Questionnaire title dynamic modification method and device, electronic equipment and storage medium

Abstract

The present application relates to artificial intelligence technology, disclose a kind of questionnaire topic dynamic modification method for financial business investigation scene, comprising: according to the selection proportion of question answer screening target to be modified selection question, using pre-training completed semantic understanding model to the question of target to be modified selection question, investigation theme and question and answer topic carries out intention recognition, obtains selection question intention label, theme intention label, question and answer topic intention label and selection question slot label;According to intention label and slot label, the modification suggestion of target to be modified selection question is constructed, according to the similarity of theme intention label and question and answer topic intention label, selection target to be modified question and answer question, and theme intention is used as the modification suggestion of target to be modified question and answer question.In addition, the present application also relates to blockchain technology, and the modification suggestion can be stored in the node of blockchain.The present application further proposes a kind of questionnaire topic dynamic modification device, electronic equipment and storage medium.The present application can improve the accuracy and efficiency of survey questionnaire modification.

Description

Technical Field

[0001] This invention relates to the field of artificial intelligence technology, and in particular to a method, apparatus, electronic device, and computer-readable storage medium for dynamically modifying questionnaire items. Background Technology

[0002] Electronic questionnaires are a common tool for collecting data online. They can be used for satisfaction surveys, feedback collection, assessments, and examinations, especially in the financial and medical fields. In order to better understand user needs, it is often necessary to send out electronic questionnaires to users. During the survey process, it is often necessary to modify the questionnaires according to the actual situation.

[0003] Modifications to questionnaire items are typically made based on the decision-maker's subjective experience. Furthermore, the uniqueness of items cannot be automatically identified between different questionnaires or versions, and the information of completed questionnaires cannot be traced, resulting in low accuracy and efficiency in questionnaire modification. Summary of the Invention

[0004] This invention provides a method, apparatus, and computer-readable storage medium for dynamically modifying questionnaire items, with the main purpose of solving the problems of low accuracy and efficiency in modifying survey questionnaires.

[0005] To achieve the above objectives, the present invention provides a method for dynamically modifying questionnaire items, comprising:

[0006] Obtain a set of completed questionnaires and identify the answers to each question in the completed questionnaire set;

[0007] Identify the question type of each question and determine the type of the question;

[0008] When the question type is a multiple choice question, target multiple choice questions to be modified are selected according to the selection ratio of the answers to each question. The pre-trained semantic understanding model is used to perform intent recognition on the target multiple choice questions to be modified, and the intent label and slot label of the target multiple choice questions to be modified are obtained.

[0009] Based on the intent tags and the slot tags, construct modification suggestions for the target multiple-choice question to be modified;

[0010] When the question type is a question-and-answer question, the survey topic of the completed questionnaire set is obtained, and the semantic understanding model is used to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent label and the question-and-answer question intent label of the survey topic;

[0011] Calculate the similarity between the topic intent tag and the question intent tag, and select the question and answer questions with a similarity less than a preset similarity threshold as target question and answer questions to be modified, and use the topic intent as a modification suggestion for the target question and answer questions to be modified;

[0012] Obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0013] Optionally, the step of filtering the target multiple-choice questions to be modified based on the selection ratio of the answers to each question includes:

[0014] Statistical analysis of the selection ratio of each option in the answers to each question;

[0015] Multiple-choice questions whose selection ratio exceeds a preset threshold are designated as target multiple-choice questions to be modified.

[0016] Optionally, the step of constructing modification suggestions for the target multiple-choice question based on the intent tag and the slot tag includes:

[0017] The intent tags are used as suggestions for modifying the question stem of the target multiple-choice question to be modified;

[0018] The target text is matched from a preset dictionary using the slot labels;

[0019] The target text is used as a modification suggestion for the options of the target multiple-choice question to be modified.

[0020] Optionally, modifying the corresponding questionnaire question in the preset question bank according to the question ID and the modification suggestion includes:

[0021] Based on the question ID, retrieve the target question from the preset question bank to modify;

[0022] The target question to be modified is added to the question modification queue, and modification prompt information is generated;

[0023] The questionnaire creator modifies the target modified questions in the preset question bank according to the prompt information and the modification suggestions, retains the target modified questions and the question ID, generates a new question ID for the modified target modified questions in the preset data, and associates the new question ID with the target to be modified multiple-choice questions or target to be modified open-ended questions in the completed questionnaire set.

[0024] Optionally, the step of using a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice question to be modified, and obtaining the intent label and slot label of the target multiple-choice question to be modified, includes:

[0025] The question stems of the target multiple-choice questions to be modified are segmented into words to obtain a word segmentation sequence of the question stems;

[0026] The question stem of the target multiple-choice question to be modified is segmented into characters to obtain the question stem character sequence;

[0027] The word segmentation sequence and the character sequence of the question stem are encoded by the encoding layer in the pre-trained semantic understanding model to obtain the word encoding sequence and the character encoding sequence.

[0028] The semantic understanding model utilizes a fusion layer to perform feature fusion on the word encoding sequence and the character encoding sequence to obtain a fusion vector.

[0029] The fusion vector is classified using the intent recognition layer in the semantic understanding model to obtain the intent label of the target multiple-choice question to be modified;

[0030] The semantic slot filling task layer in the semantic understanding model is used to predict the slots of the word encoding sequence and the character encoding sequence to obtain the slot labels of the target multiple-choice question to be modified.

[0031] Optionally, the step of using the semantic slot filling task layer in the semantic understanding model to predict slots in the word encoding sequence and the character encoding sequence to obtain the slot labels of the target multiple-choice question to be modified includes:

[0032] The word encoding sequence is decoded using the word decoder in the semantic slot filling task layer of the semantic understanding model to obtain the word hidden state of each slot;

[0033] The character encoding sequence is decoded using the character decoder in the semantic slot filling task layer to obtain the hidden state of the character in each slot;

[0034] The hidden states of characters and words are fused using a preset fusion function to obtain the hidden states of slots;

[0035] The hidden states of the slots are classified using a pre-defined classifier in the semantic slot filling task layer to obtain slot labels.

[0036] Optionally, the encoding layer in the pre-trained semantic understanding model encodes the word segmentation sequence and the character sequence of the question stem, respectively, to obtain the word encoding sequence and the character encoding sequence, including:

[0037] Add the CLS symbol to the beginning of the question stem character sequence, and add a separator after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence;

[0038] The target question stem character sequence is encoded using the character encoder in the encoding layer to obtain a character encoding sequence;

[0039] The contextual information of the word segmentation sequence in the question stem is extracted using the self-attention mechanism module of the word encoder in the encoding layer;

[0040] The sequential information of the word segmentation sequence of the question stem is obtained by utilizing the bidirectional long short-term memory network of the word encoder in the coding layer;

[0041] The context information and the sequence information are processed by a full connection to obtain the word encoding sequence.

[0042] To address the above problems, the present invention also provides a device for dynamically modifying questionnaire items, the device comprising:

[0043] The questionnaire acquisition module is used to acquire a set of completed questionnaires and identify the answers to each question in the completed questionnaire set;

[0044] The question type recognition module is used to identify the question type of each question and determine the type of the question.

[0045] The multiple-choice question modification suggestion generation module is used to filter target multiple-choice questions to be modified based on the selection ratio of the answers to each question when the question type is multiple-choice. It uses a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice questions to be modified, and obtains the intent tag and slot tag of the target multiple-choice questions to be modified. Based on the intent tag and the slot tag, it constructs modification suggestions for the target multiple-choice questions to be modified.

[0046] The question-and-answer modification suggestion generation module is used to obtain the survey topic of the completed questionnaire set when the question type is a question-and-answer question, and use the semantic understanding model to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent tag and the question intent tag of the survey topic; calculate the similarity between the topic intent tag and the question intent tag, and select the question-and-answer questions with a similarity less than a preset similarity threshold as target question-and-answer questions to be modified, and use the topic intent as a modification suggestion for the target question-and-answer questions to be modified;

[0047] The questionnaire modification module is used to obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0048] To address the above problems, the present invention also provides an electronic device, the electronic device comprising:

[0049] At least one processor; and,

[0050] A memory communicatively connected to the at least one processor; wherein,

[0051] The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the questionnaire item dynamic modification method described above.

[0052] To address the aforementioned problems, the present invention also provides a computer-readable storage medium storing at least one computer program, which is executed by a processor in an electronic device to implement the aforementioned method for dynamically modifying questionnaire items.

[0053] This invention improves the efficiency of questionnaire question modification by filtering target multiple-choice questions based on the selection ratio of answers to each question, eliminating the need for manual screening. A pre-trained semantic understanding model is used to identify the intent of the target multiple-choice questions, accurately analyzing the question intent and slot labels. Modification suggestions are then constructed based on these intent and slot labels, thereby improving the accuracy of questionnaire question modification. Furthermore, the semantic understanding model is used to identify the intent of both the survey topic and the questions, obtaining the topic intent label and the question intent label. The method involves labeling questionnaires with images, calculating the similarity between the topic intent label and the question intent label, identifying questions with similarity scores below a preset similarity threshold as target questions to be modified, and using the topic intent as a modification suggestion for the target questions to be modified. This makes the questionnaire questions more relevant to the topic, thereby improving the accuracy of questionnaire modification. The method also involves obtaining the question IDs of the target multiple-choice questions and the target question to be modified, and modifying the corresponding questionnaire questions in a preset question bank based on the question IDs and the modification suggestions. Modifying questions in the question bank using unique question IDs breaks down the boundaries between questionnaires and questionnaire versions, enabling data integration and analysis, and improving the efficiency of questionnaire modification. Therefore, the questionnaire question dynamic modification method, device, electronic device, and computer-readable storage medium proposed in this invention can solve the problem of low accuracy and efficiency in questionnaire modification. Attached Figure Description

[0054] Figure 1 A flowchart illustrating a method for dynamically modifying questionnaire items according to an embodiment of the present invention;

[0055] Figure 2 for Figure 1 The diagram shows a detailed implementation process for one step in the method for dynamically modifying questionnaire questions.

[0056] Figure 3 for Figure 1The diagram shows a detailed implementation process for another step in the method for dynamically modifying questionnaire items.

[0057] Figure 4 A functional module diagram of a questionnaire question dynamic modification device provided in an embodiment of the present invention;

[0058] Figure 5 This is a schematic diagram of the structure of an electronic device that implements the method for dynamically modifying questionnaire items, according to an embodiment of the present invention.

[0059] The realization of the objective, functional features and advantages of the present invention will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0060] It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit the invention.

[0061] This application provides a method for dynamically modifying questionnaire questions. The executing entity of this method includes, but is not limited to, at least one of the following electronic devices that can be configured to execute the method provided in this application: a server, a terminal, etc. In other words, the method for dynamically modifying questionnaire questions can be executed by software or hardware installed on a terminal device or a server device, and the software can be a blockchain platform. The server includes, but is not limited to, a single server, a server cluster, a cloud server, or a cloud server cluster. The server can be an independent server or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communication, middleware services, domain name services, security services, content delivery networks (CDNs), and big data and artificial intelligence platforms.

[0062] Reference Figure 1 The diagram shown is a flowchart illustrating a method for dynamically modifying questionnaire items according to an embodiment of the present invention. In this embodiment, the method for dynamically modifying questionnaire items includes:

[0063] S1. Obtain the set of completed questionnaires and identify the answers to each question in the set of completed questionnaires.

[0064] In this embodiment of the invention, the survey topics of the completed questionnaires can cover various aspects such as medical care, finance, and education.

[0065] In this embodiment of the invention, the completed questionnaire includes various question types such as multiple choice questions, true / false questions, and open-ended questions. The multiple choice questions may contain multiple different options, for example, one multiple choice question may contain four options: A, B, C, and D.

[0066] In this embodiment of the invention, the questions in the completed questionnaire may include basic questions and core questions. The basic questions include some basic information about the user; the core questions are the main aspects of this questionnaire, such as medical care, finance, education, etc.

[0067] In this embodiment of the invention, before identifying the answers to each question in the completed questionnaire, missing values ​​are identified for each question, and questions with missing values ​​greater than a preset threshold are removed from the completed questionnaire; and completed questionnaires of the core target group are selected based on the basic questions in the completed questionnaire.

[0068] S2. Identify the question type of each question and determine the type of the question.

[0069] In this embodiment of the invention, the question type of each question can be identified using a preset deep learning model such as the preset BERT (Bidirectional Encoder Representations from Transformer) model and the preset Bi-directional Long Short-Term Memory (BiLSTM) model.

[0070] In this embodiment of the invention, the question type is a multiple-choice question or an open-ended question.

[0071] When the question type is a multiple choice question, S3, select target multiple choice questions to be modified according to the selection ratio of the answers to each question, and use the pre-trained semantic understanding model to perform intent recognition on the target multiple choice questions to be modified, so as to obtain the intent label and slot label of the target multiple choice questions to be modified.

[0072] In this embodiment of the invention, the pre-trained semantic understanding model includes an encoding layer, a fusion layer, an intent recognition layer, and a semantic slot filling task layer. The encoding layer includes a character encoder and a word encoder. The fusion layer includes an MLP Attention module and a neural network.

[0073] The character encoder may employ a pre-trained language model with deep learning capabilities, including, but not limited to, ERNIE (Enhanced Language Representation with Informative Entities), Robertat (Robustly Optimized BERT Pretraining Approach), and BERT (Bidirectional Encoder Representations from Transformer).

[0074] The word encoder may be a self-attention encoder, which includes a self-attention module and a bidirectional long short-term memory network (BiLSTM).

[0075] In detail, the process of filtering target multiple-choice questions for modification based on the selection ratio of the answers to each question, as described in S3, includes:

[0076] Statistical analysis of the selection ratio of each option in the answers to each question;

[0077] Multiple-choice questions whose selection ratio exceeds a preset threshold are designated as target multiple-choice questions to be modified.

[0078] In this embodiment of the invention, the completed questionnaire set contains questionnaires filled out by multiple users. Because the sample size of the questionnaire set is large enough, when the proportion of users selecting the same option for a certain multiple-choice question is higher than a certain proportion, it indicates that the multiple-choice question was not comprehensively considered in a certain aspect when it was set up, and needs to be adjusted and modified.

[0079] In this embodiment of the invention, target multiple-choice questions to be modified are selected based on the selection ratio of the answers to each question, eliminating the need for manual selection and improving the efficiency of modifying questionnaire questions.

[0080] Further, see Figure 2 As shown in S3, the pre-trained semantic understanding model is used to perform intent recognition on the target multiple-choice question to be modified, obtaining the intent label and slot label of the target multiple-choice question to be modified, including:

[0081] S31. Segment the question stem of the target multiple-choice question to be modified into words to obtain the question stem word segmentation sequence;

[0082] S32. Segment the question stem of the target multiple-choice question to be modified into characters to obtain the question stem character sequence;

[0083] S33. The word segmentation sequence and the character sequence of the question stem are encoded by the encoding layer in the pre-trained semantic understanding model to obtain the word encoding sequence and the character encoding sequence.

[0084] S34. Using the fusion layer in the semantic understanding model, feature fusion is performed on the word encoding sequence and the character encoding sequence to obtain a fusion vector;

[0085] S35. Classify the fusion vector using the intent recognition layer in the semantic understanding model to obtain the intent label of the target multiple-choice question to be modified;

[0086] S36. Using the semantic slot filling task layer in the semantic understanding model, slot prediction is performed on the word encoding sequence and the character encoding sequence to obtain the slot label of the target multiple-choice question to be modified.

[0087] Furthermore, S33 includes:

[0088] Add the CLS symbol to the beginning of the question stem character sequence, and add a separator after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence;

[0089] The target question stem character sequence is encoded using the character encoder in the encoding layer to obtain a character encoding sequence;

[0090] The contextual information of the word segmentation sequence in the question stem is extracted using the self-attention mechanism module of the word encoder in the encoding layer;

[0091] The sequential information of the word segmentation sequence of the question stem is obtained by utilizing the bidirectional long short-term memory network of the word encoder in the coding layer;

[0092] The context information and the sequence information are processed by a full connection to obtain the word encoding sequence.

[0093] Specifically, the sequence of characters in the question stem is , wherein For the Nth character of the question stem character sequence, add [CLS] to the beginning of the question stem character sequence, and add the separator [SEP] after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence. For example: if the question stem character sequence is "Which financial institutions do you mainly do business with?", after adding special characters, the resulting question stem character sequence with special characters is "[CLS] Which financial institutions do you mainly do business with? [SEP]". After encoding by the BERT model, the resulting character encoding sequence is: , wherein This represents the encoding vector corresponding to the Nth character.

[0094] Specifically, the word segmentation sequence of the question stem is as follows: , wherein This represents the Mth word in the word segmentation sequence of the question stem; the context information and sequence information of the word segmentation sequence are extracted, and the context information and sequence information are processed by a fully connected layer to obtain the word encoding sequence. , wherein Let M represent the encoding vector corresponding to the Mth word, where M≤N.

[0095] In this embodiment of the invention, attention mechanism is used to obtain contextual information in the question stem, and bidirectional long short-term memory network is used to obtain sequential information in the question stem. By making full use of sequential information and contextual information, the loss of some semantic information during the parsing process is avoided, thereby improving the accuracy of understanding the question.

[0096] Furthermore, S36 includes:

[0097] The word encoding sequence is decoded using the word decoder in the semantic slot filling task layer of the semantic understanding model to obtain the word hidden state of each slot;

[0098] The character encoding sequence is decoded using the character decoder in the semantic slot filling task layer to obtain the hidden state of the character in each slot;

[0099] The hidden states of characters and words are fused using a preset fusion function to obtain the hidden states of slots;

[0100] The hidden states of the slots are classified using a pre-defined classifier in the semantic slot filling task layer to obtain slot labels.

[0101] In this embodiment of the invention, the semantic slot filling task layer includes a character decoder, a word decoder, and a classification model. The character decoder can be constructed using a Long Short-Term Memory (LSTM) network; the word decoder can be constructed using a Bi-directional Long Short-Term Memory (BiLSTM) network; and the classification model can be constructed using a pre-trained language model.

[0102] In one embodiment of the present invention, the preset fusion function may be Fuse().

[0103] In this embodiment of the invention, the pre-trained semantic understanding model is used to identify the intent of the target multiple-choice question to be modified, which can accurately analyze the question intent and slot labels, thereby improving the accuracy of questionnaire question modification.

[0104] S4. Based on the intent tag and the slot tag, construct modification suggestions for the target question to be modified.

[0105] Specifically, S4 includes:

[0106] The intent tags are used as suggestions for modifying the question stem of the target multiple-choice question to be modified;

[0107] The target text is matched from a preset dictionary using the slot labels;

[0108] The target text is used as a modification suggestion for the options of the initial target question to be modified.

[0109] In this embodiment of the invention, the preset dictionary database is a dictionary database constructed based on experience or authoritative dictionaries.

[0110] In this embodiment of the invention, for example, the question stem is "Which financial institutions do you mainly do business with?", the intent is "Select the financial institutions with which you do business", the slot label is "financial institutions", and the text result of matching the target text from the preset dictionary using the slot label is: state-owned banks, postal savings, rural credit cooperatives, microfinance companies, and village banks.

[0111] In this embodiment of the invention, modification suggestions for the target multiple-choice questions to be modified are constructed based on the intent tags and the slot tags, making the questionnaire questions more consistent with the actual situation, thereby improving the accuracy of questionnaire question modification.

[0112] When the question type is a question-and-answer question, S5, obtain the survey topic of the completed questionnaire set, and use the semantic understanding model to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively, to obtain the topic intent label and the question-and-answer question intent label of the survey topic.

[0113] In this embodiment of the invention, the method of using the semantic understanding model to perform intent recognition on the survey topic and the questions to obtain the topic intent label and the question intent label of the survey topic is similar to the method of using the pre-trained semantic understanding model to perform intent recognition on the target multiple-choice questions to be modified in S3, and will not be described in detail here.

[0114] S6. Calculate the similarity between the topic intent tag and the question intent tag, and take the question and answer questions with a similarity less than a preset similarity threshold as target question and answer questions to be modified, and take the topic intent as a modification suggestion for the target question and answer questions to be modified.

[0115] In this embodiment of the invention, common similarity algorithms can be used to calculate the similarity between the topic intent tag and the question-and-answer intent tag, such as Jaccard similarity coefficient, cosine similarity, and other algorithms.

[0116] In this embodiment of the invention, when the similarity between the topic intent tag and the question-and-answer intent tag is less than the preset similarity threshold, it indicates that the question-and-answer question does not match the survey topic and needs to be modified to make the questionnaire questions more relevant to the topic.

[0117] S7. Obtain the question IDs of the target multiple-choice questions to be modified and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0118] In this embodiment of the invention, each question in the preset question bank has a unique question ID.

[0119] For details, please refer to Figure 3 As shown, S7 describes modifying the corresponding questionnaire questions in the preset question bank according to the question ID and the modification suggestion, including:

[0120] S71. Based on the question ID, query the target question to be modified from the preset question bank;

[0121] S72. Add the target modified question to the question modification queue and generate modification prompt information;

[0122] S73. The questionnaire creator modifies the target modified question in the preset question bank according to the prompt information and the modification suggestion, retains the target modified question and the question ID, generates a new question ID for the modified target modified question in the preset data, and associates the new question ID with the target to be modified multiple choice question or the target to be modified open-ended question in the completed questionnaire set.

[0123] In this embodiment of the invention, each question in the completed questionnaire set can only be modified and edited in the preset question bank, and is treated differently according to the questionnaire status associated with the target modified question; the modified target modified question needs to be used as a new question to generate a new question ID.

[0124] In one embodiment of the present invention, the questionnaire creator can find the corresponding question in the preset question bank according to the question ID in the prompt information, and modify the question according to the modification suggestion.

[0125] In this embodiment of the invention, the unique ID of each question in the question bank is used to identify the question in different questionnaires or different versions of the same questionnaire. In subsequent questionnaire result analysis, the boundaries of questionnaires and questionnaire versions can be broken, and data integration analysis can be achieved.

[0126] This invention improves the efficiency of questionnaire question modification by filtering target multiple-choice questions based on the selection ratio of answers to each question, eliminating the need for manual screening. A pre-trained semantic understanding model is used to identify the intent of the target multiple-choice questions, accurately analyzing the question intent and slot tags. Modification suggestions are then constructed based on these intent and slot tags, thereby improving the accuracy of questionnaire question modification. Furthermore, the semantic understanding model is used to identify the intent of the survey topic and the questions, respectively, obtaining the topic intent tag and the question / answer tag. The method involves using intent tags to calculate the similarity between the topic intent tags and the question-and-answer intent tags. Questions with similarity scores below a preset similarity threshold are selected as target questions to be modified. The topic intent is used as a modification suggestion for the target questions to be modified, making the questionnaire questions more relevant to the topic and thus improving the accuracy of questionnaire modification. The method also involves obtaining the question IDs of the target multiple-choice questions and the target question-and-answer questions to be modified. Based on the question IDs and the modification suggestions, the corresponding questionnaire questions in a preset question bank are modified. Modifying questions in the question bank using unique question IDs breaks down the boundaries between questionnaires and questionnaire versions, enabling data integration and analysis and improving the efficiency of questionnaire modification. Therefore, the dynamic questionnaire question modification method proposed in this invention can solve the problem of low accuracy and efficiency in questionnaire modification.

[0127] like Figure 4 The diagram shown is a functional block diagram of a questionnaire question dynamic modification device provided in an embodiment of the present invention.

[0128] The questionnaire question dynamic modification device 100 of the present invention can be installed in an electronic device. Depending on the functions implemented, the questionnaire question dynamic modification device 100 may include a questionnaire acquisition module 101, a question type recognition module 102, a multiple-choice question modification suggestion generation module 103, an open-ended question modification suggestion generation module 104, and a questionnaire modification module 105. The module described in this invention can also be called a unit, which refers to a series of computer program segments that can be executed by the processor of an electronic device and can perform a fixed function, and are stored in the memory of the electronic device.

[0129] In this embodiment, the functions of each module / unit are as follows:

[0130] The questionnaire acquisition module 101 is used to acquire a set of completed questionnaires and identify the answers to each question in the set of completed questionnaires.

[0131] In this embodiment of the invention, the survey topics of the completed questionnaires can cover various aspects such as medical care, finance, and education.

[0132] In this embodiment of the invention, the completed questionnaire includes various question types such as multiple choice questions, true / false questions, and open-ended questions. The multiple choice questions may contain multiple different options, for example, one multiple choice question may contain four options: A, B, C, and D.

[0133] In this embodiment of the invention, the questions in the completed questionnaire may include basic questions and core questions. The basic questions include some basic information about the user; the core questions are the main aspects of this questionnaire, such as medical care, finance, education, etc.

[0134] In this embodiment of the invention, before identifying the answers to each question in the completed questionnaire, missing values ​​are identified for each question, and questions with missing values ​​greater than a preset threshold are removed from the completed questionnaire; and completed questionnaires of the core target group are selected based on the basic questions in the completed questionnaire.

[0135] The question type recognition module 102 is used to identify the question type of each question and determine the type of the question type;

[0136] In this embodiment of the invention, the question type of each question can be identified using a preset deep learning model such as the preset BERT (Bidirectional Encoder Representations from Transformer) model and the preset Bi-directional Long Short-Term Memory (BiLSTM) model.

[0137] In this embodiment of the invention, the question types are multiple choice questions, short answer questions, and true / false questions.

[0138] The multiple-choice question modification suggestion generation module 103 is used to filter target multiple-choice questions to be modified according to the selection ratio of the answers to each question when the question type is a multiple-choice question, use a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice questions to be modified, and obtain the intent tag and slot tag of the target multiple-choice questions to be modified; and construct modification suggestions for the target multiple-choice questions to be modified based on the intent tag and the slot tag.

[0139] In this embodiment of the invention, the pre-trained semantic understanding model includes an encoding layer, a fusion layer, an intent recognition layer, and a semantic slot filling task layer. The encoding layer includes a character encoder and a word encoder. The fusion layer includes an MLP Attention module and a neural network.

[0140] The character encoder may employ a pre-trained language model with deep learning capabilities, including, but not limited to, ERNIE (Enhanced Language Representation with Informative Entities), Robertat (Robustly Optimized BERT Pretraining Approach), and BERT (Bidirectional Encoder Representations from Transformer).

[0141] The word encoder may be a self-attention encoder, which includes a self-attention module and a bidirectional long short-term memory network (BiLSTM).

[0142] In detail, the multiple-choice question modification suggestion generation module 103 uses the following technical means to filter target multiple-choice questions to be modified based on the selection ratio of the answers to each question, including:

[0143] Statistical analysis of the selection ratio of each option in the answers to each question;

[0144] Multiple-choice questions whose selection ratio exceeds a preset threshold are designated as target multiple-choice questions to be modified.

[0145] In this embodiment of the invention, the completed questionnaire set contains questionnaires filled out by multiple users. Because the sample size of the questionnaire set is large enough, when the proportion of users selecting the same option for a certain multiple-choice question is higher than a certain proportion, it indicates that the multiple-choice question was not comprehensively considered in a certain aspect when it was set up, and needs to be adjusted and modified.

[0146] In this embodiment of the invention, target multiple-choice questions to be modified are selected based on the selection ratio of the answers to each question, eliminating the need for manual selection and improving the efficiency of modifying questionnaire questions.

[0147] Furthermore, the multiple-choice question modification suggestion generation module 103 utilizes a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice question to be modified, obtaining the intent label and slot label of the target multiple-choice question to be modified, including:

[0148] The question stems of the target multiple-choice questions to be modified are segmented into words to obtain a word segmentation sequence of the question stems;

[0149] The question stem of the target multiple-choice question to be modified is segmented into characters to obtain the question stem character sequence;

[0150] The word segmentation sequence and the character sequence of the question stem are encoded by the encoding layer in the pre-trained semantic understanding model to obtain the word encoding sequence and the character encoding sequence.

[0151] The semantic understanding model utilizes a fusion layer to perform feature fusion on the word encoding sequence and the character encoding sequence to obtain a fusion vector.

[0152] The fusion vector is classified using the intent recognition layer in the semantic understanding model to obtain the intent label of the target multiple-choice question to be modified;

[0153] The semantic slot filling task layer in the semantic understanding model is used to predict the slots of the word encoding sequence and the character encoding sequence to obtain the slot labels of the target multiple-choice question to be modified.

[0154] Furthermore, the multiple-choice question modification suggestion generation module 103 utilizes the encoding layer in the pre-trained semantic understanding model to encode the word segmentation sequence and the character sequence of the question stem, respectively, to obtain a word encoding sequence and a character encoding sequence, including:

[0155] Add the CLS symbol to the beginning of the question stem character sequence, and add a separator after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence;

[0156] The target question stem character sequence is encoded using the character encoder in the encoding layer to obtain a character encoding sequence;

[0157] The contextual information of the word segmentation sequence in the question stem is extracted using the self-attention mechanism module of the word encoder in the encoding layer;

[0158] The sequential information of the word segmentation sequence of the question stem is obtained by utilizing the bidirectional long short-term memory network of the word encoder in the coding layer;

[0159] The context information and the sequence information are processed by a full connection to obtain the word encoding sequence.

[0160] Specifically, the sequence of characters in the question stem is , wherein For the Nth character of the question stem character sequence, add [CLS] to the beginning of the question stem character sequence, and add the separator [SEP] after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence. For example: if the question stem character sequence is "Which financial institutions do you mainly do business with?", after adding special characters, the resulting question stem character sequence with special characters is "[CLS] Which financial institutions do you mainly do business with? [SEP]". After encoding by the BERT model, the resulting character encoding sequence is: , wherein This represents the encoding vector corresponding to the Nth character.

[0161] Specifically, the word segmentation sequence of the question stem is as follows: , wherein This represents the Mth word in the word segmentation sequence of the question stem; the context information and sequence information of the word segmentation sequence are extracted, and the context information and sequence information are processed by a fully connected layer to obtain the word encoding sequence. , wherein Let M represent the encoding vector corresponding to the Mth word, where M≤N.

[0162] In this embodiment of the invention, attention mechanism is used to obtain contextual information in the question stem, and bidirectional long short-term memory network is used to obtain sequential information in the question stem. By making full use of sequential information and contextual information, the loss of some semantic information during the parsing process is avoided, thereby improving the accuracy of understanding the question.

[0163] Furthermore, the multiple-choice question modification suggestion generation module 103 utilizes the semantic slot filling task layer in the semantic understanding model to perform slot prediction on the word encoding sequence and the character encoding sequence to obtain the slot label of the target multiple-choice question to be modified, including:

[0164] The word encoding sequence is decoded using the word decoder in the semantic slot filling task layer of the semantic understanding model to obtain the word hidden state of each slot;

[0165] The character encoding sequence is decoded using the character decoder in the semantic slot filling task layer to obtain the hidden state of the character in each slot;

[0166] The hidden states of characters and words are fused using a preset fusion function to obtain the hidden states of slots;

[0167] The hidden states of the slots are classified using a pre-defined classifier in the semantic slot filling task layer to obtain slot labels.

[0168] In this embodiment of the invention, the semantic slot filling task layer includes a character decoder, a word decoder, and a classification model. The character decoder can be constructed using a Long Short-Term Memory (LSTM) network; the word decoder can be constructed using a Bi-directional Long Short-Term Memory (BiLSTM) network; and the classification model can be constructed using a pre-trained language model.

[0169] In one embodiment of the present invention, the preset fusion function may be Fuse().

[0170] In this embodiment of the invention, the pre-trained semantic understanding model is used to identify the intent of the target multiple-choice question to be modified, which can accurately analyze the question intent and slot labels, thereby improving the accuracy of questionnaire question modification.

[0171] Furthermore, the multiple-choice question modification suggestion generation module 103 utilizes the following technical means to construct modification suggestions for the target multiple-choice question to be modified based on the intent tag and the slot tag, including:

[0172] The intent tags are used as suggestions for modifying the question stem of the target multiple-choice question to be modified;

[0173] The target text is matched from a preset dictionary using the slot labels;

[0174] The target text is used as a modification suggestion for the options of the target multiple-choice question to be modified.

[0175] In this embodiment of the invention, the preset dictionary database is a dictionary database constructed based on experience or authoritative dictionaries.

[0176] In this embodiment of the invention, for example, the question stem is "Which financial institutions do you mainly do business with?", the intent is "Select the financial institutions with which you do business", the slot label is "financial institutions", and the text result of matching the target text from the preset dictionary using the slot label is: state-owned banks, postal savings, rural credit cooperatives, microfinance companies, and village banks.

[0177] In this embodiment of the invention, modification suggestions for the target multiple-choice questions to be modified are constructed based on the intent tags and the slot tags, making the questionnaire questions more consistent with the actual situation, thereby improving the accuracy of questionnaire question modification.

[0178] The question-and-answer modification suggestion generation module 104 is used to, when the question type is a question-and-answer question, obtain the survey topic of the completed questionnaire set, use the semantic understanding model to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively, and obtain the topic intent tag and the question intent tag of the survey topic; calculate the similarity between the topic intent tag and the question intent tag, and take the question-and-answer questions with a similarity less than a preset similarity threshold as target question-and-answer questions to be modified, and use the topic intent as a modification suggestion for the target question-and-answer questions to be modified;

[0179] In this embodiment of the invention, the method of using the semantic understanding model to perform intent recognition on the survey topic and the questions to obtain the topic intent label and the question intent label of the survey topic is similar to the method of using the pre-trained semantic understanding model to perform intent recognition on the target multiple-choice questions to be modified in S3, and will not be described in detail here.

[0180] In this embodiment of the invention, common similarity algorithms can be used to calculate the similarity between the topic intent tag and the question-and-answer intent tag, such as Jaccard similarity coefficient, cosine similarity, and other algorithms.

[0181] In this embodiment of the invention, when the similarity between the topic intent tag and the question-and-answer intent tag is less than the preset similarity threshold, it indicates that the question-and-answer question does not match the survey topic and needs to be modified to make the questionnaire questions more relevant to the topic.

[0182] The questionnaire modification module 105 is used to obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0183] In this embodiment of the invention, each question in the preset question bank has a unique question ID.

[0184] In detail, the questionnaire modification module 105 uses the following technical means to modify the corresponding questionnaire questions in the preset question bank according to the question ID and the modification suggestions, including:

[0185] Based on the question ID, retrieve the target question from the preset question bank to modify;

[0186] The target question to be modified is added to the question modification queue, and modification prompt information is generated;

[0187] The questionnaire creator modifies the target modified questions in the preset question bank according to the prompt information and the modification suggestions, retains the target modified questions and the question ID, generates a new question ID for the modified target modified questions in the preset data, and associates the new question ID with the target to be modified multiple-choice questions or target to be modified open-ended questions in the completed questionnaire set.

[0188] In this embodiment of the invention, each question in the completed questionnaire set can only be modified and edited in the preset question bank, and is treated differently according to the questionnaire status associated with the target modified question; the modified target modified question needs to be used as a new question to generate a new question ID.

[0189] In one embodiment of the present invention, the questionnaire creator can find the corresponding question in the preset question bank according to the question ID in the prompt information, and modify the question according to the modification suggestion.

[0190] In this embodiment of the invention, the unique ID of each question in the question bank is used to identify the question in different questionnaires or different versions of the same questionnaire. In subsequent questionnaire result analysis, the boundaries of questionnaires and questionnaire versions can be broken, and data integration analysis can be achieved.

[0191] In detail, each module in the questionnaire question dynamic modification device 100 described in this embodiment of the invention adopts the same usage as described above. Figures 1 to 3 The method used is the same as the dynamic modification method for questionnaire questions described above, and it can produce the same technical effect, so it will not be elaborated here.

[0192] like Figure 5 The diagram shown is a structural schematic of an electronic device for implementing a method for dynamically modifying questionnaire questions, according to an embodiment of the present invention.

[0193] The electronic device 1 may include a processor 10, a memory 11, a communication bus 12 and a communication interface 13, and may also include a computer program stored in the memory 11 and capable of running on the processor 10, such as a questionnaire question dynamic modification program.

[0194] In some embodiments, the processor 10 may be composed of integrated circuits, such as a single packaged integrated circuit or multiple integrated circuits with the same or different functions, including combinations of one or more central processing units (CPUs), microprocessors, digital processing chips, graphics processors, and various control chips. The processor 10 is the control unit of the electronic device, connecting various components of the entire electronic device through various interfaces and lines. It executes programs or modules stored in the memory 11 (e.g., executing a questionnaire question dynamic modification program) and calls data stored in the memory 11 to perform various functions of the electronic device and process data.

[0195] The memory 11 includes at least one type of readable storage medium, including flash memory, portable hard drive, multimedia card, card-type memory (e.g., SD or DX memory), magnetic memory, disk, optical disk, etc. In some embodiments, the memory 11 can be an internal storage unit of an electronic device, such as a portable hard drive. In other embodiments, the memory 11 can be an external storage device of the electronic device, such as a plug-in portable hard drive, smart media card (SMC), secure digital (SD) card, flash card, etc. Furthermore, the memory 11 can include both internal and external storage units of the electronic device. The memory 11 can be used not only to store application software and various types of data installed on the electronic device, such as code for dynamically modifying questionnaire questions, but also to temporarily store data that has been output or will be output.

[0196] The communication bus 12 can be a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus, etc. This bus can be divided into an address bus, a data bus, a control bus, etc. The bus is configured to enable communication between the memory 11 and at least one processor 10, etc.

[0197] The communication interface 13 is used for communication between the aforementioned electronic device and other devices, including a network interface and a user interface. Optionally, the network interface may include a wired interface and / or a wireless interface (such as a Wi-Fi interface, Bluetooth interface, etc.), typically used to establish communication connections between the electronic device and other electronic devices. The user interface may be a display, an input unit (such as a keyboard), or optionally, a standard wired or wireless interface. Optionally, in some embodiments, the display may be an LED display, a liquid crystal display, a touch-sensitive liquid crystal display, or an OLED (Organic Light-Emitting Diode) touchscreen, etc. The display may also be appropriately referred to as a screen or display unit, used to display information processed in the electronic device and to display a visual user interface.

[0198] Figure 5 Only electronic devices with components are shown; it will be understood by those skilled in the art that... Figure 5 The structure shown does not constitute a limitation on the electronic device 1, and may include fewer or more components than shown, or combine certain components, or have different component arrangements.

[0199] For example, although not shown, the electronic device may also include a power supply (such as a battery) to power the various components. Preferably, the power supply can be logically connected to the at least one processor 10 through a power management device, thereby enabling functions such as charging management, discharging management, and power consumption management. The power supply may also include one or more DC or AC power supplies, recharging devices, power fault detection circuits, power converters or inverters, power status indicators, and other arbitrary components. The electronic device may also include various sensors, Bluetooth modules, Wi-Fi modules, etc., which will not be described in detail here.

[0200] It should be understood that the embodiments described are for illustrative purposes only and are not limited to this structure in the scope of the patent application.

[0201] The questionnaire question dynamic modification program stored in the memory 11 of the electronic device 1 is a combination of multiple instructions. When run in the processor 10, it can achieve the following:

[0202] Obtain a set of completed questionnaires and identify the answers to each question in the completed questionnaire set;

[0203] Identify the question type of each question and determine the type of the question;

[0204] When the question type is a multiple choice question, target multiple choice questions to be modified are selected according to the selection ratio of the answers to each question. The pre-trained semantic understanding model is used to perform intent recognition on the target multiple choice questions to be modified, and the intent label and slot label of the target multiple choice questions to be modified are obtained.

[0205] Based on the intent tags and the slot tags, construct modification suggestions for the target multiple-choice question to be modified;

[0206] When the question type is a question-and-answer question, the survey topic of the completed questionnaire set is obtained, and the semantic understanding model is used to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent label and the question-and-answer question intent label of the survey topic;

[0207] Calculate the similarity between the topic intent tag and the question intent tag, and select the question and answer questions with a similarity less than a preset similarity threshold as target question and answer questions to be modified, and use the topic intent as a modification suggestion for the target question and answer questions to be modified;

[0208] Obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0209] Specifically, the specific implementation method of the processor 10 for the above instructions can be referred to the description of the relevant steps in the corresponding embodiment of the accompanying drawings, and will not be repeated here.

[0210] Furthermore, if the modules / units integrated in the electronic device 1 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. The computer-readable storage medium can be volatile or non-volatile. For example, the computer-readable medium may include: any entity or device capable of carrying the computer program code, a recording medium, a USB flash drive, a portable hard drive, a magnetic disk, an optical disk, a computer memory, or a read-only memory (ROM).

[0211] The present invention also provides a computer-readable storage medium storing a computer program, which, when executed by a processor of an electronic device, can perform the following:

[0212] Obtain a set of completed questionnaires and identify the answers to each question in the completed questionnaire set;

[0213] Identify the question type of each question and determine the type of the question;

[0214] When the question type is a multiple choice question, target multiple choice questions to be modified are selected according to the selection ratio of the answers to each question. The pre-trained semantic understanding model is used to perform intent recognition on the target multiple choice questions to be modified, and the intent label and slot label of the target multiple choice questions to be modified are obtained.

[0215] Based on the intent tags and the slot tags, construct modification suggestions for the target multiple-choice question to be modified;

[0216] When the question type is a question-and-answer question, the survey topic of the completed questionnaire set is obtained, and the semantic understanding model is used to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent label and the question-and-answer question intent label;

[0217] Calculate the similarity between the topic intent tag and the question intent tag, and select the question and answer questions with a similarity less than a preset similarity threshold as target question and answer questions to be modified, and use the topic intent as a modification suggestion for the target question and answer questions to be modified;

[0218] Obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

[0219] In the several embodiments provided by this invention, it should be understood that the disclosed devices, apparatuses, and methods can be implemented in other ways. For example, the apparatus embodiments described above are merely illustrative; for instance, the division of modules is only a logical functional division, and other division methods may be used in actual implementation.

[0220] The modules described as separate components may or may not be physically separate. The components shown as modules may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs.

[0221] Furthermore, the functional modules in the various embodiments of the present invention can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or in the form of hardware plus software functional modules.

[0222] It will be apparent to those skilled in the art that the present invention is not limited to the details of the exemplary embodiments described above, and that the present invention can be implemented in other specific forms without departing from the spirit or essential characteristics of the present invention.

[0223] Therefore, the embodiments should be considered exemplary and non-limiting in all respects, and the scope of the invention is defined by the appended claims rather than the foregoing description. Thus, all variations falling within the meaning and scope of equivalents of the claims are intended to be embraced within the invention. No appended diagram markings in the claims should be construed as limiting the scope of the claims.

[0224] The blockchain referred to in this invention is a novel application model of computer technologies such as distributed data storage, peer-to-peer transmission, consensus mechanisms, and encryption algorithms. Essentially, a blockchain is a decentralized database, a chain of data blocks linked together using cryptographic methods. Each data block contains information about a batch of network transactions, used to verify the validity of the information (anti-counterfeiting) and generate the next block. A blockchain can include an underlying blockchain platform, a platform product service layer, and an application service layer.

[0225] The embodiments of this application can acquire and process relevant data based on artificial intelligence technology. Artificial intelligence (AI) refers to the theories, methods, technologies, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to obtain optimal results.

[0226] Furthermore, it is clear that the word "comprising" does not exclude other units or steps, and the singular does not exclude the plural. Multiple units or devices recited in a system claim may also be implemented by a single unit or device through software or hardware. The terms "first," "second," etc., are used to indicate names and do not indicate any specific order.

[0227] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention.

Claims

1. A method for dynamically modifying questionnaire items, characterized in that, The method includes: Obtain a set of completed questionnaires and identify the answers to each question in the completed questionnaire set; Identify the question type of each question and determine the type of the question; When the question type is a multiple choice question, target multiple choice questions to be modified are selected according to the selection ratio of the answers to each question. The pre-trained semantic understanding model is used to perform intent recognition on the target multiple choice questions to be modified, and the intent label and slot label of the target multiple choice questions to be modified are obtained. Based on the intent tags and the slot tags, construct modification suggestions for the target multiple-choice question to be modified; When the question type is a question-and-answer question, the survey topic of the completed questionnaire set is obtained, and the semantic understanding model is used to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent label and the question-and-answer question intent label of the survey topic; Calculate the similarity between the topic intent tag and the question intent tag, and select the question and answer questions with a similarity less than a preset similarity threshold as target question and answer questions to be modified, and use the topic intent as a modification suggestion for the target question and answer questions to be modified; Obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

2. The method for dynamically modifying questionnaire items as described in claim 1, characterized in that, The process of filtering target multiple-choice questions for modification based on the selection ratio of the answers to each question includes: Statistical analysis of the selection ratio of each option in the answers to each question; Multiple-choice questions whose selection ratio exceeds a preset threshold are designated as target multiple-choice questions to be modified.

3. The method for dynamically modifying questionnaire items as described in claim 1, characterized in that, The step of constructing modification suggestions for the target multiple-choice question based on the intent tag and the slot tag includes: The intent tags are used as suggestions for modifying the question stem of the target multiple-choice question to be modified; The target text is matched from a preset dictionary using the slot labels; The target text is used as a modification suggestion for the options of the target multiple-choice question to be modified.

4. The method for dynamically modifying questionnaire items as described in claim 1, characterized in that, The step of modifying the corresponding questionnaire question in the preset question bank according to the question ID and the modification suggestion includes: Based on the question ID, retrieve the target question from the preset question bank to modify; The target question to be modified is added to the question modification queue, and modification prompt information is generated; Based on the prompts and modification suggestions, the questionnaire creator modifies the target modification questions in the preset question bank, retains the target modification questions and the question IDs, generates new question IDs for the modified target modification questions in the preset data, and associates the new question IDs with the target multiple-choice questions or target open-ended questions in the completed questionnaire set.

5. The method for dynamically modifying questionnaire items as described in claim 1, characterized in that, The process involves using a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice question to be modified, obtaining the intent label and slot label of the target multiple-choice question to be modified, including: The question stems of the target multiple-choice questions to be modified are segmented into words to obtain a word segmentation sequence of the question stems; The question stem of the target multiple-choice question to be modified is segmented into characters to obtain the question stem character sequence; The word segmentation sequence and the character sequence of the question stem are encoded by the encoding layer in the pre-trained semantic understanding model to obtain the word encoding sequence and the character encoding sequence. The semantic understanding model utilizes a fusion layer to perform feature fusion on the word encoding sequence and the character encoding sequence to obtain a fusion vector. The fusion vector is classified using the intent recognition layer in the semantic understanding model to obtain the intent label of the target multiple-choice question to be modified; The semantic slot filling task layer in the semantic understanding model is used to predict the slots of the word encoding sequence and the character encoding sequence to obtain the slot labels of the target multiple-choice question to be modified.

6. The method for dynamically modifying questionnaire items as described in claim 5, characterized in that, The step of using the semantic slot filling task layer in the semantic understanding model to predict slots in the word encoding sequence and the character encoding sequence to obtain the slot labels of the target multiple-choice question to be modified includes: The word encoding sequence is decoded using the word decoder in the semantic slot filling task layer of the semantic understanding model to obtain the word hidden state of each slot; The character encoding sequence is decoded using the character decoder in the semantic slot filling task layer to obtain the hidden state of the character in each slot; The hidden states of characters and words are fused using a preset fusion function to obtain the hidden states of slots; The hidden states of the slots are classified using a pre-defined classifier in the semantic slot filling task layer to obtain slot labels.

7. The method for dynamically modifying questionnaire items as described in claim 5, characterized in that, The encoding layer in the pre-trained semantic understanding model encodes the word segmentation sequence and the character sequence of the question stem, respectively, to obtain the word encoding sequence and the character encoding sequence, including: Add the CLS symbol to the beginning of the question stem character sequence, and add a separator after the last character of each sentence corresponding to the question stem character sequence to obtain the target question stem character sequence; The target question stem character sequence is encoded using the character encoder in the encoding layer to obtain a character encoding sequence; The contextual information of the word segmentation sequence in the question stem is extracted using the self-attention mechanism module of the word encoder in the encoding layer; The sequential information of the word segmentation sequence of the question stem is obtained by utilizing the bidirectional long short-term memory network of the word encoder in the coding layer; The context information and the sequence information are processed by a full connection to obtain the word encoding sequence.

8. A device for dynamically modifying questionnaire items, characterized in that, The device includes: The questionnaire acquisition module is used to acquire a set of completed questionnaires and identify the answers to each question in the completed questionnaire set; The question type recognition module is used to identify the question type of each question and determine the type of the question. The multiple-choice question modification suggestion generation module is used to filter target multiple-choice questions to be modified based on the selection ratio of the answers to each question when the question type is multiple-choice. It uses a pre-trained semantic understanding model to perform intent recognition on the target multiple-choice questions to be modified, and obtains the intent tag and slot tag of the target multiple-choice questions to be modified. Based on the intent tag and the slot tag, it constructs modification suggestions for the target multiple-choice questions to be modified. The question-and-answer modification suggestion generation module is used to obtain the survey topic of the completed questionnaire set when the question type is a question-and-answer question, and use the semantic understanding model to perform intent recognition on the survey topic and the questions of the question-and-answer questions respectively to obtain the topic intent tag and the question intent tag of the survey topic; calculate the similarity between the topic intent tag and the question intent tag, and select the question-and-answer questions with a similarity less than a preset similarity threshold as target question-and-answer questions to be modified, and use the topic intent as a modification suggestion for the target question-and-answer questions to be modified; The questionnaire modification module is used to obtain the question IDs of the target multiple-choice questions and the target open-ended questions to be modified, and modify the corresponding questionnaire questions in the preset question bank according to the question IDs and the modification suggestions.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and, A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the questionnaire item dynamic modification method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the questionnaire item dynamic modification method as described in any one of claims 1 to 7.

Images