Natural language-based disability rating method and system

Abstract

The present application relates to the technical field of natural language processing, and more particularly to a method and device for disability grade identification based on natural language. The method constructs a term mapping library and determines a plurality of preset natural keywords that have a mapping relationship with each medical term. Then, audio information is obtained and feature extraction is performed to obtain a plurality of natural keywords. The number of natural keywords that match each preset natural keyword is counted, and based on the matching condition, it is determined whether the information extraction for the audio information meets the standard. If it is determined that the standard is not met, a corresponding instruction is generated. Based on the instruction, the relevant parameters in the information extraction process for the audio information are determined, thereby improving the information extraction efficiency, obtaining more suitable natural keywords, and providing more accurate and reliable reference credentials for subsequent disability grade identification, thereby improving the efficiency and reliability of disability grade identification.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of natural language processing, in particular to a natural language-based disability grade identification method and system. BACKGROUND

[0002] Disability grade identification is a key link in the fields of law, medicine and social security, and its accuracy directly affects the judgment of the rights and interests of the disabled. At present, the disability grade identification process still mainly relies on manual review and identification, and the identification efficiency is low and the accuracy is not high. Natural language processing technology can convert the speech content in the audio into text, realize accurate conversion from speech to text, and can be applied to disability grade identification, which can speed up the efficiency of disability grade identification and improve the accuracy of disability grade identification.

[0003] The current prior art Chinese patent CN115293229A provides a disability grade identification and evaluation basis deduction method based on natural language understanding. The technical solution analyzes the disease and injury description text of labor capacity identification, constructs an intelligent disability grade identification model based on Bert text classification and an evaluation basis matching model based on text similarity, obtains a predicted classification result through data input model, and then performs text similarity calculation and matching with the evaluation basis to obtain the evaluation basis in the disability grade identification process, thereby reducing the time of manual identification and speeding up the identification efficiency. However, the technical solution lacks a verification mechanism for the obtained predicted classification result, cannot guarantee that the predicted result is accurate, has low reliability, and lacks adjustment and optimization of the model and related parameters according to the actual result. SUMMARY

[0004] Therefore, the present application provides a natural language-based disability grade identification method and system to solve the problems of lacking a verification mechanism for the judgment process and lacking adjustment or optimization based on the judgment result in the prior art, thereby improving the accuracy of the information output for the evaluation basis of disability grade identification.

[0005] To achieve the above-mentioned purpose, the present application provides a natural language-based disability grade identification method, comprising:

[0006] Obtain a medical term set and construct a term mapping library based thereon, wherein the term mapping library stores a plurality of medical terms and a preset natural keyword having a mapping relationship with each medical term;

[0007] Collect audio information and perform conversion and recognition to obtain original speech data;

[0008] Data preprocessing is performed on the original speech data to obtain preprocessed data;

[0009] extracting features from the preprocessed data to obtain a plurality of natural keywords;

[0010] counting the number of the natural keywords matching the preset natural keywords in the term mapping library;

[0011] determining whether information extraction for the audio information meets a standard based on the matching condition, and generating a corresponding instruction when the standard is not met;

[0012] determining a screening numerical interval in a data preprocessing process, determining a signal amplification ratio in a data acquisition process, or re-determining a corresponding processing mode based on the total number of the unmatched natural keywords.

[0013] Further, the process of determining whether information extraction for the audio information meets a standard based on the matching condition comprises:

[0014] For a single natural keyword, the natural keyword matching the preset natural keyword in at least one of the term mapping libraries is recorded as a matching keyword;

[0015] counting the number of the matching keywords and the total number of the natural keywords, calculating the ratio between the number of the matching keywords and the total number of the natural keywords, and recording the ratio as a matching proportion;

[0016] determining whether information extraction for the audio information meets a standard based on a comparison result of the matching proportion and a preset matching proportion;

[0017] When it is determined that information extraction for the audio information does not meet the standard, determining the reason according to the difference between the preset matching proportion and the matching proportion.

[0018] Further, the process of determining based on the comparison result of the matching proportion and the preset matching proportion comprises:

[0019] counting the number of the mapping relationship with a single matching keyword in the term mapping library and recording it as a mapping number;

[0020] calculating a mapping average based on a plurality of mapping numbers and a plurality of matching keywords;

[0021] determining whether to preferentially send to a corresponding department based on a comparison result of the mapping average and a critical mapping average.

[0022] Further, the process of determining whether to preferentially send to a corresponding department based on the comparison result of the mapping average and the critical mapping average comprises:

[0023] traversing the term mapping library to obtain the total number of the medical terms;

[0024] determining the reduction of the preset number based on a comparison result of the total number of the medical terms and the total number of preset medical terms, wherein the reduction of the preset number is in a positive correlation with the total number of the medical terms, and the preset number is a critical number of mapping relationships;

[0025] deleting the natural keywords with the number of mapping relationships greater than the preset number.

[0026] Further, the process of determining the cause according to the difference between the preset matching proportion and the matching proportion includes:

[0027] calculating the difference between the preset matching proportion and the matching proportion and recording it as a proportion difference;

[0028] determining the cause of the non-compliance of the information extraction for the audio information based on a comparison result of the proportion difference and a preset proportion difference;

[0029] determining the corresponding processing mode based on the cause, including: determining the screening numerical interval in the data preprocessing process, determining the signal amplification multiple in the data acquisition process, or re-determining the corresponding processing mode based on the total number of the unmatched natural keywords.

[0030] Further, the process of determining the screening abnormal numerical interval in the data preprocessing process includes:

[0031] calculating the ratio between the proportion difference and the preset proportion difference and recording it as a difference ratio;

[0032] determining the reduction of the screening numerical interval based on a comparison result of the difference ratio and a preset difference ratio, wherein the reduction of the screening numerical interval is in a positive correlation with the difference ratio.

[0033] Further, the process of re-determining the corresponding processing mode based on the total number of the unmatched natural keywords includes:

[0034] counting the total number of the unmatched natural keywords and recording it as a total number of unmatched keywords;

[0035] determining the screening numerical interval in the data preprocessing process or determining the signal amplification multiple in the data acquisition process based on a comparison result of the total number of unmatched keywords and a preset total number of unmatched keywords.

[0036] Further, the process of determining the signal amplification multiple in the data acquisition process includes:

[0037] obtaining the measured signal intensity and the noise intensity in the process of converting the audio information into the original voice data, and obtaining the signal-to-noise ratio based on the ratio between the measured signal intensity and the noise intensity.

[0038] Determine the signal amplification factor and the increase amplitude of the signal amplification factor based on the comparison result of the signal-to-noise ratio and the preset signal-to-noise ratio.

[0039] Further, the original voltage value before the signal amplification factor is increased is obtained, and the increased voltage value after the signal amplification factor is increased is obtained, and the amplification ratio is obtained based on the ratio between the increased voltage value and the original voltage value.

[0040] The noise threshold value is increased based on the comparison result of the amplification ratio and the preset amplification ratio, and the increase amplitude of the noise threshold value is in a positive correlation with the amplification ratio, wherein the noise threshold value is a critical value for distinguishing effective signals and background noises.

[0041] The application further provides a natural language-based disability grade identification system, comprising:

[0042] A term mapping library construction module is configured to obtain the medical term set and construct the term mapping library based on the medical term set, wherein the term mapping library stores a plurality of medical terms and the preset natural keywords in a mapping relationship with the medical terms.

[0043] A data acquisition module is configured to acquire the audio information and perform conversion and recognition to obtain the original speech data.

[0044] A data preprocessing module is connected with the data acquisition module and configured to preprocess the original speech data to obtain the preprocessed data.

[0045] A feature extraction module is connected with the data preprocessing module and configured to extract features from the preprocessed data to obtain a plurality of natural keywords.

[0046] A statistics module is connected with the term mapping library construction module and the feature extraction module respectively and configured to count the number of natural keywords matched with the plurality of preset natural keywords.

[0047] An analysis module is connected with the statistics module and configured to determine whether the information extraction for the audio information meets the standard based on the matching condition, determine the reason when the standard is not met, and generate a corresponding instruction based on the reason.

[0048] An adjustment module is connected with the analysis module, the term mapping library construction module, the data acquisition module, the data preprocessing module, and the feature extraction module respectively and configured to determine the screening value interval in the data preprocessing process based on the instruction, determine the signal amplification factor in the data acquisition process, or determine the corresponding processing mode based on the total number of unmatched natural keywords.

[0049] Compared with the prior art, the natural language-based disability grade identification method has the beneficial effects that the method acquires a medical term set and constructs a term mapping library based on the medical term set, stores a plurality of medical terms and preset natural keywords having a mapping relationship with each medical term in the term mapping library, then collects audio information and performs conversion recognition to obtain original speech data, performs data preprocessing on the original speech data to obtain preprocessed data, performs feature extraction on the preprocessed data to obtain a plurality of natural keywords, counts the number of natural keywords matching the preset natural keywords, determines whether the information extraction for the audio information meets the standard based on the matching condition, and generates a corresponding instruction when the standard is not met, determines the screening numerical value interval in the data preprocessing process, determines the signal amplification multiple in the data collection process, or re-determines the corresponding processing mode based on the total number of unmatched natural keywords, thereby improving the information extraction efficiency for the audio information, obtaining more natural keywords that can be matched, and providing reliable and accurate reference credentials for subsequent disability grade identification, thereby improving the efficiency and credibility of disability grade identification.

[0050] Further, the application obtains a matching proportion by the number of matched keywords and the total number of natural keywords, and determines whether the information extraction for the audio information meets the standard based on the comparison between the matching proportion and the preset matching proportion, so that the reason for not meeting the standard can be determined according to the difference between the preset matching proportion and the matching proportion when the information extraction is determined not to meet the standard, thereby playing a verification role.

[0051] Further, the application also obtains a mapping mean value by obtaining the mapping number corresponding to each matched keyword, and when it is determined that a department can be preferentially sent to based on the comparison between the mapping mean value and the critical mapping mean value, the corresponding department is directly sent to, thereby improving the determination of the matched keyword and improving the efficiency in the disability grade identification process; when the department that can be preferentially sent to cannot be determined through the comparison, the total number of medical terms and the total number of preset medical terms are used to determine a preset number of reduction, and the natural keywords with a mapping number greater than the preset number are deleted and re-extracted, so that the matched natural keywords are corrected, thereby ensuring that the department that can be preferentially sent to can be determined in the subsequent comparison process, thereby improving the efficiency in the disability grade identification process.

[0052] Further, after determining the reason for the information extraction for the audio information not meeting the standard, the application can obtain a difference value ratio according to the proportion difference value and the preset proportion difference value, and determine to narrow the screening numerical value interval based on the comparison between the difference value ratio and the preset difference value ratio, thereby improving the data preprocessing effect in the information extraction process and obtaining more matched keywords.

[0053] Further, the application can also determine to increase the signal amplification ratio according to the comparison of the signal-to-noise ratio and the preset signal-to-noise ratio, thereby improving the data acquisition effect in the information extraction process, and obtaining more matching keywords.

[0054] Further, after determining to increase the signal amplification ratio, the application can obtain an amplification ratio value according to the original voltage value and the increased voltage value before and after the signal amplification ratio is increased, and determine to increase the noise threshold based on the comparison of the amplification ratio value and the preset amplification ratio value, thereby reducing the background noise pollution, and improving the data preprocessing effect in the information extraction process. BRIEF DESCRIPTION OF DRAWINGS

[0055] Figure 1 A module schematic diagram of the natural language-based injury grade identification system in the application;

[0056] Figure 2 A flowchart of the natural language-based injury grade identification method in the application;

[0057] Figure 3 A logic determination diagram for determining whether the information extraction for audio information meets the standard based on the matching proportion in the application;

[0058] Figure 4 A logic determination diagram for determining whether to preferentially send to the department and determine to reduce the preset number based on the mapping mean in the application;

[0059] Figure 5 A logic determination diagram for determining the reason why the information extraction for audio information does not meet the standard and the corresponding processing method based on the proportion difference in the application. DETAILED DESCRIPTION

[0060] In order to make the objects and advantages of the application clearer, the application will be further described below with examples; it should be understood that the specific examples described herein are only used to explain the application, and do not limit the application.

[0061] The preferred embodiments of the application will be described below with reference to the accompanying drawings. Those skilled in the art should understand that these embodiments are only used to explain the technical principles of the application, and are not intended to limit the protection scope of the application.

[0062] It should be noted that in the description of the application, unless otherwise explicitly specified and limited, the term "connection" should be understood broadly, for example, it can be fixed connection, or detachable connection, or integral connection; it can be mechanical connection, or electrical connection; it can be directly connected, or indirectly connected through an intermediate medium, or the internal communication of two elements. Those skilled in the art can understand the specific meaning of the above-mentioned term in the application according to the specific circumstances.

[0063] Referring to Figure 1 As shown in the figure, it is a module schematic diagram of the natural language-based disability grade identification system in the embodiment. It includes a term mapping library construction module, a data acquisition module, a data preprocessing module, a feature extraction module, a statistical module, an analysis module, and an adjustment module. The term mapping library construction module is used to obtain the medical term set and construct the term mapping library based on it. In the term mapping library, there are a plurality of medical terms and a plurality of preset natural keywords that have a mapping relationship with each medical term. The data acquisition module is used to acquire the audio information and perform conversion recognition to obtain the original speech data. The data preprocessing module is connected with the data acquisition module and is used to preprocess the original speech data to obtain the preprocessed data. The feature extraction module is connected with the data preprocessing module and is used to extract features from the preprocessed data to obtain a plurality of natural keywords. The statistical module is connected with the term mapping library construction module and the feature extraction module respectively, and is used to count the number of natural keywords that match the plurality of preset natural keywords. The analysis module is connected with the statistical module and is used to determine whether the information extraction for the audio information meets the standard based on the matching condition, and determine the reason when it does not meet the standard, and generate corresponding instructions based on the reason. The adjustment module is connected with the analysis module, the term mapping construction module, the data acquisition module, the data preprocessing module, and the feature extraction module respectively, and is used to determine the screening value interval in the data preprocessing process, determine the signal amplification ratio in the data acquisition process, or determine the corresponding processing mode based on the total number of unmatched natural keywords.

[0064] Referring to Figure 2 As shown in the figure, it is a flowchart of the natural language-based disability grade identification method in the embodiment. The flowchart includes at least the following steps:

[0065] S1: Obtain a medical term set and construct a term mapping library based on it. In the term mapping library, there are a plurality of medical terms and a plurality of preset natural keywords that have a mapping relationship with each medical term.

[0066] S2: Acquire audio information and perform conversion recognition to obtain original speech data.

[0067] S3: Preprocess the original speech data to obtain preprocessed data.

[0068] S4: Extract features from the preprocessed data to obtain a plurality of natural keywords.

[0069] S5: Count the number of natural keywords that match the preset natural keywords.

[0070] S6: determining whether the information extraction for the audio information meets the standard based on the matching condition, and generating a corresponding instruction when it does not meet the standard.

[0071] S7: determining the screening numerical interval in the data preprocessing process, determining the signal amplification ratio in the data collection process, or re-determining the corresponding processing mode based on the total number of unmatched natural keywords.

[0072] Specifically, in this embodiment, one preset natural keyword has a mapping relationship with one medical term. For example, the medical term includes basic terms: injury, disability, including injury degree classification terms: serious injury, minor injury, slight injury, and function assessment related terms: persistent vegetative state, quadriplegia / paraplegia, loss of joint mobility, etc. One preset natural keyword can be mapped to "injury", and another preset natural keyword can be mapped to "loss of joint mobility". When converting the collected audio information into an electrical signal, i.e., a voice input signal, through a voice input terminal, a suitable sampling frequency of 16 KHz is initially selected, and then signal conversion and recognition are performed to obtain original voice data. The audio information is information output by medical personnel based on the patient's self-report of their condition and the medical personnel's own preliminary examination and sorting. The original voice data contains environmental noise, such as equipment background noise and electromagnetic interference, which can be processed by setting a noise threshold. The preprocessing process also includes data cleaning, which can set a screening abnormal value interval to remove abnormal data that do not meet the voice characteristics or have severe interference, in order to improve the quality and reliability of the data. When the natural keyword feature extraction process is performed on the preprocessed data, for example, when the word "knee significantly limited" is extracted, a natural keyword that matches multiple preset natural keywords such as "loss of joint mobility" and "injury" can be obtained. At this time, it can be known that one natural keyword can match at least one preset natural keyword. The signal amplification ratio is determined by the signal amplifier.

[0073] Referring to FIG. 7, Figure 3 The logic determination diagram for determining whether the information extraction for the audio information meets the standard based on the matching proportion in this embodiment is shown. The process of determining whether the information extraction for the audio information meets the standard based on the matching condition includes:

[0074] For a single natural keyword, the natural keyword matching the preset natural keyword of at least one of the term mapping library is recorded as a matching keyword; the number of the matching keywords and the total number of the natural keywords are counted, the ratio between the number of the matching keywords and the total number of the natural keywords is calculated, and the ratio is recorded as a matching proportion; whether the information extraction for the audio information meets the standard is determined based on the comparison result of the matching proportion and a preset matching proportion; when it is determined that the information extraction for the audio information does not meet the standard, the cause is determined according to the difference between the preset matching proportion and the matching proportion.

[0075] Specifically, in the embodiment, the number of matching keywords can be obtained in a single disability grade identification, or can be obtained from historical data, and then the matching proportion D is calculated. A preset matching proportion D0 is set, and the comparison result of the matching proportion D and the preset matching proportion D0 is determined. It is set that D0=0.95. It should be noted that D0 can also be set to other values, and the number of matching keywords is an integer. The comparison process of the matching proportion D and the preset matching proportion D0 is as follows:

[0076] If the matching proportion D is greater than or equal to the preset matching proportion D0, it means that the number of matching keywords that can match the preset natural keywords based on the current medical term construction in the term mapping library is relatively large among the natural keywords obtained by the current information extraction for the audio information. Therefore, it can be determined that the information extraction for the audio information meets the standard, and the natural keywords obtained at this time are directly output. The natural keywords obtained at this time can be pushed to the next disability grade identification module and used as a reference for grade identification evaluation.

[0077] If the matching proportion D is less than the preset matching proportion D0, it means that the number of matching keywords that can match the preset natural keywords is relatively small at this time. The natural keywords at this time cannot be used as a reference for disability grade identification evaluation. It can be determined that the information extraction for the audio information does not meet the standard. When D is less than D0, the cause can be found based on the difference between D0 and D, and the corresponding processing mode is adopted to improve the efficiency of information extraction for the audio information, so as to obtain more matching keywords.

[0078] Please refer to Figure 4 in the embodiment. The logic determination diagram for determining whether to send to the department preferentially and determining to reduce the preset number based on the mapping average. The process of determining based on the comparison result of the matching proportion and the preset matching proportion includes:

[0079] counting the number of the medical terms mapped with the single matching keyword in the term mapping library and recording it as a mapping number; calculating a mapping average based on the mapping numbers and the number of the matching keywords; and determining whether to send to the corresponding department preferentially based on the comparison result of the mapping average and a critical mapping average.

[0080] Specifically, in the present embodiment, one matching keyword matched with the preset natural keyword can have a mapping relationship with at least one medical term, the number of the medical terms mapped with the single matching keyword is counted and recorded as a mapping number, for example, for a single matching keyword, if it has a mapping relationship with n medical terms, the mapping number of the matching keyword is recorded as n, then the mapping average S is calculated based on the total number of the obtained mapping numbers and the number of the matching keywords, because the single matching keyword has a mapping relationship with at least one medical term, the critical mapping average S0=1 can be set, and the comparison process of the mapping average S and the critical mapping average S0 is specifically as follows:

[0081] If the mapping average S is equal to the critical mapping average S0, it means that the number of the medical terms mapped with any matching keyword is one and only one, therefore, the disability grade identification can be directly performed according to the matching keyword, and the corresponding department to which the patient is sent preferentially can be determined based on the matching keyword, for example, the only one medical term matched with the matching keyword and having a mapping relationship is “loss of joint movement function”, at this time, it is directly determined to send to the orthopedics department for diagnosis and determination of the disability grade.

[0082] If the mapping average S is greater than the critical mapping average S0, it means that any matching keyword can match at least two or more preset natural keywords, i.e., the number of the medical terms having a mapping relationship is at least two or more, at this time, it can not be reliably determined to which department the patient is sent preferentially according to the matching keyword, for example, there is a single matching keyword which has a mapping relationship with three medical terms, which are “loss of joint movement function”, “mental disorder” and “complete mandibular defect”, at this time, it is not possible to accurately determine to which department the patient is sent preferentially, at this time, the extraction process of the natural keyword can be optimized to reduce the number of the matching keywords which have a mapping relationship with multiple medical terms.

[0083] Further, the process of determining whether to send to the corresponding department preferentially based on the comparison result of the mapping average and the critical mapping average includes:

[0084] traverse the term mapping library to obtain the total number of medical terms; determine a preset number of reduction based on a comparison result of the total number of medical terms and a total number of preset medical terms, and a reduction range of the preset number is in a positive correlation with the total number of medical terms, wherein the preset number is a critical mapping relationship number; and delete the natural keyword whose mapping relationship number is greater than the preset number.

[0085] Specifically, in the present embodiment, by checking each medical term in the term mapping library one by one and counting the total number H of medical terms, the preset number is a set critical mapping relationship number of medical terms in the term mapping library which simultaneously exist mapping relationship with a single matching keyword. In order to more accurately determine the reduction range of the preset number, the total number H0 of preset medical terms can be divided into the total number H1 of first preset medical terms and the total number H2 of second preset medical terms, and H1 = 40 and H2 = 45 are set. It should be noted that H1 and H2 can also be set to other values. According to the obtained medical term set, the comparison process based on the total number H of medical terms and H1 and H2 is as follows:

[0086] If the total number H of medical terms is less than or equal to the first preset total number H1 of medical terms, an instruction using a first preset number adjustment coefficient is generated, and based on the instruction, the preset number is reduced by one on the original basis; if the total number H of medical terms is greater than the first preset total number H1 of medical terms and less than or equal to the second preset total number H2 of medical terms, an instruction using a second preset number adjustment coefficient is generated, and based on the instruction, the preset number is reduced by two on the original basis; if the total number H of medical terms is greater than the second preset total number H2 of medical terms, an instruction using a third preset number adjustment coefficient is generated, and based on the instruction, the preset number is reduced by three on the original basis. It should be noted that the number of reductions of the preset number can also be set to other numerical values, which is set according to the total number of medical terms, and the minimum preset number is one. After determining the preset number, the natural keywords with a mapping relationship number greater than the preset number can be deleted, and then the extraction of natural keywords and the determination of matching keywords are performed again; by gradually narrowing down the number of matching keywords with multiple mapping relationships, the obtained matching keywords are accurately pointed, so as to accurately determine the corresponding department, and it can be clearly understood that in the case that a matching keyword has multiple mapping relationships, the department to which the matching keyword is preferentially sent can also be determined, for example, when a matching keyword has mapping relationships with "loss of joint activity function" and "serious injury" at the same time, the department to which the matching keyword is preferentially sent can be determined as the orthopedics department for diagnosis and determination of the disability grade, and it should be noted that the department to which the matching keyword is preferentially sent can be determined by gradually narrowing down the preset number. In the case that the term mapping process has problems, the medical term set can be expanded, for example, more medical-related specifications or standards are obtained and supplemented to the medical term set.

[0087] Referring to Figure 5 is a logic determination diagram for determining the reason why the information extraction for the audio information does not meet the standard and the corresponding processing mode based on the proportion difference in the embodiment. The process of determining the reason according to the difference between the preset matching proportion and the matching proportion includes:

[0088] The difference between the preset matching proportion and the matching proportion is calculated and recorded as a proportion difference; the reason why the information extraction for the audio information does not meet the standard is determined based on the comparison result of the proportion difference and a preset proportion difference; the corresponding processing mode is generated based on the reason, including: determining the screening value interval in the data preprocessing process, determining the signal amplification multiple in the data acquisition process, or re-determining the corresponding processing mode based on the total number of unmatched natural keywords.

[0089] Specifically, in the embodiment, in order to more accurately determine the reason for not meeting the standard, the preset proportion difference K0 can be divided into a first preset proportion difference K1 and a second preset proportion difference K2, a preset proportion difference standard K3=0.03 is set, K1=K3-0.01, and K2=K3+0.01. It should be noted that K1, K2, and K3 can also be set to other values according to relevant parameters, including the preset matching proportion D0, the matching proportion D, the total number H of medical terms, etc. The comparison process based on the proportion difference K and K1 and K2 is as follows:

[0090] If the proportion difference K is less than or equal to the first preset proportion difference K1, it means that the difference between the current matching proportion and the preset matching proportion is relatively small, and the number of matched natural keywords is relatively large. At this time, it can be determined that the reason for the information extraction of the audio information not meeting the standard is that there is a problem in the data preprocessing process for the original voice data. At this time, a screening numerical interval can be re-set to improve the cleaning effect of the data, thereby improving the quality and reliability of the preprocessed data and improving the subsequent analysis effect.

[0091] If the proportion difference K is greater than the first preset proportion difference K1 and less than or equal to the second preset proportion difference K2, it is not possible to determine exactly which link has a problem at this time. In order to further determine the corresponding processing method, the total number of unmatched natural keywords is obtained, and the link that has a problem is re-determined based on the total number of unmatched natural keywords, thereby re-determining the corresponding processing method.

[0092] If the proportion difference K is greater than the second preset proportion difference K2, it means that the difference between the current matching proportion and the preset matching proportion is relatively large, and the number of matched natural keywords is relatively small. At this time, it can be determined that the reason for the information extraction of the audio information not meeting the standard is that there is a problem in the data collection process for the audio information. The problem occurs at the beginning of the data collection process for the audio information, thereby causing relatively more natural keywords to be unable to match the preset natural keywords. At this time, the signal amplification ratio in the collection process can be appropriately adjusted to capture more details, thereby improving the quality of the original voice data and improving the subsequent preprocessing effect. It can be clearly understood that the data collection process is the first step of subsequent extraction of natural keywords, and if there is a problem in the data collection process, the difference between the matching proportion and the preset matching proportion will be larger.

[0093] Further, the process of determining the screening abnormal numerical interval in the data preprocessing process includes:

[0094] The ratio between the difference value and the preset difference value is calculated and recorded as a difference value ratio; and the screening numerical value interval is determined to be reduced based on a comparison result of the difference value ratio and a preset difference value ratio, and the reduction range of the screening numerical value interval is in a positive correlation with the difference value ratio.

[0095] Specifically, in the embodiment, the difference value ratio W is specifically obtained by calculating the ratio between the difference value K and the first preset difference value K1, the screening numerical value interval can be a human voice frequency band [500, 3000] Hz, the reduced screening numerical value interval is to appropriately reduce the human voice frequency band, thereby increasing the number of matched keywords, so as to increase the matching proportion D. In order to more accurately determine the reduction range of the screening numerical value interval, the preset difference value ratio W0 can be divided into a first preset difference value ratio W1 and a second preset difference value ratio W2, W1 = 0.75 and W2 = 0.9 are set. It should be noted that W1 and W2 can also be set to other values, and W2 is at most 1. The comparison process of the difference value ratio W and W1 and W2 is specifically as follows:

[0096] If the difference value ratio W is less than or equal to the first preset difference value ratio W1, a first interval adjustment coefficient instruction is generated, and based on the instruction, the screening numerical value interval is determined to be reduced to 0.96 times of the original interval. If the difference value ratio W is greater than the first preset difference value ratio W1 and less than or equal to the second preset difference value ratio W2, a second interval adjustment coefficient instruction is generated, and based on the instruction, the screening numerical value interval is determined to be reduced to 0.93 times of the original interval. If the difference value ratio W is greater than the second preset difference value ratio W2, a third interval adjustment coefficient instruction is generated, and based on the instruction, the screening numerical value interval is determined to be reduced to 0.9 times of the original interval. It can be clearly understood that the reduction rate of the screening numerical value interval can also be set to other values, and the reduction degree is also in line with the standard.

[0097] Further, the process of re-determining the corresponding processing mode based on the total number of the unmatched natural keywords includes:

[0098] The total number of the unmatched natural keywords is counted and recorded as an unmatched keyword total number; and based on a comparison result of the unmatched keyword total number and a preset unmatched keyword total number, the screening numerical value interval in the data preprocessing process is determined or the amplification multiple in the data acquisition process is determined.

[0099] Specifically, in the embodiment, the unmatched natural keyword refers to a natural keyword that does not have a mapping relationship with the medical terms. The comparison process of the unmatched keyword total number Q and the preset unmatched keyword total number Q0 is specifically as follows:

[0100] If the total number of unmatched keywords Q is less than or equal to the preset total number of unmatched keywords Q0, it indicates that the number of the current unmatched natural keywords is relatively small, and thus the number of the matched natural keywords is relatively large. At this time, it can be determined that there is a problem in the data preprocessing process for the original speech data, and a screening value interval can be re-set.

[0101] If the total number of unmatched keywords Q is greater than the preset total number of unmatched keywords Q0, it indicates that the number of the current unmatched natural keywords is relatively large, and thus the number of the matched natural keywords is relatively small. At this time, it can be determined that there is a problem in the data collection process for the audio information, and the amplification multiple in the signal collection process can be appropriately increased. Here, the total number of unmatched keywords Q0 = total number of natural keywords - number of matched keywords.

[0102] Further, the process of determining the signal amplification multiple in the data collection process includes:

[0103] The measured signal intensity and noise intensity in the process of converting the audio information into the original speech data are obtained, and a signal-to-noise ratio is obtained based on the ratio between the measured signal intensity and the noise intensity. The signal amplification multiple is increased based on the comparison result of the signal-to-noise ratio and a preset signal-to-noise ratio, and the increase amplitude of the signal amplification multiple is determined.

[0104] Specifically, in the embodiment, since the signal-to-noise ratio Y is lower than the target value Y3, the audio quality is reduced, and thus the subsequent natural keyword extraction process is problematic. At this time, the amplification multiple can be appropriately increased to improve the signal intensity. Taking a single-stage operational amplifier (signal amplifier) as an example; the preset signal-to-noise ratio Y0 can be divided into a first preset signal-to-noise ratio Y1 and a second preset signal-to-noise ratio Y2, Y1 = 65 dB and Y2 = 75 dB are set, the target value Y3 is set as 85 dB. It should be noted that Y1, Y2 and Y3 can all be set as other values meeting the requirements; the comparison process of the signal-to-noise ratio Y with Y1 and Y2 is specifically as follows:

[0105] If the signal-to-noise ratio Y is less than or equal to the first preset signal-to-noise ratio Y1 and greater than 60 dB, a first amplification ratio adjustment coefficient instruction is generated, and the signal amplification ratio is determined to be increased to 4 times the original value based on the instruction; if the signal-to-noise ratio Y is greater than the first preset signal-to-noise ratio Y1 and less than or equal to the second preset signal-to-noise ratio Y2, a second amplification ratio adjustment coefficient instruction is generated, and the signal amplification ratio is determined to be increased to 3 times the original value based on the instruction; if the signal-to-noise ratio Y is greater than the second preset signal-to-noise ratio Y2 and less than the target value Y3, a third amplification ratio adjustment coefficient instruction is generated, and the signal amplification ratio is determined to be increased to 2 times the original value based on the instruction. It can be clearly seen that the signal amplification ratio can also be set to other values, but the increased signal amplification ratio will not exceed the design requirements of the signal amplifier, and a higher amplification ratio is required for relatively weak signals, while a moderate amplification ratio is required for relatively strong signals to avoid distortion. By increasing the signal amplification ratio, the measured signal strength is improved, and the audio quality is improved, thereby improving the extraction effect of natural keywords.

[0106] Further, the original voltage value before the signal amplification ratio is increased is obtained, and the increased voltage value after the signal amplification ratio is increased is obtained, and the amplification ratio is obtained based on the ratio between the increased voltage value and the original voltage value; the noise threshold is increased based on the comparison result of the amplification ratio and the preset amplification ratio, and the increase amplitude of the noise threshold is proportional to the amplification ratio, wherein the noise threshold is a critical value for distinguishing between effective signals and background noise.

[0107] Specifically, in the embodiment, a noise threshold is set, which should be higher than the noise power. When the sound signal in the voice data is detected to be lower than the noise threshold, it is regarded as background noise, and then it is automatically muted (i.e. output is zero) to reduce the interference of background noise. At the same time, the noise threshold should be much lower than the signal power to avoid misjudging the signal as noise. When the signal-to-noise ratio is determined to be 60 dB and the noise power is 1 mW, the noise threshold can be set to 1.5 mW. The preset amplification ratio B0 can be divided into a first preset amplification ratio B1 and a second preset amplification ratio B2, and B1 = 25 and B2 = 29. B1 and B2 can also be set to other values according to related parameters. The comparison process of the amplification ratio B and B1 and B2 is as follows:

[0108] If the amplification ratio B is less than or equal to a first preset amplification ratio B1, a first noise threshold adjustment coefficient instruction is generated, and the original noise threshold is increased to 2.5 mW based on the instruction; if the amplification ratio B is greater than the first preset amplification ratio B1 and less than or equal to a second preset amplification ratio B2, a second noise threshold adjustment coefficient instruction is generated, and the original noise threshold is increased to 3.5 mW based on the instruction; if the amplification ratio B is greater than the second preset amplification ratio B2, a third noise threshold adjustment coefficient instruction is generated, and the original noise threshold is increased to 4.5 mW based on the instruction. It should be noted that the increased value of the noise threshold can also be set to other values, but the increase amplitude conforms to the standard, so that the increase of the noise threshold within the range can ensure the stability of the system performance, while avoiding affecting the signal quality.

[0109] So far, the technical solutions of the present application have been described in combination with the preferred embodiments shown in the drawings, but those skilled in the art can easily understand that the protection scope of the present application is obviously not limited to these specific embodiments. Those skilled in the art can make equivalent changes or replacements to the related technical features without departing from the principles of the present application, and the technical solutions after the changes or replacements will fall within the protection scope of the present application.

[0110] The above description is only the preferred embodiments of the present application and is not intended to limit the present application; for those skilled in the art, the present application can have various changes and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principles of the present application shall be included in the protection scope of the present application.

Claims

1. A method for disability assessment based on natural language processing, characterized in that, include: A medical terminology set is obtained and a terminology mapping library is constructed based on it. The terminology mapping library stores a number of medical terms and preset natural keywords that have a mapping relationship with each of the medical terms. Collect audio information and convert and recognize it to obtain raw speech data; The raw voice data is preprocessed to obtain preprocessed data; Feature extraction is performed on the preprocessed data to obtain several natural keywords; Count the number of natural keywords that match the preset natural keywords; Based on the matching results, determine whether the information extraction of the audio information conforms to the standard, and generate corresponding instructions when it does not conform to the standard; Based on the instructions, determine the screening value range in the data preprocessing process, determine the signal amplification factor in the data acquisition process, or redetermine the corresponding processing method based on the total number of mismatched natural keywords. The process of determining whether the information extraction from the audio information conforms to the standard based on the matching results includes: For a single natural keyword, the natural keyword that matches at least one of the preset natural keywords in the terminology mapping library is recorded as the matching keyword; The number of matching keywords is compared with the total number of natural keywords. The ratio between the number of matching keywords and the total number of natural keywords is calculated and recorded as the matching percentage. Based on the comparison result between the matching ratio and the preset matching ratio, it is determined whether the information extraction of the audio information meets the standard; When it is determined that the information extraction of the audio information does not meet the standard, the reason is determined based on the difference between the preset matching ratio and the matching ratio; The process of determining the matching ratio based on the comparison result between the matching ratio and the preset matching ratio includes: The number of times that a single matching keyword exists in the terminology mapping library is mapped is counted and recorded as the mapping number. The mean mapping value is calculated based on the number of mappings and the number of matching keywords. Based on the comparison results between the mapped mean and the critical mapped mean, it is determined whether to prioritize sending the sample to the corresponding department.

2. The method for disability assessment based on natural language according to claim 1, characterized in that, The process of determining whether to prioritize sending a patient to the corresponding department based on the comparison result between the mapped mean and the critical mapped mean includes: Traverse the terminology mapping database to obtain the total number of the medical terms; Based on the comparison between the total number of medical terms and the preset total number of medical terms, a preset number is determined to be reduced, and the reduction of the preset number is directly proportional to the total number of medical terms, wherein the preset number is the number of critical mapping relationships. Delete natural keywords whose number of mapping relationships exceeds the preset number.

3. The method for disability assessment based on natural language according to claim 1, characterized in that, The process of determining the cause based on the difference between the preset matching ratio and the actual matching ratio includes: Calculate the difference between the preset matching ratio and the matching ratio and record it as the ratio difference; Based on the comparison result between the aforementioned percentage difference and the preset percentage difference, the reasons why the information extraction of audio information does not meet the standard are determined. Based on the aforementioned reasons, a corresponding processing method is generated, including: determining the screening value range during data preprocessing, determining the signal amplification factor during data acquisition, or re-determining the corresponding processing method based on the total number of mismatched natural keywords.

4. The method for disability assessment based on natural language according to claim 3, characterized in that, The process of determining the range of outlier values ​​during data preprocessing includes: Calculate the ratio between the percentage difference and the preset percentage difference and record it as the difference ratio. The sieving value range is determined based on the comparison result between the difference ratio and the preset difference ratio, and the reduction of the sieving value range is directly proportional to the difference ratio.

5. The method for disability assessment based on natural language according to claim 3, characterized in that, The process of redetermining the corresponding processing method based on the total number of mismatched natural keywords includes: The total number of mismatched natural keywords is counted and recorded as the total number of mismatched keywords; Based on the comparison between the total number of non-matching keywords and the preset total number of non-matching keywords, the screening value range in the data preprocessing process or the signal amplification factor in the data acquisition process is determined.

6. The method for disability assessment based on natural language according to claim 3, characterized in that, The process of determining the signal amplification factor during data acquisition includes: The measured signal strength and noise intensity are obtained during the process of converting the audio information into the original speech data, and the signal-to-noise ratio is obtained based on the ratio between the measured signal strength and noise intensity. The signal amplification factor is increased based on the comparison result between the signal-to-noise ratio and the preset signal-to-noise ratio, and the increase in the signal amplification factor is determined.

7. The method for disability assessment based on natural language according to claim 6, characterized in that, Obtain the original voltage value before increasing the signal amplification factor, and obtain the increased voltage value after increasing the signal amplification factor. Based on the ratio between the increased voltage value and the original voltage value, obtain the amplification ratio. The noise threshold is increased based on the comparison result between the amplification ratio and the preset amplification ratio, and the increase in the noise threshold is proportional to the amplification ratio. The noise threshold is a critical value used to distinguish between valid signals and background noise.

8. A disability assessment system based on natural language, characterized in that, A method for performing a disability rating assessment based on natural language as described in any one of claims 1-7, comprising: A terminology mapping library construction module is used to obtain the medical terminology set and construct the terminology mapping library based on it. The terminology mapping library stores a number of the medical terms and preset natural keywords that have a mapping relationship with each of the medical terms. The data acquisition module is used to acquire the audio information and perform conversion and recognition to obtain the original voice data; A data preprocessing module, which is connected to the data acquisition module, is used to preprocess the raw voice data to obtain the preprocessed data; A feature extraction module, which is connected to the data preprocessing module, is used to extract features from the preprocessed data to obtain several natural keywords. A statistics module, which is connected to the terminology mapping library construction module and the feature extraction module respectively, is used to count the number of natural keywords that match a number of preset natural keywords; An analysis module, connected to the statistics module, is used to determine whether the information extraction of the audio information meets the standard based on the matching situation, and to determine the reason when it does not meet the standard, and to respond normally based on the reason. An adjustment module, which is connected to the analysis module, the terminology mapping construction module, the data acquisition module, the data preprocessing module, and the feature extraction module, is used to determine the screening value range in the data preprocessing process, determine the signal amplification factor in the data acquisition process, or redetermine the corresponding processing method based on the total number of mismatched natural keywords, based on the instructions.

Images