Medical questionnaire verification method and device, computer device, and storage medium

Abstract

The application discloses a medical questionnaire verification method and device, a computer device and a storage medium, and comprises the following steps: obtaining initial medical questionnaire information through a medical questionnaire verification request, wherein the initial medical questionnaire information comprises question information, an answer speech file and answerer basic information; performing speech recognition on the answer speech file to obtain answer text information and generate target question and answer information; performing coding processing on the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector; performing consistency verification on the basic information vector and the question and answer information vector through a trained medical questionnaire verification model; and generating target medical questionnaire information based on a verification result, wherein the target medical questionnaire information is used to indicate medical questionnaire information with consistent answer text information and answerer basic information. Through the artificial intelligence technology, the method can accurately verify the consistency of the medical questionnaire information, improve the reliability of the medical questionnaire information and reduce dirty data.

Description

TECHNICAL FIELD

[0001] The embodiment of the present application relates to the field of artificial intelligence, and especially relates to a medical questionnaire verification method and device, computer equipment and a storage medium. BACKGROUND

[0002] With the development of internet technology, the application of digital medicine is more and more extensive, especially in the aspect of medical information collection, intelligent medical questionnaires can improve the efficiency of medical information collection, such as patient information collection before face-to-face diagnosis, medical service evaluation after face-to-face diagnosis, etc.

[0003] The prior art can only identify the mutually exclusive answers between different questions through a series of rules for the verification of medical questionnaire information, but it is difficult to identify whether the answers of the user to the questionnaire match the basic information, therefore, the prior art cannot verify the inconsistency of the information before and after the medical questionnaire, which leads to inaccurate analysis results of the medical questionnaire. SUMMARY

[0004] The embodiment of the present application provides a medical questionnaire verification method, device, computer equipment and storage medium which can improve the accuracy of medical questionnaire verification.

[0005] To solve the above technical problems, an embodiment of the present application provides a technical solution: providing a medical questionnaire verification method, comprising:

[0006] receiving a medical questionnaire verification request and obtaining initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information comprises question information, an answer speech file and answerer basic information;

[0007] performing speech recognition on the answer speech file to obtain answer text information, and generating target question and answer information according to the question information;

[0008] encoding the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector;

[0009] performing consistency verification on the basic information vector and the question and answer information vector through a trained medical questionnaire verification model to obtain a verification result;

[0010] generating target medical questionnaire information based on the verification result, wherein the target medical questionnaire information is used to indicate the medical questionnaire information in which the answer text information is consistent with the answerer basic information.

[0011] Optionally, the receiving a medical questionnaire verification request and obtaining initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information comprises question information, an answer speech file and answerer basic information comprises:

[0012] receive a medical questionnaire verification request, and obtain question information from a preset medical questionnaire library according to a questionnaire identifier in the medical questionnaire verification request;

[0013] obtain a file storage path through the questionnaire identifier, and find an answering voice file corresponding to the voice storage path from a preset distributed file system;

[0014] obtain corresponding answerer basic information from a preset answerer information library through the questionnaire identifier, and combine the question information, the answering voice file and the answerer basic information to obtain initial medical questionnaire information.

[0015] Optionally, the voice recognition on the answering voice file is performed to obtain answer text information, and target question and answer information is generated according to the question information, and the target question and answer information includes at least one question and answer pair.

[0016] the answering voice file is windowed and framed through a preset Hamming window to obtain a plurality of voice signals, and fast Fourier transform is performed on each voice signal to obtain a power spectrum corresponding to each voice signal;

[0017] each voice signal is frequency filtered through a preset Mel spectrum filter to obtain a target Mel spectrum, and a mean square error between the target Mel spectrum and a preset voice Mel spectrum is calculated through a Gaussian function to obtain voice information corresponding to the preset voice Mel spectrum with the minimum mean square error, and the voice information is used to indicate answer text information corresponding to the answering voice file;

[0018] each question in the question information is associated with corresponding answer text information according to a question identifier in the question information to obtain target question and answer information, and the target question and answer information includes at least one question and answer pair.

[0019] Optionally, the answerer basic information and the target question and answer information are encoded to obtain a basic information vector and a question and answer information vector, and the target question and answer information includes at least one question and answer pair.

[0020] the target question and answer information is vectorized through a preset sentence vector model to obtain a question and answer information vector;

[0021] the answerer basic information is encoded through a one-hot encoding algorithm to obtain basic information encoding;

[0022] the basic information encoding is dimensionally upgraded according to a dimension of the question and answer information vector to obtain a basic information vector, and a dimension of the basic information vector is the same as that of the question and answer information vector.

[0023] Optionally, the consistency verification of the basic information vector and the question and answer information vector is performed by using the trained medical questionnaire verification model, and a verification result is obtained, which includes:

[0024] The basic information vector and the question and answer information vector are input into the trained medical questionnaire verification model, and the medical questionnaire verification model includes a bidirectional encoding transformation network.

[0025] The basic information vector and the question and answer information vector are classified for context consistency by using the bidirectional encoding transformation network in the medical questionnaire verification model, and a classification result is obtained.

[0026] Based on the classification result, a verification result corresponding to the initial medical questionnaire information is generated.

[0027] Optionally, based on the verification result, target medical questionnaire information is generated, and the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the basic information of the answerer, which includes:

[0028] Based on the verification result, noise data classification is performed on the initial medical questionnaire information, and noise question and answer information is obtained.

[0029] The noise question and answer information corresponding to the question information and the answer speech file in the initial medical questionnaire information is removed, and target medical questionnaire information is obtained, and the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the basic information of the answerer.

[0030] Optionally, after the target medical questionnaire information is generated based on the verification result, the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the basic information of the answerer, the method further includes:

[0031] The target medical questionnaire information is mapped to a text feature space by using a trained semantic recognition model, and a medical questionnaire feature vector is obtained.

[0032] The medical questionnaire feature vector is analyzed for context semantics by using a long short-term memory network layer in the semantic recognition model, and a plurality of semantic labels are obtained.

[0033] Based on each semantic label, a target medical information library is updated.

[0034] To solve the above technical problems, an embodiment of the present application further provides a medical questionnaire verification device, which includes:

[0035] receive a medical questionnaire verification request, and obtain initial medical questionnaire information through the medical questionnaire verification request, the initial medical questionnaire information comprising question information, an answer speech file and answerer basic information;

[0036] an identification module configured to perform speech recognition on the answer speech file to obtain answer text information, and generate target question and answer information according to the question information;

[0037] an encoding module configured to perform encoding processing on the answerer basic information and the target question and answer information to obtain an answerer basic information vector and a question and answer information vector;

[0038] a verification module configured to perform consistency verification on the answerer basic information vector and the question and answer information vector through a trained medical questionnaire verification model to obtain a verification result;

[0039] a generation module configured to generate target medical questionnaire information based on the verification result, the target medical questionnaire information being used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information.

[0040] Optionally, the receiving module is specifically configured to:

[0041] receive a medical questionnaire verification request, and obtain question information from a preset medical questionnaire library according to a questionnaire identifier in the medical questionnaire verification request;

[0042] obtain a file storage path through the questionnaire identifier, and find an answer speech file corresponding to the speech storage path from a preset distributed file system;

[0043] obtain corresponding answerer basic information from a preset answerer information library through the questionnaire identifier, and combine the question information, the answer speech file and the answerer basic information to obtain initial medical questionnaire information.

[0044] Optionally, the identification module is specifically configured to:

[0045] perform windowing and framing on the answer speech file through a preset Hamming window to obtain a plurality of speech signals, and perform fast Fourier transform on each speech signal to obtain a power spectrum corresponding to each speech signal;

[0046] perform frequency band filtering on the power spectrum corresponding to each speech signal through a preset Mel spectrum filter to obtain a target Mel spectrum, and calculate a mean square error between the target Mel spectrum and a preset speech Mel spectrum through a Gaussian function to obtain speech information corresponding to the preset speech Mel spectrum with the minimum mean square error, the speech information being used to indicate answer text information corresponding to the answer speech file;

[0047] According to a question identifier in the question information, each question in the question information is associated with corresponding answer text information, to obtain target question and answer information, the target question and answer information including at least one question and answer pair.

[0048] Optionally, the encoding module is specifically used for:

[0049] The target question and answer information is vectorized by a preset sentence vector model, to obtain question and answer information vectors.

[0050] The answerer basic information is encoded by a one-hot encoding algorithm, to obtain basic information encoding.

[0051] The basic information encoding is dimensionally upgraded according to the dimension of the question and answer information vectors, to obtain basic information vectors, the dimension of the basic information vectors being the same as the dimension of the question and answer information vectors.

[0052] Optionally, the verification module is specifically used for:

[0053] The basic information vectors and the question and answer information vectors are input into a trained medical questionnaire verification model, the medical questionnaire verification model including a bidirectional encoding transformation network.

[0054] The basic information vectors and the question and answer information vectors are classified for context consistency by the bidirectional encoding transformation network in the medical questionnaire verification model, to obtain classification results.

[0055] Based on the classification results, a verification result corresponding to the initial medical questionnaire information is generated.

[0056] Optionally, the generation module is specifically used for:

[0057] Based on the verification result, noise data classification is performed on the initial medical questionnaire information, to obtain noise question and answer information.

[0058] The question information and the answer speech file corresponding to the noise question and answer information are removed from the initial medical questionnaire information, to obtain target medical questionnaire information, the target medical questionnaire information being used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information.

[0059] Optionally, the medical questionnaire verification device further includes:

[0060] A mapping module is configured to map the target medical questionnaire information to a text feature space by a trained semantic recognition model, to obtain medical questionnaire feature vectors.

[0061] The parsing module is configured to perform context semantic parsing on the medical questionnaire feature vector through a long short-term memory network layer in the semantic recognition model to obtain a plurality of semantic labels.

[0062] The updating module is configured to update a preset medical information library based on the semantic labels to obtain a target medical information library.

[0063] To solve the above technical problems, the embodiment of the present application further provides a computer device, comprising a memory and a processor, the memory stores computer readable instructions, and the computer readable instructions are executed by the processor to make the processor execute the steps of the medical questionnaire verification method.

[0064] To solve the above technical problems, the embodiment of the present application further provides a computer storage medium, and the computer readable instructions are executed by one or more processors to make the one or more processors execute the steps of the medical questionnaire verification method.

[0065] The beneficial effects of the embodiment of the present application are: through voice recognition on the answer voice file in the initial medical questionnaire information, the answer text information is obtained, and then the target question and answer information is generated based on the question information in the initial medical questionnaire information, and then the vectorization processing is performed on the target question and answer information and the basic information of the answerer to obtain the basic information vector and the question and answer information vector represented by numbers, and then the consistency verification is performed on the basic information vector and the question and answer information vector by the medical questionnaire verification model, and the target medical questionnaire information is generated based on the verification result, and the target medical questionnaire information represents that the answer text information made by the answerer is consistent with the basic information of the answerer, the present application can improve the accuracy of medical questionnaire verification, thereby reducing the dirty data in the medical questionnaire information and improving the reliability of the medical questionnaire information. BRIEF DESCRIPTION OF DRAWINGS

[0066] The above and / or additional aspects and advantages of the present application will become apparent and more readily appreciated from the following description of the embodiments, taken in conjunction with the accompanying drawings, in which:

[0067] Figure 1 The basic flowchart of the medical questionnaire verification method of one embodiment of the present application;

[0068] Figure 2 The schematic diagram of one embodiment of the medical questionnaire verification method in the present application;

[0069] Figure 3 The schematic diagram of another embodiment of the medical questionnaire verification method in the present application;

[0070] Figure 4 The basic structure schematic diagram of the medical questionnaire verification device of one embodiment of the present application;

[0071] Figure 5 Basic structure block diagram of computer device of one embodiment of the present application. DETAILED DESCRIPTION

[0072] Embodiments of the present application are described in detail below with reference to the attached drawing figures, wherein the same or like component have the same or similar reference numbers and meanings throughout the several views. The embodiments described below are exemplary only and are not to be construed as limiting the present application.

[0073] As those skilled in the art will readily appreciate from the disclosure herein, ordinary reference to the singular throughout this disclosure can cover a singular or plural aspect, unless otherwise indicated. In addition, unless otherwise indicated, the use of relational terms, if any, such as first, second, third, and the like, are used solely to distinguish one from another entity or action without necessarily implying a dogmatic order or sequence. In addition, the terms "comprises," "comprising," or any other variation thereof, are intended to cover a non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements does not include only those elements but can include other elements not expressly listed or inherent to such process, method, article, or apparatus. Furthermore, no element, component, or process in the present disclosure is intended to be indispensable or crucial.

[0074] As will be understood by those familiar with the art, all terms including technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art unless otherwise defined herein. It will be further understood that terms, such as those defined in commonly used dictionaries, should be interpreted as having a meaning that is consistent with their meaning in the context of the relevant art and will not be interpreted in an idealized or overly formal sense unless expressly so defined herein.

[0075] Those skilled in the art will understand that a "terminal" as used herein includes both a device that is solely a wireless signal receiver, and a device that has both receiving and transmitting hardware capable of performing two-way communication over a two-way communication link. Such devices can include cellular or other communications devices that have single line displays or multiple line displays, or no display, Personal Communications Service (PCS) devices that can combine a voice service with a data processing, facsimile, and / or data communications capabilities, Personal Digital Assistants (PDAs) that can include a radio frequency receiver, pagers, Internet / intranet access, Web browsers, organizers, calendars, and / or a Global Positioning System (GPS) receiver, conventional laptop and / or palmtop computers or other devices that have a radio frequency receiver. A "terminal" as used herein can be portable, transportable, installed in a vehicle (aeronautical, maritime, and / or land), or adapted to be operated locally and / or in a distributed fashion, on Earth and / or in any other location in space. A "terminal" as used herein can also be a communication terminal, an Internet terminal, a music / video playing terminal, such as a PDA, a Mobile Internet Device (MID), and / or a mobile phone with music / video playing function, a smart television, a set-top box, and the like.

[0076] Referring to Figure 1 , Figure 1 is a schematic diagram of a basic flow of a medical questionnaire verification method of the present embodiment. As shown in Figure 1 , a medical questionnaire verification method includes:

[0077] S100, receiving a medical questionnaire verification request, and obtaining initial medical questionnaire information through the medical questionnaire verification request, the initial medical questionnaire information including question information, answer speech files, and answerer basic information;

[0078] It should be noted that the medical questionnaire includes a questionnaire related to the medical field, such as a questionnaire on the use of remote medical equipment, a questionnaire on the satisfaction of medical services for the elderly, a questionnaire on the medical and health situation in rural areas, a questionnaire on the purchase intention of medical insurance, and a questionnaire on information collection before consultation, and the like, and the specific questionnaire is not limited. The medical questionnaire can be a questionnaire survey for a targeted population, or a questionnaire survey randomly distributed, wherein, for the targeted questionnaire survey, the basic information of the answerer of the targeted population is usually known information, and the server obtains the basic information of the answerer through the database of the targeted population, which can improve the efficiency of subsequent medical questionnaire information verification, such as a targeted questionnaire survey for a community population, and the specific questionnaire is not limited.

[0079] In an embodiment, in order to facilitate the completion of the medical questionnaire survey by the visually impaired population or the person who cannot write (disabled), the question information in the initial medical questionnaire information is played to the answerer terminal in the form of voice, and the answerer voice file, i.e., the answerer's answer voice file, is collected through the recording device of the answerer terminal, so as to obtain the answer information of each question in the question information, so that the applicability of the medical questionnaire survey is stronger, the coverage is wider, and the information accuracy rate collected by the medical questionnaire is improved.

[0080] In an embodiment, since the basic information of the answerer, such as name, ID number, gender, address, and the like, is usually collected before the answerer answers the medical questionnaire, the basic information of the answerer in the initial medical questionnaire information is extracted through the medical questionnaire verification request, which can improve the efficiency of subsequent medical questionnaire verification. In another embodiment, the relevant medical information of the answerer is collected from the pre-set medical database through the identity in the basic information of the answerer, the medical information of the answerer is obtained, and the initial basic information of the answerer and the medical information of the answerer are combined to obtain more perfect basic information of the answerer. The embodiment can improve the accuracy of medical questionnaire verification by improving the perfection of the answerer information.

[0081] S200, performing voice recognition on the answer voice file to obtain answer text information, and generating target question and answer information according to the question information;

[0082] In an implementation, in order to verify the consistency of the medical questionnaire, the answer speech file is subjected to speech recognition by a preset speech recognition algorithm to obtain text content corresponding to the answer speech file, i.e., answer text information, wherein the preset speech recognition algorithm includes a dynamic time warping (DTW) algorithm and a hidden markov model (HMM) speech recognition algorithm, wherein the answer text information can be quickly and accurately obtained by subjecting the answer speech file to speech recognition by the dynamic time warping algorithm, thereby improving the efficiency of subsequent medical questionnaire consistency verification.

[0083] In an implementation, the answer text information is spliced according to the order of the questions to obtain target answer information, each question in the question information is spliced according to the question sequence to obtain target question information, and finally the target answer information and the target question information are merged to obtain target question and answer information. In another implementation, each question in the question information is spliced with the corresponding answer in the answer text information to generate the target question and answer information, which is not limited in detail. The generation of the target question and answer information from the question information can establish the association between the questions and the answers, thereby improving the efficiency of subsequent medical questionnaire consistency verification.

[0084] S300, encoding the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector;

[0085] It should be noted that, in order to improve the efficiency of medical questionnaire consistency verification, the answerer basic information and the target question and answer information are subjected to vectorization processing to obtain a basic information vector corresponding to the answerer basic information and a question and answer information vector corresponding to the target question and answer information. In an implementation, since the medical survey questionnaire is usually a multiple-choice question, the target question and answer information is subjected to encoding according to the answer option in the target question and answer information and the question number in the preset question bank to obtain the question and answer information vector, such as 1230010 for a question number of 123 and an answer option of C in the preset question bank, which is not limited in detail. The use of short vector encoding can improve the data processing efficiency, thereby improving the efficiency of medical questionnaire consistency verification.

[0086] In an embodiment, a word embedding algorithm is used to encode the basic information of the answerer and the target question and answer information to obtain a basic information vector and a question and answer information vector. The essence of the word embedding is a kind of mapping, and the word embedding algorithm maps the semantic space to the vector space by establishing the association between words, and at the same time, the relationship in the semantic space is also preserved as much as possible in the vector space, such as the positions of two words with close semantics in the vector space are also close. Through the encoding of the word embedding algorithm, the semantics of the basic information of the answerer and the target question and answer information can be preserved while providing verification basis for semantic consistency.

[0087] Specifically, the step S300 includes: vectorizing the target question and answer information through a preset sentence vector model to obtain a question and answer information vector; encoding the basic information of the answerer through a one-hot encoding algorithm to obtain a basic information code; and performing dimension increasing processing on the basic information code according to the dimension of the question and answer information vector to obtain a basic information vector, the dimension of the basic information vector being the same as the dimension of the question and answer information vector.

[0088] In the embodiment, the preset sentence vector model is one of natural language processing models, has a recurrent neural network structure, and continuously recurs to obtain a vector representation of the entire sentence sequence, i.e., a question and answer information vector, according to a given external topological structure (such as a constituent syntax tree). Then the basic information of the answerer is converted into a basic information code through a one-hot encoding algorithm. Since the dimension of the one-hot encoding is lower than the dimension of the sentence vector, in order to make the question and answer information vector and the basic information code be in the same dimension, the dimension increasing processing is performed on the basic information code, i.e., the basic information code is filled with 0 according to the dimension of the question and answer information vector. For example, if the question and answer information vector is 768-dimensional, the basic information code is supplemented to be 768-dimensional, so as to obtain the basic information vector.

[0089] S400, performing consistency verification on the basic information vector and the question and answer information vector through the trained medical questionnaire verification model to obtain a verification result;

[0090] It should be noted that the medical questionnaire verification model is used to verify whether the semantics of the question and answer information vector and the basic information vector are consistent. In the training stage of the medical questionnaire verification model, the labeled training sample data is obtained according to whether the basic information of the user and the questionnaire question and answer are consistent, the initialized BERT (bidirectional encoder representations from transformers) model is inputted with the training sample data for model training, and the learned model parameters are evaluated through the labeled verification sample data, so as to obtain the trained medical questionnaire verification model.

[0091] It should be noted that the structure of the medical questionnaire verification model is stacked by multiple transformer encoder encoding networks, and the consistency classification of the basic information vector and the question and answer information vector is performed through the encoding network of each transformer in the medical questionnaire verification model to obtain a consistency classification result, i.e., a verification result. For example, the data label output by the context consistent question and answer information vector is 1, and the inconsistent one is 0, which is not limited. The consistency classification refers to the semantic consistency of the question and answer information vector and the basic information vector. For example, the age of the answerer in the basic information vector is 14 years old, but the question and answer information vector indicates that the answerer is a job seeker with years of work experience. Therefore, the semantic consistency of the basic information vector and the question and answer information vector is not consistent. The embodiment can quickly verify the medical questionnaire through the language representation model BERT, and has an accuracy of more than 98.8%. The combination ability of the model and the scene is good.

[0092] Specifically, step S400 includes: inputting the basic information vector and the question and answer information vector into the trained medical questionnaire verification model, the medical questionnaire verification model including a bidirectional encoding transformation network; performing context consistency classification on the basic information vector and the question and answer information vector through the bidirectional encoding transformation network in the medical questionnaire verification model to obtain a classification result; and generating a verification result corresponding to the initial medical questionnaire information based on the classification result.

[0093] In the embodiment, the medical questionnaire verification model has a bidirectional encoding transformation network and a full connection network, such as a BERT bidirectional encoding transformation network. The bidirectional encoding transformation network in the medical questionnaire verification model is used to perform context consistency identification on the basic information vector and the question and answer information vector to obtain an identification result, and the full connection network is used to classify the identification result to obtain a classification result. In order to improve the verification efficiency of the medical questionnaire, the classification result is combined into a verification result corresponding to the initial medical questionnaire information. In one embodiment, the verification result is a classification result sequence arranged in the order of the question number in the initial medical questionnaire information, so that the verification result data is integrated, and subsequent medical questionnaire information generation is facilitated.

[0094] S500, based on the verification result, generating target medical questionnaire information, the target medical questionnaire information being used to indicate the medical questionnaire information in which the answer text information is consistent with the answerer basic information.

[0095] It should be noted that after obtaining the verification result, the initial medical questionnaire information is denoised by the verification result to eliminate the noise data in the initial medical questionnaire information. For example, for the question and answer information that is inconsistent with the basic information vector, it is marked as noise data, and for the question and answer information that is consistent with the basic information vector, it is marked as normal data, so as to obtain the target medical questionnaire information with noise type marking. When all the target medical questionnaire information is summarized, counted or analyzed subsequently, the medical questionnaire information with the answer text information consistent with the basic information of the answerer can be quickly obtained through the noise type marking, thereby improving the efficiency of subsequent medical questionnaire information processing.

[0096] Specifically, step S500 includes: based on the verification result, classifying noise data of the initial medical questionnaire information to obtain noise question and answer information; removing the question information and answer speech file corresponding to the noise question and answer information in the initial medical questionnaire information to obtain target medical questionnaire information, the target medical questionnaire information is used to indicate the medical questionnaire information with the answer text information consistent with the basic information of the answerer.

[0097] In the embodiment, since the verification result includes the consistency verification result of each question in the initial medical questionnaire information, the noise data of the initial medical questionnaire information is classified by the verification result to obtain the noise question and answer information, i.e. the noise data that the question information is inconsistent with the basic information of the answerer. Then, the question information and answer speech file corresponding to the noise question and answer information in the initial medical questionnaire information are removed, so as to obtain the target medical questionnaire information without noise data, i.e. the medical questionnaire information with the answer text information consistent with the basic information of the answerer. The embodiment can also batch generate the target medical questionnaire information, so as to improve the analysis efficiency of the target medical questionnaire information subsequently.

[0098] Further, after step S500, it further includes:

[0099] mapping the target medical questionnaire information to a text feature space by the trained semantic recognition model to obtain a medical questionnaire feature vector; performing context semantic analysis on the medical questionnaire feature vector by a long short-term memory network layer in the semantic recognition model to obtain a plurality of semantic labels; and updating a preset medical information library based on each semantic label to obtain a target medical information library.

[0100] In this embodiment, in order to effectively utilize the target medical questionnaire information, the semantic label extraction is performed on the target medical questionnaire information by using the trained semantic recognition model, wherein the trained semantic recognition model includes a convolutional neural network (CNN) layer and a long short-term memory (LSTM) layer. The target medical questionnaire information is mapped to a text feature space by using the convolutional neural network layer, and a medical questionnaire feature vector is obtained. The context semantic analysis is performed on the medical questionnaire feature vector by using the long short-term memory network layer, and a plurality of semantic labels are obtained. The data of the preset medical information database is updated based on each semantic label, and a target medical information database is obtained. This embodiment can extract effective information in the target medical questionnaire information, and the medical information database is improved.

[0101] The above embodiment can obtain answer text information by performing speech recognition on the answer speech file in the initial medical questionnaire information, generate target question and answer information based on the question information in the initial medical questionnaire information, then perform vectorization processing on the target question and answer information and the respondent basic information, and obtain a digital representation of the basic information vector and the question and answer information vector. The consistency verification is performed on the basic information vector and the question and answer information vector by using the medical questionnaire verification model, and the target medical questionnaire information is generated based on the verification result. The target medical questionnaire information indicates that the answer text information made by the respondent is consistent with the basic information of the respondent. The application can improve the accuracy of medical questionnaire verification, reduce the dirty data in the medical questionnaire information, and improve the reliability of the medical questionnaire information.

[0102] Referring to Figure 2 , one embodiment of the medical questionnaire verification method in the embodiment of the application includes:

[0103] As Figure 2 shown, S100 includes:

[0104] S110, receiving a medical questionnaire verification request, and obtaining question information from a preset medical questionnaire database according to a questionnaire identifier in the medical questionnaire verification request;

[0105] In an embodiment, different respondents correspond to different question information, in the process of answering the questions by the respondents, the next question information is automatically generated according to the respondent basic information and / or the answer (the question language file) corresponding to the last question information, when the respondent completes the answer to all the question information, the initial medical questionnaire is generated, and the question identifier corresponding to each question information and the questionnaire identifier of the initial medical questionnaire are stored in the preset medical questionnaire library in association, when the medical questionnaire verification request is received, the question identifier corresponding to the medical questionnaire identifier is found from the preset medical questionnaire library through the questionnaire identifier in the medical questionnaire verification request, and the corresponding question information is determined through the question identifier, so that the flexible medical questionnaire related information is quickly obtained, the storage structure is optimized, and each medical questionnaire does not need to be stored separately, so that the demand amount of the storage space is reduced.

[0106] S120, a file storage path is obtained through the questionnaire identifier, and a respondent voice file corresponding to the voice storage path is found from the preset distributed file system;

[0107] In the embodiment, since the respondent voice file needs a larger storage space relative to the text form file, if all the respondent voice files are stored in the same file system, the data access will be slow and the processing efficiency will be low, therefore, in order to improve the searching efficiency of the voice file, the respondent voice file is stored in the distributed file system, and each respondent voice file is associated with the questionnaire identifier of the corresponding respondent, when the respondent voice file needs to be obtained, the file storage path of the respondent voice file in the distributed file system is obtained through the questionnaire identifier, and the respondent voice file is obtained from the file storage path, so that the reading efficiency of the voice file is improved, and the verification efficiency of the subsequent medical questionnaire is improved.

[0108] S130, the respondent basic information corresponding to the questionnaire identifier is obtained from the preset respondent information library, and the question information, the respondent voice file and the respondent basic information are combined to obtain the initial medical questionnaire information.

[0109] It should be noted that, in order to improve the accuracy of medical questionnaire verification and reduce dirty data in the medical questionnaire, the corresponding respondent basic information is obtained from the pre-set respondent information library through the questionnaire identifier. In an embodiment, the respondent provides answers to the related questions of the respondent basic information in the process of answering the questions. Then, the corresponding respondent basic information can be directly obtained from the pre-set respondent information library through the questionnaire identifier. In another embodiment, the respondent only provides the respondent basic information related to the identity identifier. Then, the respondent basic information is collected from the pre-set respondent information library through the identity identifier, so as to obtain more complete respondent basic information. Finally, the question information, the answering voice file and the respondent basic information are combined to obtain the initial medical questionnaire information. The embodiment can obtain complete respondent basic information, thereby improving the accuracy of medical questionnaire verification and improving the precision of the medical questionnaire.

[0110] The above embodiment improves the speech recognition performance of the answering voice file in the initial medical questionnaire information through the distributed file system, thereby obtaining the answer text information. Then, the target question and answer information is generated based on the question information in the initial medical questionnaire information. Next, the vectorization processing is performed on the target question and answer information and the respondent basic information to obtain the digital representation of the basic information vector and the question and answer information vector. Then, the consistency verification is performed on the basic information vector and the question and answer information vector through the medical questionnaire verification model. The target medical questionnaire information is generated based on the verification result. The target medical questionnaire information represents that the answer text information made by the respondent is consistent with the respondent basic information. The embodiment can improve the accuracy and efficiency of medical questionnaire verification, thereby reducing the dirty data in the medical questionnaire information and improving the reliability of the medical questionnaire information.

[0111] Please refer to Figure 3 An embodiment of the medical questionnaire verification method in the embodiment of the present application includes:

[0112] As Figure 3 shown, S200 includes:

[0113] S210, the answering voice file is windowed and framed through the pre-set Hamming window to obtain a plurality of voice signals, and the fast Fourier transform is performed on each voice signal to obtain the power spectrum corresponding to each voice signal;

[0114] In this embodiment, in order to frame the answer speech file smoothly, reduce the influence of discontinuous signal on the accuracy of subsequent fast Fourier transform, and obtain a plurality of speech signals by windowing and framing processing of the answer speech file through the Hamming window, wherein the preset parameters of the Hamming window include window length, offset and shape, based on the preset parameters of the Hamming window, the edge smoothness is set to 0, that is, a plurality of speech signals with smooth transition can be collected. Therefore, after multiplying the Hamming window, the fast Fourier transform is performed on each speech signal to obtain the energy distribution in the frequency spectrum, that is, the power spectrum corresponding to each speech signal. The embodiment can improve the smoothness of the answer speech file framing, and further improve the accuracy of speech recognition.

[0115] In this embodiment, in order to improve the performance of speech recognition, the power spectrum corresponding to each speech signal is mapped to the mel frequency scale through the preset mel spectrum filter, and the harmonics are eliminated to highlight the formants of the original sound, and the target mel spectrum is obtained. Finally, the mean square error between the target mel spectrum and the preset speech mel spectrum is calculated through the Gaussian function, the target speech mel spectrum with the minimum mean square error is obtained, and the speech information corresponding to the target speech mel spectrum is set as the speech information corresponding to the answer speech file, that is, the answer text information. The embodiment can simulate the sensitivity of human ear hearing to different frequency bands through the mel spectrum filter to improve the performance of speech recognition, and further improve the efficiency of medical questionnaire verification.

[0116] In this embodiment, in order to improve the performance of speech recognition, the power spectrum corresponding to each speech signal is mapped to the mel frequency scale through the preset mel spectrum filter, and the harmonics are eliminated to highlight the formants of the original sound, and the target mel spectrum is obtained. Finally, the mean square error between the target mel spectrum and the preset speech mel spectrum is calculated through the Gaussian function, the target speech mel spectrum with the minimum mean square error is obtained, and the speech information corresponding to the target speech mel spectrum is set as the speech information corresponding to the answer speech file, that is, the answer text information. The embodiment can simulate the sensitivity of human ear hearing to different frequency bands through the mel spectrum filter to improve the performance of speech recognition, and further improve the efficiency of medical questionnaire verification.

[0117] S230, according to the question identifier in the question information, each question in the question information is associated with the corresponding answer text information to obtain target question and answer information, and the target question and answer information includes at least one question and answer pair.

[0118] In this embodiment, in order to improve the verification efficiency of each question information in the initial medical questionnaire information, each question information is associated and bound with the corresponding answer text information to obtain target question and answer information, wherein the target question and answer information includes at least one question and answer pair, that is, the question information-answer text information is formed according to the question identifier, such as the binding of question 1 and the corresponding answer, the binding of question 2 and the corresponding answer, which is not limited.

[0119] The above embodiment performs speech recognition on the answer speech file in the initial medical questionnaire information through a Hamming window and a Mel spectrum filter to obtain answer text information, generates target question and answer information based on the question information in the initial medical questionnaire information, then performs vectorization processing on the target question and answer information and the answerer basic information to obtain a digital representation of the basic information vector and the question and answer information vector, performs consistency verification on the basic information vector and the question and answer information vector through the medical questionnaire verification model, and generates target medical questionnaire information based on the verification result, wherein the target medical questionnaire information indicates that the answer text information made by the answerer is consistent with the basic information of the answerer. The application can improve the accuracy of medical questionnaire verification, thereby reducing dirty data in the medical questionnaire information and improving the reliability of the medical questionnaire information.

[0120] To solve the above technical problems, the embodiment of the application further provides a medical questionnaire verification device. For details, please refer to Figure 4 , Figure 4 The basic structure of the medical questionnaire verification device is shown in the figure.

[0121] As Figure 4 shown, a medical questionnaire verification device includes a receiving module 1100, an identification module 1200, an encoding module 1300, a verification module 1400, and a generation module 1500. The receiving module 1100 is configured to receive a medical questionnaire verification request and obtain initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information includes question information, an answer speech file, and answerer basic information. The identification module 1200 is configured to perform speech recognition on the answer speech file to obtain answer text information and generate target question and answer information according to the question information. The encoding module 1300 is configured to perform encoding processing on the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector. The verification module 1400 is configured to perform consistency verification on the basic information vector and the question and answer information vector through a trained medical questionnaire verification model to obtain a verification result. The generation module 1500 is configured to generate target medical questionnaire information based on the verification result, wherein the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information.

[0122] The medical questionnaire verification device obtains answer text information by performing voice recognition on the answer voice file in the initial medical questionnaire information, generates target question and answer information based on the question information in the initial medical questionnaire information, then performs vectorization processing on the target question and answer information and the basic information of the answerer to obtain a basic information vector and a question and answer information vector in a digital form, performs consistency verification on the basic information vector and the question and answer information vector through the medical questionnaire verification model, and generates target medical questionnaire information based on the verification result, wherein the target medical questionnaire information indicates that the answer text information made by the answerer is consistent with the basic information of the answerer, and the medical questionnaire verification accuracy can be improved, so that the dirty data in the medical questionnaire information is reduced and the reliability of the medical questionnaire information is improved.

[0123] In some embodiments, the receiving module 1100 is specifically configured to receive a medical questionnaire verification request, and acquire question information from a preset medical questionnaire library according to a questionnaire identifier in the medical questionnaire verification request; acquire a file storage path through the questionnaire identifier, and find an answer voice file corresponding to the voice storage path from a preset distributed file system; acquire corresponding basic information of an answerer from a preset answerer information library through the questionnaire identifier, and combine the question information, the answer voice file and the basic information of the answerer to obtain initial medical questionnaire information.

[0124] In some embodiments, the identification module 1200 is specifically configured to perform windowing and framing on the answer voice file through a preset Hamming window to obtain a plurality of voice signals, and perform fast Fourier transform on each voice signal to obtain a power spectrum corresponding to each voice signal; perform frequency band filtering on the power spectrum corresponding to each voice signal through a preset Mel spectrum filter to obtain a target Mel spectrum, and calculate the mean square error between the target Mel spectrum and a preset voice Mel spectrum through a Gaussian function to obtain voice information corresponding to the preset voice Mel spectrum with the minimum mean square error, wherein the voice information is used to indicate answer text information corresponding to the answer voice file; according to a question identifier in the question information, each question in the question information is associated with corresponding answer text information to obtain target question and answer information, wherein the target question and answer information includes at least one question and answer pair.

[0125] In some embodiments, the encoding module 1300 is specifically configured to perform vectorization processing on the target question and answer information through a preset sentence vector model to obtain a question and answer information vector; encode the basic information of the answerer through a one-hot encoding algorithm to obtain a basic information code; and perform dimension upgrading processing on the basic information code according to the dimension of the question and answer information vector to obtain a basic information vector, wherein the dimension of the basic information vector is the same as the dimension of the question and answer information vector.

[0126] In some embodiments, the verification module 1400 is specifically used to: input the basic information vector and the question-answer information vector into a trained medical questionnaire verification model, the medical questionnaire verification model including a bidirectional encoding transformation network; perform context consistency classification on the basic information vector and the question-answer information vector through the bidirectional encoding transformation network in the medical questionnaire verification model to obtain a classification result; based on the classification result, generate a verification result corresponding to the initial medical questionnaire information.

[0127] In some embodiments, the generation module 1500 is specifically used to: based on the verification result, classify the initial medical questionnaire information into noise data to obtain noise question and answer information; remove the question information and answer voice file corresponding to the noise question and answer information in the initial medical questionnaire information to obtain target medical questionnaire information, and the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the basic information of the respondent.

[0128] In some embodiments, the medical questionnaire verification device also includes: a mapping module 1600, which is used to map the target medical questionnaire information to a text feature space through a trained semantic recognition model to obtain a medical questionnaire feature vector; a parsing module 1700, which is used to perform contextual semantic parsing on the medical questionnaire feature vector through the long short-term memory network layer in the semantic recognition model to obtain multiple semantic labels; and an updating module 1800, which is used to update the status of a preset medical information library based on each semantic label to obtain a target medical information library.

[0129] To solve the above technical problems, the embodiment of the present invention also provides a computer device. Figure 5 , Figure 5 This is a basic structural block diagram of the computer device in this embodiment.

[0130] like Figure 5 As shown, a schematic diagram of the internal structure of a computer device. The computer device includes a processor, a non-volatile storage medium, a memory and a network interface connected via a system bus. Among them, the non-volatile storage medium of the computer device stores an operating system, a database and computer-readable instructions, and the database may store a control information sequence. When the computer-readable instructions are executed by the processor, the processor can implement a medical questionnaire verification method. The processor of the computer device is used to provide computing and control capabilities to support the operation of the entire computer device. The memory of the computer device may store computer-readable instructions. When the computer-readable instructions are executed by the processor, the processor can execute a medical questionnaire verification method. The network interface of the computer device is used to connect and communicate with the terminal. Those skilled in the art will understand that Figure 5The structure shown in the figure is only a block diagram of part of the structure related to the scheme of the present application, and does not constitute a limitation on the computer device to which the scheme of the present application is applied. The specific computer device can include more or fewer components than those shown in the figure, or combine certain components, or have a different arrangement of components.

[0131] The processor in the embodiment is configured to execute Figure 4 The specific functions of the receiving module 1100, the identifying module 1200, the encoding module 1300, the verifying module 1400 and the generating module 1500 in the embodiment are stored in the memory together with program codes and various data required for executing the above modules. The network interface is used for data transmission between the user terminal and the server. The memory in the embodiment stores program codes and data required for executing all sub-modules in the verification device of the medical questionnaire, and the server can call the program codes and data of the server to execute the functions of all sub-modules.

[0132] The computer device performs speech recognition on the answer speech file in the initial medical questionnaire information to obtain answer text information, generates target question and answer information based on the question information in the initial medical questionnaire information, then performs vectorization processing on the target question and answer information and the basic information of the answerer to obtain a basic information vector and a question and answer information vector represented in digital form, performs consistency verification on the basic information vector and the question and answer information vector through the medical questionnaire verification model, and generates target medical questionnaire information based on the verification result. The target medical questionnaire information indicates that the answer text information provided by the answerer is consistent with the basic information of the answerer. The present application can improve the accuracy of medical questionnaire verification, thereby reducing dirty data in the medical questionnaire information and improving the reliability of the medical questionnaire information.

[0133] The present application also provides a computer storage medium, wherein computer readable instructions are executed by one or more processors to make the one or more processors execute the steps of the medical questionnaire verification method of any one of the above embodiments.

[0134] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiments can be completed by a computer program instructing related hardware. The computer program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiments. The storage medium can be a non-volatile storage medium such as a magnetic disc, an optical disc, a read-only memory (ROM) or a random access memory (RAM).

[0135] The application further provides a computer storage medium, computer readable instructions executed by one or more processors cause the one or more processors to perform the steps of the verification method of the medical questionnaire of any of the above embodiments.

[0136] Those skilled in the art can understand that all or part of the processes in the above-mentioned embodiment methods can be completed by instructing the relevant hardware through a computer program, and the computer program can be stored in a computer readable storage medium. When the program is executed, it can include the processes of the above-mentioned embodiment methods. The storage medium can be a non-volatile storage medium such as a magnetic disk, an optical disk, a read-only memory (ROM), or a random access memory (RAM).

[0137] Those skilled in the art can understand that the steps, measures, and schemes in the various operations, methods, and processes discussed in the present application can be alternated, changed, combined, or deleted. Further, other steps, measures, and schemes in the various operations, methods, and processes discussed in the present application can also be alternated, changed, rearranged, decomposed, combined, or deleted. Further, the steps, measures, and schemes in the various operations, methods, and processes in the prior art can also be alternated, changed, rearranged, decomposed, combined, or deleted.

[0138] The above only describes some embodiments of the present application. It should be pointed out that those skilled in the art can make several improvements and refinements without departing from the principles of the present application, and these improvements and refinements should also be considered as the protection scope of the present application.

Claims

1. A method of validating a medical questionnaire, characterized by, The method comprises the following steps: receiving a medical questionnaire verification request and obtaining initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information comprises question information, answer speech files and answerer basic information; performing speech recognition on the answer speech files to obtain answer text information, and generating target question and answer information according to the question information; encoding the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector; performing consistency verification on the basic information vector and the question and answer information vector through a trained medical questionnaire verification model to obtain a verification result; generating target medical questionnaire information based on the verification result, wherein the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information; the encoding of the answerer basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector comprises: vectorizing the target question and answer information through a preset sentence vector model to obtain a question and answer information vector; encoding the answerer basic information through a one-hot encoding algorithm to obtain a basic information code; dimensionally upgrading the basic information code to obtain a basic information vector according to the dimension of the question and answer information vector, wherein the dimension of the basic information vector is the same as that of the question and answer information vector; the generation of target medical questionnaire information based on the verification result, wherein the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information comprises: classifying noise data of the initial medical questionnaire information based on the verification result to obtain noise question and answer information, wherein the question and answer information that is inconsistent with the basic information vector is noise question and answer information; removing question information and answer speech files corresponding to the noise question and answer information in the initial medical questionnaire information to obtain target medical questionnaire information, wherein the target medical questionnaire information is used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information.

2. The medical questionnaire verification method according to claim 1, characterized by, the receiving of a medical questionnaire verification request and the obtaining of initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information comprises question information, answer speech files and answerer basic information comprises: receiving a medical questionnaire verification request and obtaining question information from a preset medical questionnaire library according to a questionnaire identifier in the medical questionnaire verification request; obtaining a file storage path through the questionnaire identifier, and searching for an answer speech file corresponding to the file storage path from a preset distributed file system; obtaining corresponding answerer basic information from a preset answerer information library through the questionnaire identifier, and merging the question information, the answer speech files and the answerer basic information to obtain initial medical questionnaire information.

3. The medical questionnaire verification method of claim 1, wherein, the performing of speech recognition on the answer speech files to obtain answer text information, and the generation of target question and answer information according to the question information comprises: The answer speech file is windowed and framed by a preset Hamming window to obtain multiple speech signals, and fast Fourier transform is performed on each speech signal to obtain a power spectrum corresponding to each speech signal; A target mel spectrum is obtained by performing frequency band filtering on the power spectrum corresponding to each speech signal through a preset mel spectrum filter, and a preset speech mel spectrum corresponding to a speech information with a minimum mean square error between the target mel spectrum and the preset speech mel spectrum is obtained by calculating the mean square error between the target mel spectrum and the preset speech mel spectrum through a Gaussian function, wherein the speech information is used to indicate the answer text information corresponding to the answer speech file; According to the question identifier in the question information, each question in the question information is associated with the corresponding answer text information to obtain target question and answer information, and the target question and answer information includes at least one question and answer pair.

4. The medical questionnaire verification method of claim 1, wherein, The consistency of the basic information vector and the question and answer information vector is verified by the trained medical questionnaire verification model to obtain a verification result, which includes: The basic information vector and the question and answer information vector are input into the trained medical questionnaire verification model, and the medical questionnaire verification model includes a bidirectional encoding transformation network; The context consistency of the basic information vector and the question and answer information vector is classified by the bidirectional encoding transformation network in the medical questionnaire verification model to obtain a classification result; Based on the classification result, a verification result corresponding to the initial medical questionnaire information is generated.

5. The method of validating a medical questionnaire according to any one of claims 1 to 4, characterized in that, In addition to the medical questionnaire information corresponding to the answer text information consistent with the answerer's basic information, the target medical questionnaire information is used to indicate that the answer text information is consistent with the answerer's basic information, and the target medical questionnaire information is generated based on the verification result. The target medical questionnaire information is mapped to a text feature space by a trained semantic recognition model to obtain a medical questionnaire feature vector; The context semantic analysis of the medical questionnaire feature vector is performed by a long short-term memory network layer in the semantic recognition model to obtain multiple semantic labels; Based on each semantic label, the state of the preset medical information library is updated to obtain a target medical information library.

6. A medical questionnaire verification apparatus characterized by comprising: It includes: The receiving module is used for receiving a medical questionnaire verification request and obtaining initial medical questionnaire information through the medical questionnaire verification request, wherein the initial medical questionnaire information includes question information, answer speech file and answerer's basic information; The recognition module is used for performing speech recognition on the answer speech file to obtain answer text information, and generating target question and answer information according to the question information; The encoding module is used for encoding the answerer's basic information and the target question and answer information to obtain a basic information vector and a question and answer information vector; The verification module is used for verifying the consistency of the basic information vector and the question and answer information vector by the trained medical questionnaire verification model to obtain a verification result; The generation module is used for generating target medical questionnaire information based on the verification result, and the target medical questionnaire information is used to indicate the medical questionnaire information consistent with the answer text information and the answerer's basic information; The encoding of the answerer's basic information and the target question and answer information includes: vectorize the target question and answer information through a preset sentence vector model to obtain question and answer information vectors; encode the answerer basic information through a one-hot encoding algorithm to obtain basic information encoding; dimensionally upgrade the basic information encoding according to the dimensions of the question and answer information vectors to obtain basic information vectors, the dimensions of the basic information vectors being the same as the dimensions of the question and answer information vectors; based on the verification result, generate target medical questionnaire information, the target medical questionnaire information being used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information, and the target medical questionnaire information comprising: based on the verification result, classify noise data of the initial medical questionnaire information to obtain noise question and answer information, wherein the question and answer information that is inconsistent with the basic information vectors is noise question and answer information; remove question information and answer speech files corresponding to the noise question and answer information in the initial medical questionnaire information to obtain target medical questionnaire information, the target medical questionnaire information being used to indicate medical questionnaire information in which the answer text information is consistent with the answerer basic information.

7. A computer device, comprising: a device comprising a memory and a processor, the memory storing computer readable instructions, the computer readable instructions being executed by the processor to cause the processor to perform the steps of the medical questionnaire verification method according to any one of claims 1 to 5.

8. A computer storage medium, characterized in that computer readable instructions being executed by one or more processors to cause the one or more processors to perform the steps of the medical questionnaire verification method according to any one of claims 1 to 5.

Images