker, the status description information, and the type text with the end marker are then concatenated to obtain the text fragment corresponding to the status description information.

[0163] The start marker indicates that the text with the start marker is typed text, and that the text with the start marker is the beginning of a text segment, meaning it is located at the beginning of the text segment. The end marker indicates that the text with the end marker is typed text, and that the text with the end marker is the end of a text segment, meaning it is located at the end of the text segment.

[0164] Since the computer device will subsequently concatenate multiple text fragments to obtain the target text, in order to distinguish different text fragments within the target text, the computer device adds start and end markers to the type text. Type text with a start marker is placed at the beginning of the text fragment, and type text with an end marker is placed at the end of the text fragment. Status description information is placed in the middle of the text fragment. The computer device recognizes the start and end markers; the text between one start marker position and the next end marker position belongs to one text fragment, thus identifying a text fragment.

[0165] Therefore, by adding start and end markers to typed text, computer devices can distinguish not only typed text and state description information in text fragments, but also different text fragments in a target text by detecting the start and end markers, further reducing the processing complexity of computer devices.

[0166] Among them, the start tag and the end tag can be tags of any type. For example, the start tag and the end tag are symbols of different types of text formats. For example, the start tag is "<>", the end tag is "", the type text is "gender", and the status description information is "female". Then, after adding the start tag to the type text, the computer device gets "<gender>", and after adding the end tag to the type text, it gets "< / gender>". Then, the type text with the start tag added, the status description information, and the type text with the end tag added are concatenated in sequence, and the resulting text segment is "<gender>female< / gender>".

[0167] In another possible implementation, when the status description information includes a parameter, the computer device maps the parameter to a parameter description word according to the type text corresponding to the status description information, and the parameter description word is used to describe the parameter belonging to this type. The computer device combines the parameter description word and the type text corresponding to the status description information to obtain the text segment corresponding to the status description information.

[0168] Since different parameters have different physical meanings, and in the case of different physical meanings, parameters of the same size may also contain different information. Therefore, when the status description information includes a parameter, if the parameter is directly used as a part of the text segment, it will be difficult for the computer device to capture the information contained in the parameter during subsequent status detection. For the convenience of subsequent processing, the computer device maps the parameter to a parameter description word according to the type text corresponding to the status description information, and then combines the parameter description word with the type text. Since the type text corresponding to the status description information represents the type to which the status description information belongs, when the status description information includes a parameter, this type text represents the type to which the parameter belongs, that is, this type text can represent the physical meaning of the parameter. Therefore, when the computer device maps the parameter according to the type text, the obtained parameter description word can describe the parameter belonging to this type.

[0169] For example, if the type text corresponding to the status description information is "age", then the status description information includes a parameter representing age. According to the size of the parameter representing age, the computer device can map the parameter to parameter description words such as "infant", "toddler", "child", "youth", "middle-aged", and "elderly", so as to divide the parameter representing age into different age stages.

[0170] In another possible implementation, when the type text corresponding to the status description information represents the type of physical examination, the computer device maps the status description information to a physical examination result description word, and the physical examination result description word is used to describe the physical examination result. The computer device combines the physical examination result description word and the type text corresponding to the status description information to obtain the text segment corresponding to the status description information.

[0171] When the type text corresponding to the status description information represents the physical examination type, the type to which the status description information belongs is the physical examination type, meaning the status described by the status description information is the state obtained from the physical examination. However, the form of the status description information will differ for different types of physical examinations. Directly incorporating multiple forms of status description information as part of a text fragment would increase the complexity of subsequent status detection by the computer device. To facilitate subsequent processing, the computer device maps the status description information to physical examination result description terms, and then combines these terms with the type text. By converting the status description information into terms that are easily understood by the computer device, the complexity of subsequent status detection processing by the computer device is reduced.

[0172] Optionally, if the status description information includes at least one status description entry, the computer device maps each status description entry to a physical examination result description word to obtain at least one physical examination result description word.

[0173] Each status description entry includes a text indicating the type of medical examination item and information indicating the examination result. The text indicates the type of medical examination item, and the information indicates the result of that medical examination item. The description of the medical examination result corresponding to the status description entry is used to describe the result of that medical examination item. After obtaining at least one description of the medical examination result, the computer device combines this description with the text indicating the type of medical examination to obtain a text fragment.

[0174] In cases where the status description information includes multiple status description entries, each corresponding to a different medical examination type, and the computer device receives multiple medical examination result description terms, the computer device first concatenates these multiple medical examination result description terms, and then combines the concatenated multiple medical examination result description terms with the type text. Optionally, to facilitate the subsequent differentiation of different medical examination result description terms within the same text segment, the computer device first separates the multiple medical examination description terms using punctuation marks, such as a comma, before concatenating them.

[0175] Optionally, the medical examination result information in the status description entry is a medical examination result marker, which indicates whether the medical examination result of the medical examination item is negative or positive. The computer device then maps the status description entry to medical examination result description terms according to the medical examination result marker. For example, the medical examination result marker includes a negative marker and a positive marker. A negative marker indicates that the medical examination result of the medical examination item is negative, and a positive marker indicates that the medical examination result of the medical examination item is positive. If the status description entry includes a negative marker, the computer device concatenates the medical examination item type text and "negative" in the status description entry to obtain the medical examination result description terms. If the status description entry includes a positive marker, the computer device concatenates the medical examination item type text and "positive" in the status description entry to obtain the medical examination result description terms.

[0176] For example, the physical examination results are marked with text-formatted symbols, with "(-)" for negative and "(+)" for positive. For instance, if the status description entry is "Pregnancy test (-)", where "pregnancy test" is a text-type physical examination item, then the corresponding physical examination result description is "pregnancy test negative"; if the status description entry is "bilateral Babinski sign (+)", where "bilateral Babinski sign" is a text-type physical examination item, then the corresponding physical examination result description is "bilateral Babinski sign positive".

[0177] Optionally, the medical examination result information in the status description entry is a medical examination result parameter, which is the medical examination parameter obtained by performing the medical examination item. The computer device then maps the status description entry to medical examination result description terms according to the medical examination result parameters. For example, if the status description entry includes medical examination item type text and medical examination result parameters, the computer device obtains the parameter threshold corresponding to the medical examination item type text, and determines the medical examination result description terms corresponding to the status description entry based on the magnitude of the parameter threshold and the medical examination result parameters.

[0178] For example, the status description entry is "blood pressure 149 / 86 mmHg", where "blood pressure" is the text of the physical examination item type and "149 / 86 mmHg" is the physical examination result parameter. The computer equipment determines the parameter threshold corresponding to "blood pressure", compares "149 / 86 mmHg" with the parameter threshold corresponding to blood pressure, and thus determines that the physical examination result description word corresponding to this status description entry is "high blood pressure".

[0179] In another possible implementation, the state description information describes the state of an object. The computer device identifies the state description information when the type text corresponding to the state description information represents a symptom type, obtaining a target symptom description statement within the state description information. This target symptom description statement describes a symptom where the object does not exist. The computer device then obtains other statements from the state description information besides the target symptom description statement, and combines these other statements with the type text corresponding to the state description information to obtain the text fragment corresponding to the state description information.

[0180] Since state description information describes the state of an object, and the absence of any symptom also constitutes a state for that object, the state description information corresponding to the symptom type text will include a target symptom description statement. This target symptom description statement is a statement used to describe a symptom that the object does not have. However, if this state description information is directly used as part of the text segment corresponding to the symptom type, the computer device may mistakenly treat a symptom that the object does not have as a present symptom during subsequent state detection, leading to incorrect state detection results. Therefore, to avoid this situation, the computer device identifies the target symptom description statement in the state description information and combines the other statements in the state description information (excluding the target symptom description statement) with the type text, thereby removing the target symptom description statement. This avoids interference from the target symptom description statement during state detection and improves the accuracy of state detection.

[0181] Optionally, after obtaining the target symptom description statement, the computer device extracts the symptom words from the target symptom description statement and determines the target type text, which represents the symptom type for which the object does not exist. The computer device combines the symptom words and the target type text to obtain the text fragment corresponding to the status description information.

[0182] Since the absence of any symptom can also constitute the state of the object, to maximize the amount of information referenced during state detection, the computer device removes the target symptom description from the state description information and combines it into a text fragment. Based on this target symptom description, another text fragment can be generated. Because the target type text identifies the symptom type that the object does not possess, combining the symptom words and the target type text to obtain a text fragment allows the computer device to detect the target type text within that fragment and determine that the symptom words in the text fragment represent a symptom that the object does not possess. By separating the target symptom description from other statements into two text fragments, the amount of information in the state description information is preserved, and the possibility of treating a symptom that does not exist as a present symptom is avoided, further improving the accuracy of subsequent state detection.

[0183] For example, if the status description information is "The patient experienced abdominal pain 2 days ago, which was dull and aching, without nausea or vomiting, but with flatulence, and without fever, chills or other discomfort, and reported being allergic to cefazolin", then the target symptom description statement would be "without nausea or vomiting, and without fever, chills or other discomfort". The symptom words in the target symptom description statement include "nausea", "vomiting", "fever" and "chills".

[0184] 304. The computer device splices together the multiple text fragments obtained to obtain the target text.

[0185] After receiving multiple text fragments, the computer device concatenates these fragments end-to-end to obtain the target text. Therefore, the target text comprises multiple text fragments, and each text fragment includes state description information and the corresponding type text.

[0186] Since the descriptive information includes multiple state descriptions of different types, it is difficult to directly perform state detection on the descriptive information. However, the embodiments of this application use information fusion to integrate multiple state descriptions into a single target text, thereby fusing the diverse features in the descriptive information into a unified text feature. This unified text feature is more suitable for state detection using semantic recognition technology.

[0187] Furthermore, the target text is composed of concatenated text fragments, each of which includes type text and state description information belonging to that type. Therefore, each text fragment corresponds to a type text. By combining the state description information with the corresponding type text into a text fragment, it is possible to classify different types of state description information based on the text fragment. Thus, based solely on the target text, it is possible to distinguish different types of state description information within the target text. While integrating multiple state description information, it also ensures that different types of state description information can be distinguished, resulting in a greater amount of information provided by the target text.

[0188] Taking medical information as an example, medical information includes status description information such as age, gender, chief complaint, medical history, and physical examination. Age information can be divided into discrete features such as infant, toddler, child, youth, middle-aged, and elderly. Gender information is also a discrete feature, while chief complaint and medical history information are text features.

[0189] In related technologies, such as Figure 4As shown, for text features such as chief complaint information and medical history information, a feature extraction model is used to process them and obtain the feature vectors corresponding to the text features. For discrete features such as gender information and age information, one-hot encoding is used to obtain the corresponding feature vectors. Then, the multiple feature vectors obtained are fused and a state detection model is used to perform state detection on the fused feature vectors. However, since the feature vectors corresponding to text features are much larger than the feature vectors corresponding to discrete features, the state detection model has difficulty learning the feature vectors corresponding to discrete features, resulting in poor prediction performance of the state detection model.

[0190] In the method provided in the embodiments of this application, such as Figure 5 As shown, a complete set of medical information 501 includes multiple fields, such as age, gender, chief complaint, medical history, and physical examination. Each field has a corresponding field name and includes status description information. The computer device uses the method provided in this embodiment to convert the medical information 501 into target text 502. Through information fusion, the various types of status description information in the medical information are merged into a single target text, which facilitates subsequent state detection of the target text using semantic analysis technology, thereby improving the accuracy of state detection.

[0191] 305. The computer device performs state detection on the target text and obtains the predicted probabilities corresponding to multiple reference state words.

[0192] In this context, each reference state word represents a state, and the prediction probability corresponding to each reference state word indicates the likelihood that the reference state word is the state word corresponding to the descriptive information. The higher the prediction probability of a reference state word, the greater the likelihood that the reference state word is the state word corresponding to the target text; conversely, the lower the prediction probability of a reference state word, the lower the likelihood that the reference state word is the state word corresponding to the target text.

[0193] In one possible implementation, a computer device extracts features from the target text to obtain text features corresponding to the target text, performs feature transformation on the text features, and obtains the predicted probabilities corresponding to multiple reference state words. Here, the text features corresponding to the target text represent the features contained in the target text; optionally, these text features are a text feature matrix or a text feature vector.

[0194] Optionally, the computer device performs feature transformation on the text features to obtain reference features. The reference features include feature values ​​in multiple dimensions, with each dimension corresponding to a reference state word. The feature value of each dimension in the reference features is determined as the prediction probability corresponding to the reference state word for each dimension.

[0195] 306. Among multiple reference state terms, the computer device determines the state term corresponding to the highest predicted probability.

[0196] The predicted probability corresponding to each reference state word represents the likelihood that the reference state word is the state word corresponding to the descriptive information. Therefore, the reference state word with the highest predicted probability is most likely to be the state word corresponding to the target text. The computer device then determines the state word corresponding to the highest predicted probability, and this state word is used to represent the state described by the descriptive information corresponding to the target text.

[0197] Because multiple predicted probabilities obtained from state detection of target text may contain errors, for example in the medical field, as shown in Table 1 below, the state detection results contain obvious errors. Such detection results can mislead doctors and affect their judgment.

[0198] Table 1

[0199] Real state gender age Chief complaint Medical history Predicted state Should I refuse? hypertension 2 6 lower extremity edema …… abnormal liver function no dermatitis 2 5 itchy skin …… Nasal septal distortion no diarrhea 2 6 bloating and nausea …… intestinal obstruction no tonsillitis 2 4 Follow-up visit …… Cerumen impaction no Refractive error 2 6 Follow-up visit …… hemangioma no Stroke 2 6 Follow-up visit …… High blood lipids no High blood lipids 1 6 Headache for several days …… vascular headache no High blood lipids 1 7 Follow-up visit, improved …… hypertension no

[0200] To reduce the error rate of the final determined state words, the computer device employs a rejection strategy to evaluate whether multiple predicted probabilities are incorrect, and rejects the detection result if an error is found. This rejection strategy includes threshold-based rejection strategies and confidence-based prediction model-based rejection strategies.

[0201] Threshold-based rejection strategy: The computer device determines a first prediction probability, which is the maximum prediction probability among multiple reference state words. If the first prediction probability is greater than a first threshold, the state word corresponding to the first prediction probability is determined among the multiple reference state words.

[0202] If the first predicted probability is greater than the first threshold, it means that the reference state word corresponding to the first predicted probability is likely to be the state word corresponding to the target text. Therefore, the state detection result is considered accurate, and the computer device determines the reference state word corresponding to the first predicted probability as the state word corresponding to the target text.

[0203] In one possible implementation, if the first predicted probability is not greater than a first threshold, a second predicted probability is determined. The second predicted probability is the highest predicted probability among multiple predicted probabilities other than the first predicted probability. A difference parameter between the first and second predicted probabilities is determined. If the difference parameter is not less than the second threshold, the state word corresponding to the first predicted probability is determined from multiple reference state words. Here, the difference parameter represents the difference between the first and second predicted probabilities; a larger difference parameter indicates a larger difference between the two probabilities, and a smaller difference parameter indicates a smaller difference between the two probabilities.

[0204] Considering that a high similarity between the reference state words corresponding to the first predicted probability and those corresponding to the second predicted probability would lower the first predicted probability (e.g., reference state words like diarrhea, gastroenteritis, enteritis, irritable bowel syndrome), even if the first predicted probability is not greater than the first threshold, it does not necessarily mean the state detection result is inaccurate. In this case, the difference between the first and second predicted probabilities can be considered. If the difference is sufficiently large, it indicates that the reference state words corresponding to the first and second predicted probabilities can be distinguished during state detection. Therefore, the state detection result can be considered accurate, and the reference state word corresponding to the first predicted probability is determined as the state word corresponding to the target text. In this embodiment, the difference parameter and the second threshold are used to determine whether the difference between the first and second predicted probabilities is sufficiently large. If the difference parameter is not less than the second threshold, the difference between the first and second predicted probabilities is considered sufficiently large.

[0205] Optionally, the computer device determines that the state detection has failed if the difference parameter is less than a second threshold. A difference parameter less than the second threshold indicates that the difference between the first predicted probability and the second predicted probability is not large enough. This means that it is difficult to distinguish the reference state word corresponding to the first predicted probability and the reference state word corresponding to the second predicted probability during the state detection process. Therefore, the computer device considers the state detection result to be incorrect and determines that the state detection has failed. Optionally, the computer device outputs a state detection failure message.

[0206] Optionally, the computer device determines the absolute value of the difference between the first predicted probability and the second predicted probability; determines the sum of the first predicted probability and the second predicted probability, and defines the ratio between the absolute value and the sum as the difference parameter. Alternatively, the computer device may use other methods to determine the difference parameter between the first predicted probability and the second predicted probability, which is not limited in this embodiment.

[0207] Rejection Strategy Based on Confidence Prediction Model: The computer device uses a confidence prediction model to predict the probabilities of multiple reference state words based on the target text. The prediction result indicates whether the predicted probabilities of the multiple reference state words are accurate. If the prediction result indicates that the predicted probabilities of the multiple reference state words are accurate, the computer device determines the state word with the highest predicted probability among the multiple reference state words.

[0208] The confidence prediction model is used to determine whether the predicted probability of state detection is accurate. This confidence prediction model can determine whether the predicted probability obtained from state detection of the target text is accurate based on the target text. Optionally, this confidence prediction model is a model trained by the computer device, or it can be a model trained by another device and then sent to the computer device.

[0209] Optionally, the training data for the confidence prediction model includes sample text, predicted probabilities obtained by state detection of the sample text, and manually labeled real results. These real results indicate the accuracy of the predicted probabilities obtained by state detection of the sample text. Based on the confidence prediction model, the predicted probabilities are predicted according to the sample text to obtain prediction results. The confidence prediction model is then trained based on the loss value between the predicted results and the real results, enabling the model to learn automatically and achieve confidence assessment.

[0210] By performing steps 305-306 above, state detection is achieved on the target text, yielding state words that represent the state described by the descriptive information. It should be noted that the process of determining state words in steps 305-306 is merely an exemplary process; other methods can also be used to perform state detection on the target text and obtain state words.

[0211] In the method provided in this application embodiment, the description information includes multiple state description information and corresponding type text. The state description information and the corresponding type text are combined into a text fragment, and then the multiple text fragments are spliced ​​together to form the target text. This not only realizes the fusion of multiple state description information into a single target text, but also distinguishes different types of state description information based on the type text in the text fragment. Therefore, performing state detection on the target text can comprehensively consider multiple types of state description information, which is beneficial to improving the accuracy of state detection.

[0212] Furthermore, by adding target markers to the type text, the type text in the text fragment can be identified by recognizing the target markers in the text fragment, thereby distinguishing between state description information and type text in the text fragment. This eliminates the need to distinguish between state description information and type text through semantic recognition, reducing processing complexity and improving the efficiency of subsequent state detection.

[0213] Furthermore, by adding start and end markers to typed text, it is possible to distinguish not only typed text and state description information in text fragments by detecting the start and end markers, but also different text fragments in a target text, further reducing the processing complexity of computer devices.

[0214] Furthermore, by mapping parameters to parameter description terms, or mapping status description information to physical examination result description terms, different types of status description information can be converted into terms that are easy for computer devices to understand, thereby reducing the processing complexity of status detection by computer devices.

[0215] Furthermore, combining the statements other than the target symptom description statement in the state description information with the type text, thereby removing the target symptom description statement, can avoid interference from the target symptom description statement during the state detection process, which is beneficial to improving the accuracy of state detection.

[0216] Furthermore, by separating the target symptom description statement into two text segments with other statements, the amount of information in the state description information is preserved, and the situation of treating non-existent symptoms as existing symptoms is avoided, which can further improve the accuracy of subsequent state detection.

[0217] In another embodiment, the computer device stores a state detection model for performing state detection. In this case, steps 305-306 above can be replaced by the following steps: the computer device performs state detection on the target text based on the state detection model to obtain state words.

[0218] The training process of this state detection model is detailed below. Figure 8 Examples of this implementation will not be described here. Optionally, the state detection model may be a model trained by the computer device, or a model trained by another device and sent to the computer device.

[0219] In one possible implementation, the state detection model includes multiple reference state words. The computer device performs state detection on the target text based on the state detection model to obtain the predicted probabilities corresponding to the multiple reference state words. Among the multiple reference state words, the state word corresponding to the highest predicted probability is determined.

[0220] Optionally, the state detection model includes a feature extraction network and a feature transformation network, whereby the feature transformation network includes multiple reference state terms. For example... Figure 6 As shown, the process by which a computer device performs state detection on a target text based on a state detection model to obtain the predicted probabilities corresponding to multiple reference state words includes:

[0221] 601. The computer device extracts features from the target text based on the feature extraction network in the state detection model, thereby obtaining the text features corresponding to the target text.

[0222] The feature extraction network in the state detection model is used to extract text features. After the computer device obtains the target text, it calls the feature extraction network to extract features from the target text and obtain the text features corresponding to the target text. These text features are used to represent the features contained in the target text. Optionally, the text features are a text feature matrix or a text feature vector.

[0223] The feature extraction network in this embodiment can be implemented using various network structures, such as BERT (Bidirectional Encoder Representation from Transformers). Taking BERT as an example, the input is the target text, which can be represented as S = {x1, x2, ..., x...}. n}, where n is the length of the target text, x n The characters in the target text are represented by Y. After feature extraction using BERT, the corresponding text features of the target text are obtained. The text features can be represented as Y. enc (S).

[0224] 602. The computer device performs feature transformation on text features based on the feature transformation network in the state detection model to obtain the predicted probabilities corresponding to multiple reference state words.

[0225] In the state detection model, the feature transformation network is used to transform text features into predicted probabilities corresponding to multiple reference state words. After obtaining the text features corresponding to the target text, the computer device performs feature transformation on the text features based on the feature transformation network to obtain multiple predicted probabilities.

[0226] The feature transformation network in this embodiment can be implemented using various network structures, such as a Softmax (normalized) neural network. Optionally, the normalized neural network is represented by the following formula:

[0227] p=σ(W r h+b r )

[0228] Where p represents the predicted probability corresponding to multiple reference state words, h represents the text feature, σ represents the normalization function, and W r and b r These are the parameters in the Softmax neural network.

[0229] Figure 7 This is a schematic diagram of a state detection method based on a state detection model provided in an embodiment of this application, as shown below. Figure 7As shown, the state detection model includes a feature extraction network 701 and a feature transformation network 702. After the computer device obtains the target text 703 corresponding to the description information, it inputs the target text 703 into the feature extraction network 701. The feature extraction network 701 extracts features from the target text 703 to obtain text features 704. Then, the text features 704 are input into the feature transformation network 702. The feature transformation network 702 transforms the text features 704 to obtain the prediction probability 705 corresponding to the reference state word. Then, the rejection strategy in step 306 above is used to determine whether the prediction probability 705 is accurate. If it is accurate, the state word 706 corresponding to the maximum prediction probability is determined.

[0230] The state detection method provided in this application can determine the disease type that best matches the patient's current state based on medical information, thereby assisting doctors in diagnosis and reducing misdiagnosis and missed diagnosis. This method can be applied to Clinical Decision Support Systems (CDSS), which are human-computer interaction-based medical information application systems that provide clinical decision support to medical professionals, assisting in clinical decision-making through data and models. This medical information-based assisted diagnostic decision-making method, through operations such as mining, classifying, and regressing medical information, can continuously discover new knowledge to help doctors make better decisions during the diagnostic process.

[0231] This application provides a state detection method based on information fusion and a rejection strategy. It can fuse multiple state descriptions into a single target text, which is then detected by a state detection model, improving the accuracy of the end-to-end state detection model. Furthermore, addressing the instability and poor adversarial capabilities of state detection models, a rejection strategy is employed to determine the accuracy of the model's predictions. This automatically identifies potential erroneous predictions, reducing the likelihood of incorrect predictions and improving the robustness of the state detection model.

[0232] To compare the state detection methods of this scheme, experiments and analyses were conducted from two aspects: model performance and instance data.

[0233] Table 2

[0234] Model Top 1 accuracy Top 3 accuracy Top 5 accuracy False positive rate Based on machine learning 0.47 0.62 0.83 5％ Based on discrete feature fusion 0.75 0.87 0.916 3％ The method of this application 0.76 0.89 0.930 0.6％

[0235] Among them, the machine learning-based method refers to structuring the descriptive information and using the structured data to train a machine learning model for state detection. The discrete feature fusion-based method refers to fusing the feature vectors corresponding to different state descriptions in the descriptive information before performing state detection. Top1 accuracy refers to the probability that the predicted probability corresponding to an accurate state word is the highest predicted probability; Top3 accuracy refers to the probability that the predicted probability corresponding to an accurate state word is among the top three predicted probabilities; and Top5 accuracy refers to the probability that the predicted probability corresponding to an accurate state word is among the top five predicted probabilities. As can be seen from Table 2, the method in this application has higher accuracy and lower false positive rate.

[0236] Figure 8 This is a flowchart illustrating a training method for a state detection model provided in an embodiment of this application. The execution entity of this embodiment is a computer device; see [link to relevant documentation]. Figure 8 The method includes:

[0237] 801. Computer equipment acquires sample text and corresponding sample status words.

[0238] The sample state words corresponding to the sample text represent the state described by the sample text. This sample text is related to the above. Figure 3 The target text in the embodiments is similar. Optionally, the sample state words corresponding to the sample text are accurate state words determined manually.

[0239] 802. Computer equipment uses a state detection model to perform state detection on sample text and obtain the sample prediction probability corresponding to multiple reference state words.

[0240] This step 802 is related to the above. Figure 6 The embodiments are similar and will not be described in detail here.

[0241] 803. The computer equipment trains the state detection model based on the sample prediction probabilities corresponding to multiple reference state words and the sample state words.

[0242] Since the sample prediction probabilities corresponding to multiple reference state words are used to represent the probability that the corresponding reference state word is the state word corresponding to the sample text, and the sample state word is the state word corresponding to the sample symptom description statement, the accuracy of the state detection model can be determined by the sample prediction probabilities of the multiple reference state words and the sample state word, so as to train the state detection model and improve its accuracy.

[0243] In one possible implementation, the computer device determines the target value as the true probability corresponding to the sample state word, determines the loss value of the state detection model based on the true probability corresponding to the sample state word and the sample predicted probability corresponding to the same reference state word, and trains the state detection model based on the loss value.

[0244] A reference state word that is identical to the sample state word is identified. The closer the predicted probability of the sample corresponding to the reference state word is to the true probability, the higher the accuracy of the state detection model. Therefore, based on the difference between the predicted probability and the true probability of the sample, the loss value of the state detection model can be determined. The state detection model is then trained using this loss value to improve its accuracy.

[0245] The target value can be any value. For example, the target value is 1, which means the true probability is 1. Optionally, the computer device trains the state detection model using the results of state detection on multiple sample texts. Then, when the true probability is 1, the computer device determines the loss value of the state detection model using the following formula:

[0246]

[0247] Where H(x) represents the loss value of the state detection model, i represents the index of the sample text, n represents the total number of sample texts, and p(x) i ) represents the sample prediction probability corresponding to the i-th sample text. This sample prediction probability refers to the prediction probability corresponding to the reference state word that is the same as the sample state word.

[0248] The method provided in this application embodiment trains a state detection model using sample text and corresponding sample state words. Subsequently, the trained state detection model is used to perform state detection on any text and obtain the corresponding state words. Therefore, state detection is automated by training the model.

[0249] Figure 9 This is a schematic diagram of the structure of a state detection device provided in an embodiment of this application. See also... Figure 9 The device includes:

[0250] The information acquisition module 901 is used to acquire description information, which includes multiple state description information and type text corresponding to each state description information. The type text corresponding to the state description information indicates the type to which the state description information belongs.

[0251] The combination module 902 is used to combine each state description information and the corresponding type text to obtain a text fragment;

[0252] The splicing module 903 is used to splice multiple text fragments to obtain the target text;

[0253] The state detection module 904 is used to perform state detection on the target text to obtain state words, which represent the state described by the descriptive information.

[0254] In the state detection device provided in this application embodiment, the description information includes multiple state description information and corresponding type text. The state description information and the corresponding type text are combined into a text fragment, and then the multiple text fragments are spliced ​​together to form the target text. This not only realizes the fusion of multiple state description information into a single target text, but also distinguishes different types of state description information based on the type text in the text fragment. Therefore, state detection of the target text can comprehensively consider multiple types of state description information, which is beneficial to improving the accuracy of state detection.

[0255] Optionally, see Figure 10 The description information includes multiple fields, and the device also includes:

[0256] The information and field determination module 905 is used to obtain the status description information of any field among multiple fields, and determine the field name of any field as the type text corresponding to the status description information of any field.

[0257] Optionally, see Figure 10 The combined module 902 includes:

[0258] The tag adding unit 9021 is used to add a target tag to the type text corresponding to each state description information. The target tag is used to indicate that the text with the target tag added is type text.

[0259] Combination unit 9022 is used to combine type text with target markers and state description information to obtain text fragments corresponding to the state description information.

[0260] Optionally, see Figure 10 The target marker includes a start marker and an end marker. The marker adding unit 9021 is used to add a start marker and an end marker to the type text respectively, so as to obtain type text with a start marker added and type text with an end marker added.

[0261] Combination unit 9022 is used to sequentially concatenate type text with a start marker, state description information, and type text with an end marker to obtain the text fragment corresponding to the state description information.

[0262] Optionally, see Figure 10 The combined module 902 includes:

[0263] The parameter mapping unit 9023 is used to map the parameters to parameter description words according to the type text corresponding to the state description information when the state description information includes parameters. The parameter description words are used to describe the parameters belonging to the type.

[0264] The combination unit 9022 is used to combine the parameter description words and the type text corresponding to the state description information to obtain the text fragment corresponding to the state description information.

[0265] Optionally, see Figure 10 The combined module 902 includes:

[0266] The information mapping unit 9024 is used to map the status description information into physical examination result description words when the type text corresponding to the status description information represents the physical examination type. The physical examination result description words are used to describe the physical examination results.

[0267] The combination unit 9022 is used to combine the descriptive words of the physical examination results and the type text corresponding to the status description information to obtain the text fragment corresponding to the status description information.

[0268] Optionally, see Figure 10 The status description information includes at least one status description entry, and each status description entry includes a physical examination item type text and physical examination result information. The physical examination item type text indicates the type of physical examination item. The information mapping unit 9024 is used to map each status description entry into a physical examination result description word to obtain at least one physical examination result description word.

[0269] Optionally, see Figure 10 The state description information describes the state of the object. The combination module 902 includes:

[0270] The statement recognition unit 9025 is used to recognize the state description information when the type text corresponding to the state description information represents the symptom type, and to obtain the target symptom description statement in the state description information. The target symptom description statement is used to describe the symptom that does not exist in the object.

[0271] The statement acquisition unit 9026 is used to acquire statements other than the target symptom description statement in the status description information;

[0272] Combination unit 9022 is used to combine other statements and type text corresponding to the state description information to obtain text fragments corresponding to the state description information.

[0273] Optionally, see Figure 10 The combined module 902 also includes:

[0274] Symptom word acquisition unit 9027 is used to acquire symptom words in the target symptom description statement;

[0275] The type text determination unit 9028 is used to determine the target type text, which represents the symptom type where the object does not exist;

[0276] Combination unit 9022 is used to combine symptom words and target type text to obtain text fragments corresponding to state description information.

[0277] Optionally, see Figure 10 The status detection module 904 includes:

[0278] The prediction probability detection unit 9041 is used to perform state detection on the target text and obtain the prediction probability corresponding to multiple reference state words. The prediction probability corresponding to each reference state word represents the possibility that the reference state word is the state word corresponding to the descriptive information.

[0279] The state word determination unit 9042 is used to determine the state word with the highest predicted probability among multiple reference state words.

[0280] Optionally, see Figure 10 The prediction probability detection unit 9041 is used for:

[0281] Feature extraction is performed on the target text to obtain the text features corresponding to the target text;

[0282] By performing feature transformation on the text features, the predicted probabilities corresponding to multiple reference state words are obtained.

[0283] Optionally, see Figure 10 Status word determination unit 9042, used for:

[0284] Determine the first prediction probability, which is the maximum prediction probability among the prediction probabilities corresponding to multiple reference state words;

[0285] If the first predicted probability is greater than the first threshold, the state word corresponding to the first predicted probability is determined from multiple reference state words.

[0286] Optionally, see Figure 10 The device also includes:

[0287] The second prediction probability determination module 906 is used to determine a second prediction probability when the first prediction probability is not greater than a first threshold. The second prediction probability is the maximum prediction probability among multiple prediction probabilities other than the first prediction probability.

[0288] The difference parameter determination module 907 is used to determine the difference parameter between the first predicted probability and the second predicted probability;

[0289] The state detection module 904 is also used to determine the state word corresponding to the first predicted probability among multiple reference state words, provided that the difference parameter is not less than the second threshold.

[0290] Optionally, see Figure 10 The device also includes:

[0291] The detection failure determination module 908 is used to determine that the state detection has failed when the difference parameter is less than the second threshold.

[0292] Optionally, see Figure 10 The difference parameter determination module 907 includes:

[0293] The absolute value determination unit 9071 is used to determine the absolute value of the difference between the first predicted probability and the second predicted probability;

[0294] The sum determination unit 9072 is used to determine the sum of the first predicted probability and the second predicted probability;

[0295] The difference parameter determination unit 9073 is used to determine the ratio between the absolute value and the sum as the difference parameter.

[0296] Optionally, see Figure 10 Status word determination unit 9042, used for:

[0297] Based on the confidence prediction model, the prediction probabilities of multiple reference state words are predicted according to the target text, and the prediction results are obtained. The prediction results indicate whether the prediction probabilities of multiple reference state words are accurate.

[0298] If the prediction results indicate that the predicted probabilities of multiple reference state terms are accurate, then among the multiple reference state terms, determine the state term with the highest predicted probability.

[0299] Optionally, see Figure 10 The status detection module 904 includes:

[0300] The state detection unit 9043 is used to perform state detection on the target text based on the state detection model to obtain state words.

[0301] Optionally, see Figure 10 The state detection model includes multiple reference state terms, and the state detection unit 9043 is used for:

[0302] Based on the state detection model, the state of the target text is detected to obtain the predicted probabilities of multiple reference state words.

[0303] Among multiple reference state terms, determine the state term corresponding to the highest predicted probability.

[0304] Optionally, see Figure 10 The status detection unit 9043 is used for:

[0305] Based on the feature extraction network in the state detection model, feature extraction is performed on the target text to obtain the text features corresponding to the target text;

[0306] Based on the feature transformation network in the state detection model, feature transformation is performed on the text features to obtain the predicted probabilities corresponding to multiple reference state words.

[0307] Optionally, see Figure 10 The device also includes:

[0308] The sample acquisition module 909 is used to acquire sample text and corresponding sample status words, where the sample status words represent the state described by the sample text.

[0309] The state detection module 904 is also used to perform state detection on sample text based on the state detection model to obtain the sample prediction probability corresponding to multiple reference state words.

[0310] The model training module 910 is used to train the state detection model based on the sample prediction probabilities and sample state words corresponding to multiple reference state words.

[0311] Optionally, see Figure 10 Model training module 910 includes:

[0312] The true probability determination unit 9101 is used to determine the target value as the true probability corresponding to the sample state word;

[0313] The loss value determination unit 9102 is used to determine the loss value of the state detection model based on the true probability corresponding to the sample state word and the sample prediction probability corresponding to the reference state word that is the same as the sample state word.

[0314] Model training unit 9103 is used to train the state detection model based on the loss value.

[0315] It should be noted that the state detection device provided in the above embodiments is only illustrated by the division of the above functional modules when performing state detection. In practical applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the computer device can be divided into different functional modules to complete all or part of the functions described above. In addition, the state detection device and the state detection method embodiments provided in the above embodiments belong to the same concept, and the specific implementation process can be found in the method embodiments, which will not be repeated here.

[0316] This application also provides a computer device, which includes a processor and a memory. The memory stores at least one computer program, which is loaded and executed by the processor to implement the operations performed in the state detection method of the above embodiments.

[0317] Optionally, the computer device is provided as a terminal. Figure 11 A schematic diagram of the structure of a terminal 1100 provided in an exemplary embodiment of this application is shown. The terminal 1100 includes a processor 1101 and a memory 1102.

[0318] Processor 1101 may include one or more processing cores, such as a quad-core processor, an octa-core processor, etc. Processor 1101 may be implemented using at least one hardware form selected from DSP (Digital Signal Processing), FPGA (Field Programmable Gate Array), and PLA (Programmable Logic Array). Processor 1101 may also include a main processor and a coprocessor. The main processor, also known as a CPU (Central Processing Unit), is used to process data in the wake-up state; the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, processor 1101 may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing the content to be displayed on the screen. In some embodiments, processor 1101 may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning.

[0319] The memory 1102 may include one or more computer-readable storage media, which may be non-transitory. The memory 1102 may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory devices. In some embodiments, the non-transitory computer-readable storage media in the memory 1102 are used to store at least one computer program, which is used by the processor 1101 to implement the state detection method provided in the method embodiments of this application.

[0320] In some embodiments, the terminal 1100 may also optionally include: a peripheral device interface 1103 and at least one peripheral device. The processor 1101, memory 1102, and peripheral device interface 1103 can be connected via a bus or signal line. Each peripheral device can be connected to the peripheral device interface 1103 via a bus, signal line, or circuit board. Optionally, the peripheral device includes at least one of: a radio frequency circuit 1104, a display screen 1105, a camera assembly 1106, an audio circuit 1107, and a power supply 1109.

[0321] Peripheral device interface 1103 can be used to connect at least one I / O (Input / Output) related peripheral device to processor 1101 and memory 1102. In some embodiments, processor 1101, memory 1102 and peripheral device interface 1103 are integrated on the same chip or circuit board; in some other embodiments, any one or two of processor 1101, memory 1102 and peripheral device interface 1103 can be implemented on separate chips or circuit boards, which is not limited in this embodiment.

[0322] The radio frequency (RF) circuit 1104 is used to receive and transmit RF (Radio Frequency) signals, also known as electromagnetic signals. The RF circuit 1104 communicates with communication networks and other communication devices via electromagnetic signals. The RF circuit 1104 converts electrical signals into electromagnetic signals for transmission, or converts received electromagnetic signals back into electrical signals. Optionally, the RF circuit 1104 includes: an antenna system, an RF transceiver, one or more amplifiers, a tuner, an oscillator, a digital signal processor, a codec chipset, a user identity module card, etc. The RF circuit 1104 can communicate with other devices via at least one wireless communication protocol. This wireless communication protocol includes, but is not limited to: metropolitan area networks (MANs), various generations of mobile communication networks (2G, 3G, 4G, and 5G), wireless local area networks (WLANs), and / or WiFi (Wireless Fidelity) networks. In some embodiments, the RF circuit 1104 may also include circuitry related to NFC (Near Field Communication), which is not limited in this application.

[0323] Display screen 1105 is used to display a UI (User Interface). This UI may include graphics, text, icons, videos, and any combination thereof. When display screen 1105 is a touch display screen, it also has the ability to collect touch signals on or above its surface. These touch signals can be input as control signals to processor 1101 for processing. In this case, display screen 1105 can also be used to provide virtual buttons and / or a virtual keyboard, also known as soft buttons and / or a soft keyboard. In some embodiments, there may be one display screen 1105, disposed on the front panel of terminal 1100; in other embodiments, there may be at least two display screens, disposed on different surfaces of terminal 1100 or in a folded design; in still other embodiments, display screen 1105 may be a flexible display screen, disposed on a curved or folded surface of terminal 1100. Furthermore, display screen 1105 may be configured as a non-rectangular, irregular shape, i.e., a non-rectangular screen. The display screen 1105 can be made of materials such as LCD (Liquid Crystal Display) and OLED (Organic Light-Emitting Diode).

[0324] The camera assembly 1106 is used to acquire images or videos. Optionally, the camera assembly 1106 includes a front-facing camera and a rear-facing camera. The front-facing camera is disposed on the front panel of the terminal 1100, and the rear-facing camera is disposed on the back of the terminal 1100. In some embodiments, there are at least two rear-facing cameras, which are any one of a main camera, a depth-sensing camera, a wide-angle camera, and a telephoto camera, to achieve background blurring by fusion of the main camera and the depth-sensing camera, panoramic shooting by fusion of the main camera and the wide-angle camera, VR (Virtual Reality) shooting, or other fusion shooting functions. In some embodiments, the camera assembly 1106 may also include a flash. The flash may be a single-color temperature flash or a dual-color temperature flash. A dual-color temperature flash refers to a combination of a warm light flash and a cool light flash, which can be used for light compensation at different color temperatures.

[0325] The audio circuit 1107 may include a microphone and a speaker. The microphone is used to collect sound waves from the user and the environment, converting the sound waves into electrical signals that are input to the processor 1101 for processing, or input to the radio frequency circuit 1104 for voice communication. For stereo sound acquisition or noise reduction purposes, multiple microphones may be used, each positioned at a different location on the terminal 1100. The microphone may also be an array microphone or an omnidirectional microphone. The speaker is used to convert electrical signals from the processor 1101 or the radio frequency circuit 1104 into sound waves. The speaker may be a conventional diaphragm speaker or a piezoelectric ceramic speaker. When the speaker is a piezoelectric ceramic speaker, it can convert electrical signals not only into audible sound waves but also into inaudible sound waves for purposes such as distance measurement. In some embodiments, the audio circuit 1107 may also include a headphone jack.

[0326] Power supply 1109 is used to power the various components in terminal 1100. Power supply 1109 can be AC ​​power, DC power, a disposable battery, or a rechargeable battery. When power supply 1109 includes a rechargeable battery, the rechargeable battery can support wired charging or wireless charging. The rechargeable battery can also be used to support fast charging technology.

[0327] Those skilled in the art will understand that Figure 11 The structure shown does not constitute a limitation on terminal 1100 and may include more or fewer components than shown, or combine certain components, or use different component arrangements.

[0328] Optionally, the computer device is provided as a server. Figure 12 This is a schematic diagram of a server structure provided in an embodiment of this application. The server 1200 can vary significantly due to different configurations or performance. It may include one or more Central Processing Units (CPUs) 1201 and one or more memories 1202. The memories 1202 store at least one computer program, which is loaded and executed by the processor 1201 to implement the methods provided in the above-described method embodiments. Of course, the server may also have wired or wireless network interfaces, a keyboard, and input / output interfaces for input and output. The server may also include other components for implementing device functions, which will not be elaborated upon here.

[0329] This application also provides a computer-readable storage medium storing at least one computer program, which is loaded and executed by a processor to implement the operations performed in the state detection method of the above embodiments.

[0330] This application also provides a computer program product or computer program, which includes computer program code stored in a computer-readable storage medium. The processor of a computer device reads the computer program code from the computer-readable storage medium and executes the computer program code, causing the computer device to perform the operations performed in the state detection method of the above embodiments.

[0331] Those skilled in the art will understand that all or part of the steps of the above embodiments can be implemented by hardware or by a program instructing related hardware. The program can be stored in a computer-readable storage medium, such as a read-only memory, a disk, or an optical disk.

[0332] The above description is only an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present application should be included within the protection scope of the present application.

Claims

1. A state detection method, characterized in that, The method includes: Obtain description information, which includes multiple fields, each field being used to record status description information, and the types of status description information recorded in different fields are different; For any one of the multiple fields, obtain the status description information in the field, and determine the field name of the field as the type text corresponding to the status description information in the field, wherein the type text corresponding to the status description information indicates the type to which the status description information belongs; For each state description information, a start marker and an end marker are added to the type text corresponding to the state description information to obtain type text with the start marker added and type text with the end marker added. The start marker is used to indicate that the text with the start marker added is type text, and the end marker is used to indicate that the text with the end marker added is type text. For each state description information, the state description information is converted into a state description text corresponding to the state description information, and the state described by the state description text matches the state described by the state description information; the type text with the start mark added, the state description text corresponding to the state description information, and the type text with the end mark added are concatenated in sequence to obtain the text fragment corresponding to the state description information. The obtained text fragments are concatenated to obtain the target text; The target text is subjected to state detection to obtain state words, which represent the state described by the descriptive information.

2. The method according to claim 1, characterized in that, The step of converting the state description information into state description text corresponding to the state description information includes: When the status description information includes parameters, the parameters are mapped to parameter description words according to the type text corresponding to the status description information. The parameter description words are used to describe the parameters belonging to the type, and the status description text includes the parameter description words.

3. The method according to claim 1, characterized in that, The step of converting the state description information into state description text corresponding to the state description information includes: When the type text corresponding to the status description information represents the physical examination type, the status description information is mapped to physical examination result description terms, the physical examination result description terms are used to describe the physical examination results, and the status description text includes the physical examination result description terms.

4. The method according to claim 1, characterized in that, The state description information describes the state of an object, and the step of converting the state description information into state description text corresponding to the state description information includes: When the type text corresponding to the state description information represents the symptom type, the state description information is identified to obtain the target symptom description statement in the state description information. The target symptom description statement is used to describe the symptoms that do not exist in the object. Obtain other statements in the status description information besides the target symptom description statement, and the status description text includes the other statements.

5. The method according to claim 4, characterized in that, The method further includes: Obtain symptom words from the target symptom description statement; Determine the target type text, which represents the symptom type where the object does not exist; The symptom words and target type text are combined to obtain the text fragment corresponding to the state description information.

6. The method according to claim 1, characterized in that, The state detection of the target text to obtain state words includes: State detection is performed on the target text to obtain the predicted probabilities corresponding to multiple reference state words. The predicted probability corresponding to each reference state word represents the probability that the reference state word is the state word corresponding to the description information. Among the plurality of reference state words, the state word corresponding to the highest predicted probability is determined.

7. The method according to claim 6, characterized in that, Determining the state word with the highest predicted probability among the plurality of reference state words includes: Determine a first prediction probability, which is the maximum prediction probability among the prediction probabilities corresponding to the plurality of reference state words; If the first predicted probability is greater than the first threshold, the state word corresponding to the first predicted probability is determined from the plurality of reference state words.

8. The method according to claim 7, characterized in that, The method further includes: If the first prediction probability is not greater than the first threshold, a second prediction probability is determined, wherein the second prediction probability is the maximum prediction probability among a plurality of prediction probabilities other than the first prediction probability. Determine the difference parameter between the first predicted probability and the second predicted probability; If the difference parameter is not less than the second threshold, the state word corresponding to the first predicted probability is determined from the plurality of reference state words.

9. The method according to claim 1, characterized in that, The state detection of the target text to obtain state words includes: Based on the state detection model, the target text is subjected to state detection to obtain the state words.

10. The method according to claim 9, characterized in that, Before performing state detection on the target text based on the state detection model to obtain the state words, the method further includes: Obtain sample text and corresponding sample state words, where the sample state words represent the state described by the sample text; Based on the state detection model, state detection is performed on the sample text to obtain the sample prediction probability corresponding to multiple reference state words; The state detection model is trained based on the sample prediction probabilities corresponding to the multiple reference state words and the sample state words.

11. A state detection device, characterized in that, The device includes: The information acquisition module is used to acquire descriptive information, which includes multiple fields. Each field is used to record state description information, and the types of state description information recorded in different fields are different. The information acquisition module is further configured to acquire the status description information in any field among the plurality of fields, and determine the field name of the field as the type text corresponding to the status description information in the field, wherein the type text corresponding to the status description information indicates the type to which the status description information belongs; The combination module is used to add a start mark and an end mark to the type text corresponding to each state description information, so as to obtain type text with the start mark and type text with the end mark. The start mark is used to indicate that the text with the start mark is type text, and the end mark is used to indicate that the text with the end mark is type text. The combination module is further configured to, for each state description information, convert the state description information into a state description text corresponding to the state description information, wherein the state described by the state description text matches the state described by the state description information; and sequentially concatenate the type text with the start mark, the state description text corresponding to the state description information, and the type text with the end mark to obtain the text fragment corresponding to the state description information. The concatenation module is used to concatenate multiple text fragments to obtain the target text; The state detection module is used to perform state detection on the target text to obtain state words, which represent the state described by the descriptive information.

12. The apparatus according to claim 11, characterized in that, The combined module includes: The parameter mapping unit is used to map the parameters into parameter description words according to the type text corresponding to the status description information when the status description information includes parameters. The parameter description words are used to describe the parameters belonging to the type. The status description text includes physical examination result description words.

13. The apparatus according to claim 11, characterized in that, The combined module includes: An information mapping unit is used to map the status description information into physical examination result description terms when the type text corresponding to the status description information represents the physical examination type. The physical examination result description terms are used to describe the physical examination results, and the status description text includes the physical examination result description terms.

14. The apparatus according to claim 11, characterized in that, The state description information describes the state of the object, and the combination module includes: The statement recognition unit is used to recognize the state description information when the type text corresponding to the state description information represents the symptom type, and to obtain the target symptom description statement in the state description information, wherein the target symptom description statement is used to describe the symptom that does not exist in the object; The statement acquisition unit is used to acquire statements other than the target symptom description statement in the status description information, and the status description text includes the other statements.

15. The apparatus according to claim 14, characterized in that, The combined module further includes: The symptom word acquisition unit is used to acquire symptom words from the target symptom description statement; A type text determination unit is used to determine a target type text, wherein the target type text represents a symptom type in which the object does not exist; The combination unit is used to combine the symptom words and the target type text to obtain the text fragment corresponding to the state description information.

16. The apparatus according to claim 11, characterized in that, The status detection module includes: The prediction probability detection unit is used to perform state detection on the target text and obtain the prediction probability corresponding to multiple reference state words. The prediction probability corresponding to each reference state word represents the probability that the reference state word is the state word corresponding to the description information. The state word determination unit is used to determine the state word with the highest predicted probability among the plurality of reference state words.

17. The apparatus according to claim 16, characterized in that, The state word determination unit is used for: Determine a first prediction probability, which is the maximum prediction probability among the prediction probabilities corresponding to the plurality of reference state words; If the first predicted probability is greater than the first threshold, the state word corresponding to the first predicted probability is determined from the plurality of reference state words.

18. The apparatus according to claim 17, characterized in that, The device further includes: The second prediction probability determination module is used to determine a second prediction probability when the first prediction probability is not greater than the first threshold. The second prediction probability is the maximum prediction probability among a plurality of prediction probabilities other than the first prediction probability. The difference parameter determination module is used to determine the difference parameter between the first predicted probability and the second predicted probability; The state detection module is further configured to determine the state word corresponding to the first predicted probability among the plurality of reference state words, provided that the difference parameter is not less than the second threshold.

19. The apparatus according to claim 11, characterized in that, The status detection module includes: The state detection unit is used to perform state detection on the target text based on the state detection model to obtain the state words.

20. The apparatus according to claim 19, characterized in that, The device further includes: The sample acquisition module is used to acquire sample text and corresponding sample state words, wherein the sample state words represent the state described by the sample text. The state detection module is also used to perform state detection on the sample text based on the state detection model to obtain the sample prediction probability corresponding to multiple reference state words. The model training module is used to train the state detection model based on the sample prediction probabilities corresponding to the multiple reference state words and the sample state words.

21. A computer device, characterized in that, The computer device includes a processor and a memory, the memory storing at least one computer program, which is loaded and executed by the processor to perform the operations performed in the state detection method as described in any one of claims 1 to 10.

22. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one computer program, which is loaded and executed by a processor to perform the operations performed in the state detection method as described in any one of claims 1 to 10.

23. A computer program product, characterized in that, The computer program product includes computer program code stored in a computer-readable storage medium. A processor of a computer device reads the computer program code from the computer-readable storage medium and executes the computer program code, causing the computer device to perform the operations performed in the state detection method as described in any one of claims 1 to 10.

Images