Market investigation method based on AI big language model, construction method of simulation human instances and market investigation system

Abstract

The invention relates to cross application of an artificial intelligence technology in the field of market investigation, in particular to a technology for generating simulated human equipment by using an AI (artificial intelligence) big language model and performing dynamic interactive investigation, which comprises the following steps of: acquiring and processing original data containing user characteristic information to form a standardized, high-quality and low-noise data set; defining a human device parameter system, and based on the data set and by utilizing an AI large language model, generating a plurality of simulated human devices which have corresponding attribute characteristics and can carry out context association dialogue according to each group of human device parameters, so as to form a simulated human device library; generating an investigation questionnaire based on to-be-investigated target content, selecting a target simulation person from the simulation person library, sending the investigation questionnaire to the simulation person to simulate a real interview, and performing multiple rounds of investigation to obtain interaction data; and generating an investigation report based on the interaction data, so that a target market user group can be dynamically simulated in a large-scale and low-cost manner, and new products, new concepts and new schemes can be quickly fed back.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of artificial intelligence and market research, in particular to a market research method based on an AI large language model, a construction method of a simulated person setup instance and a market research system. BACKGROUND

[0002] Traditional market research methods mainly include: (1) questionnaire survey method: through online or offline channels to target groups to issue questionnaires, collect feedback. (2) Focus group method: invite a group of target users to discuss under the guidance of the moderator. (3) User interview method: in-depth communication with individual users to obtain detailed insights. With the development of artificial intelligence and big data technology, user portrait technology based on data mining has emerged, which can extract user feature tags from massive data and simulate a batch of simulated users (virtual users) corresponding to real users in the real world, give these simulated users the characteristics and personality of real users, and simulate the thinking characteristics and answer results of real users. Then help users to efficiently and low-costly complete the collection of research data.

[0003] However, these methods generally have the following defects: (1) high cost and long cycle: traditional interviews require recruiting real users, paying compensation, and investing a lot of manpower and time. (2) Sample bias: limited sample size and coverage range, and the results may be biased. (3) Compliance risk: collecting a large amount of personal data of real users may lead to violations.

[0004] Therefore, a market research method based on an AI large language model, a construction method of a simulated person setup instance and a market research system are provided to solve the above problems. SUMMARY

[0005] In order to solve the above-mentioned problems, the present application provides a market research method based on an AI large language model, a construction method of a simulated person setup instance and a market research system, which can overcome the defects of high cost, long cycle, sample bias and other defects of traditional market research methods, and create a dynamic, large-scale, low-cost simulation of target market user groups, which can quickly feedback new products and new concepts.

[0006] To achieve the above purpose, the present application provides a market research method based on an AI large language model, a construction method of a simulated person setup instance and a market research system, comprising the following steps: S1: obtaining and processing original data containing user feature information to form a standardized data set; S2: defining a person setup parameter system, based on the standardized data set, and using an AI large language model, generating a plurality of simulated person setup instances with corresponding attribute characteristics and capable of context-related dialogue according to each group of person setup parameters, forming a simulated person setup library; S3: generating a survey questionnaire based on the target content to be investigated, selecting a target simulated persona from the simulated persona library, and sending the survey questionnaire to the virtual persona to simulate a real interview, and conducting multiple rounds of follow-up questions to obtain interaction data; S4: generating a survey report based on the interaction data.

[0007] Preferably, S1 specifically comprises: S11: collecting original data containing user feature information through API interface, web crawler or third-party data platform; S12: preprocessing the original text data using the natural language understanding and processing technology built-in AI large language model; S13: automatically labeling the processed data, which includes automatically identifying the names, places, brand names and product names in the text, and adding interest topic labels to the text based on the text classification model to form a standardized data set.

[0008] Preferably, the preprocessing in S12 includes denoising filtering, text normalization and data standardization processing. Preferably, the persona parameters in S2 include demographic attributes and behavioral preferences.

[0009] Preferably, S2 specifically comprises: According to the persona parameter system, parameter sampling is performed from the standardized data set to generate multiple sets of persona parameter combinations that meet the preset distribution conditions; Each set of persona parameter combinations is managed and indexed with an AI large language model to generate simulated persona instances with corresponding attribute characteristics and the ability to conduct context-related dialogues, forming a simulated persona library for large-scale parallel investigation, wherein each simulated persona instance has a corresponding ID and persona parameters.

[0010] Preferably, S3 specifically comprises: S31: automatically generating a survey questionnaire based on the target content to be investigated; S32: selecting a matching simulated persona as a target simulated persona from the simulated persona library based on the investigation target; S33: sending the questions of the survey questionnaire to each target simulated persona in turn, and driving the target simulated persona to simulate a real user interview through an AI large language model, conducting contextually coherent multiple rounds of follow-up questions and answers until all questions are completed, and generating interaction data.

[0011] Preferably, S4 specifically comprises: S41: performing data aggregation and quantitative analysis on the interaction data, including statistical satisfaction, calculating net recommendation value and cross analysis; S42: Insight mining on the open question text feedback in the interaction data, including keyword extraction, topic extraction, and opinion clustering; S43: Based on the quantitative analysis and insight mining results, calling a report template to automatically generate a visual comprehensive survey report.

[0012] A market research system based on an AI large language model adopts a market research method based on an AI large language model, comprising: A data acquisition and processing module is configured to acquire and process original data containing user feature information to form a standardized data set. A persona library construction module is configured to define a persona parameter system, generate a plurality of simulated personas with corresponding attribute characteristics and context association dialogues based on the standardized data set and AI large language model according to each group of persona parameters, and form a simulated persona library. A research interaction module is configured to generate a research questionnaire based on target content to be researched, select target simulated personas from the simulated persona library, send the research questionnaire to the virtual personas to simulate real interviews, conduct multiple rounds of follow-up questions, and obtain interaction data. A report generation module is configured to generate a research report based on the interaction data.

[0013] A method for constructing a simulated persona instance, comprising: Obtaining a group of persona parameter combinations generated based on real data sampling; Inputting the persona parameter combinations into an AI large language model; Simulating a virtual individual with corresponding attributes and behavioral preferences based on the persona parameter combinations by the AI large language model to serve as an interactive simulated persona instance; Assigning a unique ID to the simulated persona instance and storing it in a simulated persona library for research task invocation.

[0014] Therefore, the present application adopts the above-mentioned market research method based on an AI large language model, the method for constructing a simulated persona instance, and the market research system, which has the following advantages: (1) The present application has cost / efficiency advantages: no real users need to be recruited, and the work that takes weeks or months in traditional methods can be completed in hours or days, greatly reducing costs.

[0015] (2) The present application can provide unbiased samples: the generated simulated personas cover a wider range, avoiding sample bias.

[0016] (3) The present application is safe and compliant: the entire process does not involve real user privacy data, avoiding data compliance risks.

[0017] The technical solutions of the present application are described in further detail below with reference to the accompanying drawings and examples. BRIEF DESCRIPTION OF DRAWINGS

[0018] Fig. 1 A flowchart of a market research method based on an AI large language model in the present application is shown. Fig. 2 A structural diagram of a market research method based on an AI large language model in an embodiment of the present application is shown. DETAILED DESCRIPTION

[0019] The following detailed description of the embodiments of the present application provided in the accompanying drawings is not intended to limit the scope of the claimed present application, but merely represents selected embodiments of the present application. All other embodiments obtained by those of ordinary skill in the art based on the embodiments in the present application without making creative efforts fall within the scope of the present application.

[0020] Unless otherwise defined, the technical terms or scientific terms used in the present application should be understood in their ordinary meaning by those of ordinary skill in the art to which the present application belongs.

[0021] The terms such as "include" or "contain" and the like used in the present application mean that the elements before the terms encompass the elements listed after the terms, and do not exclude the possibility of also encompassing other elements. The orientations or positional relationships indicated by the terms "in", "on", "upper", "lower", etc. are based on the orientations or positional relationships shown in the drawings, and are merely for the convenience of describing the present application and simplifying the description, and therefore cannot be understood as indicating or implying that the devices or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and therefore cannot be understood as limiting the present application. When the absolute position of the described object changes, the relative positional relationship may also change accordingly. In the present application, unless otherwise explicitly specified and limited, the term "attached" and the like should be understood broadly, for example, it can be fixedly connected, or it can be detachably connected, or it can be integrated; it can be directly connected, or it can be indirectly connected through an intermediate medium; it can be the internal connection of two elements or the interaction relationship between two elements. For those of ordinary skill in the art, the specific meaning of the above terms in the present application can be understood according to the specific circumstances.

[0022] EMBODIMENT A market research method based on an AI large language model, as shown in Figs. 1-2 includes the following steps: S1: Obtain raw data containing user feature information and process it to form a standardized data set; S1 specifically includes: S11: Collect raw data containing user feature information through API interface, web crawler or third-party data platform; S12: Preprocessing the original text data using natural language processing techniques; The preprocessing includes denoising filtering, text normalization, and data standardization processing; S13: Automatically labeling the processed data, including automatically identifying names, places, brand names, product names in the text, and adding interest topic labels such as "makeup", "childcare" based on text classification models to form a standardized data set.

[0023] S2: Define a person parameter system, based on the standardized data set, and use AI large language models to generate multiple simulated person parameters with corresponding attribute characteristics and context-related dialogues according to each group of person parameters, forming a simulated person parameter library; The person parameters in S2 include demographic attributes (gender, etc.) and behavioral preferences (consumption habits, etc.).

[0024] S2 specifically includes: According to the person parameter system, parameter sampling is performed from the standardized data set to generate multiple groups of person parameter combinations that meet the preset distribution conditions; For example: To investigate "Z generation makeup consumption in first-tier cities", the system will extract multiple parameters such as {age: 22, city: Shanghai, interest: makeup, consumption ability: medium-high, personality: willing to share}.

[0025] Each group of person parameter combinations is managed and indexed with an AI large language model to generate simulated person instances with corresponding attribute characteristics and context-related dialogues, forming a simulated person library for large-scale parallel research, where each simulated person instance has a unique ID and a complete set of corresponding person parameters.

[0026] S3: Based on the target content to be investigated, generate a survey questionnaire, select target simulated person from the simulated person library, send the survey questionnaire to the virtual person to simulate real interviews, conduct multiple rounds of follow-up questions, and obtain interaction data; S3 specifically includes: S31: Automatically generate a survey questionnaire based on the target content to be investigated; the target content to be investigated can be user-inputted investigation content.

[0027] S32: Based on the investigation target, filter matching simulated persons from the simulated person library as target simulated persons; For example, to test a high-end milk powder, only extract "housewives / husbands with 0-3 year-old children" and other person parameters.

[0028] S33: Send the questions of the survey questionnaire to each target avatar in turn, and simulate the real user interview by driving the target avatar with the AI large language model, and perform multi-round follow-up questions and answers with coherent context until all questions are completed, to generate interaction data.

[0029] S4: Generate a survey report based on the interaction data.

[0030] S4 specifically includes: S41: Data aggregation and quantitative analysis of the interaction data, including satisfaction statistics, net recommendation value calculation, and cross analysis; S42: Insight mining on open-ended text feedback in the interaction data, including keyword extraction, theme extraction, and opinion clustering; S43: Based on the results of quantitative analysis and insight mining, call a report template to automatically generate a visual comprehensive survey report.

[0031] Embodiment 1 A market research system based on an AI large language model, which adopts a market research method based on an AI large language model, including: A data collection and processing module for obtaining and processing original data containing user feature information to form a standardized data set; A persona library construction module for defining a persona parameter system, generating a plurality of simulated personas with corresponding attribute characteristics and context-related dialogues based on the standardized data set and using an AI large language model according to each group of persona parameters, and forming a simulated persona library; A research interaction module for generating a survey questionnaire based on target content to be researched, selecting a target simulated persona from the simulated persona library, sending the survey questionnaire to the virtual persona to simulate a real interview, and performing multi-round follow-up questions to obtain interaction data; A report generation module for generating a survey report based on the interaction data.

[0032] Embodiment 2 A construction method of a simulated persona instance, including: Obtain a set of persona parameter combinations generated based on real data sampling; Input the persona parameter combinations into an AI large language model; Simulate a virtual individual with corresponding attributes and behavioral preferences based on the persona parameter combinations by the AI large language model, as an interactive simulated persona instance; Assign a unique ID to the simulated persona instance and store it in a simulated persona library for research task calling.

[0033] Therefore, the application adopts the market research method based on the AI large language model, the construction method of the simulated person instance, and the market research system, and by using the AI large language model, the static user portrait parameters are changed into the simulated person instance with dynamic and interactive characteristics. The construction method of “data-> parameter-> person instance” is adopted to ensure the diversity and authenticity of the simulated person instance. The market research task is converted into the interaction process with the simulated person instance library, the real users do not need to be recruited, the work of the traditional method which needs weeks or months can be completed in hours or days, the cost is greatly reduced, the generated simulated person instance covers more comprehensively, and the sample deviation is avoided.

[0034] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application but not to limit it, although the present application has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that: the technical solutions of the present application can still be modified or replaced by the equivalent, and these modifications or equivalent replacements cannot make the modified technical solutions deviate from the spirit and scope of the technical solutions of the present application.

Claims

1. A market research method based on an AI large language model, characterized in that, Includes the following steps: S1: Obtain and process raw data containing user feature information to form a standardized dataset; S2: Define a character parameter system, based on the standardized data set, and use an AI large language model to generate multiple simulated characters with corresponding attribute features and capable of context-related dialogue according to each set of character parameters, forming a simulated character library; S3: Generate a survey questionnaire based on the target content to be surveyed, select the target simulated persona from the simulated persona library, send the survey questionnaire to the virtual persona to simulate a real interview, conduct multiple rounds of follow-up questions, and obtain interactive data; S4: Generate a survey report based on the interaction data.

2. The market research method based on an AI large language model as described in claim 1, characterized in that, S1 specifically includes: S11: Collect raw data containing user characteristic information through API interfaces, web crawlers, or third-party data platforms; S12: Preprocess the raw text data using natural language processing techniques; S13: Automatically label the processed data. The labeling includes automatically identifying personal names, place names, brand names, and product names in the text, and adding interest-related topic tags to the text based on a text classification model to form a standardized data set.

3. The market research method based on an AI large language model as described in claim 2, characterized in that, Preprocessing in S12 includes noise reduction filtering, text normalization, and data standardization.

4. The market research method based on an AI large language model as described in claim 3, characterized in that, The character parameters in S2 include: demographic attributes and behavioral preferences.

5. The market research method based on an AI large language model as described in claim 4, characterized in that, S2 specifically includes: Based on the aforementioned character parameter system, parameters are sampled from the standardized dataset to generate multiple sets of character parameter combinations that meet preset distribution conditions. Each set of character parameters is managed and indexed with an AI large language model to generate simulated character instances with corresponding attribute features and capable of context-related dialogue, forming a simulated character library for large-scale parallel research. Each simulated character instance has a corresponding ID and character parameters.

6. The market research method based on an AI large language model as described in claim 5, characterized in that, S3 specifically includes: S31: Automatically generate a survey questionnaire based on the target content to be surveyed; S32: Based on the research objectives, select matching simulation characters from the simulation character database as target simulation characters; S33: Send the survey questionnaire questions to each target simulation persona in sequence, and drive the target simulation persona to simulate real user interviews through an AI large language model, conduct multiple rounds of follow-up questions and answers in a coherent context until all questions are completed, and generate interactive data.

7. The market research method based on an AI large language model as described in claim 1, characterized in that, S4 specifically includes: S41: Perform data aggregation and quantitative analysis on the interaction data, including statistical satisfaction, calculation of net recommender value and cross analysis; S42: Perform insight mining on the open-ended text feedback in the interactive data, including keyword extraction, topic extraction, and opinion clustering; S43: Based on the quantitative analysis and insight mining results, call the report template to automatically generate a visualized comprehensive survey report.

8. A market research system based on an AI large language model, employing a market research method based on an AI large language model as described in any one of claims 1-7, characterized in that, include: The data acquisition and processing module is used to acquire and process raw data containing user characteristic information to form a standardized dataset; The character library construction module is used to define the character parameter system. Based on the standardized data set, and using the AI ​​large language model, it generates multiple simulated characters with corresponding attribute features and capable of context-related dialogue according to each set of character parameters, thus forming a simulated character library. The survey interaction module is used to generate a survey questionnaire based on the target content to be surveyed, select a target simulated persona from the simulated persona library, send the survey questionnaire to the virtual persona to simulate a real interview, conduct multiple rounds of follow-up questions, and obtain interactive data. The report generation module is used to generate a survey report based on the interactive data.

9. A method for constructing a simulated character instance, characterized in that, include: Obtain a set of character parameter combinations generated based on real data sampling; The aforementioned character setting parameters are input into the AI ​​large language model; The AI ​​large language model simulates and generates virtual individuals with corresponding attributes and behavioral preferences based on the combination of the character parameters, serving as interactive simulation character instances; Assign a unique ID to each simulated persona instance and store it in the simulated persona library for use in research tasks.

Images