Research data generation method and device, electronic equipment and storage medium

Abstract

The embodiment of the invention relates to an investigation data generation method and device, electronic equipment and a storage medium. The method comprises the steps of obtaining input configuration data; according to the feature data of the investigation objects, generating portrait data corresponding to virtual investigation objects matched with the investigation objects in number; generating virtual investigation objects matched with the investigation objects in quantity through an investigation object generation model according to the portrait data corresponding to each virtual investigation object; inputting the to-be-investigated content to each generated virtual investigation object, analyzing the to-be-investigated content through each virtual investigation object, and outputting investigation reply content corresponding to the to-be-investigated content; and performing statistical analysis on the investigation reply content output by each virtual investigation object to obtain a target investigation result. According to the survey data generation method and device, the electronic equipment and the storage medium, a large amount of survey reply content can be obtained through the virtual survey object, so that the time cost and the economic cost of social survey are reduced.

Description

Technical Field

[0001] This application relates to the technical field of data processing, and particularly to a method, apparatus, electronic device, and storage medium for generating survey data. Background Art

[0002] In order to obtain accurate analysis results of social phenomena, a large amount of survey data often needs to be collected during a survey. The quantity of survey data needs to be large enough to fully reflect the characteristics and trends of social phenomena and ensure the accuracy and reliability of the analysis results. However, in the case of real people participating in a survey, to obtain a sufficient quantity of survey data, it is necessary to invest a large amount of time waiting for the participation of real people or pay a substantial fee to encourage real people to provide survey data. Therefore, when conducting a comprehensive and in-depth survey by using the method of real people participating in the survey, to obtain a sufficient quantity of survey data with statistical significance, it may require a large amount of time and economic costs. Summary of the Invention

[0003] Embodiments of this application disclose a method, apparatus, electronic device, and storage medium for generating survey data, which can generate virtual survey objects with a sufficient number of survey objects to obtain a large amount of survey response content, thereby reducing the time cost and economic cost of social surveys.

[0004] Embodiments of this application disclose a method for generating survey data, including:

[0005] Obtain input configuration data, where the configuration data includes characteristic data of survey objects and the number of survey objects;

[0006] According to the characteristic data of the survey objects, generate portrait data corresponding to the virtual survey objects respectively that match the number of survey objects; the portrait data is used to describe the virtual survey objects;

[0007] Generate virtual survey objects that match the number of survey objects through a survey object generation model according to the portrait data corresponding to each of the virtual survey objects;

[0008] Input the content to be surveyed into each of the generated virtual survey objects, and analyze the content to be surveyed through each of the virtual survey objects, and output the survey response content corresponding to the content to be surveyed;

[0009] Conduct statistical analysis on the survey response content output by each of the virtual survey objects to obtain the target survey result.

[0010] As an alternative implementation, the feature data includes a plurality of user features and parameter ranges corresponding to each of the user features; generating portrait data respectively corresponding to virtual research objects that match the number of the research objects according to the feature data of the research object includes: based on a truncated normal distribution, generating a plurality of feature samples corresponding to each of the user features according to the parameter ranges corresponding to each of the user features; wherein, the number of feature samples corresponding to the same user feature is the same as the number of the research objects; combining the plurality of feature samples corresponding to each of the user features to obtain portrait data respectively corresponding to virtual research objects that match the number of the research objects, and the portrait data includes one feature sample corresponding to each of the plurality of user features.

[0011] As an alternative implementation, the generating a plurality of feature samples corresponding to each of the user features according to the parameter ranges corresponding to each of the user features based on a truncated normal distribution includes: calculating a target mean value and a target standard deviation corresponding to a first user feature according to the parameter range corresponding to the first user feature; the first user feature being any one of the user features; generating a plurality of first feature samples corresponding to the first user feature based on a truncated normal distribution according to the parameter range corresponding to the first user feature; calculating a first mean value and a first standard deviation according to the current plurality of first feature samples corresponding to the first user feature; calculating a deviation value according to a first difference between the first mean value and the target mean value and a second difference between the first standard deviation and the target standard deviation; if the deviation value is greater than or equal to a deviation threshold, adjusting N first feature samples according to an optimization function to obtain a new plurality of first feature samples corresponding to the first user feature, incrementing N by one, and re-executing the step of calculating a first mean value and a first standard deviation according to the current plurality of first feature samples corresponding to the first user feature until N is equal to an iteration threshold; N being a positive integer; if the deviation value is less than the deviation threshold, using the current plurality of first feature samples as the plurality of feature samples corresponding to the first user feature.

[0012] As an alternative implementation, after the model generated from the research objects generates virtual research objects that match the number of the research objects according to the portrait data corresponding to each of the virtual research objects, the method further includes: obtaining behavioral sample research content, where the behavioral sample research content includes a sample research scenario and behavioral research questions, and the behavioral research questions are open-ended questions; obtaining, according to the portrait data corresponding to the first virtual research object, the sample behavioral content that should be output after the first virtual research object analyzes the behavioral sample research content; the first virtual research object is any one of the virtual research objects; inputting the sample research scenario and the behavioral research questions into the first virtual research object, and having the first virtual research object analyze the sample research scenario and the behavioral research questions to output the research behavioral content corresponding to the behavioral sample research content; calculating the semantic similarity between the research behavioral content and the sample behavioral content; if the semantic similarity is less than the similarity threshold, then generating a new first virtual research object by the model generated from the research objects according to the portrait data corresponding to the first virtual research object.

[0013] As an alternative implementation, after the model generated from the research objects generates virtual research objects that match the number of the research objects according to the portrait data corresponding to each of the virtual research objects, the method further includes: obtaining decision-making sample research content, where the decision-making sample research content includes a sample research scenario and decision-making research questions, and the decision-making research questions are closed-ended questions; obtaining, according to the portrait data corresponding to the first virtual research object, the sample options that should be output after the first virtual research object analyzes the decision-making sample research content; the first virtual research object is any one of the virtual research objects; inputting the sample research scenario and the decision-making research questions into the first virtual research object, and having the first virtual research object analyze the sample research scenario and the decision-making research questions to output the research options corresponding to the decision-making sample research content; if the research options are not the same as the sample options, then generating a new first virtual research object by the model generated from the research objects according to the portrait data corresponding to the first virtual research object.

[0014] As an alternative implementation, the content to be investigated includes the questions to be investigated and multiple investigation conditions; the method further includes: combining the multiple investigation conditions to obtain multiple different sets of investigation conditions; each set of investigation conditions includes at least one investigation condition; according to the multiple different sets of investigation conditions, grouping virtual investigation objects matching the number of investigation objects to obtain virtual investigation object groups corresponding to each set of investigation conditions; inputting the content to be investigated into each of the generated virtual investigation objects, and analyzing the content to be investigated by each of the virtual investigation objects to output the investigation response content corresponding to the content to be investigated, including: inputting the first set of investigation conditions and the questions to be investigated into each of the second virtual investigation objects included in the virtual investigation object group corresponding to the first set of investigation conditions, and analyzing the first set of investigation conditions and the questions to be investigated by each of the second virtual investigation objects to output the investigation response content corresponding to the questions to be investigated; the first set of investigation conditions is any set of investigation conditions.

[0015] As an alternative implementation, the portrait data includes a feature sample corresponding to each of the multiple user features; the grouping of the virtual investigation objects matching the number of investigation objects according to the multiple different sets of investigation conditions to obtain virtual investigation object groups corresponding to each set of investigation conditions includes: dividing the parameter range corresponding to the target user feature according to the multiple different sets of investigation conditions to obtain parameter sub-ranges corresponding to each set of investigation conditions; the target user feature is one of the multiple user features; determining the parameter sub-range to which each virtual investigation object belongs according to the parameter sub-ranges corresponding to each set of investigation conditions and the feature samples corresponding to each virtual investigation object and the target user feature; determining the set of investigation conditions corresponding to each virtual investigation object according to the parameter sub-range to which each virtual investigation object belongs to obtain virtual investigation object groups corresponding to each set of investigation conditions.

[0016] An embodiment of the present application discloses a device for generating investigation data, including:

[0017] An acquisition module, configured to acquire input configuration data, where the configuration data includes feature data of an investigation object and the number of investigation objects;

[0018] A portrait module, configured to generate portrait data corresponding to virtual investigation objects matching the number of investigation objects according to the feature data of the investigation object; the portrait data is used to describe the virtual investigation objects;

[0019] A construction module for generating, according to the portrait data corresponding to each virtual research object, virtual research objects that match the number of the research objects through a research object generation model;

[0020] A research module for inputting the content to be researched into each of the generated virtual research objects, analyzing the content to be researched through each of the virtual research objects, and outputting the research reply content corresponding to the content to be researched;

[0021] A statistical module for statistically analyzing the research reply content output by each of the virtual research objects to obtain a target research result.

[0022] An embodiment of the present application discloses an electronic device, including a memory and a processor. A computer program is stored in the memory. When the computer program is executed by the processor, the processor implements the method described in any one of the above embodiments.

[0023] An embodiment of the present application discloses a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0024] An embodiment of the present application discloses a computer program product, including a computer program. When the computer program is executed by a processor, the method described in any one of the above embodiments is implemented.

[0025] The method, apparatus, electronic device, and storage medium for generating survey data disclosed in the embodiments of the present application obtain the input configuration data, where the configuration data includes the characteristic data of the survey objects and the number of survey objects; generate portrait data corresponding to the virtual survey objects respectively that match the number of survey objects according to the characteristic data of the survey objects; the portrait data is used to describe the virtual survey objects; generate virtual survey objects that match the number of survey objects through the survey object generation model according to the portrait data corresponding to each virtual survey object; input the content to be surveyed into each generated virtual survey object, and analyze the content to be surveyed through each virtual survey object, and output the survey response content corresponding to the content to be surveyed; perform statistical analysis on the survey response content output by each virtual survey object to obtain the target survey result. In the embodiments of the present application, through the configuration data, portrait data of virtual survey objects that match the required number is generated, and the portrait data can accurately describe each virtual survey object, so that the virtual survey objects generated through the portrait data can be more real. When the content to be surveyed is input into these virtual survey objects, the analysis and reaction processes of real people can be simulated, and accurate survey response content can be output to improve the accuracy rate; moreover, each virtual survey object can analyze the content to be surveyed in parallel, that is, there is no need to wait for the participation and feedback time of real people, thereby reducing the time cost of obtaining survey data and improving the efficiency of the survey; in addition, by using virtual survey objects to generate survey response content, there is no need to recruit, motivate, and manage the participation of real people, reducing the economic cost consumed by the entire survey, thereby realizing the efficient and low-cost generation of survey data. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following will briefly introduce the drawings required in the embodiments. Obviously, the drawings in the following description are only some embodiments of the present application. For those of ordinary skill in the art, without creative efforts, other drawings can be obtained based on these drawings.

[0027] Figure 1 It is a flowchart of the method for generating survey data in one embodiment;

[0028] Figure 2 It is a flowchart of generating portrait data in one embodiment;

[0029] Figure 3 It is a flowchart of the method for generating survey data in one embodiment;

[0030] Figure 4 It is a flowchart of the method for generating survey data in another embodiment;

[0031] Figure 5Flowchart of a method for generating survey data in another embodiment;

[0032] Figure 6 Block diagram of a device for generating survey data in one embodiment;

[0033] Figure 7 Block diagram of the structure of an electronic device in one embodiment. Detailed implementation manners

[0034] Next, the technical solutions in the embodiments of the present application will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the present application. Obviously, the described embodiments are only a part of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present application.

[0035] It should be noted that the terms "including" and "having" in the embodiments of the present application and any variations thereof are intended to cover non-exclusive inclusions. For example, a process, method, system, product, or device that includes a series of steps or units is not limited to the listed steps or units, but optionally further includes steps or units not listed, or optionally further includes other steps or units inherent to these processes, methods, products, or devices.

[0036] It can be understood that the terms "first", "second", etc. used in the present application can be used in this article to describe various elements, but these elements are not limited by these terms. These terms are only used to distinguish the first element from another element. For example, without departing from the scope of the present application, the first virtual survey object can be called the second virtual survey object, and similarly, the second virtual survey object can be called the first virtual survey object. The first virtual survey object and the second virtual survey object can be the same virtual survey object or different virtual survey objects.

[0037] In relevant survey scenarios, real people participate in the survey. If a sufficient sample size is to be obtained, a large amount of time needs to be invested. Because each participant may need to go through processes such as being recruited, introduced to the survey purpose, completing a questionnaire or participating in an interview, etc., and these processes will accumulate into a relatively long time cost.

[0038] Moreover, in order to motivate real people to actively participate in filling out questionnaires or participating in other forms of survey activities, certain economic rewards or other forms of incentives need to be provided. This will also increase the economic cost of the survey. Especially on some online survey platforms, such as website surveys, it often costs a lot to collect a certain number of questionnaires.

[0039] In addition, some of the research content may involve relatively sensitive topics of participants in the real population or topics that they are unwilling to participate in, resulting in fewer participants who can provide research data and insufficient research data. In this case, to increase the research data, more time costs and / or economic costs are required.

[0040] The embodiments of the present application disclose a method, device, electronic device, and storage medium for generating research data, which can generate virtual research objects with a sufficient number of research objects to obtain a large amount of research response content, thereby reducing the time cost and economic cost of social research.

[0041] As Figure 1 shown, in one embodiment, a method for generating research data is provided, which can be applied to an electronic device. The electronic device may include, but is not limited to, a mobile phone, a smart wearable device, a tablet computer, a PC (Personal Computer), a computer, etc. The method may include the following steps 110 to 150.

[0042] Step 110, obtain the input configuration data.

[0043] When conducting market research or similar data collection activities, users need to obtain statistically significant research data. However, since the number of research objects is usually large, it is difficult for users to describe each virtual research object individually. Therefore, by inputting configuration data into the electronic device, the electronic device can portrait each virtual research object through the configuration data to generate virtual research objects that match the number of research objects.

[0044] In some embodiments, the configuration data refers to the relevant data used to configure and describe the research objects; the configuration data includes the characteristic data of the research objects and the number of research objects. The number of research objects refers to the total number of virtual objects planned for research (such as interviewees, user groups, etc.).

[0045] The characteristic data of the research objects refers to the information used to describe the attributes of each research object. The characteristic data includes multiple user characteristics and the parameter ranges corresponding to each user characteristic. User characteristics refer to the attributes that can reflect the characteristics of different research objects, such as age, gender, occupation, income level, hobbies, educational background, etc.

[0046] It should be noted that the parameter range corresponding to the user characteristic not only includes the numerical range, but also can include various specific contents of the user characteristic and the probability distribution corresponding to each specific content; for example, when the user characteristic is gender, the corresponding parameter range can include the specific contents male or female, and the quantity ratio of male to female.

[0047] Exemplarily, the characteristic data of the research objects may include: the parameter range corresponding to the age characteristic may include different single or multiple age groups such as 18 - 24 years old and / or 25 - 34 years old; the parameter range corresponding to the gender characteristic may be clearly defined as male or female, or the quantity ratio of male to female; the parameter range corresponding to the occupation characteristic may be specific occupation types such as teachers, doctors, engineers, etc., or broader occupation classifications such as white - collar workers, blue - collar workers, and so on.

[0048] In some embodiments, the electronic device may have a display screen for displaying a research configuration interface. The user can select different user characteristics and set parameter ranges for each user characteristic in this research configuration interface; the electronic device obtains the configuration data of this research by identifying the key information and text information of the research configuration interface. For example, the user can select "age" as a user characteristic in the research configuration interface and set the parameter range corresponding to this user characteristic as "18 - 35 years old".

[0049] In some embodiments, when the configuration data is provided in the form of text input by the user, the electronic device can obtain the configuration data by performing text recognition and analysis on a piece of input text. Specifically, the electronic device can use text parsing technologies (such as regular expressions, natural language processing, etc.) to identify the user characteristics and parameter ranges in the text. For example, by identifying delimiters such as semicolons (;), commas (,), and colons (:) in the text, multiple user characteristics and corresponding parameter ranges are determined.

[0050] Specifically, the electronic device can distinguish the user characteristics and the corresponding parameter ranges by colons, different user characteristics by commas, and the characteristic data of the research objects and the number of research objects by semicolons. For example, the text input into the electronic device can be "Age: 18 - 35, Gender ratio: 50%, Income level: 5000 - 10000, Educational background: Undergraduate students and undergraduate graduates; Number of research objects: 300", and the electronic device can identify and parse the corresponding configuration data.

[0051] Step 120, generate portrait data corresponding to each of the virtual research objects that match the number of research objects according to the characteristic data of the research objects.

[0052] The portrait data is used to describe the virtual research objects. Each virtual research object has a corresponding portrait data. The portrait data includes a characteristic sample corresponding to each of the multiple user characteristics. The user characteristics refer to the user characteristics in the characteristic data of the research objects; the characteristic sample corresponding to the user characteristic refers to the specific value or specific instance of the user characteristic included in the virtual research object.

[0053] It should be noted that the number of user characteristics included in the portrait data is the same as the number of user characteristics included in the characteristic data of the research object. This enables the portrait data to fully describe the corresponding virtual research object, and the characteristic samples corresponding to multiple user characteristics in the portrait data of each virtual research object together constitute the complete image of each virtual research object.

[0054] In some embodiments, the electronic device can create an empty portrait data structure for each virtual research object, and all user characteristics are included in the empty portrait data structure; then select a value from the parameter range corresponding to each user characteristic as the characteristic sample corresponding to each user characteristic in the portrait data of the virtual research object.

[0055] Optionally, for different user characteristics, different methods are used to select the characteristic samples. For continuous user characteristics (such as age, income, etc.), the electronic device can randomly select a value within the parameter range corresponding to the user characteristic. For example, if the parameter range corresponding to the user characteristic of age is 20 - 60 years old, then a random age value within this range can be selected, such as 46 years old.

[0056] For discrete user characteristics (such as gender, occupation, etc.), the electronic device can randomly select a specific content as the characteristic sample corresponding to the user characteristic according to the specific content included in the parameter range corresponding to the user characteristic and the corresponding probability distribution. For example, if the user characteristic is gender and there are two specific contents (male, female), and their corresponding probability distribution is 50%, then the electronic device can randomly select a gender, and among the virtual research objects matching the number of research objects, the number of males and females is equal. Or, if the user characteristic is occupation and the corresponding parameter range is a probability distribution table, which includes various specific occupation types and the occurrence probabilities corresponding to each occupation type, then the electronic device can randomly select an occupation according to this probability distribution table, and among the virtual research objects matching the number of research objects, the occurrence probabilities of various occupation types conform to this probability distribution table.

[0057] In some embodiments, after the electronic device generates the portrait data corresponding to the virtual research objects respectively matching the number of research objects, it can also perform a rationality verification on the portrait data corresponding to each virtual research object. Specifically, the electronic device can determine whether the relationship between the characteristic samples corresponding to multiple user characteristics of each virtual research object is reasonable. If it is detected that the characteristic sample of a certain virtual research object is unreasonable, then the portrait data corresponding to the virtual research object is regenerated. For example, it is unreasonable for a 16-year-old virtual research object to have a driver's license.

[0058] Based on the characteristic data of the research objects, a unique and compliant portrait data is generated for each virtual research object, making the generated virtual research objects that match the number of research objects more realistic. When the content to be researched is input into these virtual research objects, it can simulate the analysis and reaction processes of the real population.

[0059] Step 130: According to the portrait data corresponding to each virtual research object, the research object generation model generates virtual research objects that match the number of research objects.

[0060] In some embodiments, after generating the portrait data of each virtual research object, the electronic device can use the research object generation model to sequentially convert multiple portrait data into corresponding virtual research objects.

[0061] The research object generation model is a complex, generative artificial intelligence model used to generate virtual research objects with corresponding attributes and behaviors by analyzing feature samples (such as age, gender, occupation, hobbies, etc.) in the portrait data, using machine learning algorithms and deep learning techniques.

[0062] It should be noted that each portrait data corresponds to a virtual research object. Assuming the number of research objects is K, the electronic device needs to use the research object generation model K times. Each time, a virtual research object is generated according to one portrait data; generating virtual research objects that match the number of research objects is equivalent to an artificial intelligence model that generates K virtual research objects.

[0063] In the embodiments of the present application, the generated virtual research objects not only have feature samples similar to those of the real population but also can simulate the analysis and reaction processes of the real population, enabling the virtual research objects to think and judge based on the content to be researched, and thus outputting research reply content that conforms to logic and the actual situation.

[0064] Step 140: Input the content to be researched into each of the generated virtual research objects, and each virtual research object analyzes the content to be researched and outputs the research reply content corresponding to the content to be researched.

[0065] The content to be researched refers to the specific information or topic that needs to be researched and analyzed. The content to be researched includes the research questions and multiple research conditions.

[0066] The research questions refer to the questions that specifically need to be asked to the virtual research objects, which can cover multiple aspects, such as consumer preferences, product satisfaction, market trends, etc.

[0067] Optionally, the questions to be investigated can be open-ended questions that allow the respondents to freely express their opinions, or closed-ended questions that provide limited options for the respondents to choose from. The type of questions to be investigated depends on the purpose of the investigation and the data to be collected.

[0068] Specifically, the types of questions to be investigated can include open-ended questions and closed-ended questions. For example, open-ended questions can be behavioral research questions used to obtain information about specific behaviors, and closed-ended questions can be decision-making research questions used to obtain the option results of specific decisions.

[0069] Investigation conditions refer to a series of restrictions, constraints, or preconditions to be considered during the investigation, which can include but are not limited to specific time ranges (such as a certain year or quarter), geographical restrictions (such as a certain country or region), target groups (such as specific age groups or professional groups), scenarios (such as online shopping or offline experience), known information (such as known market data or consumer behavior patterns), etc.

[0070] For example, the content to be investigated can include investigation conditions: reading the advertising slogans of brand A and brand B, and the question to be investigated is: whether there is an idea to purchase brand A or brand B after reading.

[0071] Investigation response content refers to the responses or data given by virtual respondents after analyzing the content to be investigated, which can include but are not limited to text descriptions (such as specific evaluations of products by consumers), numerical scores (such as scores for product satisfaction), selection items (such as answers selected in closed-ended questions), etc. The specific form of the investigation response content can depend on the type of question to be investigated and the purpose of this investigation.

[0072] Exemplarily, when the type of question to be investigated is an open-ended question, the specific form of the investigation response content can be investigation behavior content. When the type of question to be investigated is a closed-ended question, the specific form of the investigation response content can be investigation options.

[0073] In some embodiments, after the electronic device inputs the content to be investigated into each generated virtual respondent, each virtual respondent analyzes and judges according to its own attributes and behavior patterns, and reasons and makes decisions according to the key information in the content to be investigated (such as the question to be investigated and multiple investigation conditions, etc.), and gives the investigation response content corresponding to the content to be investigated. By using virtual respondents to simulate the thinking and reaction process of real people when facing investigation questions, valuable and practical investigation data can be obtained through virtual respondents.

[0074] Step 150: Statistically analyze the survey response content output by each virtual survey object to obtain the target survey result.

[0075] In some embodiments, the electronic device can extract useful information or potential patterns from a large amount of response data through statistical analysis, providing solid data support for subsequent decision-making, market strategy planning, product improvement, and other aspects.

[0076] The target survey result refers to the specific and quantifiable conclusions or trends expected to be obtained through statistical analysis, which may include, but are not limited to, the preference analysis results of the issues to be surveyed, such as brand loyalty, product satisfaction, functional requirements, etc.

[0077] In some embodiments, the statistical analysis methods that the electronic device can use may include, but are not limited to, descriptive statistics (used to summarize and describe the basic characteristics of the survey response content output by each virtual survey object, such as mean, median, mode, variance, etc.), inferential statistics (inferring the overall survey response content based on the survey response content output by individual data virtual survey objects, such as hypothesis testing, confidence interval estimation, etc.), correlation analysis (analyzing the linear or non-linear relationship between two or more survey conditions, such as Pearson correlation coefficient, Spearman rank correlation coefficient, etc.), cluster analysis (dividing virtual survey objects with similar portrait data into the same group to better analyze the common characteristics of their survey response content), etc.

[0078] In some embodiments, the electronic device can perform statistical analysis based on the survey response content output by each virtual survey object and the corresponding portrait data to obtain the target survey result, so that the target survey result includes the statistical analysis results of the portrait data corresponding to each virtual survey object. Specifically, the target survey result may further include the overall characteristics description of the survey objects (such as age distribution, gender ratio, geographical distribution, etc.), the behavior patterns of the survey objects (such as purchase habits, consumption preferences, usage frequency, etc.), etc., combining the survey response content of the virtual survey objects with the portrait data to obtain more accurate and effective survey data.

[0079] Exemplarily, in a satisfaction survey for a certain new product, the electronic device can obtain the survey response content and the corresponding portrait data of a large number of virtual survey objects; through statistical analysis, the target survey result obtained by the electronic device may include: young people aged between 25 and 35 have the highest satisfaction with the product and generally believe that the product has a fashionable appearance and practical functions; in terms of gender, female survey objects are more picky about the color selection of the product and tend to choose softer and warmer colors; in terms of geographical distribution, the satisfaction of survey objects in first-tier cities is generally higher than that in second- and third-tier cities, which may be related to the market acceptance and consumption ability in first-tier cities.

[0080] In the embodiments of the present application, the electronic device generates portrait data of virtual research objects that matches the required quantity through configuration data. The portrait data can accurately describe each virtual research object, making the virtual research objects generated through the portrait data more realistic. When inputting the content to be researched into these virtual research objects, it can simulate the analysis and reaction processes of real people, output accurate research response content, so as to improve the accuracy rate. Moreover, each virtual research object can analyze the content to be researched in parallel, that is to say, there is no need to wait for the participation and feedback time of real people, thereby reducing the time cost of obtaining research data and improving the efficiency of the research. In addition, by generating research response content through virtual research objects, there is no need to recruit, motivate, and manage the participation of real people, reducing the economic cost consumed by the entire research, thus realizing the efficient and low-cost generation of research data.

[0081] As Figure 2 shown, in some embodiments, the step of generating portrait data corresponding to virtual research objects that matches the number of research objects according to the characteristic data of the research objects may include the following steps 202 to step 204.

[0082] Step 202, based on the truncated normal distribution, generate multiple characteristic samples corresponding to each user characteristic according to the parameter range corresponding to each user characteristic. Among them, the number of characteristic samples corresponding to the same user characteristic is the same as the number of research objects.

[0083] In some embodiments, the electronic device may determine the mean and standard deviation of the first user characteristic according to the first parameter range corresponding to the first user characteristic, where the first user characteristic is any user characteristic; use a random number generator for the truncated normal distribution to generate multiple random numbers according to the first parameter range, and these random numbers will be used as the values of the characteristic samples; and ensure that the number of generated characteristic samples matches the number of research objects. The random number generator for the truncated normal distribution is a tool specifically used to generate random numbers for the truncated normal distribution.

[0084] Exemplarily, for the user characteristic of "age", the corresponding parameter range is 20 to 40 years old, its mean is 30, and the standard deviation is 5; then the electronic device may use a random number generator for the truncated normal distribution to generate multiple random numbers within the range of [20, 40] as characteristic samples.

[0085] In order to make the multiple characteristic samples corresponding to each user characteristic generated closer to the actual distribution, thereby improving the authenticity of the virtual research objects, therefore, the electronic device may perform deviation tests on the multiple characteristic samples corresponding to each user characteristic generated.

[0086] In some embodiments, the electronic device may calculate the target average value and the target standard deviation corresponding to the first user feature according to the parameter range corresponding to the first user feature; the first user feature is any user feature; based on the truncated normal distribution, according to the parameter range corresponding to the first user feature, generate a plurality of first feature samples corresponding to the first user feature; calculate the first average value and the first standard deviation according to the current plurality of first feature samples corresponding to the first user feature; calculate the deviation value according to the first difference between the first average value and the target average value and the second difference between the first standard deviation and the target standard deviation; if the deviation value is greater than or equal to the deviation threshold, adjust the N first feature samples according to the optimization function to obtain a new plurality of first feature samples corresponding to the first user feature, and increment N by one, and re-execute the step of calculating the first average value and the first standard deviation according to the current plurality of first feature samples corresponding to the first user feature until N is equal to the iteration threshold; N is a positive integer; if the deviation value is less than the deviation threshold, use the current plurality of first feature samples as the plurality of feature samples corresponding to the first user feature.

[0087] Optionally, the electronic device may add the square of the first difference to the square of the second difference to obtain the deviation value. If the deviation value is greater than or equal to the preset deviation threshold, it indicates that there is a large difference between the currently generated plurality of first feature samples and the parameter range corresponding to the first user feature, and adjustment is required. In order to reduce the loss and complexity of the adjustment, the electronic device may first adjust the N first feature samples according to the optimization function. If the newly generated plurality of first feature samples after adjustment still need to be adjusted, the electronic device then adjusts (N + 1) first adjustment samples, and so on, until the deviation value is less than the deviation threshold, that is, the currently generated plurality of first feature samples conform to the parameter range corresponding to the first user feature and no adjustment is required, or until N is equal to the iteration threshold.

[0088] In some embodiments, if when N is equal to the iteration threshold, the deviation value calculated from the plurality of first feature samples obtained by the electronic device is still greater than or equal to the preset deviation threshold, the electronic device may use the current plurality of first feature samples as the plurality of feature samples corresponding to the first user feature, or may also re-execute the step of generating a plurality of first feature samples corresponding to the first user feature based on the truncated normal distribution according to the parameter range corresponding to the first user feature.

[0089] In the embodiments of the present application, when generating a plurality of feature samples corresponding to each user feature according to the parameter range corresponding to each user feature, using the truncated normal distribution can ensure that the plurality of feature samples are within the specified parameter range and maintain the characteristics of the normal distribution, which helps to simulate the probability distribution of each user feature in the real world and avoid generating extreme values outside the parameter range.

[0090] Step 204: Combine multiple feature samples corresponding to each user feature to obtain portrait data corresponding to virtual research objects that match the number of research objects.

[0091] It should be noted that the portrait data of each virtual research object should include multiple user features, and each user feature corresponds to a feature sample selected from the corresponding multiple feature samples. For example, if the user feature is age and the multiple feature samples are 18, 19, 20... then in the portrait data corresponding to a certain virtual research object, when the user feature is age, the corresponding feature sample can be 20.

[0092] In some embodiments, the combination methods that can be used by the electronic device include, but are not limited to, the random sampling method, the stratified sampling method, etc.

[0093] Specifically, for each user feature, the electronic device can randomly select a feature sample from the corresponding multiple feature samples, and combine the selected feature samples in the order of user features or according to a certain preset rule to form the portrait data of a virtual research object. Repeat this process until the portrait data corresponding to the virtual research objects that match the required number of research objects are generated.

[0094] Optionally, in the portrait data corresponding to different virtual research objects, the feature samples corresponding to the same user feature can be the same or different.

[0095] Exemplarily, assume that there are three user features: age, gender, and income level, and the number of research objects is 100. Then the electronic device can first generate 100 feature samples according to the parameter ranges corresponding to the three user features respectively; then randomly select one from the 100 feature samples corresponding to age, randomly select one from the 100 feature samples of gender, and randomly select one from the 100 feature samples of income level; combine these three selected samples into the portrait data of a virtual research object; repeat this process 100 times to obtain the portrait data corresponding to 100 virtual research objects respectively.

[0096] In the embodiments of the present application, the electronic device using the truncated normal distribution can ensure that multiple feature samples are within the specified parameter range and maintain the characteristics of the normal distribution, which helps to simulate the probability distribution of each user feature in the real world and avoid generating extreme values outside the parameter range.

[0097] It not only ensures that the feature samples are within the specified parameter range but also maintains the characteristics of the normal distribution, which helps to more realistically simulate the probability distribution of each user feature in the actual world. At the same time, through the adjustment of the optimization function and the combination of feature samples, the portrait data corresponding to the multiple virtual research objects finally obtained has accuracy and authenticity.

[0098] As Figure 3 shown, in one embodiment, a method for generating research data is provided, which can be applied to the above-mentioned electronic device. The method may include the following steps 302 to 310.

[0099] Step 302, obtain the behavioral sample research content.

[0100] The behavioral sample research content refers to the information used to analyze specific behaviors; the behavioral sample research content includes the sample research scenario and the behavioral research questions. The sample research scenario refers to the specific environment and background when conducting the behavioral sample research. The behavioral research questions refer to the questions used to obtain information about specific behaviors.

[0101] Optionally, the behavioral research questions are open-ended questions; open-ended questions refer to questions that cannot be simply answered with "yes" or "no" or other simple words or numbers. Open-ended questions can better display the deep behavioral characteristics such as the thinking mode, needs, motivations, and attitudes of the research objects, so as to confirm whether the generated virtual research objects are more real and accurate.

[0102] In some embodiments, the electronic device can obtain the sample research scenario and the behavioral research questions input by the user; it can also obtain the sample research scenario and the behavioral research questions through the behavioral sample database.

[0103] Step 304, according to the portrait data corresponding to the first virtual research object, obtain the sample behavioral content that should be output after the first virtual research object analyzes the behavioral sample research content. The first virtual research object is any virtual research object.

[0104] The sample behavioral content refers to the specific content of the behavioral characteristics or patterns that a real person with the portrait data corresponding to the first virtual research object may show after analyzing the behavioral sample research content in the sample research scenario.

[0105] In some embodiments, the electronic device can obtain the sample behavioral content corresponding to the behavioral sample research content and the portrait data of the first virtual research object from the behavioral sample database. Specifically, the electronic device can find multiple sample behavioral contents that match the behavioral sample research content in the behavioral sample database, and then find the sample behavioral content corresponding to the portrait data that is most similar to the portrait data of the first virtual research object from the multiple sample behavioral contents, as the sample behavioral content that should be output after the first virtual research object analyzes the behavioral sample research content.

[0106] Step 306: Input the sample research scenario and behavioral research questions into the first virtual research object, and analyze the sample research scenario and behavioral research questions through the first virtual research object to output the research behavioral content corresponding to the behavioral sample research content.

[0107] For the relevant description of step 306, reference can be made to the relevant description of step 140 in the above embodiment, which will not be elaborated here.

[0108] Step 308: Calculate the semantic similarity between the research behavioral content and the sample behavioral content.

[0109] In some embodiments, the electronic device can represent the research behavioral content as a first word embedding vector; represent the sample behavioral content as a second word embedding vector; calculate the distance between the first word embedding vector and the second word embedding vector, and determine the semantic similarity between the research behavioral content and the sample behavioral content according to this distance.

[0110] Optionally, the methods for calculating the distance that the electronic device can use include but are not limited to cosine similarity (the closer the cosine value is to 1, the more similar the semantics between the research behavioral content and the sample behavioral content, and the higher the semantic similarity), Euclidean distance or Manhattan distance (the smaller the distance, the more similar the semantics between the research behavioral content and the sample behavioral content, and the higher the semantic similarity), Pearson correlation coefficient, etc.

[0111] Step 310: If the semantic similarity is less than the similarity threshold, generate a new first virtual research object through the research object generation model according to the portrait data corresponding to the first virtual research object.

[0112] The semantic similarity being less than the similarity threshold indicates that the semantic matching degree between the behavior of the first virtual research object and the expected or standard behavior does not meet the standard, that is, the behavior evaluation is unqualified. Specifically, the semantic similarity being less than the similarity threshold means that the performance of the first virtual research object in a certain behavior or language is quite different from that of real humans or the preset standard, so it is regarded as abnormal behavior. To correct this abnormal behavior, the electronic device can generate a new first virtual research object through the research object generation model according to the portrait data corresponding to the first virtual research object, and re-execute step 306 to conduct a behavior evaluation on the newly generated first virtual research object. This process will be repeated until the semantic similarity corresponding to the new first virtual research object is greater than or equal to the similarity threshold.

[0113] A semantic similarity greater than or equal to the similarity threshold indicates that the behavior of the first virtual research object conforms to the normal behavior standard or expectation, that is, the behavior evaluation is qualified. Specifically, the first virtual research object generated by the electronic device can realistically imitate real humans in terms of behavior performance, and the semantics of its behavior and language are highly consistent with the real situation or preset standard. This enables the first virtual research object to be used in various scenarios that require simulating real human behavior, such as market research, user behavior analysis, product testing, etc., to improve the accuracy and effectiveness of the research data output by the first virtual research object.

[0114] In the embodiments of the present application, the electronic device can accurately evaluate whether the behavior of the generated virtual research object conforms to the expectation or standard behavior by means of the behavior sample research content and calculating the semantic similarity between the research behavior content and the sample behavior content, making the behavior and language of each virtual research object closer to real humans and being able to more realistically reflect the needs and preferences of users; moreover, compared with traditional real-person research, using the generated virtual research object for research can also avoid the biases and subjectivities that may occur in real-person research, making the obtained research data more objective and reliable.

[0115] As Figure 4 shown, in one embodiment, a method for generating research data is provided, which can be applied to the above-mentioned electronic device. The method may include the following steps 402 to step 408.

[0116] Step 402, obtain the decision sample research content.

[0117] The decision sample research content refers to the information used to analyze specific decision results. The decision sample research content includes the sample research scenario and the decision research question. The sample research scenario refers to the specific environment and background when conducting the sample research. The decision research question refers to the question used to obtain the option results of a specific decision.

[0118] In some embodiments, the sample research scenario included in the behavior sample research content in step 302 and the sample research scenario included in the decision sample research content may be the same research scenario, but the behavior research question included in the behavior sample research content in step 302 is different from the decision research question included in the decision sample research content.

[0119] Optionally, the decision research question belongs to a closed-ended question. A closed-ended question refers to a question for which various possible answers have been pre-designed, and the participant can only select one or several ready-made answer options. By setting the decision research question as a closed-ended question, variables can be more effectively controlled, interference factors can be reduced, and thus the accuracy of the decision-making judgment of the virtual research object can be more directly determined.

[0120] Step 404: Obtain the sample options that should be output after analyzing the decision sample research content based on the portrait data corresponding to the first virtual research object. The first virtual research object is any virtual research object.

[0121] The sample options refer to the decision-making judgment options that a real person with the portrait data corresponding to the first virtual research object in the sample research scenario may make after analyzing the decision sample research content.

[0122] For the relevant description of obtaining the sample options in Step 404, reference can be made to the relevant description of obtaining the sample behavior content in Step 304 in the above-mentioned embodiment, which will not be elaborated here.

[0123] Step 406: Input the sample research scenario and the decision research question into the first virtual research object, and have the first virtual research object analyze the sample research scenario and the decision research question to output the research options corresponding to the decision sample research content.

[0124] For the relevant description of Step 404, reference can be made to the relevant description of Step 140 in the above-mentioned embodiment, which will not be elaborated here.

[0125] Step 408: If the research options are different from the sample options, generate a new first virtual research object according to the portrait data corresponding to the first virtual research object through the research object generation model.

[0126] The fact that the research options are different from the sample options indicates that there is a significant difference between the actual research results made by the first virtual research object and the expected sample options, that is, the decision evaluation is unqualified. Specifically, the difference between the research options and the sample options may mean that the first virtual research object has a difference in making decisions from real humans or the preset standards. Therefore, the electronic device can generate a new first virtual research object according to the portrait data corresponding to the first virtual research object through the research object generation model, and re-execute Step 406 to conduct a decision evaluation on the newly generated first virtual research object. This process will be repeated until the research options corresponding to the new first virtual research object are the same as the sample options.

[0127] The fact that the research options are the same as the sample options indicates that the actual research results made by the first virtual research object are consistent with the expected sample options, that is, the decision evaluation is qualified. Specifically, the first virtual research object generated by the electronic device can more realistically imitate real humans in making decisions, so as to improve the accuracy and authenticity of the research data output by the first virtual research object.

[0128] In some embodiments, steps 302 to 310 and steps 402 to 408 can both be used to evaluate virtual research objects so that the virtual research objects are closer to real people, and there is no interference between steps 302 to 310 and steps 402 to 408. Therefore, in the actual process of generating virtual research objects, the electronic device can only execute steps 302 to 310; or only execute steps 402 to 408; or execute steps 302 to 310 and steps 402 to 408 simultaneously, and the execution order is not limited. Thus, in terms of behavior and decision-making, the virtual research objects are closer to real people, so as to obtain more real and accurate research data through the virtual research objects.

[0129] In the embodiments of the present application, the electronic device can accurately evaluate whether the generated virtual research objects conform to the decisions of real people in terms of decision-making through the research content of decision samples and sample options, making the behaviors and languages of each virtual research object closer to real humans, being able to more truly reflect the needs and preferences of users, and making the obtained research data more real and accurate.

[0130] As Figure 5 shown, in one embodiment, a method for generating research data is provided, which can be applied to the above-mentioned electronic device. The method may include the following steps 502 to 506.

[0131] Step 502, combine multiple research conditions to obtain multiple different research condition sets. Each research condition set includes at least one research condition.

[0132] In some embodiments, the content to be surveyed includes survey questions and multiple research conditions.

[0133] By combining multiple research conditions, multiple different research condition sets can be formed, and each set represents a specific research perspective or scenario. This enables the survey to cover a wider range of situations and details, thereby improving the comprehensiveness and accuracy of this survey.

[0134] Exemplarily, assume that the content to be surveyed may include research conditions: reading the advertising slogan of brand A, reading the advertising slogan of brand B, and reading the advertising slogan of brand C. Then, at most 7 different research condition sets can be formed, namely: only reading the advertising slogan of brand A, only reading the advertising slogan of brand B, only reading the advertising slogan of brand C, reading the advertising slogans of brand A and B simultaneously, reading the advertising slogans of brand A and C simultaneously, reading the advertising slogans of brand B and C simultaneously, and reading the advertising slogans of brand A, B, and C simultaneously.

[0135] Step 504: Group the virtual research objects that match the number of research objects according to multiple different sets of research conditions to obtain virtual research object groups corresponding to each set of research conditions.

[0136] In some embodiments, since the number of research conditions is limited, the number of sets of research conditions combined is limited. Usually, the number of sets of research conditions is less than the number of research objects. Therefore, the electronic device can group the virtual research objects that match the number of research objects, so that the sets of research conditions corresponding to different virtual research object groups are different, but different virtual research objects may be the same set of research conditions or different sets of research conditions.

[0137] Optionally, the electronic device can randomly group the virtual research objects according to multiple different sets of research conditions to ensure that the probability of each virtual research object being assigned to each virtual research object group is equal, thereby reducing bias to a certain extent. Specifically, the methods for the electronic device to randomly group can include but are not limited to random number generators, shuffling algorithms, etc.

[0138] In some embodiments, the electronic device can divide the parameter range corresponding to the target user feature according to multiple different sets of research conditions to obtain parameter sub-ranges corresponding to each set of research conditions; the target user feature is one of multiple user features; according to the parameter sub-ranges corresponding to each set of research conditions and the feature samples corresponding to each virtual research object and the target user feature, determine the parameter sub-range to which each virtual research object belongs; according to the parameter sub-range to which each virtual research object belongs, determine the set of research conditions corresponding to each virtual research object to obtain virtual research object groups corresponding to each set of research conditions.

[0139] Optionally, the target user feature is one of the multiple user features included in the portrait data. Specifically, the target user feature can be a random one of the multiple user features or a user-specified user feature.

[0140] In some embodiments, multiple virtual research objects included in the same virtual research object group respectively belong to the same parameter sub-range; the number of virtual research objects included in different virtual research object groups can be equal or not equal.

[0141] Exemplarily, assume that there are 3 sets of research conditions (set a, set b, and set c), the target user characteristic is "age", and the corresponding parameter range is 20 to 35 years old; therefore, the electronic device can divide the parameter range corresponding to the target user characteristic "age" according to the three different sets of research conditions, and obtain that the parameter sub-range corresponding to set a is 20 to 25 years old, the parameter sub-range corresponding to set b is 25 to 30 years old, and the parameter sub-range corresponding to set c is 30 to 35 years old; then, according to the parameter sub-range to which each virtual research object belongs, determine the set of research conditions corresponding to each virtual research object, and obtain the group of virtual research objects corresponding to each set of research conditions. Assume that there are 40 virtual research objects aged 20 to 25 years old, 130 virtual research objects aged 25 to 30 years old, and 30 virtual research objects aged 30 to 35 years old. Then, the electronic device can confirm that the group of virtual research objects corresponding to set a contains 40 virtual research objects, the group of virtual research objects corresponding to set b contains 130 virtual research objects, and the group of virtual research objects corresponding to set c contains 30 virtual research objects.

[0142] Step 506: Input the first set of research conditions and the research question to each second virtual research object included in the group of virtual research objects corresponding to the first set of research conditions, and analyze the first set of research conditions and the research question through each second virtual research object, and output the research reply content corresponding to the research question. The first set of research conditions is any set of research conditions.

[0143] For the relevant description of step 506, reference can be made to the relevant description of step 140 in the above embodiment, which will not be repeated here.

[0144] In the embodiment of the present application, the electronic device combines multiple sets of research conditions to obtain multiple different sets of research conditions, so that each set can represent a specific research perspective or scenario, enabling the research to cover a wider range of situations and details, thereby improving the comprehensiveness of the research data.

[0145] As Figure 6 shown, in one embodiment, a research data generation device 600 is provided, which can be applied to the above-mentioned electronic device. The research data generation device 600 may include an acquisition module 610, a portrait module 620, a construction module 630, a research module 640, and a statistics module 650.

[0146] The acquisition module 610 is configured to acquire the input configuration data, where the configuration data includes the characteristic data of the research object and the number of research objects.

[0147] An image module 620, configured to generate image data corresponding to virtual research objects that match the number of the research objects according to the feature data of the research object; the image data is used to describe the virtual research objects.

[0148] A construction module 630, configured to generate virtual research objects that match the number of the research objects according to the image data corresponding to each of the virtual research objects through a research object generation model.

[0149] A research module 640, configured to input the content to be researched into each of the generated virtual research objects, analyze the content to be researched through each of the virtual research objects, and output the research response content corresponding to the content to be researched.

[0150] A statistics module 650, configured to perform statistical analysis on the research response content output by each of the virtual research objects to obtain a target research result.

[0151] Optionally, the feature data includes multiple user features and the parameter ranges corresponding to each of the user features.

[0152] In some embodiments, the image module 620 is further configured to generate multiple feature samples corresponding to each of the user features based on a truncated normal distribution according to the parameter ranges corresponding to each of the user features; wherein, the number of feature samples corresponding to the same user feature is the same as the number of the research objects; combine the multiple feature samples corresponding to each of the user features to obtain the image data corresponding to the virtual research objects that match the number of the research objects, and the image data includes one feature sample corresponding to each of the multiple user features.

[0153] Optionally, the image module 620 is further configured to calculate a target average value and a target standard deviation corresponding to the first user feature according to a parameter range corresponding to the first user feature; the first user feature is any user feature; generate a plurality of first feature samples corresponding to the first user feature based on a truncated normal distribution according to the parameter range corresponding to the first user feature; calculate a first average value and a first standard deviation according to the current plurality of first feature samples corresponding to the first user feature; calculate a deviation value according to a first difference between the first average value and the target average value and a second difference between the first standard deviation and the target standard deviation; if the deviation value is greater than or equal to a deviation threshold, adjust the N first feature samples according to an optimization function to obtain a new plurality of first feature samples corresponding to the first user feature, increment N by one, and re-execute the step of calculating the first average value and the first standard deviation according to the current plurality of first feature samples corresponding to the first user feature until N is equal to an iteration threshold; N is a positive integer; if the deviation value is less than the deviation threshold, use the current plurality of first feature samples as the plurality of feature samples corresponding to the first user feature.

[0154] In some embodiments, the research data generation device 600 further includes a behavior evaluation module and a decision evaluation module.

[0155] The behavior evaluation module is configured to obtain behavior sample research content, where the behavior sample research content includes a sample research scenario and a behavior research question, and the behavior research question is an open-ended question; obtain sample behavior content that should be output after the first virtual research object analyzes the behavior sample research content according to the portrait data of the first virtual research object; the first virtual research object is any virtual research object; input the sample research scenario and the behavior research question into the first virtual research object, and have the first virtual research object analyze the sample research scenario and the behavior research question to output research behavior content corresponding to the behavior sample research content; calculate a semantic similarity between the research behavior content and the sample behavior content; if the semantic similarity is less than a similarity threshold, generate a new first virtual research object according to the portrait data of the first virtual research object through the research object generation model.

[0156] A decision evaluation module, configured to obtain the content of a decision sample survey, where the content of the decision sample survey includes a sample survey scenario and a decision survey question, and the decision survey question belongs to a closed question; obtain the sample options that should be output after the first virtual survey object analyzes the content of the decision sample survey according to the portrait data corresponding to the first virtual survey object; the first virtual survey object is any virtual survey object; input the sample survey scenario and the decision survey question into the first virtual survey object, and through the first virtual survey object analyzing the sample survey scenario and the decision survey question, output the survey options corresponding to the content of the decision sample survey; if the survey options are different from the sample options, then generate a new first virtual survey object according to the portrait data corresponding to the first virtual survey object through the survey object generation model.

[0157] Optionally, the content to be surveyed includes survey questions and multiple survey conditions.

[0158] In some embodiments, the survey data generation device 600 further includes a grouping module.

[0159] The grouping module is configured to combine the multiple survey conditions to obtain multiple different sets of survey conditions; each set of survey conditions includes at least one survey condition; group the virtual survey objects matching the number of survey objects according to the multiple different sets of survey conditions to obtain virtual survey object groups corresponding to each set of survey conditions.

[0160] Optionally, the survey module 640 is further configured to input the first set of survey conditions and the survey question to each second virtual survey object included in the virtual survey object group corresponding to the first set of survey conditions, and through each second virtual survey object analyzing the first set of survey conditions and the survey question, output the survey response content corresponding to the survey question; the first set of survey conditions is any set of survey conditions.

[0161] Optionally, the portrait data includes a feature sample corresponding to each of multiple user features.

[0162] In some embodiments, the grouping module is further configured to divide the parameter range corresponding to the target user feature according to the multiple different sets of research conditions, so as to obtain parameter sub-ranges corresponding to each set of research conditions; the target user feature is one of the multiple user features; determine the parameter sub-range to which each virtual research object belongs according to the parameter sub-ranges corresponding to each set of research conditions and the feature samples corresponding to the target user feature of each virtual research object; determine the set of research conditions corresponding to each virtual research object according to the parameter sub-range to which each virtual research object belongs, so as to obtain a group of virtual research objects corresponding to each set of research conditions.

[0163] In the embodiments of the present application, the electronic device generates portrait data of virtual research objects that matches the required quantity through configuration data. The portrait data can accurately describe each virtual research object, making the virtual research objects generated through the portrait data more realistic. When inputting the content to be surveyed into these virtual research objects, it can simulate the analysis and reaction process of real people and output accurate survey response content to improve the accuracy rate; moreover, each virtual research object can analyze the content to be surveyed in parallel, that is, there is no need to wait for the participation and feedback time of real people, thereby reducing the time cost of obtaining survey data and improving the efficiency of the survey; in addition, by generating survey response content through virtual research objects, there is no need to recruit, motivate, and manage the participation of real people, reducing the economic cost consumed by the entire survey, thus realizing the efficient and low-cost generation of survey data.

[0164] Figure 7 It is a structural block diagram of an electronic device in an embodiment. The electronic device can be a device such as a mobile phone, a tablet computer, or a smart wearable device. As Figure 7 shown, the electronic device 700 may include one or more of the following components: a processor 710, and a memory 720 coupled to the processor 710. The memory 720 may store one or more computer programs, and the one or more computer programs may be configured to be executed by one or more processors 710 to implement the methods described in the above embodiments.

[0165] The processor 710 may include one or more processing cores. The processor 710 connects various parts within the entire electronic device 700 through various interfaces and lines. By running or executing instructions, programs, code sets, or instruction sets stored in the memory 720, and by invoking data stored in the memory 720, it performs various functions of the electronic device 700 and processes data. Optionally, the processor 710 may be implemented in at least one hardware form of digital signal processing (DSP), field-programmable gate array (FPGA), or programmable logic array (PLA). The processor 710 may integrate a combination of one or several of a central processing unit (CPU), a graphics processing unit (GPU), and a modem, etc. Among them, the CPU mainly processes the operating system, user interface, application programs, etc.; the GPU is responsible for rendering and drawing display content; the modem is used to process wireless communications. It can be understood that the above-mentioned modem may not be integrated into the processor 710 and may be implemented separately through a communication chip.

[0166] The memory 720 may include random access memory (RAM) and may also include read-only memory. The memory 720 can be used to store instructions, programs, code, code sets, or instruction sets. The memory 720 may include a program storage area and a data storage area. Among them, the program storage area may store instructions for implementing the operating system, instructions for implementing at least one function (such as touch function, sound playback function, image playback function, etc.), instructions for implementing the above-mentioned various method embodiments, etc. The data storage area may also store data created during the use of the electronic device 700.

[0167] It can be understood that the electronic device 700 may include more or fewer structural elements than those shown in the above structural block diagram. For example, it includes a power supply, input keys, a camera, a speaker, a screen, an RF (radio frequency) circuit, a Wi-Fi (wireless fidelity) module, a Bluetooth module, sensors, etc., and these are not limited here.

[0168] An embodiment of this application discloses a computer-readable storage medium that stores a computer program. When the computer program is executed by a processor, it implements the methods described in the above various embodiments.

[0169] An embodiment of the present application discloses a computer program product, including a computer program, and when the computer program is executed by a processor, it can implement the methods described in the above embodiments.

[0170] Those of ordinary skill in the art can understand that all or part of the processes in the methods of the above embodiments can be completed by instructing relevant hardware through a computer program. The program can be stored in a non-volatile computer-readable storage medium. When the program is executed, it can include the processes of the embodiments of the above methods. Among them, the storage medium can be a magnetic disk, an optical disk, a read-only memory

[0171] (Read-Only Memory, ROM), etc.

[0172] Any reference to a memory, storage, database, or other medium as used herein may include non-volatile and / or volatile memory. Suitable non-volatile memory may include ROM, programmable ROM (Programmable ROM, PROM), erasable PROM (Erasable PROM, EPROM), electrically erasable PROM (Electrically Erasable PROM, EEPROM), or flash memory. Volatile memory may include random access memory (random access memory, RAM), which is used as an external cache. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (Static RAM, SRAM), dynamic RAM (Dynamic Random Access Memory, DRAM), synchronous DRAM (synchronous DRAM, SDRAM), double data rate SDRAM

[0173] (Double Data Rate SDRAM, DDR SDRAM), enhanced SDRAM (Enhanced Synchronous DRAM, ESDRAM), synchronous link DRAM (Synchlink DRAM, SLDRAM), memory bus direct RAM (Rambus DRAM, RDRAM), and direct memory bus dynamic RAM (Direct Rambus DRAM, DRDRAM).

[0174] It should be understood that the "one embodiment" or "an embodiment" mentioned throughout the specification means that the specific features, structures or characteristics related to the embodiment are included in at least one embodiment of the present application. Therefore, the appearances of "in one embodiment" or "in an embodiment" throughout the specification do not necessarily refer to the same embodiment. In addition, these specific features, structures or characteristics can be combined in one or more embodiments in any suitable manner. Those skilled in the art should also be aware that the embodiments described in the specification are all optional embodiments, and the actions and modules involved are not necessarily essential to the present application.

[0175] In various embodiments of the present application, it should be understood that the magnitudes of the sequence numbers of the above processes do not necessarily mean the inevitable sequence of execution. The execution sequence of each process should be determined by its function and internal logic, and should not constitute any limitation to the implementation process of the embodiments of the present application.

[0176] The technical features of the above embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered to be within the scope described in this specification.

[0177] The units described as separate components above may or may not be physically separated. The components shown as units may or may not be physical units, that is, they can be located in one place, or can be distributed to multiple network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0178] In addition, each functional unit in the embodiments of the present application can be integrated into one processing unit, or each unit can exist physically alone, or two or more units can be integrated into one unit. The above integrated unit can be implemented in the form of hardware or in the form of a software functional unit.

[0179] If the above integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-accessible memory. Based on this understanding, the technical solution of the present application, in essence, or the part that contributes to the prior art, or all or part of this technical solution, can be embodied in the form of a software product. This computer software product is stored in a memory and includes several instructions for causing a computer device (which can be a personal computer, a server or a network device, etc., specifically, the processor in the computer device) to execute some or all of the steps of the above methods in the various embodiments of the present application.

[0180] The above has introduced in detail a method, apparatus, electronic device, and storage medium for generating research data disclosed in the embodiments of the present application. Specific examples are used in this article to elaborate on the principle and implementation manner of the present application. The description of the above embodiments is only used to help understand the method of the present application and its core idea. At the same time, for those of ordinary skill in the art, according to the idea of the present application, there will be changes in the specific implementation manner and application scope. In summary, the content of this specification should not be construed as a limitation on the present application.

Claims

1. A method for generating survey data, characterized in that: The method comprises: Obtaining input configuration data, wherein the configuration data includes characteristic data of the survey object and the number of survey objects; Generate, based on the characteristic data of the survey object, portrait data corresponding to virtual survey objects that match the number of survey objects; the portrait data is used to describe the virtual survey objects; Generate virtual research objects matching the number of the research objects according to the portrait data corresponding to each of the virtual research objects through a research object generation model; Inputting the content to be investigated into each of the generated virtual survey subjects, analyzing the content to be investigated through each of the virtual survey subjects, and outputting survey response content corresponding to the content to be investigated; Statistical analysis is performed on the survey response content output by each of the virtual survey subjects to obtain target survey results.

2. The method according to claim 1, characterized in that The feature data includes a plurality of user features and a parameter range corresponding to each of the user features; the generating, based on the feature data of the survey subjects, portrait data corresponding to virtual survey subjects that match the number of survey subjects, includes: Based on the truncated normal distribution, according to the parameter range corresponding to each of the user characteristics, a plurality of feature samples corresponding to each of the user characteristics are generated; wherein the number of feature samples corresponding to the same user characteristic is the same as the number of the survey subjects; The multiple feature samples corresponding to each of the user features are combined to obtain portrait data corresponding to virtual research objects that match the number of research objects, wherein the portrait data includes a feature sample corresponding to each of the multiple user features.

3. The method according to claim 2, characterized in that The generating a plurality of feature samples corresponding to each of the user features based on the truncated normal distribution and according to the parameter range corresponding to each of the user features comprises: Calculating a target mean value and a target standard deviation corresponding to a first user feature according to a parameter range corresponding to the first user feature; the first user feature is any user feature; Based on a truncated normal distribution, generating a plurality of first feature samples corresponding to the first user feature according to a parameter range corresponding to the first user feature; Calculate a first average value and a first standard deviation according to a plurality of current first feature samples corresponding to the first user feature; Calculate a deviation value according to a first difference between the first average value and the target average value, and a second difference between the first standard deviation and the target standard deviation; If the deviation value is greater than or equal to the deviation threshold, adjust N first feature samples according to the optimization function to obtain multiple new first feature samples corresponding to the first user feature, add one to N, and re-execute the step of calculating the first average value and the first standard deviation according to the current multiple first feature samples corresponding to the first user feature until N is equal to the iteration threshold; N is a positive integer; If the deviation value is less than the deviation threshold, the current plurality of first feature samples are used as a plurality of feature samples corresponding to the first user feature.

4. The method according to claim 1, characterized in that: After generating virtual research subjects matching the number of research subjects according to the portrait data corresponding to each virtual research subject through the research subject generation model, the method further includes: Obtaining behavior sample survey content, wherein the behavior sample survey content includes sample survey scenarios and behavior survey questions, and the behavior survey questions are open questions; According to the portrait data corresponding to the first virtual research object, the sample behavior content that should be output after the first virtual research object analyzes the behavior sample research content is obtained; the first virtual research object is any virtual research object; Inputting the sample survey scenario and the behavior survey question into the first virtual survey object, analyzing the sample survey scenario and the behavior survey question through the first virtual survey object, and outputting survey behavior content corresponding to the behavior sample survey content; Calculating the semantic similarity between the survey behavior content and the sample behavior content; If the semantic similarity is less than the similarity threshold, a new first virtual research subject is generated according to the portrait data corresponding to the first virtual research subject by using the research subject generation model.

5. The method according to claim 1, characterized in that After generating the virtual research objects matching the number of the research objects according to the portrait data corresponding to each of the virtual research objects through the research object generation model, the method further includes: Obtaining decision sample survey content, wherein the decision sample survey content includes sample survey scenarios and decision survey questions, and the decision survey questions are closed questions; According to the portrait data corresponding to the first virtual research object, the sample options that should be output by the first virtual research object after analyzing the decision sample research content are obtained; the first virtual research object is any virtual research object; Inputting the sample survey scenario and the decision survey question into the first virtual survey object, analyzing the sample survey scenario and the decision survey question through the first virtual survey object, and outputting survey options corresponding to the decision sample survey content; If the survey option is different from the sample option, a new first virtual survey subject is generated according to the portrait data corresponding to the first virtual survey subject by using the survey subject generation model.

6. The method according to claim 1, characterized in that The content to be investigated includes questions to be investigated and multiple investigation conditions; the method further includes: Combining the multiple survey conditions to obtain multiple different survey condition sets; each of the survey condition sets includes at least one survey condition; According to the multiple different survey condition sets, virtual survey subjects matching the number of survey subjects are grouped to obtain virtual survey subject groups corresponding to the respective survey condition sets; The step of inputting the content to be investigated into each of the generated virtual investigation objects, analyzing the content to be investigated by each of the virtual investigation objects, and outputting investigation reply content corresponding to the content to be investigated includes: The first survey condition set and the question to be surveyed are input into each second virtual survey subject included in the virtual survey subject group corresponding to the first survey condition set, the first survey condition set and the question to be surveyed are analyzed by each second virtual survey subject, and survey answer content corresponding to the question to be surveyed is output; the first survey condition set is any survey condition set.

7. The method according to claim 6, characterized in that The portrait data includes a feature sample corresponding to each of the plurality of user features; the virtual survey subjects matching the number of survey subjects are grouped according to the plurality of different survey condition sets to obtain virtual survey subject groups corresponding to each of the survey condition sets, including: According to the multiple different survey condition sets, the parameter range corresponding to the target user feature is divided to obtain the parameter sub-range corresponding to each of the survey condition sets; the target user feature is one of the multiple user features; Determine the parameter sub-range to which each virtual survey object belongs according to the parameter sub-range corresponding to each survey condition set and the feature samples corresponding to each virtual survey object and the target user feature; According to the parameter sub-range to which each virtual research object belongs, a research condition set corresponding to each virtual research object is determined, and a virtual research object group corresponding to each research condition set is obtained.

8. A device for generating survey data, characterized in that: The device comprises: An acquisition module, used to acquire input configuration data, wherein the configuration data includes characteristic data of the survey object and the number of survey objects; A portrait module, used to generate portrait data corresponding to virtual research objects that match the number of research objects according to the characteristic data of the research objects; the portrait data is used to describe the virtual research objects; A construction module, configured to generate virtual research subjects matching the number of the research subjects according to the portrait data corresponding to each of the virtual research subjects through a research subject generation model; A survey module, used for inputting the content to be surveyed into each of the generated virtual survey objects, analyzing the content to be surveyed through each of the virtual survey objects, and outputting survey reply content corresponding to the content to be surveyed; The statistical module is used to perform statistical analysis on the survey response content output by each of the virtual survey subjects to obtain the target survey results.

9. An electronic device, characterized in that: The method comprises a memory and a processor, wherein a computer program is stored in the memory, and when the computer program is executed by the processor, the processor implements the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer-readable storage medium stores a computer program, and when the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images