Questionnaire survey analysis processing system based on different industries
By integrating questionnaire keyword extraction, practitioner feature database and correlation calculation module, combined with the improved YOLO network and ResNet module, accurate matching of practitioners and questionnaire topics is achieved, solving the problem of sample matching bias and improving the efficiency and data quality of questionnaire surveys.
Patent Information
- Application Number
- CN202510806093.5
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-06-17
- Publication Date
- 2025-09-12
Smart Images
Figure CN120633652A_ABST
Abstract
Description
Technical Field
[0001] The present invention relates to the technical field of data analysis and processing, and more particularly to a questionnaire analysis and processing system based on different industries. Background Art
[0002] In today's data-driven era, questionnaires are widely used across a wide range of industries, including education, healthcare, finance, and manufacturing, as a highly efficient means of obtaining information. By conducting questionnaire surveys of practitioners across various industries, we can gain a deeper understanding of the industry's current state, their needs, and their pain points, providing crucial data support for corporate strategic decision-making, product optimization, and policy formulation.
[0003] However, existing questionnaire analysis and processing systems have numerous limitations. The primary limitation is data distortion caused by sample matching bias. Existing methods often screen samples based on basic practitioner attributes (such as industry and job title), without in-depth analysis of their semantic relevance to the questionnaire topic. Traditional systems also underutilize practitioners' historical response data, failing to identify the depth of their engagement with specific topics. These issues lead to widespread resource waste (high proportion of invalid samples) and low confidence in conclusions (data that fails to truly reflect the needs of the target group) in traditional surveys. According to market research organizations, approximately 38% of companies need to repeat surveys due to sample matching bias, incurring an additional 25%-40% in costs. In policymaking, inaccurate samples lead to policy adjustments in 29% of cases, severely impacting implementation efficiency. Therefore, developing a questionnaire processing system that can accurately match target practitioners based on semantic relevance analysis between questionnaire topics and practitioners' historical behavior is of great practical significance for improving data quality, reducing survey costs, and supporting scientific decision-making.
[0004] Therefore, how to provide a questionnaire analysis system that can accurately match target practitioners is an urgent problem that those skilled in the art need to solve. Summary of the Invention
[0005] In view of this, the present invention provides a questionnaire survey analysis and processing system based on different industries, which can accurately match target practitioners based on the semantic correlation analysis between questionnaire topics and practitioners' historical behaviors, thereby increasing the proportion of valid samples in the questionnaire survey.
[0006] In order to achieve the above object, the present invention adopts the following technical solutions:
[0007] A questionnaire survey analysis and processing system based on different industries includes: a questionnaire keyword extraction module, a practitioner feature database, a central control module, an element matching module, and a correlation calculation module, wherein the questionnaire keyword extraction module extracts keyword data of a target questionnaire; the practitioner feature database pre-stores practitioner feature elements and keyword elements involved in historical questionnaires by the practitioners; the keyword data and keyword elements are sent to the element matching module through the central control module, and the practitioner feature elements are sent to the correlation calculation module; the keyword data and keyword elements are matched by the element matching module to obtain a matching result; the matching result is sent to the correlation calculation module through the central control module, and the correlation calculation module calculates the correlation between the practitioners and the target questionnaire keyword data based on the matching result and the practitioner feature elements, and screens the target practitioners based on the correlation.
[0008] Preferably, a keyword extraction model is set up inside the questionnaire keyword extraction module, including an attention module, an expansion normalization module, a coupling strategy module and a keyword extraction module. The keyword extraction model takes the questionnaire title text, question description text and option text as input; the attention module adopts the ResNet basic network, and an attention block is set in each residual block, and adjacent residual blocks are separated by at least two network layers, and outputs a feature matrix with attention weights; the expansion normalization module obtains normalized features of uniform dimension through expansion convolution with different expansion rates; the coupling strategy module performs differential fusion based on field type and outputs fusion features; the keyword extraction module maps the fusion features to the industry keyword library dimension, and finally obtains keyword data through the Softmax function; the keyword extraction model uses classification cross entropy as the loss function, uses Softmax as the activation function, and uses the Adam optimizer to optimize parameters during the training process.
[0009] Preferably, the keyword extraction model adopts an improved YOLO network model, the improved YOLO network model adopts the Mosaic data enhancement algorithm to enhance the input data samples, and the improved YOLO network model uses the GSPDARKNET53 network to extract sample data features and generate multi-scale feature maps.
[0010] Preferably, the improved YOLO network model inserts an ECA component after multiple CSP1-X structures. The ECA component uses global average pooling to convert the input feature map into a one-dimensional vector, then performs one-dimensional convolution and SIGMOID activation in the C dimension, and finally re-inputs the obtained one-dimensional vector into the feature map.
[0011] Preferably, the characteristic elements of practitioners in the practitioner characteristic database include the practitioners' participation in historical questionnaire keywords; it also includes obtaining historical questionnaires, the historical questionnaires including the industry to which the questionnaire belongs, the questionnaire content, the practitioner identification and the corresponding answers; by statistically analyzing the historical questionnaires, determining the participation of each practitioner in the keyword elements, and the industry attribute elements of each keyword element; for each practitioner, storing their keyword element participation and the industry attribute elements of the keyword elements involved; obtaining the practitioner's characteristic elements and keyword elements includes reading the participation of the specified target practitioner in the keyword elements and the keyword elements of the target practitioner from the stored practitioner participation in the keyword elements and industry attribute elements.
[0012] Preferably, the reading of the keyword element participation of the designated target practitioner includes: the number of times the target practitioner mentioned the keyword in the historical questionnaire; or the number of mentions and the corresponding time period; or the number of mentions, the time period and the keyword context attributes.
[0013] Preferably, the element matching module specifically includes: extracting the industry field scope and time period from the keyword data of the target questionnaire; the characteristic elements of the practitioners include the industry activity trajectory information of the practitioners; determining the correlation degree based on the matching results and the characteristic elements of the practitioners, judging whether the industry activity trajectory of the practitioners matches the industry field scope and time period, if so, determining the first correlation parameter as the first preset value; if not, determining it as the third preset value; calculating the correlation degree based on the first correlation parameter and the matching result.
[0014] Preferably, obtaining the industry activity track information of the practitioner specifically includes: obtaining the questionnaire submission record of the practitioner, and generating the industry activity track information based on the submission location and timestamp;
[0015] And / or, determine the industry activity trajectory information based on the practitioner's industry registration information or project participation records.
[0016] Preferably, the determining whether the industry activity trajectory of the practitioner matches the scope and time period of the industry field to which he belongs, and if so, determining the first correlation parameter to be a first preset value, specifically includes:
[0017] If the industry activity trajectory matches the industry field and time period to which the target questionnaire belongs, the first correlation parameter is determined to be the first preset value;
[0018] If the industry activity trajectory does not match the industry field and time period of the target questionnaire, but matches the extended scope, determining the first correlation parameter to be the second preset value; wherein the extended scope includes the core scope;
[0019] If none of them match, the third preset value is determined, wherein the first preset value>the second preset value>the third preset value.
[0020] Through the above technical solutions, it can be seen that compared with the existing technology, the present invention discloses a questionnaire survey analysis and processing system based on different industries. The system integrates questionnaire keyword extraction, practitioner feature database, factor matching and correlation calculation modules, and realizes seamless data flow through the central control module, forming a closed-loop process from keyword extraction to sample screening, thereby improving processing efficiency; adopting an improved YOLO network model and ResNet attention module, combined with Mosaic data enhancement and multi-scale dilated convolution, accurately extracting industry keywords in questionnaire titles, questions and options, supporting industry customized vocabulary, and significantly improving the depth and robustness of keyword extraction; the practitioner feature database stores the number of keyword mentions, time periods, industry attributes and contextual attributes, and the factor matching module combines industry fields, time periods and activity trajectories for multi-dimensional verification, realizing "semantic-industry-time" three-dimensional correlation analysis, and avoiding the one-sidedness of traditional attribute matching. BRIEF DESCRIPTION OF THE DRAWINGS
[0021] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the following briefly introduces the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are merely embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying any creative work.
[0022] Figure 1 This is a structural diagram provided by the present invention. DETAILED DESCRIPTION
[0023] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. 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 making creative efforts are within the scope of protection of the present invention.
[0024] The embodiment of the present invention discloses a questionnaire analysis and processing system based on different industries, such as Figure 1As shown, it includes: a questionnaire keyword extraction module, a practitioner feature database, a central control module, an element matching module, and a correlation calculation module. The questionnaire keyword extraction module extracts the keyword data of the target questionnaire; the practitioner feature database pre-stores the practitioner feature elements and the keyword elements involved in the practitioners in the historical questionnaires; the keyword data and the keyword elements are sent to the element matching module through the central control module, and the practitioner feature elements are sent to the correlation calculation module; the keyword data and the keyword elements are matched by the element matching module to obtain the matching result; the matching result is sent to the correlation calculation module through the central control module, and the correlation calculation module calculates the correlation between the practitioners and the target questionnaire keyword data according to the matching result and the practitioner feature elements, and screens the target practitioners based on the correlation.
[0025] In a specific embodiment, a keyword extraction model is set up inside the questionnaire keyword extraction module, including an attention module, an expansion normalization module, a coupling strategy module and a keyword extraction module. The keyword extraction model takes the questionnaire title text, question description text and option text as input; the attention module adopts the ResNet basic network, and an attention block is set in each residual block. Adjacent residual blocks are separated by at least two network layers, and a feature matrix with attention weights is output; the expansion normalization module obtains normalized features of uniform dimension through expansion convolution with different expansion rates; the coupling strategy module performs differential fusion based on field type and outputs fusion features; the keyword extraction module maps the fusion features to the industry keyword library dimension, and finally obtains keyword data through the Softmax function; the keyword extraction model uses classification cross entropy as the loss function, Softmax as the activation function, and the Adam optimizer to optimize the parameters during the training process.
[0026] In a specific embodiment, the keyword extraction model adopts an improved YOLO network model, the improved YOLO network model adopts the Mosaic data enhancement algorithm to enhance the input data samples, and the improved YOLO network model uses the GSPDARKNET53 network to extract sample data features and generate multi-scale feature maps.
[0027] The number of residual components in the CSP1-X structure of the improved YOLO network model is greater than the number of residual components in the CSP1-X structure of the YOLO network model before the improvement. By increasing the number of residual components in the CSP1-X structure of the YOLO network model, the robustness of the model in the keyword extraction process is increased, and the depth and accuracy of keyword extraction are improved.
[0028] In a specific embodiment, the improved YOLO network model inserts ECA components after multiple CSP1-X structures. The ECA component uses global average pooling to convert the input feature map into a one-dimensional vector, then performs one-dimensional convolution and SIGMOID activation in the C dimension, and finally re-inputs the obtained one-dimensional vector into the feature map.
[0029] In a specific embodiment, the characteristic elements of practitioners in the practitioner characteristic database include the practitioners' participation in historical questionnaire keywords; it also includes obtaining historical questionnaires, which include the industry to which the questionnaire belongs, the questionnaire content, the practitioner identification and the corresponding answers; by statistically analyzing the historical questionnaires, determining the participation of each practitioner in the keyword elements, as well as the industry attribute elements of each keyword element; for each practitioner, storing their keyword element participation and the industry attribute elements of the keyword elements involved; obtaining the practitioner's characteristic elements and keyword elements includes reading the participation of the specified target practitioner in the keyword elements and the keyword elements of the target practitioner from the stored practitioner participation in the keyword elements and industry attribute elements.
[0030] In a specific embodiment, reading the participation of the keyword elements of the designated target practitioner includes: the number of times the target practitioner mentioned the keyword in the historical questionnaire; or the number of mentions and the corresponding time period; or the number of mentions, the time period and the keyword context attributes.
[0031] In a specific embodiment, the factor matching module specifically includes: extracting the industry field scope and time period from the keyword data of the target questionnaire; the characteristic factors of the practitioners include the industry activity trajectory information of the practitioners; determining the correlation degree based on the matching results and the characteristic factors of the practitioners, judging whether the industry activity trajectory of the practitioners matches the industry field scope and time period; if so, determining the first correlation parameter as the first preset value; if not, determining it as the third preset value; calculating the correlation degree based on the first correlation parameter and the matching results.
[0032] In a specific embodiment, the questionnaire feature vector B can be generated based on the keyword data of the determined target questionnaire. i The key words of the practitioners’ participation history questionnaire can be integrated to generate the practitioners’ feature vector A i Then, the following formula can be used to calculate the matching results between the keyword elements of the practitioners’ historical questionnaires and the keyword data of the target questionnaires:
[0033]
[0034] Wherein, cosθ represents the matching result, n represents the number of types of keyword elements or keyword data, i represents the identifier of keyword elements or keyword data, A irepresents the i-th keyword element of the practitioners’ participation history questionnaire, which is also the above-mentioned practitioner feature vector, B i represents the i-th keyword data of the target questionnaire, that is, the above questionnaire feature vector, μ i represents the correlation factor corresponding to the i-th keyword element or keyword data, μ i This is a set value, and the specific value is not limited.
[0035] In a specific embodiment, the first association parameter of the practitioner is determined based on the practitioner's industry activity trajectory information; the second association parameter of the practitioner is determined based on the practitioner's participation in historical questionnaires; and the degree of association between the practitioner and the target questionnaire is determined based on the first association parameter, the second association parameter and the matching result.
[0036] In one embodiment, the degree of association between the person to be associated and the event to be associated can be determined based on the following three aspects: 1. The practitioner's participation in historical questionnaires; 2. The matching of the practitioner's industry activity trajectory information with the industry field scope and time period in the keyword data of the target questionnaire; 3. The matching results of the practitioner's participation in historical questionnaires and the keyword data of the target questionnaire. The following formula can be used to calculate the degree of association between the practitioner and the target questionnaire:
[0037] Score=k1*α+k2*β+k3*cosθ;
[0038] Among them, Score represents the correlation degree, k1, k2, and k3 represent the weight coefficients of the above three aspects respectively; α represents the second correlation parameter related to "practitioners' participation in historical questionnaires"; β represents the first correlation parameter related to "the matching of practitioners' industry activity trajectory information with the industry field scope and time period in the keyword data of the target questionnaire", and cosθ represents "the matching results of practitioners' participation in historical questionnaires and keyword data of the target questionnaire"; k1, k2, and k3 are set values, and the specific values are not limited.
[0039] In other implementations, the correlation between the practitioner and the target questionnaire can be determined based on the first correlation parameter and the matching result. In this implementation, the correlation between the practitioner and the target questionnaire can be calculated using the following formula:
[0040] Score=k4*β+k5*cosθ;
[0041] Among them, Score represents the correlation degree, k4 represents the weight coefficient of the first correlation parameter, and k5 represents the weight coefficient of the matching result; k4 and k5 are set values, and the specific values are not limited.
[0042] Alternatively, in other implementations, the degree of association between the practitioner and the target questionnaire may be determined based on the second association parameter and the matching result. In this implementation, the degree of association between the practitioner and the target questionnaire may be calculated using the following formula:
[0043] Score=k6*α+k7*cosθ;
[0044] Wherein, Score represents the correlation degree, k6 represents the weight coefficient of the second correlation parameter, and k7 represents the weight coefficient of the matching result.
[0045] In a specific embodiment, obtaining the industry activity trajectory information of the practitioner specifically includes: obtaining the practitioner's questionnaire submission record, and generating the industry activity trajectory information based on the submission location and timestamp;
[0046] and / or, determining industry activity trajectory information based on the practitioner's industry registration information or project participation records.
[0047] In a specific embodiment, determining whether the industry activity trajectory of the practitioner matches the scope and time period of the industry field to which he belongs, and if so, determining the first correlation parameter to be a first preset value specifically includes:
[0048] If the industry activity trajectory matches the industry field and time period to which the target questionnaire belongs, the first correlation parameter is determined to be the first preset value;
[0049] If the industry activity trajectory does not match the industry field and time period of the target questionnaire, but matches the extended scope, the first correlation parameter is determined to be the second preset value; wherein the extended scope includes the core scope;
[0050] If none of them match, the third preset value is determined, wherein the first preset value>the second preset value>the third preset value.
[0051] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. Reference can be made to the common and similar parts between the various embodiments. For the devices disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple, and the relevant parts can be referred to the method description.
[0052] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present invention. Various modifications to these embodiments will be readily 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 is not limited to the embodiments shown herein but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.
Claims
1. A questionnaire survey analysis and processing system based on different industries, characterized by: include: Questionnaire keyword extraction module, practitioner feature database, central control module, element matching module, correlation calculation module, the questionnaire keyword extraction module extracts keyword data of the target questionnaire; the practitioner feature database pre-stores practitioner feature elements and keyword elements involved in historical questionnaires by the practitioners; the keyword data and keyword elements are sent to the element matching module through the central control module, and the practitioner feature elements are sent to the correlation calculation module; the keyword data and keyword elements are matched by the element matching module to obtain matching results; the matching results are sent to the correlation calculation module through the central control module, and the correlation calculation module calculates the correlation between the practitioners and the target questionnaire keyword data according to the matching results and the practitioner feature elements, and screens the target practitioners based on the correlation.
2. The questionnaire analysis and processing system based on different industries according to claim 1 is characterized in that: The questionnaire keyword extraction module internally sets a keyword extraction model, including an attention module, an expansion normalization module, a coupling strategy module, and a keyword extraction module. The keyword extraction model takes the questionnaire title text, question description text, and option text as input; the attention module adopts the ResNet basic network, and sets an attention block in each residual block. Adjacent residual blocks are separated by at least two network layers, and outputs a feature matrix with attention weights; The expansion normalization module obtains normalized features of unified dimension through expansion convolution with different expansion rates; the coupling strategy module performs differential fusion based on field types and outputs fusion features; The keyword extraction module maps the fusion features to the industry keyword library dimension and finally obtains the keyword data through the Softmax function; the keyword extraction model uses classification cross entropy as the loss function, Softmax as the activation function, and Adam optimizer to optimize the parameters during the training process.
3. The questionnaire analysis and processing system based on different industries according to claim 2 is characterized in that: The keyword extraction model adopts an improved YOLO network model, which uses a Mosaic data enhancement algorithm to enhance the input data samples. The improved YOLO network model uses a GSPDARKNET53 network to extract sample data features and generate multi-scale feature maps.
4. The questionnaire analysis and processing system based on different industries according to claim 3 is characterized in that: The improved YOLO network model inserts ECA components after multiple CSP1-X structures. The ECA component uses global average pooling to convert the input feature map into a one-dimensional vector, then performs one-dimensional convolution and SIGMOID activation in the C dimension, and finally re-inputs the obtained one-dimensional vector into the feature map.
5. The questionnaire survey analysis and processing system based on different industries according to claim 1 is characterized in that: The characteristic elements of practitioners in the practitioner characteristic database include the practitioners' participation in historical questionnaire keywords; it also includes obtaining historical questionnaires, which include the industry to which the questionnaire belongs, the questionnaire content, the practitioner identification and the corresponding answers; by statistically analyzing the historical questionnaires, determining the participation of each practitioner in the keyword elements, and the industry attribute elements of each keyword element; for each practitioner, storing their keyword element participation and the industry attribute elements of the keyword elements involved; obtaining the practitioner's characteristic elements and keyword elements includes reading the participation of the specified target practitioner in the keyword elements and the keyword elements of the target practitioner from the stored practitioner participation in the keyword elements and industry attribute elements.
6. The questionnaire analysis and processing system based on different industries according to claim 5 is characterized in that: The reading of the keyword elements of the designated target practitioner includes: the number of times the target practitioner mentioned the keyword in the historical questionnaire; or the number of mentions and the corresponding time period; or the number of mentions, the time period and the keyword context attributes.
7. The questionnaire analysis and processing system based on different industries according to claim 6 is characterized in that: The element matching module specifically includes: extracting the industry field scope and time period from the keyword data of the target questionnaire; the characteristic elements of the practitioners include the industry activity trajectory information of the practitioners; determining the correlation degree based on the matching results and the characteristic elements of the practitioners, judging whether the industry activity trajectory of the practitioners matches the industry field scope and time period; if so, determining the first correlation parameter as the first preset value; if not, determining it as the third preset value; calculating the correlation degree based on the first correlation parameter and the matching results.
8. The questionnaire analysis and processing system based on different industries according to claim 7 is characterized in that: Acquiring the industry activity track information of the practitioner specifically includes: acquiring the questionnaire submission record of the practitioner, and generating the industry activity track information based on the submission location and timestamp; And / or, determine the industry activity trajectory information based on the practitioner's industry registration information or project participation records.
9. The questionnaire analysis and processing system based on different industries according to claim 7 is characterized in that: The determining whether the industry activity trajectory of the practitioner matches the scope and time period of the industry field to which the practitioner belongs, and if so, determining the first correlation parameter to be the first preset value specifically includes: If the industry activity trajectory matches the industry field and time period to which the target questionnaire belongs, the first correlation parameter is determined to be the first preset value; If the industry activity trajectory does not match the industry field and time period of the target questionnaire, but matches the extended scope, determining the first correlation parameter to be the second preset value; wherein the extended scope includes the core scope; If none of them match, the third preset value is determined, wherein the first preset value>the second preset value>the third preset value.