Nurse-friendly consultation question and answer method and system

Abstract

The invention discloses a nurse-friendly consultation question and answer method and system, and belongs to the technical field of artificial intelligence, and the method comprises the steps: obtaining a consultation record, carrying out the data analysis of the consultation record, obtaining question and answer data, and building a question and answer data set; constructing a question-answer model based on question-answer data of the question-answer data set; acquiring a question of the patient, searching an answer to the question based on the question and answer model, and if the answer is searched, returning the searched answer; and if the answer is not searched, feeding back to the nurse station, and storing the question and the answer into the question and answer data set. By introducing the automatic consultation question and answer model, the working efficiency and the service quality of nurses are remarkably improved.

Description

Technical Field

[0001] The present invention relates to the field of artificial intelligence technology, and more specifically to a nurse-friendly consultation question-answering method and system. Background Art

[0002] In the modern medical system, nurses, as an indispensable part of the medical team, not only bear the heavy responsibility of directly caring for patients, but also frequently communicate with patients and their families to answer various medical consultations. However, with the continuous advancement of medical technology and the improvement of patients' health awareness, the content of patient consultations has become increasingly complex and diverse, covering disease knowledge, treatment plans, medication guidance, rehabilitation advice, and psychological support. This trend has put forward higher requirements on the professional quality and coping ability of nurses, and at the same time exposed many deficiencies in the current consultation and question-answering processing mechanism.

[0003] Faced with numerous inquiries from patients and their families, nurses often need to manually look through materials, recall professional knowledge, or use limited online resources for searches. This process is time-consuming and laborious, which not only reduces work efficiency, but also may take up time that should be used for direct patient care due to long information retrieval, resulting in hidden waste of medical resources. In addition, knowledge in the medical field is updated rapidly, and new therapies and new drugs emerge in an endless stream. Faced with new problems or the latest medical advances, nurses may not be able to provide accurate answers in a timely manner due to information lags, affecting patient satisfaction and trust.

[0004] Furthermore, the traditional consultation question-and-answer method lacks a systematic recording mechanism, resulting in a large amount of valuable patient consultation data not being effectively collected and utilized. This not only limits the hospital's ability to quantitatively evaluate and continuously improve service quality, but also hinders the development of personalized medical services based on big data analysis.

[0005] Therefore, how to provide an efficient, intelligent, and nurse-friendly consultation and question-answering method and system is an urgent problem that technical personnel in this field need to solve. Summary of the invention

[0006] In view of this, the present invention provides a nurse-friendly consultation question-and-answer method and system. By constructing a question-and-answer system, an automated consultation question-and-answer process is realized, thereby improving the work efficiency and service quality of nurses.

[0007] In order to achieve the above object, the present invention adopts the following technical solution:

[0008] In one aspect, the present invention provides a nurse-friendly consultation question-answering method, comprising:

[0009] Obtain consultation records, perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set;

[0010] Building a question-answering model based on the question-answering data of the question-answering dataset;

[0011] The patient's question is obtained, and the answer to the question is searched based on the question-answering model. If the answer is found, the searched answer is returned; if the answer is not found, it is fed back to the nurse station, and the question and answer are stored in the question-answering data set.

[0012] Preferably, performing data analysis on the consultation records to obtain question and answer data, and forming a question and answer data set, includes:

[0013] Dividing the consultation record into a plurality of dialogue segments according to the response language text library;

[0014] Analyze the conversation segments respectively to obtain questions and answers of the question-and-answer segments;

[0015] Constructing a correspondence between the question and the answer;

[0016] A question-answering dataset is constructed according to the question, the answer, and the corresponding relationship between the question and the answer.

[0017] Preferably, the consultation record is segmented, and the consultation record is divided into a plurality of dialogue segments according to the response language library, including:

[0018] The consultation record is divided into question information A1, answer information B1, question information A2, answer information B2, ..., question information A1, answer information B1, answer information B2, answer information B2, answer information B3, answer information B4, answer information B5, answer information B6, answer information B7, answer information B8, answer information B9, answer information B11, answer information B12, answer information B13, answer information B14, answer information B15, answer information B16, answer information B17, answer information i 、Answer information B i , …;

[0019] Using the answer text library to determine the question information A i Whether to include a response, if the question information A i If a response word is included, the question information and answer information between the previous question information including the response word and the current question information including the response word constitute a dialogue segment.

[0020] Preferably, the conversation segments are analyzed separately to obtain the questions and answers of the question and answer segments, including:

[0021] Using the answer text library to determine the question and answer information A i Whether to include a response, if the question and answer information A i Contains the answer, then the question information A i Not a question, answer information B i-1 The answer to the first question in the conversation segment.

[0022] Preferably, constructing a question-answering model based on the question-answering data of the question-answering dataset includes:

[0023] The question and answer data in the question and answer data set are stored as key-value pairs in a hash table data structure of the question and answer database;

[0024] Setting the correspondence between keywords and tags in the question and answer database, that is, key-value pair information of the question and answer data;

[0025] Create a scheduled task, classify and traverse the hash table of the question and answer database, store each question and answer data, and complete the construction of the question and answer model.

[0026] Preferably, a scheduled task is created to classify and traverse the hash table of the question and answer database, and store each question and answer data, including:

[0027] Obtain keywords of the question and answer data, and search among the tag keywords in the question and answer database to see whether there is a tag corresponding to the keyword;

[0028] If there is a corresponding tag, directly search the hash table corresponding to the tag in the question and answer database; if there is a corresponding hash table, store the question and answer data in the hash table corresponding to the tag; if there is no corresponding hash table, create a new hash table and store the question and answer data;

[0029] If there is no corresponding tag, obtain the remaining keywords of the question and answer data, and check whether there are tags corresponding to the remaining keywords in the tag keywords of the question and answer database. If so, store them; if not, list them as questions that cannot be archived.

[0030] On the other hand, the present invention provides a system for implementing any one of the above-mentioned nurse-friendly consultation question-answering methods, comprising:

[0031] A data recording module is used to obtain consultation records, perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set;

[0032] A model building module, used to build a question-answering model based on the question-answering data of the question-answering dataset;

[0033] The question-and-answer module is used to obtain the patient's questions, search for answers to the questions based on the question-and-answer model, and return the answers if they are found; if they are not found, the answers are fed back to the nurse station, and the questions and answers are stored in the question-and-answer data set.

[0034] Preferably, the data recording module includes:

[0035] An input unit for obtaining consultation records;

[0036] A division unit, used for dividing the consultation record into a plurality of dialogue segments according to the response language library;

[0037] A screening unit, used to analyze the conversation segments respectively to obtain questions and answers of the question and answer segments;

[0038] A relationship mapping unit, used to construct a corresponding relationship between the question and the answer;

[0039] The data set unit constructs a question-answering data set according to the question, the answer, and the corresponding relationship between the question and the answer.

[0040] It can be seen from the above technical solutions that compared with the prior art, the present invention discloses a nurse-friendly consultation question-and-answer method and system. By introducing an automated consultation question-and-answer model, the work efficiency and service quality of nurses have been significantly improved. This model relies on data analysis technology to build a question-and-answer model, which can respond to various inquiries quickly and accurately, and can get timely and effective answers to both routine questions and professional consultations. In addition, the data recording and optimization mechanism of the present invention can continuously collect user feedback and behavior data, continuously optimize the performance of the question-and-answer model, ensure the continuous optimization and upgrading of the user experience, and further promote the intelligent and efficient process of medical consultation services. BRIEF DESCRIPTION OF THE DRAWINGS

[0041] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.

[0042] Figure 1 A schematic diagram of the process flow of the inquiry question and answer method provided by the present invention;

[0043] Figure 2 It is a graph showing the proportion of questions and answers in the actual application process of the present invention;

[0044] Figure 3 It is a schematic diagram of the change of the manual participation rate during the practical application of the present invention;

[0045] Figure 4 This is a framework diagram of the consultation question and answer system provided by the present invention. DETAILED DESCRIPTION

[0046] The following will be combined with the drawings in the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0047] The embodiment of the present invention discloses a nurse-friendly consultation question-answering method, referring to Figure 1 ,include:

[0048] Obtain consultation records, perform data analysis on them, obtain question-and-answer data, and build a question-and-answer data set. Consultation records include nurses' daily work records and patient consultation records recorded by electronic devices, etc. Data analysis includes converting voice data into text data, extracting questions and corresponding answers from text data.

[0049] Build a question-answering model based on the question-answering data of the question-answering dataset;

[0050] Get the patient's question, search for the answer to the question based on the question-answering model, and if the answer is found, return the searched answer; if the answer is not found, feedback is given to the nurse station, and the question and answer are stored in the question-answering data set. Specifically, when the patient calls the nurse through the calling machine, the question-answering model automatically identifies the patient's question and generates an answer based on it. If the question already has an answer, it will be returned directly; if it is a new question, the system will repeat the patient's question information to the nurse station, and the staff at the nurse station can click the answer button to record the data. The system will record the nurse's answer, store the data in the database, and then simulate the nurse's reply and repeat the answer. In the actual application process, for the questions raised by patients, the proportion of answers from the question-answering model and manual answers is shown in the figure below. Figure 2 shown.

[0051] Furthermore, the system will record the process and results of each consultation and Q&A session, analyze and optimize the data, and continuously improve the accuracy and efficiency of the Q&A model. Figure 3 As shown in the figure, with continuous optimization, the manual participation rate gradually decreases.

[0052] Furthermore, we analyze the consultation records to obtain question and answer data and build a question and answer data set, including:

[0053] The consultation record is divided into multiple dialogue segments according to the answer language text library; specifically, the consultation record is divided into question information A1, answer information B1, question information A2, answer information B2, ..., question information A i 、Answer information B i , ...; Use the answer language library to determine the question information A iWhether it contains a response. If the question information A i contains a response, then the question information and answer information between the previous question information containing a response and the current question information containing a response form a dialogue segment.

[0054] Analyze each dialogue segment separately to obtain the questions and answers of the Q&A segments;

[0055] Construct the correspondence between questions and answers;

[0056] Construct a Q&A dataset based on the questions, answers, and the correspondence between questions and answers.

[0057] Extract the questions and answers in the Q&A segments and construct the correspondence between questions and answers; further, analyze the question information A i to determine whether the text content of A is a question or a response. In this process, use a response text library to determine whether it contains {"oh", "that's fine", "then like this", "got it", "ok", "thanks", "thank you", "thank", "touched", "thank you so much"}. If it contains the above information, then the current A i is not a question but a closing statement. Then A i the previous B i-1 is the outermost A i-1 (or the outermost response A n the first A after that n-1 , where n < i) is the correct answer. Thus, the correspondence between questions and correct answers is completed.

[0058] Construct a Q&A dataset based on the questions, answers, and the correspondence between questions and answers.

[0059] Furthermore, when constructing the Q&A dataset, it is also necessary to organize the questions and answers. The specific steps include:

[0060] (1) Organize the questions and corresponding answers. For those that cannot generate the correspondence between questions and answers according to the data analysis process, they need to be played to the back-end operators of the software system, and the operators will input and supplement the data questions and answers.

[0061] (2) Extract keywords. The steps use Apache Lucene or Elasticsearch in the Java open-source toolkit to implement the keyword extraction of the text content of the TF-IDF algorithm)

[0062] (3) Classification of similar questions is actually keyword classification. Questions and answers are manually classified using keywords in the questions and answers. Questions or answers related to medicines are classified into the medicines category label, and questions related to wards are classified into the ward category. Similarly, there are nursing, hygiene, outings, complaints, and condition categories. After the above classification, 3-5 keywords are set for each category, for example: medicine category: dressing change, medication, tonic, addition of medicine, drip. After the subsequent questions are sent out, after the voice-to-text conversion and the keywords are extracted, the tag keyword matching is performed first when the question library data is checked. After the match is found, the corresponding tag will be searched first.

[0063] (4) Organize the questions and store them in a database. The system will store the questions and answers generated within 24 hours of the day into the overall question and answer database. Before storing them in the database, keyword extraction and data matching are required to filter out duplicate or similar questions.

[0064] Among them, semantic analysis uses Java open source tools such as ANTLR, JavaCC or uses more general natural language processing libraries such as OpenNLP for semantic analysis, and then extracts keywords from the analysis results and then searches them in the question answer library.

[0065] In another embodiment, building a question-answering model based on question-answering data of a question-answering dataset includes:

[0066] The question and answer data in the question and answer dataset are taken as key-value pairs and stored in the hash table data structure of the question and answer database; the key-value format of Redis String format is used to store Jin Xiaoning tag data, and the (comprehensive) HashMap data structure in Redis is used as the storage medium for all question answers, and the (tag-named) HashMap data structure in Redis is used as another storage location for new question answer information.

[0067] Set the corresponding relationship between keywords and tags in the question-answer database, that is, the key-value pair information of the question-answer data. For example: key (drug change)-value (drug category), key (drug addition)-value (drug category)..., which needs to be maintained manually at the beginning because there is no historical data accumulation. Several main categories need to be maintained. If there are no such categories in the subsequent matching process, they will be automatically created by the program.

[0068] Create a scheduled task, classify and traverse the hash table of the question and answer database, store each question and answer data, and complete the construction of the question and answer model.

[0069] Furthermore, a scheduled task is created to classify and traverse the hash table of the question-and-answer database and store each question-and-answer data, including:

[0070] Get the keywords of the question and answer data, and search for tags corresponding to the keywords in the tag keywords in the question and answer database. In the search process, the keywords are used first to search for tags in Redis, and then the keywords are used to search for data in the HashMap of the corresponding tags to prevent duplication. If the data cannot be found, the (comprehensive) HashMap is searched to prevent duplication and avoid failure to search.

[0071] If there is a corresponding label, directly search the hash table corresponding to the label in the question and answer database. If there is a corresponding hash table, store the question and answer data in the hash table corresponding to the label. If there is no corresponding hash table, create a new hash table and store the question and answer data.

[0072] If there is no corresponding tag, obtain the remaining keywords of the question and answer data and check whether there is a tag corresponding to the remaining keywords in the tag keyword of the question and answer database. If yes, store it; if not, list it as a question that cannot be archived. Record the questions that cannot be archived and show them to the system operation and maintenance personnel. The operation and maintenance personnel regularly hand over the unclassified questions to the nurse station and ask whether to create a new question tag or classify it into an existing tag.

[0073] On the other hand, the present invention provides a system for implementing any one of the above-mentioned nurse-friendly consultation question-answering methods, referring to Figure 4 ,include:

[0074] The data recording module is used to obtain consultation records, perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set;

[0075] A model building module, used to build a question-answering model based on the question-answering data of the question-answering dataset;

[0076] The question-and-answer module is used to obtain patients' questions and search for answers to the questions based on the question-and-answer model. If the answer is found, the answer is returned; if the answer is not found, it is fed back to the nurse station and the question and answer are stored in the question-and-answer dataset.

[0077] Specifically, the data recording module includes:

[0078] An input unit for obtaining consultation records;

[0079] A division unit, used for dividing the consultation record into a plurality of dialogue segments according to the response language library;

[0080] A screening unit, used to analyze the dialogue segments respectively to obtain questions and answers of the question-and-answer segments;

[0081] A relation mapping unit, used to construct the corresponding relationship between questions and answers;

[0082] The data set unit constructs a question-answering data set based on questions, answers, and the corresponding relationship between questions and answers.

[0083] In this specification, each embodiment is described in a progressive manner, and each embodiment focuses on the differences from other embodiments. The same or similar parts between the embodiments can be referred to each other. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part.

[0084] The above description of the disclosed embodiments enables one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to the embodiments shown herein, but rather to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A nurse-friendly consultation question-answering method, characterized in that: include: Obtain consultation records, perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set; Building a question-answering model based on the question-answering data of the question-answering dataset; Obtaining the patient's question, searching for an answer to the question based on the question-answering model, and returning the searched answer if an answer is found; If no answer is found, the answer is fed back to the nurse station, and the question and answer are stored in the question and answer data set.

2. A nurse-friendly consultation question-answering method according to claim 1, characterized in that: Perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set, including: Dividing the consultation record into a plurality of dialogue segments according to the response language text library; Analyze the conversation segments respectively to obtain questions and answers of the question-and-answer segments; Constructing a correspondence between the question and the answer; A question-answering dataset is constructed according to the question, the answer, and the corresponding relationship between the question and the answer.

3. A nurse-friendly consultation question-answering method according to claim 2, characterized in that: The consultation record is segmented into multiple dialogue segments according to the response language library, including: The consultation record is divided into question information A1, answer information B1, question information A2, answer information B2, ..., question information A1, answer information B1, answer information B2, answer information B2, answer information B3, answer information B4, answer information B5, answer information B6, answer information B7, answer information B8, answer information B9, answer information B11, answer information B12, answer information B13, answer information B14, answer information B15, answer information B16, answer information B17, answer information i 、Answer information B i , …; Using the answer text library to determine the question information A i Whether to include a response, if the question information A i If a response word is included, the question information and answer information between the previous question information including the response word and the current question information including the response word constitute a dialogue segment.

4. A nurse-friendly consultation question-answering method according to claim 3, characterized in that: Analyze the conversation segments respectively, and obtain the questions and answers of the question and answer segments, including: Using the answer text library to determine the question and answer information A i Whether to include a response, if the question and answer information A i Contains the answer, then the question information A i Not a question, answer information B i-1 The answer to the first question in the conversation segment.

5. A nurse-friendly consultation question-answering method according to claim 1, characterized in that: Building a question-answering model based on the question-answering data of the question-answering dataset includes: The question and answer data in the question and answer data set are stored as key-value pairs in a hash table data structure of the question and answer database; Setting the correspondence between keywords and tags in the question and answer database, that is, key-value pair information of the question and answer data; Create a scheduled task, classify and traverse the hash table of the question and answer database, store each question and answer data, and complete the construction of the question and answer model.

6. A nurse-friendly consultation question-answering method according to claim 5, characterized in that: Create a scheduled task to classify and traverse the hash table of the question-and-answer database and store each question-and-answer data, including: Obtain keywords of the question and answer data, and search among the tag keywords in the question and answer database to see whether there is a tag corresponding to the keyword; If there is a corresponding tag, directly search the hash table corresponding to the tag in the question and answer database; if there is a corresponding hash table, store the question and answer data in the hash table corresponding to the tag; if there is no corresponding hash table, create a new hash table and store the question and answer data; If there is no corresponding tag, obtain the remaining keywords of the question and answer data, and check whether there are tags corresponding to the remaining keywords in the tag keywords of the question and answer database. If so, store them; if not, list them as questions that cannot be archived.

7. A system for implementing a nurse-friendly consultation question-answering method as claimed in any one of claims 1 to 6, characterized in that: include: A data recording module is used to obtain consultation records, perform data analysis on the consultation records to obtain question and answer data, and form a question and answer data set; A model building module, used to build a question-answering model based on the question-answering data of the question-answering dataset; A question-and-answer module is used to obtain the patient's questions, search for answers to the questions based on the question-and-answer model, and return the answers if they are found; If no answer is found, the answer is fed back to the nurse station, and the question and answer are stored in the question and answer data set.

8. A nurse-friendly consultation question-answering system according to claim 7, characterized in that: The data recording module comprises: An input unit for obtaining consultation records; A division unit, used for dividing the consultation record into a plurality of dialogue segments according to the response language library; A screening unit, used to analyze the conversation segments respectively to obtain questions and answers of the question and answer segments; A relationship mapping unit, used to construct a corresponding relationship between the question and the answer; The data set unit constructs a question-answering data set according to the question, the answer, and the corresponding relationship between the question and the answer.

Images