CN112948559B - Audio identification method and system for assisting pharmacy service staff - Google Patents
Audio identification method and system for assisting pharmacy service staff Download PDFInfo
- Publication number
- CN112948559B CN112948559B CN202110302111.8A CN202110302111A CN112948559B CN 112948559 B CN112948559 B CN 112948559B CN 202110302111 A CN202110302111 A CN 202110302111A CN 112948559 B CN112948559 B CN 112948559B
- Authority
- CN
- China
- Prior art keywords
- symptoms
- symptom
- sentence
- sound
- keyword
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 39
- 208000024891 symptom Diseases 0.000 claims abstract description 149
- 239000003814 drug Substances 0.000 claims abstract description 89
- 229940079593 drug Drugs 0.000 claims abstract description 64
- 208000037265 diseases, disorders, signs and symptoms Diseases 0.000 claims abstract description 30
- 201000010099 disease Diseases 0.000 claims abstract description 28
- 239000011159 matrix material Substances 0.000 claims description 12
- 238000004458 analytical method Methods 0.000 claims description 11
- 238000012545 processing Methods 0.000 claims description 7
- 238000004590 computer program Methods 0.000 claims description 6
- 230000008569 process Effects 0.000 claims description 6
- 238000005259 measurement Methods 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 2
- 239000000284 extract Substances 0.000 abstract description 3
- 238000004422 calculation algorithm Methods 0.000 description 11
- 238000012549 training Methods 0.000 description 11
- 230000011218 segmentation Effects 0.000 description 7
- 238000000605 extraction Methods 0.000 description 5
- 230000000694 effects Effects 0.000 description 4
- 230000036541 health Effects 0.000 description 4
- 238000003745 diagnosis Methods 0.000 description 3
- 206010037660 Pyrexia Diseases 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007547 defect Effects 0.000 description 2
- 208000035475 disorder Diseases 0.000 description 2
- 230000002452 interceptive effect Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 238000012216 screening Methods 0.000 description 2
- 238000001228 spectrum Methods 0.000 description 2
- 208000017667 Chronic Disease Diseases 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000004364 calculation method Methods 0.000 description 1
- 238000009223 counseling Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000013499 data model Methods 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 239000000645 desinfectant Substances 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000018109 developmental process Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 230000000737 periodic effect Effects 0.000 description 1
- 230000005236 sound signal Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/332—Query formulation
- G06F16/3329—Natural language query formulation or dialogue systems
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F16/00—Information retrieval; Database structures therefor; File system structures therefor
- G06F16/30—Information retrieval; Database structures therefor; File system structures therefor of unstructured textual data
- G06F16/33—Querying
- G06F16/3331—Query processing
- G06F16/334—Query execution
- G06F16/3343—Query execution using phonetics
-
- G—PHYSICS
- G10—MUSICAL INSTRUMENTS; ACOUSTICS
- G10L—SPEECH ANALYSIS TECHNIQUES OR SPEECH SYNTHESIS; SPEECH RECOGNITION; SPEECH OR VOICE PROCESSING TECHNIQUES; SPEECH OR AUDIO CODING OR DECODING
- G10L15/00—Speech recognition
- G10L15/02—Feature extraction for speech recognition; Selection of recognition unit
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H20/00—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance
- G16H20/10—ICT specially adapted for therapies or health-improving plans, e.g. for handling prescriptions, for steering therapy or for monitoring patient compliance relating to drugs or medications, e.g. for ensuring correct administration to patients
-
- G—PHYSICS
- G16—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR SPECIFIC APPLICATION FIELDS
- G16H—HEALTHCARE INFORMATICS, i.e. INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR THE HANDLING OR PROCESSING OF MEDICAL OR HEALTHCARE DATA
- G16H70/00—ICT specially adapted for the handling or processing of medical references
- G16H70/40—ICT specially adapted for the handling or processing of medical references relating to drugs, e.g. their side effects or intended usage
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Health & Medical Sciences (AREA)
- Theoretical Computer Science (AREA)
- Computational Linguistics (AREA)
- General Health & Medical Sciences (AREA)
- Public Health (AREA)
- Databases & Information Systems (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Medicinal Chemistry (AREA)
- General Engineering & Computer Science (AREA)
- Acoustics & Sound (AREA)
- Epidemiology (AREA)
- Human Computer Interaction (AREA)
- Medical Informatics (AREA)
- Primary Health Care (AREA)
- Audiology, Speech & Language Pathology (AREA)
- Data Mining & Analysis (AREA)
- Mathematical Physics (AREA)
- Artificial Intelligence (AREA)
- Toxicology (AREA)
- Pharmacology & Pharmacy (AREA)
- Multimedia (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Treatment And Welfare Office Work (AREA)
Abstract
The invention discloses an audio recognition method and system for assisting drugstore service personnel, wherein the method comprises the following steps: collecting onsite voice, and identifying and extracting a keyword set corresponding to symptoms, diseases or medicines; searching a keyword set corresponding to symptoms, symptoms or medicines in a symptom-medicine administration database to obtain more than one symptoms and corresponding medicine administration data; more than one condition and corresponding medication data are presented to the pharmacy attendant for reference by the pharmacy attendant. The method and the system can accurately extract the keywords in the inquiry voice conversation and automatically inquire and give corresponding medication data, thereby providing help for the pharmacy service staff on the premise of not increasing the operation burden for the pharmacy service staff.
Description
Technical Field
The invention relates to the field of auxiliary systems for new medicine retail, in particular to an audio identification method and system for assisting pharmacy service personnel.
Background
Drug stores are stores and drug pads that sell drugs. The retail drug stores in China have a short development time but grow fast. According to statistics, the number of Chinese retail drug stores in 2017 is 45.4 thousands. These pharmacies have millions of pharmacist attendants, most of which are pharmacist attendants, and these pharmacist attendants are generally required to afford services such as advice of medicine purchase and counseling for medication for mild or common chronic diseases to residents of the community who arrive at the pharmacy. According to statistics, about 30-40 common diseases of a pharmacy, about 40-50 common diseases, and about 50-60 other diseases which are not particularly frequent are provided. At present, at least 800-1200 Chinese and western patent medicines sold in general drugstores are available, and large-scale drugstores can even provide 3000-8000 products; also comprises about 200-500 products of instruments, 150-450 products of disinfectants, 200-500 health products and the like. It is also difficult for the pharmaceutical specialist to have a comprehensive understanding of all categories. The available medicines of 70-350 types are generally available in clinics and community medical sites, and the medicines need to be provided by service personnel of drugstores; in addition, pharmacies often provide some related services for health science popularization education, overall health preliminary programs, and health tracking management.
However, most pharmacy attendants are not pharmacy specialists, as the pharmacy attendant of a pharmacy is usually neither a professional nor a general practitioner. Pharmacy service personnel of a plurality of basic stores only go on duty through training, so that medicine taking and consultation services of a plurality of 140-160 diseases cannot be mastered.
In order to avoid the occurrence of wrong medication guidance, the pharmacy service staff is usually assisted by a medication data database; however, the current medical database is too professional, the query time is long, the vocabulary understanding difficulty is high, the knowledge requirement on a user is the level of a specialist, and the knowledge level of pharmacy service personnel is uneven, so that the application in a pharmacy shop is difficult.
Disclosure of Invention
The invention provides an audio recognition method and system for assisting pharmacy service personnel, which are used for solving the technical problem that pharmacy service personnel in a pharmacy cannot comprehensively and accurately provide medication consultation and suggestion of excessive diseases due to incomplete knowledge.
In order to solve the technical problems, the technical scheme provided by the invention is as follows:
an audio identification method and system for assisting pharmacy service personnel comprises the following steps:
collecting onsite voice, and identifying and extracting a keyword set corresponding to symptoms, symptoms or medicines;
searching a keyword set corresponding to symptoms, symptoms or medicines in a symptom-medicine administration database to obtain more than one symptoms and corresponding medicine administration data;
more than one condition and corresponding medication data are presented to the pharmacy attendant for reference by the pharmacy attendant.
Preferably, live speech is collected and recognized, including:
the method comprises the steps of collecting voice audios of a site, carrying out frequency domain analysis on the audio of each picked sound segment, and clustering the sound segment with the previous sound segment according to the frequency spectrum and the amplitude of the sound segment so as to match the sound segment to a corresponding sound source.
Preferably, extracting a set of keywords corresponding to a symptom, a condition, or a drug comprises:
classifying the sound segments according to question sentences and negative sentences by adopting a classifier, and processing according to the following conditions:
a: if the current sentence pattern is a negative sentence:
if the previous sentence pattern is a question sentence, removing the keywords identified in the previous sentence, and finishing the identification; if the last sentence pattern is not a question sentence and the last sentence pattern is a non-negative sentence, ending the identification;
b: if the current sentence pattern is a positive sentence:
if the last sentence pattern is a question sentence, identifying a keyword from the current sentence pattern or the last sentence pattern; if the last sentence pattern is a non-question sentence, ending the identification;
c: if the current sentence pattern is a statement sentence or a question sentence, the current sentence pattern is identified to obtain a keyword.
Preferably, the identifying of the keyword comprises: the method comprises the steps of segmenting words of a sentence, carrying out dependency syntactic analysis on the segmented words to obtain a syntactic structure, obtaining a syntactic structure tree, and extracting keywords related to symptoms, symptoms or medicines from the syntactic structure tree.
Preferably, after the keyword is identified, the keyword attributes are matched, and the keyword attributes include: disease name, symptom name, and drug name; and adds the keyword to the set of keywords corresponding to the attribute.
Preferably, according to the historical data of the sound source, a clerk and a customer are matched, and according to the matching result of the sound source of the clerk and the customer, the collected onsite voice is converted into the arrangement of the sentence pattern; each switching or change of the sound emission source is judged as the end of one sentence pattern and the beginning of the next sentence pattern.
Preferably, the search in the symptom-medication database is performed according to the corresponding category of symptom, symptom or medication of each group of the set of spoken keywords in the following manner:
a: and searching according to the keyword set related to the symptom:
searching according to the keyword set related to the symptoms to obtain more than one symptoms corresponding to the symptoms and corresponding medication data; matching with a keyword set related to symptoms according to a plurality of symptom weights corresponding to each symptom in more than one symptom, calculating a matching degree, and performing descending order arrangement of more than one symptom according to the matching degree;
b: and searching according to the keyword set related to the disease:
searching according to the keyword set related to the symptoms to obtain a symptom list corresponding to the symptoms and medication data corresponding to the symptoms in the symptoms; acquiring a keyword set related to symptoms and a plurality of symptom lists corresponding to symptoms for matching, calculating the matching degree, and performing descending order arrangement on the medication data according to the matching degree;
c: and searching according to the keyword set related to the medicine:
and searching according to the keyword set related to the medicine to obtain a disease state and a symptom list corresponding to the medicine, acquiring a plurality of symptom lists corresponding to the keyword set related to the symptom and the disease state for matching, calculating the matching degree, and performing descending order arrangement of the medication data according to the matching degree.
Preferably, when the pharmacy service staff is presented with the one or more symptoms and the corresponding medication data, the pharmacy service staff is further provided with an option of determining the diagnosis, and when the pharmacy service staff chooses no, the method further comprises:
displaying a symptom list corresponding to more than one disease symptoms to an operation end of a pharmacy service staff; acquiring positive and negative selection results of the pharmacy service personnel for each symptom in the symptom list, adding the symptom with the positive selection result into a keyword set related to the symptom, and deleting the symptom with the negative selection result from the keyword set when the symptom with the negative selection result exists in the keyword set;
searching in the symptom-medication data database again according to the updated keyword set related to the symptom to obtain more than one symptom and corresponding medication data;
and repeating the steps until the pharmacy service personnel selects yes in the option of determining the diagnosis.
Preferably, the collecting of the voice of the site includes acquiring the voice of the site through a quad-matrix microphone set composed of microphones arranged in at least four directions; matching the sound segments to corresponding sound sources, and determining the directions of the sound sources according to a triangular distance measurement principle through phase deviations of frequencies collected by corresponding microphones in the four-matrix microphone group;
during the process of collecting the live voice of each conversation, clustering is carried out according to the positions of the sound sources so as to cluster the sound sources of the areas into one conversation so as to convert the voice into the arrangement of the sentence patterns.
The present invention also provides a computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps of any of the methods described above when executing the computer program.
The invention has the following beneficial effects:
1. the audio identification method for assisting the pharmacy service staff assists the pharmacy service staff in service, can make up the shortage of professional knowledge of the pharmacy service staff, avoids providing unprofessional opinions for customers, and reduces the risk of the pharmacy; moreover, on-site voice is automatically collected by voice, so that the operation requirement on the pharmacy service staff is low, the inquiry can be realized while the automatic inquiry is carried out, the inquiry result can be quickly given, the disease and drug data can be inquired from the keywords in the aspects of symptoms, diseases or drugs, and the pharmacy service staff can be helped on the premise of not increasing the operation burden on the pharmacy service staff.
2. In a preferred scheme, the audio recognition method for assisting pharmacy service personnel can extract keywords according to the possible sentence patterns of the use scene and the actual use occasion of pharmacy inquiry, can remove the interfering keywords in the inquiry process by combining the matching of the sound emitting sources, and ensures that the extracted voice data is correct, thereby improving the accuracy of final inquiry.
3. The system is used for the audio identification method and the system for assisting the pharmacy service staff, and the extraction of the keywords is more accurate and the final recommendation of the medication data is more accurate through the positioning of the sound production source, the identity matching and the sentence pattern judgment.
In addition to the objects, features and advantages described above, other objects, features and advantages of the present invention are also provided. The present invention will be described in further detail below with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this application, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification. In the drawings:
fig. 1 is a flow chart illustrating an audio recognition method for pharmacy attendant assistance according to a preferred embodiment of the present invention.
Detailed Description
The embodiments of the invention will be described in detail below with reference to the drawings, but the invention can be implemented in many different ways as defined and covered by the claims.
Referring to fig. 1, the audio recognition method and system for pharmacy attendant assistance of the present invention comprises the following steps:
s1, collecting onsite voice, and identifying and extracting a keyword set corresponding to symptoms, symptoms or medicines; during implementation, the voice can be converted into characters, and keywords are extracted through grammar analysis, word segmentation and keyword lexicon matching; in this embodiment, the keyword set is divided into three groups according to symptoms, disorders or drugs.
S2, searching the keyword set corresponding to the symptom, the symptom or the medicine in a symptom-medicine data database to obtain more than one symptom and corresponding medicine data; the symptom-medication data base is formed by adopting various existing medical classics and past cases through data comparison and screening, and the screening can be performed by combining computer comparison manually.
And S3, displaying the more than one disease symptoms and the corresponding medication data to the pharmacy service staff for the reference of the pharmacy service staff. In implementation, the display can be displayed to the drugstore service staff through the handheld terminal of the drugstore service staff or the display terminal fixed in the drugstore.
The steps can assist the pharmacy service personnel in service, make up the shortage of professional knowledge of the pharmacy service personnel, avoid providing unprofessional opinions to customers and reduce the risk of the pharmacy; moreover, on-site voice is automatically collected by voice, so that the operation requirement on the pharmacy service personnel is low, automatic query can be realized while query is carried out, a query result can be quickly given, the disease and medicine data can be queried from the keywords of the symptoms, the diseases or the medicines, and the pharmacy service personnel can be helped without increasing the operation burden on the (medicine instruction) pharmacy service personnel.
In the implementation, the method collects the voice of the site and identifies, and also comprises the steps of positioning and matching of the sound source, division of sentence patterns, division of words and sentences and the like, wherein the following processing can be carried out for the sound source:
s11, collecting the voice frequency of the site, carrying out frequency domain analysis to the voice frequency of each picked sound segment, clustering the sound segment with the previous sound segment according to the phase and amplitude of the sound segment, and matching the sound segment to the corresponding sound source. And analyzing the original voice signal of the human voice, and processing the digital signal of the voice to judge the character role of the pronunciation. The basic principle is followed that each person pronounces in a certain frequency segment, frequency domain analysis is carried out on the audio frequency of the picked sound of each segment, frequency distribution conditions are counted to carry out one-to-one matching, then classification is carried out, and finally the category of the sound source is obtained, because the sound source is usually 2-3 persons, the audio frequency can be classified only by simply distinguishing each person from each other. By the method, the sound sources are classified firstly, and then the semantic understanding of the speech characters is combined, so that the accuracy and the stability of scene understanding can be enhanced better.
In addition, when implemented, the division of the sentence pattern can also be determined as follows:
s12, matching the clerk and the customer according to the historical data of the sound source, and converting the collected onsite voice into sentence arrangement according to the matching result of the sound source of the clerk and the customer; each switching or change of the sound emission source is judged as the end of one sentence pattern and the beginning of the next sentence pattern. Of course, on the premise that the clerk and the customer are not matched, the end of one sentence and the start of the next sentence may be determined only by the judgment of the sentence.
In the existing common sales scenes, 1-to-1 dialogue question-answer scenes are mostly adopted, the number of participating characters is small in most cases, and the voice source character information is lost after the voice source character information is converted into text through voice recognition. The human can distinguish the pronunciation objects through sound, and if each sentence or question-answer sentence can distinguish the pronunciation objects, the alternative question-answer sentences can distinguish character sources, so that the auxiliary system can provide great help for scene understanding. The general solution is to understand the scene dialogue directly through text characters and a semantic analysis algorithm to distinguish the character roles, and the method has the advantages of being not easily affected by environmental factors and strong robustness on scene environment switching. The method has the defects that the algorithm stability is not enough, semantic understanding is easy to make mistakes, and the wrong scene discrimination can bring completely opposite results to keyword extraction, recognition and medicine analysis of an auxiliary system of pharmacy service personnel. The application can solve the defect through steps S11 and/or S12.
The sentence pattern is divided to extract key words from the inquiry conversation more accurately; in practice, extracting the keyword set corresponding to the symptom, disorder or drug may include the following steps:
s13, classifying the sound segments according to the question sentences and the negative sentences by adopting an Xgboost classifier, and processing according to the following conditions:
a: if the current sentence pattern is a negative sentence:
if the previous sentence pattern is a question sentence, removing the keywords identified in the previous sentence, and finishing the identification; if the last sentence pattern is not a question sentence and the last sentence pattern is a non-negative sentence, ending the identification;
b: if the current sentence pattern is a non-negative sentence:
if the last sentence pattern is a question sentence, identifying a keyword from the current sentence pattern or the last sentence pattern; if the last sentence pattern is a non-question sentence, ending the identification;
c: and if the current sentence pattern is a non-question sentence and a non-negative sentence, identifying the current sentence pattern to obtain a keyword.
In the semantic understanding scene of pharmacy question answering, more question answering conversations exist between pharmacy service personnel and customers, so that the key role is played in scene understanding and keyword extraction, and the semantic understanding effect can be better improved. For example: when the pharmacy service staff relate to fever in the process of communicating with the customers, after the system analyzes grammatical components, semantically understands and extracts keywords, and finally carries out keyword similarity matching calculation to obtain the final symptom fever. 1. If the pharmacy service staff expresses that the semantic meaning of the symptom is a non-question scene, whether the target symptom of the symptom is uncertain or not needs to be judged through semantic scene understanding before and after the fact; 2. if the pharmacy attendant expresses that the symptom is an interrogative sentence, whether the symptom is the target symptom or not only needs to judge whether the symptom is the target symptom or not by judging whether the answer after the interrogative sentence is a positive sentence or a negative sentence. Therefore, scene understanding can be quickly positioned and key target symptoms can be quickly positioned through the question sentences and the negative sentences.
The symptom keywords can be accurately acquired through the ABC steps, and invalid information is removed. The ABC step can be implemented by training the model: setting a format of a training corpus, dividing sentence contents into question sentences and non-question sentences, judging the question sentences as a two-classification solution, and setting labels to be 1 and 0; preprocessing an input sentence, performing part-of-speech tagging and word segmentation algorithm based on Viterbi algorithm, extracting keywords based on tfidf algorithm and textrank algorithm model, performing word segmentation processing on the sentence, and dividing the long sentence into a plurality of word segments; constructing a feature project, performing TFIDF feature extraction on the linguistic data after word segmentation, and constructing a training feature matrix vector; and comparing the model parameters by using k-fold cross validation to obtain the optimal parameters and the corresponding iteration times. And (3) model training, setting parameters and iteration times, and training a sample by using an xgboost classifier. After the iterative training is carried out, an xgboost optimal parameter result is obtained, and a parameter model is saved.
After model training is completed, the models can be used for classification: the xgboost model parametric model data is loaded. And performing sentence division processing on the text sentences after the voice recognition. And (5) preprocessing the sentences, wherein the steps are consistent with those of the training stage. Inputting the word segmentation sentences into an xgboost model to obtain a classification result; the result category is 1 and is an interrogative sentence, and 0 is a non-interrogative sentence. (in this embodiment, only question sentences and short negative sentences are identified, and positive sentence statement sentences are not distinguished), if there is no, or positive sample of the training model, and other sentence patterns are negative samples, training is performed by using the xgboost model, and result type 1 is represented as a negative sentence, and 0 is a non-negative sentence.
Negative sentences can be judged by using similar classification modes. Therefore, scene sentence pattern judgment is obtained, the semantic understanding effect in a question and answer scene can be enhanced by combining the sentence pattern judgment, the system operation efficiency is improved, and the target characteristic keywords are quickly positioned and obtained.
In implementation, identifying the obtained keyword may further include:
s14, segmenting the sentence, performing dependency syntactic analysis on the segmented words to obtain a syntactic structure, acquiring a syntactic structure tree, and extracting keywords related to symptoms, symptoms or medicines from the syntactic structure tree. During implementation, sentence pattern judgment is carried out by adopting the Xgboost-based Chinese question sentence judgment model, word segmentation is carried out by adopting an HMM-based Chinese word segmentation model, and keywords are extracted by adopting a word2vec word vector medical data model.
In implementation, after the keyword is identified and obtained, the keyword attributes are matched in S15, and the keyword attributes include: disease name, symptom name, and drug name; and adds the keyword to the set of keywords corresponding to the attribute.
In practice, the search in the symptom-medication data database is performed according to the corresponding category of symptom, symptom or medication of each group of the set of spoken keywords in the following manner:
a: and searching according to the keyword set related to the symptom:
searching according to the keyword set related to the symptoms to obtain more than one symptoms corresponding to the symptoms and corresponding medication data; matching with a keyword set related to symptoms according to a plurality of symptom weights corresponding to each symptom in more than one symptom, calculating a matching degree, and performing descending order arrangement of more than one symptom according to the matching degree;
b: and searching according to the keyword set related to the disease:
searching according to the keyword set related to the symptoms to obtain a symptom list corresponding to the symptoms and medication data corresponding to the symptoms in the symptoms; acquiring a keyword set related to symptoms and a plurality of symptom lists corresponding to symptoms for matching, calculating the matching degree, and performing descending order arrangement on the medication data according to the matching degree;
c: and searching according to the keyword set related to the medicine:
in the retrieval, the retrieval can be carried out according to the keyword set related to the medicine to obtain the disease symptoms and the symptom lists corresponding to the medicine, the keyword set related to the symptom and the symptom lists corresponding to the disease symptoms are obtained to be matched, the matching degree is calculated, and the medication data are arranged in a descending order according to the matching degree.
In practice, when the at least one disease condition and the corresponding medication data are displayed to the pharmacy service staff, the pharmacy service staff is provided with an option of determining whether to diagnose, and when the pharmacy service staff selects no, the method further comprises:
displaying a symptom list corresponding to more than one disease symptoms to an operation end of a pharmacy service staff; acquiring positive and negative selection results of the pharmacy service staff for each symptom in the symptom list, adding the symptom with the positive selection result into a keyword set related to the symptom, and deleting the symptom with the negative selection result from the keyword set when the symptom with the negative selection result exists in the keyword set;
searching in the symptom-medication data database again according to the updated keyword set related to the symptom to obtain more than one symptom and corresponding medication data;
and repeating the steps until the pharmacy service personnel selects yes in the option of determining the diagnosis.
In implementation, the collecting of the live voice comprises acquiring the live voice through a four-matrix microphone group consisting of microphones arranged in at least four directions; matching the sound segments to corresponding sound sources, and determining the directions of the sound sources according to a triangular distance measurement principle through the phase deviation of the audio collected by corresponding microphones in the four-matrix microphone group; during the process of collecting the live voice of each conversation, clustering is carried out according to the orientation of the sound source so as to cluster the sound sources of the region into one conversation so as to convert the voice into the arrangement of the sentence pattern.
Due to the fact that sound sources of a pharmacy scene are complex, various noises are easily mixed, sound and audio are jumped, the audio distribution of character pronunciation is changed, and therefore classification results are affected. In the invention, the problem can be solved by applying the four-matrix microphone, and the four-matrix microphone can be used for positioning and spacing the sound in the scene, and in the medical scene, the sales inquiry dialogue is usually relatively close. Most of noise and background sound in a scene can be removed through positioning and orientation of the four-matrix microphone, and a cleaner target sound effect is obtained.
The method for analyzing the sound information comprises a plurality of methods, such as an AGC algorithm and a hidden Markov model algorithm, and in recent years, the deep learning algorithm for comparing fire and heat obtains a good effect, the method has high real-time requirements on the algorithm by selling 1 to 1 simple special scenes in a pharmacy, and combines four-matrix microphone hardware to classify the sound sources, and the specific steps of using audio analysis comprise the following flows:
a small sound signal with fixed time length t is obtained through a matrix microphone, and most of noise and far sound are filtered through an array microphone. And acquiring the sound digital signal in the time domain in the scene. And performing Gaussian filtering on the signal to filter out residual white Gaussian noise. The time domain f (t) signal is converted to a frequency domain f (ω) signal using an FFT fast transform. After obtaining the signal in the frequency domain, the signal spectrum obtained by superposing a plurality of different frequencies is obtained, the phase and the amplitude of each periodic frequency signal are obtained, and the phase and the amplitude are normalized.
In a two-dimensional space formed by the phase and the frequency of each corresponding frequency band of each fixed time length t, the phase is taken as x, the amplitude is taken as y, the x and the y of each frequency band of each frequency domain are calculated, the x and the y of each time period t are clustered by using a kmean algorithm, k is usually set to be 3, a main sound emitting source in a scene is 2 persons, in addition, 1 type is noise or can not be distinguished, and the audio data segments t falling in the center of the same cluster are the same sound emitting source. Thereby obtaining sound segments belonging to the same sound source. The influence of ambient noise on the interrogation session for extracting the sound source can be avoided.
The present invention also provides a computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the steps of any of the above embodiments being implemented when the computer program is executed by the processor.
In summary, the method and the system have the advantages that the voice on the spot is automatically acquired by adopting the voice, so the operation requirement on the pharmacy service personnel is low, the inquiry and the automatic inquiry can be realized at the same time, the inquiry result can be quickly given, the information including symptoms and medication data can be inquired from the keywords in the aspects of symptoms, symptoms or medicines, and the assistance can be provided for the pharmacy service personnel on the premise of not increasing the operation burden on the pharmacy service personnel. In addition, keywords can be prepared and extracted according to the possible sentence patterns of the use scene and the actual use occasion of pharmacy inquiry, and the interfering keywords in the inquiry process can be removed by combining the matching of the sound emitting sources, so that the extracted voice data is ensured to be correct, and the accuracy of final inquiry is improved. Through sound source positioning, identity matching and sentence pattern judgment, extraction of key words is more accurate, and recommendation of final medication data is more accurate.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (6)
1. An audio recognition method for pharmacy attendant assistance, comprising the steps of:
collecting onsite voice, and identifying and extracting a keyword set corresponding to symptoms, symptoms or medicines;
wherein, gather the speech and discernment on-the-spot, include:
collecting the voice frequency of a site, carrying out frequency domain analysis on the voice frequency of each picked sound segment, and clustering the sound segment with the previous sound segment according to the phase and amplitude of the sound segment so as to match the sound segment to a corresponding sound production source;
matching a clerk and a customer according to the historical data of the sound source, and converting the collected onsite voice into sentence arrangement according to the matching result of the sound source of the clerk and the customer; each switching or changing of the sound emitting source is judged as the end of one sentence pattern and the beginning of the next sentence pattern;
extracting a set of keywords corresponding to a symptom, condition, or drug, comprising:
classifying the sound segments by adopting a classifier according to question sentences and negative sentences, and processing according to the following conditions:
a: if the current sentence pattern is a negative sentence:
if the previous sentence pattern is a question sentence, removing the keywords identified in the previous sentence, and finishing the identification; if the last sentence pattern is not a question sentence and the last sentence pattern is a non-negative sentence, ending the identification;
b: if the current sentence pattern is a non-negative sentence:
if the last sentence pattern is a question sentence, identifying a keyword from the current sentence pattern or the last sentence pattern; if the last sentence pattern is a non-question sentence, ending the identification;
c: if the current sentence pattern is a non-question sentence and a non-negative sentence, identifying the current sentence pattern to obtain a keyword;
searching the keyword set corresponding to the symptom, the disease or the medicine in a symptom-disease-medicine database to obtain more than one disease and corresponding medicine data;
and displaying the more than one disease symptoms and the corresponding medication data to a pharmacy attendant for the pharmacy attendant to refer to.
2. The audio recognition method for pharmacy attendant assistance as recited in claim 1, wherein said recognizing results in keywords comprising: the method comprises the steps of segmenting words of a sentence, carrying out dependency syntactic analysis on the segmented words to obtain a syntactic structure, obtaining a syntactic structure tree, and extracting keywords related to symptoms, symptoms or medicines from the syntactic structure tree.
3. The method of claim 2, wherein the identifying keywords matches keyword attributes, the keyword attributes comprising: disease name, symptom name, and drug name; and adds the keyword to the set of keywords corresponding to the attribute.
4. A method of audio recognition for pharmacy attendant assistance according to any of claims 1 to 3, wherein the retrieval in the symptom-medication-data database is performed according to the corresponding category of symptom, symptom or medication of each set of spoken keywords as follows:
a: and searching according to the keyword set related to the symptom:
searching according to the keyword set related to the symptoms to obtain more than one symptoms corresponding to the symptoms and corresponding medication data; matching the keyword set related to the symptoms according to a plurality of symptom weights corresponding to each symptom in the more than one symptoms, calculating a matching degree, and performing descending order arrangement of the more than one symptoms according to the matching degree;
b: and searching according to the keyword set related to the disease:
searching according to the keyword set related to the symptoms to obtain a symptom list corresponding to the symptoms and medication data corresponding to the symptoms in the symptoms; acquiring a keyword set related to symptoms, matching the keyword set with a plurality of symptom lists corresponding to the symptoms, calculating the matching degree, and performing descending order arrangement on the medication data according to the matching degree;
c: and searching according to the keyword set related to the medicine:
searching according to the keyword set related to the medicine to obtain the symptoms and the symptom lists corresponding to the medicine, obtaining the keyword set related to the symptoms, matching with the symptom lists corresponding to the symptoms, calculating the matching degree, and performing descending order arrangement on the medication data according to the matching degree.
5. The audio recognition method for pharmacy attendant assist as claimed in any one of claims 1 to 3, wherein the capturing of the speech of the site comprises capturing the speech of the site by a quad-matrix microphone set of microphones arranged in at least four directions; matching the sound segments to corresponding sound sources, and determining the directions of the sound sources according to a triangular distance measurement principle through the phase deviation of the audio collected by the corresponding microphones in the four-matrix microphone group;
and in the process of collecting the live voice for each conversation, clustering according to the position of the sound source so as to cluster the sound source at the position into one conversation so as to convert the voice into the arrangement of the sentence pattern.
6. A computer system comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the steps of the method of any of the preceding claims 1 to 5 are performed when the computer program is executed by the processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110302111.8A CN112948559B (en) | 2021-03-22 | 2021-03-22 | Audio identification method and system for assisting pharmacy service staff |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110302111.8A CN112948559B (en) | 2021-03-22 | 2021-03-22 | Audio identification method and system for assisting pharmacy service staff |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN112948559A CN112948559A (en) | 2021-06-11 |
| CN112948559B true CN112948559B (en) | 2022-05-20 |
Family
ID=76227536
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110302111.8A Active CN112948559B (en) | 2021-03-22 | 2021-03-22 | Audio identification method and system for assisting pharmacy service staff |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN112948559B (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110489566A (en) * | 2019-08-22 | 2019-11-22 | 上海软中信息系统咨询有限公司 | A kind of hospital guide's method of intelligence hospital guide's service robot |
| CN112238460A (en) * | 2020-10-26 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Application of intelligent robot in clinic |
| CN112242199A (en) * | 2020-10-10 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Method for guiding consultation by robot |
| CN112241451A (en) * | 2020-10-27 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Application of intelligent robot in pharmacy |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106557653B (en) * | 2016-11-15 | 2017-09-22 | 合肥工业大学 | A kind of portable medical intelligent medical guide system and method |
-
2021
- 2021-03-22 CN CN202110302111.8A patent/CN112948559B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110489566A (en) * | 2019-08-22 | 2019-11-22 | 上海软中信息系统咨询有限公司 | A kind of hospital guide's method of intelligence hospital guide's service robot |
| CN112242199A (en) * | 2020-10-10 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Method for guiding consultation by robot |
| CN112238460A (en) * | 2020-10-26 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Application of intelligent robot in clinic |
| CN112241451A (en) * | 2020-10-27 | 2021-01-19 | 北京小乔机器人科技发展有限公司 | Application of intelligent robot in pharmacy |
Also Published As
| Publication number | Publication date |
|---|---|
| CN112948559A (en) | 2021-06-11 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US11714861B2 (en) | Query selection method and system | |
| CN108877921B (en) | Medical intelligent triage method and medical intelligent triage system | |
| US10515292B2 (en) | Joint acoustic and visual processing | |
| US7580835B2 (en) | Question-answering method, system, and program for answering question input by speech | |
| US7650283B2 (en) | Dialogue supporting apparatus | |
| US9043206B2 (en) | System and methods for matching an utterance to a template hierarchy | |
| CN111401066B (en) | Artificial intelligence-based word classification model training method, word processing method and device | |
| CN112786194A (en) | Medical image diagnosis guide inspection system, method and equipment based on artificial intelligence | |
| CN106326640A (en) | Medical speech control system and control method thereof | |
| CN107305541A (en) | Speech recognition text segmentation method and device | |
| KR102216768B1 (en) | System and Method for Analyzing Emotion in Text using Psychological Counseling data | |
| Sagha et al. | Enhancing multilingual recognition of emotion in speech by language identification | |
| CN113569023A (en) | Chinese medicine question-answering system and method based on knowledge graph | |
| US20140255886A1 (en) | Systems and Methods for Content Scoring of Spoken Responses | |
| Kumar et al. | A deep learning approaches and fastai text classification to predict 25 medical diseases from medical speech utterances, transcription and intent | |
| CN111145903A (en) | Method and device for acquiring vertigo inquiry text, electronic equipment and inquiry system | |
| US8498859B2 (en) | Voice processing system, method for allocating acoustic and/or written character strings to words or lexical entries | |
| Wagner et al. | Applying cooperative machine learning to speed up the annotation of social signals in large multi-modal corpora | |
| CN109036506B (en) | Internet medical inquiry supervision method, electronic device and readable storage medium | |
| CN110069614A (en) | A kind of question and answer exchange method and device | |
| CN112948559B (en) | Audio identification method and system for assisting pharmacy service staff | |
| CN112053760B (en) | Medication guide method, medication guide device, and computer-readable storage medium | |
| Nikadon et al. | BERTAgent: The development of a novel tool to quantify agency in textual data | |
| CN111967235B (en) | Form processing method, form processing device, computer equipment and storage medium | |
| CN113761899A (en) | Medical text generation method, device, equipment and storage medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |